Double Case: Agreement by Suffixaufnahme

Double Case

This page intentionally left blank

Double Case Agreement by Suffixaufnahme

EDITED BY

Frans Plank

New York Oxford OXFORD UNIVERSITY PRESS 1995

Oxford University Press Oxford New York Athens Auckland Bangkok Calcutta Cape Town Dar es Salaam Delhi Florence Hong Kong Istanbul Karachi Kuala Lumpur Madras Madrid Melbourne Mexico City Nairobi Paris Singapore Taipei Tokyo Toronto and associated companies in Berlin Ibadan

Copyright © 1995 by Oxford University Press, Inc. Published by Oxford University Press, Inc. 198 Madison Avenue, New York, New York 10016 Oxford is a registered trademark of Oxford University Press, Inc. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of Oxford University Press. Library of Congress Cataloging-in-Publication Data Double case : agreement by Suffixaufnahme / edited by Frans Plank. p. cm. Rev. papers of the Franz Nikolaus Finck Memorial Symposium, held Sept. 21-23, 1991, in Konstanz. Includes bibliographical references and indexes. ISBN 0-19-508775-5 1. Grammar, Comparative and general—Case—Congresses. 2. Grammar, Comparative and general—Agreement—Congresses. 3. Grammar, Comparative and general—Noun phrase—Congresses. 4. Typology (Linguistics)—Congresses. I. Plank, Frans. II. Franz Nikolaus Finck Memorial Symposium (1991 : Konstanz, Germany) P240.6.D68 1995 415—dc20 94-9485

246897531 Printed in the United States of America on acid-free paper

PREFACE

Cases come in paradigms but are, as it were, lone wolves when on syntactic duty. For a noun phrase, "having two cases is as bad as having none at all," and, since languages are generally well-behaved (aren't they?), "there is no evidence that this situation [in which a given noun phrase is assigned more than one case] arises in natural languages: i.e. it is typically the case that noun phrases in natural language are morphologically marked for only one case." This is textbook teaching, quoted respectively from Peter Sells's Lectures on Contemporary Syntactic Theories (Stanford: CSLI, 1985, p. 53) and Andrew Radford's Transformational Syntax (Cambridge: Cambridge University Press, 1981, p. 318); but more advanced reading is likelier to confirm than to disconfirm that this is the received wisdom in this matter. In actual fact, however, cases are not quite as unsociable. Ask a case aficionado and (s)he will oblige you with a handful or two of examples such as Evenki amut-tu-la '(in)to the lake', where a dative-locative suffix is followed by an allative-illative one, or Limbu a-nduzum-l -n-ille 'with that of my friend', my-friend-Genitive-Absolutive-Instrumental. Typically, or atypically, local cases are the ones most fond of teaming up with each other, and genitives are especially willing to host the whole lot of their paradigmatic companions. The focus of this volume is on another kind of double, or indeed triple or further case marking, though one bearing a certain resemblance to the case inflection of genitives as just illustrated from Limbu: the use of case (plus perhaps further inflectional categories, in particular number) for purposes of agreement, linking a noun already carrying a case (plus perhaps number), typically a genitive, to another nominal to which it is syntagmatically related, typically (or so it often seems) as an attribute; an example is Hurrian Tessop-peai tev-ai 'by the word of Tessob', where Tessop, the weather god, is in the genitive (-pe) and agrees with tev 'word' by copying its instrumental suffix (-ai). Known traditionally, if not widely nor entirely accurately, as Suffixaufnahme, such case (plus perhaps number) agreement of case-marked nouns is not widespread but is distributed in intriguing areal patterns across the Old World and southerly parts of the New, crying out for typological and/or genealogical explanation. Our collective ambition in this volume is to offer a comprehensive account of such double case marking due to agreement. We do not expect to be superseded in the near future. Ostensibly marginal, Suffixaufnahme in fact ramifies widely and deeply, involving several major theoretical issues in morphology and syntax. First and second, Suffixaufnahme is a challenge to theories of case and of agreement, representing an apparent oddity in either respect. Third and fourth, the differ-

vi

Preface

entiation of parts of speech and of inflection and derivation are high on the Suffixaufnahme agenda: the anomaly of this pattern depends on the caseagreeing words being nouns (or at any rate palpably nouny) rather than adjectives, and on the genitive (or other relevant case) in such case-agreeing words being inflectional rather than derivational; case-agreeing derived adjectives would hardly arouse much excitement. Fifth, grammatical relations are implicated insofar as case agreement has variously been supposed to encode attribution (whether of an ordinarily tight kind or of a loose or even extraposed kind), apposition, or (secondary or co-)predication—not exactly the bestunderstood subset of relations to begin with. Sixth, the key question here is indeed a larger one, concerning noun phrase constituency and the "depth" or "flatness" of syntax, long deemed to be a core parameter of typological variation. Conspicuous among the possible correlates of this parameter is that of agglutinative versus flective morphology, and the contribution of morphological typology itself to the milieus that are conducive or hostile to Suffixaufnahme is a seventh issue. It must have been in the mid-1970s that I was first attracted to the present topic, if perhaps subconsciously. Ergativity was the subject of a collection then in the making, and curiously, the group of languages qualifying for consideration more or less included those practicing Suffixaufnahme; however, at the time that trait seemed too inconspicuous for anybody to notice much. The Hurrian and Old Georgian classes that I had the almost exclusive privilege to sit in on at Konstanz in the mid-1980s did made me notice at last. Two things were gradually dawning on me then: Suffixaufnahme had a considerable history behind it, and one prone to repeat itself; and, if more time were to be spent on it profitably, this was best done in concert, such were its real dimensions. Planning on the collaboration whose results are now documented in this volume began in earnest in late 1989, when I suggested Suffixaufnahme as the possible topic of a conference to a few friends and acquaintances whose expertise in relevant languages surpassed mine (despite all my educational efforts). In the spring of 1990 preparatory materials were circulated more widely, and the response was encouraging, even if some addressees first needed convincing that there really was something to confer on. Generously funded by the Deutsche Forschungsgemeinschaft, the European Science Foundation (through its EUROTYP program, in which some of us are participating), the Land Baden-Wurttemberg, and the Universitat Konstanz, the Franz Nikolaus Finck Memorial Symposium could then be convened at Konstanz on 21-23 September 1991. It was so called in honor of the man primarily responsible for introducing Suffixaufnahme into the typological discourse. Strange though it may seem to the uninitiated, such was the fascination of Suffixaufnahme over those assembled, minded like the weather, most unquietly, that the daily proceedings could only with difficulty be interrupted as the wee hours were approaching. "Things that love night love not such nights as these" (Lear, III, iii). All contributions included in the present volume underwent revision in the

Prefacee

vvii

wake of this conference. (Robert Hetzron's was only commissioned afterwards.) Some chapters were completed commendably early, while others took a little more time to mature, with the irrevocably final deadline period expiring sometime in the summer of 1993. My thanks are due to all contributors for the efforts they put into this joint venture, and especially to Edith Moravcsik for summing up the proceedings on location at Konstanz and for acting as co-commentator ever after. Thanks too to Katefina Hladka and Wolfgang Schellinger for much-appreciated organizational assistance, and for sharing the blame for the occasional unspotted typo (and for a few that had, perhaps, better gone unspotted—such as abnominal, admonimal, or abdominal). We all owe a great debt of gratitude, especially heartfelt on the part of the editor, to our publisher; the cooperation with Oxford University Press (New York), at every stage of the production of this book, could not have been smoother. F. P.

This page intentionally left blank

CONTENTS

Contributors, xi Abbreviations, xiii I PROLOGUE 1. (Re-)Introducing Suffixaufnahme, 3 Frans Plank II THE ANCIENT NEAR EAST 2. Suffixaufnahme in Hurrian and Urartian, 113 Gernot Wilhelm 3. Suffixaufnahme in Hurrian: Normal Cases and Special Cases, 136 Ilse Wegner III THE CAUCASUS 4. Suffixaufnahme in Kartvelian, 151 Winfried Boeder 5. Direct-Oblique Agreement of Attributes in Daghestanian, 216 Aleksandr E. Kibrik 6. Genitives and Adjectives as Attributes in Daghestanian, 230 Ol'ga Ju. Boguslavskaja IV INDO-EUROPEAN 7. Indo-European o-Stems and Feminine Stems in -I, 243 Francisco Villar 8. Slavonic's Closest Approach to Suffixaufnahme: The Possessive Adjective, 265 Greville G. Corbett 9. Inflecting Postpositions in Indie and Kashmiri, 283 John R. Payne V CHUKCHI-KAMCHATKAN 10. Possessive and Relational Forms in Chukchi, 301 Maria Koptjevskaja-Tamm

x

Contents

VI CUSHITIC 11. Genitival Agreement in Awngi: Variation on an Afroasiatic Theme, 325 Robert Hetzron VII AUSTRALIAN 12. Suffixaufnahme and Related Case Marking Patterns in Australian Languages, 339 Fritz Schweiger 13. Double Case Marking in Kanyara and Mantharta Languages, Western Australia, 363 Peter Austin 14. Suffixaufnahme and Apparent Ellipsis in Martuthunira, 380 Alan Dench 15. Multiple Case in Kayardild: Anti-iconic Suffix Ordering and the Diachronic Filter, 396 Nick Evans VIII DIACHRONY 16. Binder-Anaphors and the Diachrony of Case Displacement, 431 Anthony Rodrigues Aristar IX EPILOGUE 17. Summing up Suffixaufnahme, 451 Edith A. Moravcsik Subject Index, 485 Language Index, 489 Author Index, 495

CONTRIBUTORS

Anthony R. Aristar Department of English Texas A&M University USA Peter Austin Department of Linguistics La Trobe University Australia

Aleksandr E. Kibrik Department of Theoretical and Applied Linguistics Moscow State University Russia . Maria Koptjevskaja-Tamm Department of Linguistics Stockholm University Sweden

Winfried Boeder Department of English University of Oldenburg Germany

Edith A Moravcsik Department of Linguistics University of Wisconsin—Milwaukee USA

Ol'ga Ju. Boguslavskaja Institute of Russian Russian Academy of Sciences, Moscow Russia

John R. Payne Department of Linguistics University of Manchester England T, n, , Frans Plank Department of Linguistics . University of Konstanz Germany Fritz Schweiger Department of Mathematics University of Salzburg Austria

Greville G. Corbett Department of Linguistic and International Studies University of Surrey England Alan Dench Centre for Linguistics University of Western Australia Australia Nick Evans Department of Linguistics and Language Studies University of Melbourne Australia Robert Hetzron Department of Germanic, Oriental, and Slavic University of California at Santa Barbara USA

Francisco Villar Department of Classical Philology and Indo-European University of Salamanca Spain Ilse Use w Wegner Institute for for Ancient Ancient Near Near Eastern Eastem Institute Studies Free University of Berlin Germany Gernot Wilhelm Institute for Oriental Philology University of Wiirzburg Germany

This page intentionally left blank

ABBREVIATIONS

A Abl Abs A.Case Ace Act Adj Adjct AdjP Adv All Anim Antipass Aor Appr Art Ass Attr Aug Avers Ben Carr Caus C.Case Class Coll Com Comp Conj Conseq Const Cont CV Dat Dct Def Dem Deriv Des

transitive-subject form ablative case absolutive case "associating" use of a case accusative case actual mood adjective; adjective marker adjectivizing marker adjective phrase adverbial case allative case animate antipassive voice aorist tense/aspect apprehensive mood article associative case attribute; attributive marker augmentative marker aversive case benefactive case carrier affix causative case; causative verb form "complementizing" use of a case classifier collective comitative case (also "having" case or proprietive) comparative degree; comparative case conjunction consequential case construct state continuous aspect connecting vowel dative case direct case or stem defmitizing marker; definite demonstrative pronoun or marker derivational marker desiderative mood

xiv

Det Dir Ditrans Don DS Du Eff Emph Equ Erg Ess Excl Ext Fern Fut Gen Genenrl Hort I11 Imp Imperf Inanim Inch Incl Ind Indef Indiv Inf Ins Intent Intrans IObj Irr Lat Lig Loc Masc Masd M.Case Mid N Neg Neut Nom Nomin NP

Abbreviations

determiner directional case ditransitive verbal donative case different-subject form dual number effector case emphatic equative case ergative case essive case exclusive extended intransitive feminine gender future tense genitive case general genitive case hortative mood illative case imperative mood imperfect tense/aspect inanimate inchoative aspect/aktionsart inclusive indicative mood indefinite individualizing marker infinitive instrumental case intentive mood intransite indirect object irrealis mood lative case ligative/ligature locative case masculine gender masdar "modal" use of a case middle voice noun negation neuter gender nominative case nominalizing marker noun phrase

Abbreviations

Obj Obl Oper Ord Orient Orig P Pass Per Perf P1 Poss Postp Pot PP Pre Prep Pres Pret prim Priv Pro Prom Prop Prtcpl Purp Qu Quot R RC Redup Ref Rel Res R1t Rltn S Sbj sec Sg Spec SS Stat Subj T.Case

object case oblique case or stem operative case ordinal-numeral marker orientational marker origin case object (patient) form passive voice pergressive/perlative case perfect tense/aspect plural number possessive pronominal marker; possessive (case or derivational) form postposition potential mood/tense prepositional/postpositional phrase prepositive case preposition present tense preterite tense marker of "primary" relation privative (or "lacking") case pronominal marker prominence marker proprietive case participle purposive (noun or verb) form question marker quotative restrictivity marker root-complement reduplication referential article relative-clause marker resultative nominalization relator (article) relational (case or derivational) form sentence; intransitive-subject form subject (case) marker of "secondary" relation singular number specific same-subject form stative subjunctive mood "T-complementizing" use of a case

xv

xvi Top Trans Transl Util Usit V Voc 1/2/3 I/II/III/IV. . . -

+ =

Abbreviations topic marker transitive translative case utilitive case usitative (or past-habitual) tense/aspect verb vocative case lst/2nd/3rd person noun classes morpheme boundary (unless stated otherwise) clitic boundary components in gloss not segmented or segmentable in example (unless stated otherwise) Juxtaposition of category labels without intervening blank or boundary symbol (e.g. NomP1, 2Sg) indicates cumulative exponence of the respective categories.

I PROLOGUE

This page intentionally left blank

1 (Re-)Introducing Suffixaufnahme Frans PJank

1. Avant-Propos Although not a household term (yet), Suffixaufnahme has a typological record that is astonishingly long for a rare and seemingly marginal phenomenon. This unsung record merits retracing. As will emerge from the historical narrative to follow (Section 2), the morphological and syntactic issues that Suffixaufnahme has now and then been seen to raise are not really so marginal, and indeed are issues that contemporary typology is increasingly recognizing as formidable challenges. They involve the limits on case marking and on agreement; the word-class distinction of nouns and adjectives, the difference between inflection and derivation, and the status of genitives and other possessive forms in these two respects; the nature of attribution; noun-phrase constituency and, generally, the depth or flatness of syntax; and systematic interdependencies among all kinds of further structural features. In the remaining sections the approach will be systematic rather than historical, presenting the comparative anatomy of Suffixaufnahme in broad outline, to be fleshed out in the rest of this volume. 2. A Double Case History

2.1. A Genitive Vanishes: Bopp The first comparatist to take note of Suffixaufnahme, not yet so christened, was none other than Franz Bopp (1791-1876). On 11 December 1842, addressing the Berlin Academy of Sciences, Bopp saw what he could do to establish the Sanskritic pedigree, as was his custom, of Georgian and its "Iberian" relatives. Finding much that served his purpose, he was somewhat taken aback by the curious habit of Georgian genitives to add, as a rule, the case and perhaps the number ending of their governing noun, as exemplified in a few examples quoted from his source (1848: 275).1

4

Prologue

(1) a. gwam-isa krist-es-isa body-Gen Christ-Gen-Gen 'of the body of Christ' b. cqoba-sa mter-ta-sa attack-Dat enemy-OblPl-Dat 'at the attack of the enemies' c. qeli-ta mocikul-ta-ta hand-OblPl apostle-Ob1P1-Ob1P1 'through the hands of the apostles' d. cinamsrbol-n-i laskar-ta-n-i forerunner-Pl-Nom army-Ob1P1-Pl-Nom 'the vanguard of the armies'

Bopp noted that this pattern, visible of course only when heads were not in the nominative (or absolutive) singular and thus lacking an overt ending, was peculiar to genitives following their heads rather than to those preceding them. What is here glossed as 'Ob1[ique] P1[ural]' was nothing to worry about; Indo-European too knew paradigms with such syncretic rather than specifically genitive forms. Outlandish though it looked on the face of it, this Georgian phenomenon could upon reflection easily be squared with sound Indo-European practice. What gave initial offense was that it was a noun, of all parts of speech, that agreed with its head in case and sometimes also in number; in the Boppian universe such agreement behavior was rather something for adjectives to engage in. But Bopp was not at a loss for ways and means of IndoEuropeanizing the Georgian Sprachgeist: he declared the attributive nouns agreeing in case (and number) really to be possessive adjectives, and the genitive accordingly to be derivational rather than inflectional. Bopp did not bother to adduce any independent evidence in support of the allegedly adjectival nature of agreeing genitives. It was not inconsistent with his adjective theory that agreeing genitives literally copied the actual endings of their heads, whereas in Indo-European languages adjectival agreement inflection could be formally distinct from the inflection of the corresponding nouns; in Georgian adjectives inflected just like nouns in the first place. But when Bopp went on to observe that attributive adjectives in Georgian could either precede or follow their heads, agreeing with them or failing to inflect in either position, he might have remembered that genitives, by contrast, tended not to agree when prenominal and to agree when postnominal. Another unresolved issue was whether the genitive was also supposed to be deriving adjectives from nouns in other than attributive functions, as in the case of the governing noun in (1a), differing in this respect from other cases used to encode arguments of verbs or adpositions. Disregarding such difficulties, the adjective theory enabled Bopp, who was wont to adopt a diachronic perspective, to assume that Georgian-style attribu-

(Re-)Introducing Suffixaufnahmmee

5

tion could have evolved from Indo-European origins, both structurally and formally (compare the Sanskrit genitive desinence -sya with Georgian -(i)sa!), and that it had stayed within the limits of what was Indo-Europeanly possible. The scenario that he envisaged had genitival case inflection reinterpreted as a suffix deriving possessive adjectives from nouns, after which metamorphosis these attributes could agree in case and number with their preceding head nouns. What Georgian had lost in the process, at least in postnominal position, were straight inflectional genetives, something retained by the offspring of Sanskrit. It was no concern of Bopp's whether the fate of the genitive in Georgian was perhaps tied up with any other twists of the Iberian branch as it diverged from the Sanskritic stem. Systematic interdependences of structural traits were something for typologists to ponder—only for quite a while there was none to ponder over case-agreeing genitives in particular. And then came Finck.

2.2. The Group-inflecting Haupttyp: Finck Throughout his life, cut short by a heart attack (or, as rumor had it, a duel), Franz Nikolaus Finck (1867-1910) appears to have been torn between the prosaic and the poetic, as if unable to reconcile his businessman father's stern realism with his beloved mother's genteel sensibilities and fancies. Young Finck pursued a military career, but after five years of active service he abandoned it to publish poetry and a plea for a new ethics. Taking up jurisprudence at university, he soon turned to the study of language. He trained as an Indo-Europeanist and specialized in historical Balto-Slavic accentology and Armenian philology, but he would not be confined to what was tried and tested, ancient and Aryan. He went on to write grammars of Aran Irish, Eastern Armenian, and the German and Armenian dialects of Romany, some including dictionaries and all researched in the field, as well as a batch of articles on the syntax of Samoan, Greenlandic, Georgian and other Caucasian tongues, and to take stock of the genetic affiliations within the Bantu and the Polynesian families, and indeed globally—all within a decade. Wide though they ranged, his descriptive and historical-comparative productions were impeccably sober, conscientious, and austere. At the same time, his most cherished ambition was to advance typology, and he was not the first nor the last seeker of system to succumb to the temptation of giving free rein to his imagination and indeed prejudice, tarnishing the reputation of the typological enterprise as well as his own. Finck's frame of reference here was the Humboldtian tradition, especially as lately developed in Franz Misteli's revision of Heymann Steinthal's Charakteristik der hauptsdchlichsten Typen des Sprachbaues (1860/1893) and in James Byrne's awesomely (or awfully) idiosyncratic General Principles of the Structure of Language (1885/1892). Notably in his Der deutsche sprachbau ah ausdruck deutscher Weltanschauung (1899), extolling his native sprachbau and Weltanschauung to the skies, Finck came up with a grand system of all kinds of

6

Prologue

structural traits, including most importantly the subjectivity of the verb,2 that were supposedly motivated by racial temperaments and other psychological predispositions to do with the dominance of cognition and sensation over emotion and with mental excitability. However, the last typological, or "characterological," work that he was able to complete saw him return from such wild and sometimes ludicrous speculation to paying regard to plain structural facts in their own right, albeit facts selected a trifle impressionistically.3 Die Haupttypen des Sprachbaus (1910), only about 150 pages long, offered descriptive sketches of eight living languages, intended to represent the gamut of structural types that could be distinguished in Finck's classificatory scheme, based on the mental operations of analysis and synthesis. In terminology partly reminiscent of Steinthal and Misteli, the eight types were referred to as root-isolating, incorporating, juxtaposing (anreihend), subordinating, stemisolating, root-inflecting, stem-inflecting, and group-inflecting; Mandarin Chinese, Greenlandic Eskimo, Subya (South Western Bantu), Turkish (Osmanli), Samoan, Egyptian Arabic, Modern Greek (Dhimotiki), and Georgian were chosen as their respective representatives on account of their supposed typological purity. In light of the mental operation of analysis, the first of Finck's two overall dimensions of variation, typological diversity was a question of the extent to which the perceptual and cognitive complexes expressed by sentences were analytically fragmented, with words being regarded as the units expressing the fragments. The words of Greenlandic and Subya accordingly represented the opposite extremes of maximal and minimal semantic comprehensiveness, with those of Turkish, Georgian, Arabic, Mandarin (representing the ideal state), Greek, and Samoan being increasingly fragmentary intermediates. In regard to synthesis, the second typological dimension, the differences involved the extent to which the basic fragments were related to one another in their recombination, as manifested by overt connective forms (whose variety supplied Finck's preferred names for his eight types). Finck's most isolating specimens were Mandarin and—a little less so, owing to lexical stem formatives— Samoan; Turkish, Greenlandic, and Subya were intermediate, intimating syntagmatic interrelatedness by relatively loosely-attached grammatical morphology; and the apogee of grammaticalized interrelating was reached with Arabic, Greek, and Georgian, where inflections were tightly fused with roots, stems, and word-groups (i.e., phrases), respectively. However plausible or otherwise, this two-pronged typology of Finck's was the first to recognize, if not unequivocally, group-inflection as the hallmark of a separate type. Owing to Finck's choice of Georgian as the language illustrating this type, the structural trait that he named Suffixaufnahmmee (1910: 141)4 was destined to make its reappearance, this time within the typological rather than the historical-comparative discourse.5 Focusing on the traits that were allegedly most characteristic, Finck was struck by what seemed to him a most peculiar mode of encoding attributive constructions that had been prevalent in Georgian up to the eighteenth century but had since been declining. The short illustrative narrative which fol-

(Re-)lntroducing Suffixaufnahmmee

1

lowed his grammar sketch actually provided an appropriate example (with the relevant part in boldface):6 (2) ra turpa prinvelia, camoi axa ert-nia bavsv-ta-gan-ma what wonderful bird=is, exclaimed one-Erg child-Ob1P1-of-Erg ' "What a wonderful bird!" exclaimed one of the children'

As in some of Bopp's examples, the attributive noun here carried a cumulative case and number suffix, -ta Oblique Plural, comparable to an IndoEuropean-style genitive plural, which might have seemed to suffice to identify this noun as an attribute (although perhaps not quite unmistakably, owing to its syncretic nature as a general oblique rather than a specifically genitive form). The attribute further carried an enclitic postposition with ablative meaning, gan, clearing up all doubts about the grammatical relation borne by its host. Remarkably, however, the attribute further carried an ergative case suffix, -ma, which was not really justified by the grammatical relation of the attribute itself, since the ergative was normally reserved for transitive and some intransitive subjects. Of course the ergative was relationally appropriate for the head of this attribute, ert, functioning as subject of the verb 'exclaim'; and ert did indeed show ergative marking. The ergative suffix on the attribute thus could only be made sense of as being a copy of that found on the head, repeating on the attributive constituent the relational marking of the phrase containing it and located on the head of that phrase. Finck suggested the term "Suffixaufnahme" for such uniform case marking on heads and attributive nouns already carrying genitival and perhaps further attributive relational marking. What Finck did not care to comment on were two further examples in his sample text where Suffixaufnahme might have been expected to apply but did not: (3) ert-s didi kac-is ezo-si one-Dat great man-Gen yard-Loc 'inside a yard of a rich man' (4) saxl-is patron-ma house-Gen mistress-Erg 'the mistress of the house'

In the first, the relational marker of the head, -si(na), was arguably still more an enclitic postposition than a case suffix (although it governed the dative only on the numeral, or indefinite article, with the noun itself appearing in the bare stem form, as with genuine case suffixes), and its at least partly preserved postpositional status perhaps exempted it from Aufnahme. In the second, however, the head carried what was unquestionably a genuine case, and the very same as in (2), but the attribute's genitive here was not followed by a copy of the head's ergative.

8

Prologue

Attentive readers could thus have inferred that Suffixaufnahme was merely optional at the stage of Georgian illustrated by Finck's narrative, with this originally Old Georgian pattern perhaps already in a state of decline. They were in no position, however, to be certain whether genitives with and without copies of the case of their heads were alternating randomly or systematically, perhaps depending on the relative order of head and attribute, as Franz Bopp had observed some time before. Remarkable though Suffixaufnahme seemed to Finck, he did not think it strange that he encountered it in Georgian, of all languages. His reason was not that he considered Georgian to be at root Indo-European; being groupinflecting, Georgian was precisely the type of language in which to expect it, if anywhere. This at least was Finck's conviction. Apparently the extent of analytic fragmentation was not thought to be a direct influence on Suffixaufnahme. It was not the parameter of analysis but that of synthesis, the manner of recombining basic units, that would or would not create the conditions for Suffixaufnahme to blossom, however modestly and transiently. Parts that were syntagmatically combined had to be overtly interrelated, which ruled out the isolating types as possible Suffixaufnahme territory. For the three types where parts were interrelated by more or less loosely attached grammatical morphology (subordinating, incorporating, or juxtaposing), it would not have been logically impossible to connect heads and nominal attributes in the manner of Suffixaufnahme. Indeed, all three of Finck's type specimens (Turkish, Greenlandic, and Subya) did show agreement of one kind or another, most extensively the Bantu representative; two knew morphological cases, including an attributive one (Subya did not); all distinguished, at least rudimentarily, a word class of nouns (not very well delimited in Turkish, though, and subsuming what was elsewhere expressed by adjectives in Subya); and all had attributive constructions (with those of Subya rather reminding Finck of appositions). The three inflecting types, fusing grammatical relationship indicators more tightly with basic building blocks, likewise met all these logical conditions; but in Finck's opinion, it was the type that had word-groups or phrases rather than roots or stems as the units interrelated by inflections that was alone destined to let Suffixaufnahme develop. His reasoning here was not exactly straightforward. The essence of groupinflection was that the elements which formally attracted inflections did so by virtue of being part of a potentially complex group of meaningful elements, which in its entirety was to be syntagmatically interrelated with other such groups by the inflections. By implication, entire groups ought to be inflected only once, rather than inflections being distributed over their components by agreement; but this implication was not highlighted by Finck. What he emphasized instead was that, unlike the situation in root- and stem-inflection where roots or stems plus inflections were tightly-knit words secondarily also forming phrases, in group-inflection words were at best embryonic units, with the groups themselves as the morphosyntactic primes. The internal cohesion of such groups was not at all tight, permitting diverse interpolations. Suffix-

(Re-)Introducing Suffixaufnahmmee

9

aufnahme, then, was but one manifestation of this looseness of phraseinternal combination. In (2), the connection between the two parts of the subject phrase, ert 'one' and bavsv-ta-gan 'of the children', was sufficiently loose for both of them to merit being related to the rest of the clause individually, by receiving identical subject case marking. Further supposed manifestations were (a) the ability of nouns in the genitive to take further case marking and to function as NPs in their own right (as in tqav-isa-s hide-Gen-Dat 'from that of the [horse's] hide'); (b) the ongoing process of attaching loose relational appendages more tightly, i.e., of transforming postpositions into case suffixes (as exemplified by -gan and -si(na) in (2) and (3)); and (c) the possibility of separating case suffixes from noun bases by a plural suffix (with Georgian, however, cumulating rather than separating the expression of plural and of a subset of its cases). Finck's justification of such interdependencies was, to say the least, less than conclusive. On the more obvious interpretation of the principle of groupinflection, Georgian was not its purest possible representative, insofar as adjectives and articles could agree with their nouns in case and number, with more than one component of NPs thus inflecting for these categories. Subordinating Turkish would have fitted the bill much better, as can hardly have escaped Finck. In fact, Suffixaufnahme itself is counter to this principle thus interpreted, since both the head and the attributive noun are case-marked for the relation of what would seem to constitute one phrase; (2'), with subject marking only on the final constituent, in this case a genitive, would correspond better to our current notion of phrase-inflection. (2')

. . . camoi axa ert bavsv-ta-ma . . . exclaimed one child-OblPl-Erg

Something else must have been foremost in Finck's mind, then, when he claimed Suffixaufnahme for the group-inflecting type. He appears to have assumed that what ostensibly was a single if complex NP really was not one in Georgian, at least not in the sense of an attributive construction as close-knit as those of root- or stem-inflecting types. If what seemed to be attributive nouns, and perhaps also adjectives and determiners, really formed phrases and indeed referential units of their own that were either independently related to the rest of the clause in essentially the same way as their coreferential nouns were, or were in some kind of loose apposition with these nouns, it was not surprising that they took the same relational marking as these nouns themselves. This made Suffixaufnahme, and perhaps other repeated relational marking in Georgian, less an instance of genuine case agreement in root- or stem-inflecting style than of the same case more or less independently assigned twice. This reading of Finck raises the question of what typological significance Suffixaufnahme can have had for him. With what other structural traits could it reasonably be supposed to have been linked? The identical case marking of different but referentially-related phrases

10

Prologue

that was termed Suffixaufnahme presupposed that nominal phrases had what might be called a flat or nonconfigurational syntax; and such syntax, rather than phrase-inflection in the more obvious sense, apparently seemed to Finck to be a pervasive characteristic of Georgian.7 It was a characteristic, though, that he had also perceived in juxtaposing Subya. On Finck's implicit assumptions, this Bantu language might have shown Suffixaufnahme too, if only it had been endowed with cases. Finck's conjecture that hypostasis formations—i.e., nouns in the genitive undergoing further case marking and acting as independent NPs—were a trait akin to Suffixaufnahme, differing from it in that there is no overt NP with identical case marking present (schematically '[inside a yard] inside that of a rich man'), was not implausible in view of his idea about the independence of phrases undergoing Suffixaufnahme. He probably assumed that Suffixaufnahme implies the possibility of hypostasis and vice versa. Finck's two further correlates of Suffixaufnahme—relational markers with a status intermediate between those of postpositions and case suffixes, and a predominance of suffixes not cumulating case with number—were not predicated on the syntactic relationship between nominal phrases. Instead, they were regarded as indicators of the looseness of the connection between inflected elements and their inflections. These two traits, therefore, ought to have been considered characteristics setting apart the subordinating, incorporating, and juxtaposing types as a group from the inflecting types, and they actually were so treated in some of Finck's relevant sketches. But no conclusions were drawn from the fact that of the three inflecting specimens, it was group-inflecting Georgian that resembled these non-inflecting or agglutinative types most closely in this respect. Finck's scheme effectively had no principled reasons to offer as to why Suffixaufnahme should have anything to do with agglutinative morphology—implying it, or being implied by it, or both. In Finck's eight grammar sketches a host of further structural traits made their appearance—in that of Georgian, for example, the availability of distinct syntactic frames for verbs of action and of experience, occasioning a partly ergative alignment of core actants.8 Some such traits were only tenuously related to the basic parameters of analysis and synthesis, and the characterological method in general was not one to encourage the rigorous examination of the alleged clustering of traits, however plausible, across a wide range of languages. As for Suffixaufnahme, if it was interdependent with anything, as was Finck's tenet, this had not yet been established by induction. As far as Die Haupttypen des Sprachbaus was concerned, Suffixaufnahme might at worst have been a curiosity unique to (Old) Georgian, not systematically related to anything else and without any wider significance. Strangely, the adjective theory of so distinguished a predecessor as Franz Bopp was passed over in complete silence. Perhaps it impressed Finck as gratuitously circular. More likely it was inconsistent with his alternative way of explaining away attributive nouns agreeing in case (and number) with their heads through his theory of group-inflection and flat syntax. If Suffixauf-

(Re-)Introducing Suffixaufnahmeme

11

nahme was an ordinary instance of attributive adjectives agreeing with their head nouns, the stem- and perhaps the root-inflecting types would have been its natural home, not group-inflecting (Old) Georgian. Besides, Finck knew how to tell an adjective from a noun, didn't he? In Romany, whose German dialect he spoke reasonably well and described so that others would be able to learn it too (few did, through no fault of Finck's), the attributive forms of nouns were adjectives (1903: 17, 37), like those of Old Georgian in Bopp's view. For Finck, the suffix -kar-j-gar- added to oblique stems of nouns was derivational rather than being a genitive case inflection, although this was how it had often been labeled. The adjectives derived by means of this suffix agreed with their heads in gender (compare (5a) and (5b)), number (compare (5b) and (5c)), and directness/obliqueness (compare (5a) and (5d)), as did regular adjectives. (5) a. i tsaw-es-kar-i dai the=DctSgFem boyMasc-Ob1Sg-Adjct-DctSgFem motherFem (NomSg) 'the boy's mother' b. o tsaw-es-kar-o dad the=DctSgMasc boyMasc-Ob1Sg-Adjct-DctSgMasc fatherMasc (NomSg) 'the boy's father' c. i tsaw-en-g r-e dad-a the=DctP!Masc boyMasc-Ob1P1-Adjct-DctPlMasc fatherMasc-DctP1 'the boy's fathers' (also 'the boys'/children's fathers', see below) d. i tsaw-es-k r-a daj-a the=Ob1SgFem boyMasc-Ob1Sg-Adjct-DctSgFem motherFem-OblSg 'the boy's mother' (object)

But, although Finck did not elaborate any more than Bopp had done, the fact that these attributes agreed with their heads presumably was not his only reason for being so categorical about their not being nouns in the genitive. Independent evidence was provided by definite articles: they belonged with the head nouns, with which they agreed, subject to certain neutralizations, not with the attributes, despite the definiteness of these (or so Finck's translations of his examples suggested); and if the attributes had been genitives, one might have expected them to definitize their heads (as in the translations), rendering definite articles superfluous. Moreover, unlike nouns, such attributes could not be modified by an adjective in regular syntactic construction; instead, modifying adjectives were, in invariabe form, incorporated by them as the first part of compounds (1903: 37): (6)

a. O bar-o tsawo the=DctSgMasc big-DctSgMasc boyMasc (NomSg) 'the big boy'

12

Prologue b. i bare-tsaw-es-k r-i dai the=DctSgFem big-boyMasc-OblSg-Adjct-DctSgFem motherFem 'the big boy's mother'

On the other hand, was Finck not struck by the formal parallelism between purportedly derivational -k r-/-g r- and the dative, ablative, instrumental, and prepositional cases? Like denominal attributes, they were all formed from singular or plural oblique stems (cf. dative tsaw-es-ke, tsaw-en-ge). Moreover, like these case-inflected nouns, and unlike denominal adjectives in languages such as German which could merely agree in number, the denominal attributes could distinguish inherent number by virtue of separate oblique suffixes for singular and plural—at least when their heads were singular. Thus, compare (5a) with (7), where agreement number is the same (singular, final suffix), but inherent number differs. (7) i tsaw-en-g r-i dai the=DctSgFem boy-Ob1P1-Adjct-DctSgFem motherFem (NomSg) 'the boys'/children's mother'

The limitation of inherent number distinction was that with plural heads the oblique suffix of the attribute was obligatorily plural (1903: 27), as seen in (5c), where one might have expected the contrastive singular form tsaw-es-kar-e.9 In the German dialect of Romany, as codified by Finck and taking into account only the information provided in Finck's grammar, the distinction between an adjective and a genitival noun thus was not quite as clearcut as Finck (and Bopp) would have it. Plainly, attributes in -k r-/-g r- shared some properties with bona fide adjectives, but also at least two (or one and a half) with what seemed to be case-inflected nouns. One might have wished to know which party they aligned with on further criteria such as compatibility with the indefinite article and with possessive, demonstrative, and interrogative pronouns; but alas, Finck did not tell us how to say 'a boy's mother' or 'my/this/ which boy's mother' in German Romany. On the evidence that he had and shared with us, German Romany would have had to be acknowledged as showing Suffixaufnahme at least to a degree—to the degree that caseagreeing attributes in -k r-/-g r- were not adjectives but nouns. So, was German Romany perhaps also group-inflecting to a degree? Finck's grammar was strictly off limits to any such typologizing, but the impression one gets is that he would have classed this exile from India as stem-inflecting. There was no mention of morphological marking accruing to whole phrases rather than to words individually, nor of flat, appositive syntax. However, his supposed correlates of Suffixaufnahme in particular were indeed there. First, case suffixes were not always easily distinguishable from postpositions. There was the contrast of direct or basic forms (functioning as subjects and, in the case of inanimates, also as objects) and oblique stems (the object form for animates), and there was a contrast of four cases that all

(Re-)Introducing Suffixaufnahmmee

13

took oblique stems as base. Given two such paradigmatic contrasts and the syntagmatic combination of forms from the two sets, the difference here might have passed for one between two cases and four postpositions. Second, relational marking was only cumulated with number within the direct-oblique system (tsaw-o DctSg, tstiw-es OblSg, tsaw-e DctPl, tsaw-en Ob1Pl); the four other cases (or postpositions or ex-postpositions) were not bound up with number marking, this being handled by the oblique suffixes obligatorily preceding them. Third, attributive forms in -k r-/-g r- could occur as independent NPs, taking further case inflection, either within syntactic contexts sanctioning the ellipsis of heads or otherwise to create new lexical items, such as wast-es-k r-o 'the one of the hand, i.e. glove', graj-en-g r-o 'the one of the horses, i.e. horse-dealer', mas-es/en-k/g r-o 'the one of the meat, i.e. butcher', and virtually all other occupational terms.

2.3. The Spread of Suffixaufnahme:

Bork I, Bork II

Followers of Franz Nikolaus Finck in his interpretation of Suffixaufnahme might have been expected to keep their eyes peeled for further specimens of group-inflection so as to be able to ascertain that it was indeed a characteristic of this type rather than being uniquely (Old) Georgian. Die Haupttypen des Sprachbaus would have been of little help as a guide, though, giving no indication at all where else to look. However, while Finck's book was still in the making, emerging from lectures given at the universities of Marburg (over ten years prior to publication, according to the preface) and Berlin (since 1903), the incidence of Suffixaufnahme-like patterns had in fact already been shown to be higher, beyond Romany. In May 1905 the Orientalistische Litteratur-Zeitung carried a short notice by Ferdinand Bork (1871-1962), pointing to affinities between the two groups of languages in which the schoolteacher from Konigsberg was specializing in his spare time, Caucasian and Ancient Near Eastern.10 Chief among these was the curiosity that Finck was to call, or was already calling, Suffixaufnahme. Bork's point of departure was Elamite (spoken until the first century CE in the Luristan and Khozistan areas of today's Iran, of no known affiliation—but Bork knew better), whose genitive had been claimed to have an almost adjective-like character. Given an areal or genetic propensity for adjectival rather than nominal attribution, it would only be natural, in languages richer in cases than Elamite (where the genitive, if it was a case, was a loner), for attributive genitives to be turned into real adjectives by the copying of endings from their heads. Far from merely speculating, Bork could cite examples. In the language preserved in the Mittanl Letter of the second millenium BCE and originally spoken in northern Mesopotamia (but the Hurrian dominions at times extended far into southeastern Anatolia, northern Syria, and also eastward), appropriate passages such as (8) had recently been deciphered in which an attribute's genitive suffix and a "lengthening syllable" -ne were followed by a copy of the head's genitive.11

14

Prologue

(8) Immoria-ve KUR Mizirre-ve-ne-ve evri-ve Immoria-Gen land Egypt-Gen-ne-Gen lord-Gen 'of Immoria, of the land of Egypt's lord'

For Hittite and Luwian of Anatolia, both Indo-European (not necessarily for Bork, though), analogous patterns had allegedly been demonstrated beyond doubt. (Some doubts rearose later.) Moving north, Bork did not neglect Old Georgian, where what had been called Suffixanreihungg ("suffix-stringing") in orientalist circles was so exuberant as to occur repeatedly on the last, though not on the middle, of a series of attributes: (9) klite-n-i sasupevel-isa ca-ta-jsa-n-i key-P1-Nom kingdom-Gen heaven-OblPl-Gen-Pl-Nom '(the) keys of the kingdom of (the) heavens'

Even postpositions on their way toward casehood, such as gan with ablative meaning and governing the genitive, could be resumed on attributes, although not in the form of a true copy; rather, they were substituted for by the dative, originally a general local case and perhaps also a source of genitive -isa: (10)

pir-isa-gan uvmrto-ta-sa face-Gen-in infide1-OblPl-Dat 'from the face of the infidels'

Concerning Tsakhur, a member of Lezgian, the southernmost branch of Northeast Caucasian, Bork had specialist information that attributive genitives had formerly copied the plural suffix of their heads in like manner, with the examples that he had seen lacking case marking on heads: (11)

jak-bi dekhk-in-bi axe-Pi father-GenSg-Pl 'the axes of the father'

Some years later, in a brief sequel to his notice, Bork (1913) drew attention also to Bats, a member of Northeast (or North-Central) Caucasian but spoken in the South Caucasian area, which according to his source12 likewise permitted attributive genitives and ablatives to resume the cases of their preceding heads, either directly or after adjectivalization by means of -co: (12)

a. bakhe-v thhe dad Daivth-e-v mouth-Instructive our father David-Gen-Instructive 'through the mouth of our father David' (Daivthe being the genitive of Davith) b. chana-v bhestak-re-co-v one-Instructive warrior-Abl-Adjct-Instructive 'through one of the warriors'

(Re-)Introducing Suffixaufnahmmee

15

Bork now also announced the discovery of such agreement in Elamite. Thus Suffixaufnahme could no longer be presumed unique to (Old) Georgian. According to Bork, its spread was not random but areally defined, extending (a) to the Caucasus, especially in the south but including more Caucasian languages than merely the Kartvel ones, and (b) the Ancient Near East, especially around Mesopotamia and Anatolia, including what by latterday standards were two genetic groups (Hurrian-Urartian, Indo-European Anatolian) plus one isolate (Elamite). It represented a major affinity between what Bork termed the "Caucasian" and "Hittite clusters" (Sprachkreise), but whether it was due to common origin or to diffusion was best left undecided. Although Bork introduced Suffixaufnahme as an ostensibly weird and wonderful pattern, he explained it as, in principle, something quite ordinary. Genitival attributes agreeing in case and/or number with their governing nouns were genuine adjectives; if anything was out of the ordinary, it was that there were no two kinds of attributive constituents coexisting. It was only a few days before writing the sequel on Bats that Bork discovered that in this he was concurring with Bopp. Now, if Bork's story was about straightforward adjective agreement, one wonders why he found it worth telling. Agreeing attributive adjectives were nothing peculiar to the Caucasian and "Hittite" clusters, and he could have concluded with Bopp that what he was dealing with were natural developments of hard-core Indo-European. In fact, unlike Bopp, Bork did not speculate about the reanalysis of genitive inflection as denominal derivation. What he regarded as crucial was a deep-rooted need in the Caucasian and "Hittite" clusters overtly to relate the rectum to the regens (1905: 185); but he had no reasons to offer for the special urgency of this need in these clusters. In principle he could have done without the adjective theory altogether. He would only have had to abandon the axiom that adjectives were the only words ever able to agree in case and/or number with nouns. If the need of the old Caucasians and "Hittites" to relate the rectum to the regens was really so urgent, why should they have bothered which word class a rectum happened to belong to? After all, so far as Bork knew, only in Bats could genitives be manifestly adjectivalized before agreeing. In actual fact, Bork tacitly abandoned the adjective theory even before his discovery that Bopp had anticipated him. Now borrowing his terms from Heinrich Winkler, especially from his demonstration that Caucasian was morphosyntactically special vis-a-vis other eastern languages (1896), he characterized the sentences of Hurrian in the first thorough interpretation of the Mittani Letter (1909) as agglomerations of essentially autonomous parts, loosely and non-obligatorily related to one another by the resumption of suffixes. At the heart of this essentially appositive mode of syntax were predicates which were not really verbal but participial (or adjectival or nominal), which were indifferent to any distinction of verbal voices, which did not really establish a relational frame in the sense of valencies to be filled by nominals in particular verb-determined syntactic relations, and which accordingly did not establish a clear contrast between complements (subject, object) and circum-

16

Prologue

stances. With predicates failing inherently to bind any nominals, these needed to be indicated on the predicate by pronominal elements. And this technique of having elements on a phrase that pointed to its partner, establishing dependencies where there was mere juxtaposition, was generalized to all syntactic levels. Whether pronominal or otherwise, affixes carried by one constituent were most conveniently used for such pointing; they only needed to be repeated on the phrase that was to be related to the original carrier phrase. An instance of this were nominal groups in juxtaposition, one denoting the possessed and the other the possessor, where the latter could be bracketed with the former by simply resuming its suffixes. Despite its own relational marking, the possessor itself was not inherently related to the possessed as its attribute; it was syntactically independent like all phrases, but it took some general locative (rather than specifically genitive) marker roughly reflecting its semantic contribution to the whole clause—schematically 'the roof, that on the house' rather than 'the roof of the house'. The fullest statement of Winkler's own relational typology had actually been given in his Zur sprachgeschichte (1887). In the chapter on adnominal constructions Winkler (1887: 245-274) distinguished two functions of overt marking: to bind two nominals to one another, and to identify their mutual relationship as one of possession or more generally mere relatedness. He further distinguished the syntactic relationships thus marked as ones of subordination or of loosely appositive or explicative juxtaposition. In the latter case—schematically 'the house, this one, (of) the king' or 'the lion, that is, his tail'—there was not, strictly speaking, an adnominal relation between the two nominals concerned, since syntactically they were of the same rank; their overt encoding, if any other than plain sequence, was accordingly entrusted to binding rather than relationship-identifying elements. It was demonstrative elements that typically served as such juxtapositional binders, often indiscriminately binding nominals as well as adjectives and relative clauses. The genitive case was the prototypical pure relationship-identifier; but subordination could also be encoded, if less purely, by cross-referential (possessive) pronominals linking the possessor/rectum with its following regens. Diachronically, subordination could develop from juxtaposition, and genitives in particular could evolve from demonstrative binders, from expressions of local relations, or from full words designating 'property'. Genitives could in turn be reinterpreted as derivational, changing the word class of nouns to be related to a head; or they could be lost for phonetic or other reasons, heralding a return to the juxtapositional stage. According to Winkler, Africa and Melanesia, Polynesia, and Malaya were domains of juxtaposition, and most of Eurasia and the Americas of adnominal subordination. Australia struck him as special insofar as there was a genitive, suggesting adnominal subordination, but this genitive seemed to be formed from locative cases, which were a dominant trait in the generally appositive syntax of Australian languages. Suffixaufnahme might have put him in a dilemma: the resumption of case and perhaps other marking was an obvious instance of binding, pointing to loose juxtaposition; but the resuming nominals were in the genitive, like

(Re-)Introducing Suffixaufnahmmee

17

attributes in tight subordination. No wonder Winkler did not recognize a Suffixaufnahme when he saw one: double genitives in Old Georgian, one resumed from the possession, were mistaken by him for instances of emphatic suffix iteration ('of the man, yes indeed, of the man'), supposedly underlining the identity of the possessor (Winkler 1907). Encountering Suffixaufnahme in the Caucasus and the Near East, and (generally) recognizing it for what it was, Winkler's acquaintance Bork allayed his uneasiness about such contradictorily motivated double marking by seizing on Winkler's Australian speculations about the primarily local rather than strictly attributive role of these remote genitives. Bork's new Winklerian rationale behind the Mittanians' need to relate by means of identical case and perhaps other marking what ostensibly were rectum and regens, thus, was quite un-Boppian but reminiscent of Finck. There was no more mention of suffix-resuming genitival attributes being possessive adjectives (the possible product of metamorphosis from genitives in Winkler's scheme, presupposing subordinative attribution); in this type of language, strictly speaking, there was no tight syntactic relation of attribution in the first place, nor was there a genuine genitive, and what seemed to be rectum and regens were more or less independent juxtaposed phrases in need of overt bracketing. Subsequently Bork did become aware of Finck's Haupttypen, and he immediately endorsed the notion of group-inflection. It was on the grounds of their being predominantly group-inflecting that Bork, following in Winkler's footsteps, included virtually all Caucasian languages and those of the Ancient Near East, and also Basque and perhaps other ancient languages between Mesopotamia and the Gulf of Biscay, in one "Caucasian" family, conceding that it was perhaps the result of a mixture of two older families, his former "clusters." Suffixaufnahme was the grammatical key fossil of Bork's "Caucasian" (1924: 174) and had already proven widespread throughout its several branches, although on different premises. Conveniently omitting Basque, Bork (1924: 174) confirmed Suffixaufnahme for Elamite and further added Sumerian on the evidence of examples such as these, respectively: (13)

puhu-ri sijan Insusinak-mi-ra issue-Defpf!rson temple Insusmak-Defthing-Defperson 'the issue of the temple of Insusinak'

(14)

e dNin Girsu-k-ak-e house lord Girsu-Gen-Gen-Dem 'this house of the lord of Girsu'

According to Bork, nouns in Elamite (13) had different definite suffixes for persons (-ri/-ra) and things (-me/-mi, omissible from the noun itself), which were resumed by nouns, or at least by the last in a sequence of nouns, to be bracketed with a definite person or thing designation; thus, -mi and -ra on

18

Prologue

Insusinak pointed to a thing and a person to which this name was to be related.13 What happened in Sumerian (14) was that the genitive intended to relate the whole phrase dNin Girsu-k to e was not placed after its core member, dNin, but after this entire phrase, following upon the primary genitive of Girsu. Although it was in the spirit of group-inflection (recall (2') above), this Sumerian pattern was in fact not identical to Suffixaufnahme in the Old Georgian or Hurrian sense because there was no agreement: the genitive of d Nin was only marked once, phrase-finally, rather than on both constituents to be bracketed (dNin-ak Girsu-k-ak). Bork glossed over this difference, offering "suffix-accumulation" (Suffixhaufung) as a term supposedly equivalent to Finck's Suffixaufnahme. Elamite Suffixaufnahme conformed to Finck's pattern, except that the suffixes resumed were not suffixes of case (and number). Although Bork would occasionally observe that group-inflection was like pre-agglutination, with affixes not yet glued to their hosts firmly enough to resist being played around with, and that appositive clause structures required the incorporation of pronominals into the predicate, the relationship of his group-inflecting "Caucasian" to the agglutinative (or, in Finck's terminology, subordinating) and the incorporating types did not come up for discussion. One presumes he would have been surprised to encounter the key fossil of "Caucasian" anywhere else except between Mesopotamia and the Gulf of Biscay. Perhaps he would not have been surprised at the Roma practicing Suffixaufnahme to a degree, since they might have picked it up, incompletely, during their wanderings.

2.4. Miscegenation: Bourgeois For Henri Bourgeois of Brussels, who does not figure in any history of linguistics or biographical dictionary, the Ibero-Caucasian family was fantasy, not science. In the manner of Winkler or Bork any language could have been proved a first cousin of any other, and in particular of illiterate Abkhaz and Circassian and time worn Elamite, Mittanian, Etruscan, and Ligurian. For sober Bourgeois, what structural similarities existed were in the first instance due to typological affinity (or chance), not to genetic affiliation. Like Basque, Finnish, Estonian, Hungarian, Turkish, indeed Ural-Altaic in toto, Malay, etc., and unlike Aryan influences on it such as Armenian and Greek, Old Georgian was agglutinative, and so was Modern Georgian, since agglutinative languages were less changeable than flective14 ones (witness Greek and Armenian) if not interfered with. For Bourgeois (1909a), unaware of group-inflection, the chief diagnostic of agglutination was affixes, in particular suffixes, that did not cumulate several categories and that were strung out mechanically after their roots. Further morphological characteristics were (a) that inflection was essentially uniform for all words of a class, (b) that nouns, adjectives, and to some extent pronouns did not much differ in inflection, and (c) that attributes were invariable instead of co-varying with their primaries. As to syntax, agglutinative languages could either arrange words in their natural order, which was determine before determinant, and largely dispense

(Re-)lntroducing Suffixaufnahmmee

19

with relation-identifying markers, or they could invert natural order and add relational markers. Malay did the former and Ural-Altaic the latter, respectively rendering 'le chien du voisin de mon frere est beau' as 'chien-voisinfrere-moi-beau' or 'de moi-de frere-de voisin-chien-beau-est' (1909b: 294).15 The objective of Bourgeois's article in the Revue de linguistique et de philologie comparee of 1909, published in two instalments despite its brevity, was to reduce to order Old Georgian nominal inflection—as attested in translations of the Bible from the fifth and the seventh century, the oldest texts available—in line with what agglutinative morphology could be expected to be like. As he saw it, there were bare roots, being indeterminate denotations of concepts and of an adjectival nature (such as kac 'man, to be man, like man'), and three layers of accretions: first, case suffixes simultaneously rendering roots determinate (e.g. genitive kac-is(a) 'of (the) one who is man'); second, relational appendages taking case-inflected forms as bases (e.g. ablative kac-is(a)-gan 'of/from (the) one who is man'); and third, though virtually indistinguishable from these, postpositions or adverbs governing local cases or the genitive. Not quite living up to the agglutinative ideal, case was cumulated with number in the first series of accretions (cf. genitive singular kac-is(a) and plural kac-ta, with -ta the plural also of locative-dative); the later introduction of a separate plural suffix -eb (thus kac-eb-is(a) man-P1-Gen) was to be welcomed from this perspective. Nouns proper arose only from the case and number marking plus determination of adjectival roots. Apart from being syntactic primaries, these nouns could also serve as secondaries, as attributes to principal nouns. When following their primaries, attributes agreed with them through resuming their casenumber-cum-determination suffixes; and the agglutinative structure lent itself to repetition of this process: (15)

asul-ni-mk vidr-ta kalak-isa-ta-ni daughter-NomP1-? inhabitant-GenPl town-GenSg-GenPl-NomPl 'the daughters of the inhabitants, those of those of the town'

(The first attribute here was in fact skipped in this process. Second-layer appendages such as ablative -gan generally did not participate in it either, according to Bourgeois.) Owing to inflectional similarities, there was little point in distinguishing determinate forms as nouns and adjectives; all that needed to be said was that when determinate forms were syntactic secondaries and followed their primaries, they agreed with these. For almost all of this Bourgeois could cite parallels in other agglutinative languages such as Basque, Hungarian, and Finnish. It was only the agreement of attributes with their principals that was not intrinsic to the agglutinative type. Bourgeois was inclined to believe that any inconsistencies in this respect were due to influences from languages of a type that permitted or even required agreement. For Georgian the culprits clearly were Greek and Armenian, providing the blueprint for the earliest biblical translations. Although

20

Prologue

Bourgeois did not elaborate, Greek was especially pertinent since, apart from having agreeing adjectives, its postnominal genitival and prepositional attributes resumed the case-number(-gender)-inflecting definite article of their principals: (16)

hai thugater-es hai ton polit-on the=NomP1Fem daughterFem-NomP1 the=NomP1Fem the=GenP1Masc citizenMasc-GenP1 'the daughters [the ones] of the citizens'

The Greek models of Old Georgian translations which Bourgeois cited did not actually have this article-Aufnahme construction, but rather its alternative with genitival attributes between the principals' definite article and the principal noun ('the of the citizens daughters'). At any rate, once Georgian had freed itself of pernicious agreeing influences, its attributes were reverting to the original agglutinative mode, preceding their primaries, carrying markers identifying them as secondaries (such as -is(a)), and shunning Suffixaufnahme—just as in Ural-Altaic. For Henri Bourgeois, then, Suffixaufnahme was a kind of mongrel issuing from the meeting of malleable agglutinative morphology with the flective type and not destined to survive. Nonetheless, considering the frequency of such meetings, at least in Eurasia, such hybrids, if short-lived, might have been more numerous. 2.5. Why Not Armenian: Vogt Armenian had not been granted membership in Bork's "Caucasian," despite its location. Its Indo-European credentials were by now too incontrovertible. Its genitives, however, were not quite as well-behaved as one would have expected from a member of that worthy stem-inflecting family. As first pointed out by Heinrich Hiibschmann in a very short notice (1906), which had obviously escaped Henri Bourgeois's attention, Classical Armenian could replace the ordinary genitive on attributes with the case of the head, especially if this was an ablative or instrumental: (17)

a. i knoj-e t'agawor-i-n by wife-Ab1Sg king-GenSg-Def b. i knoj-e t'agawor-e-n by wife-AblSg king-AblSg-Def 'by the wife of the king'

It was Hans Vogt (1903-1986) who assigned to such sporadic case attractions more than curiosity value. An expert on Kartvel but apparently unaware of the typological considerations this family had occasioned in Finck's

(Re-)Introducing Suffixaufnahmeme

21

Haupttypen, Vogt (1932) recognized Classical Armenian case attractions as functionally analogous to Suffixaufnahme in Old Georgian, itself touched on again in a later publication of Vogt's (1947: 129-131).16 As was seen when comparing (17b) to its Old Georgian equivalent (18), with an enclitic ablative postposition governing the genitive in lieu of the instrumental, the sole difference was that Armenian attributive nouns dropped their original genitive when resuming the case of the head while those of Old Georgian retained it. (18) col-isa-gan m-is mep-isa-jsa wife-Gen-by the-Gen king-Gen-Gen 'by the wife of the king'

Unwittingly echoing Heinrich Winkler (who was in the same article chided for his ridiculous interpretation of double genitives as emphatic iteration), Vogt ascribed a dual function to case inflection in Old Georgian: (a) that of identifying the role of a word or group of words within its phrase, which in fact was to some extent already achieved by linear order, and (b) that of indicating which words belonged with which others. By Suffixaufnahme the attribute—for Vogt self-evidently a noun—was stamped as a member of the NP that also contained the carrier of the case copied. The repetition of prepositions (including that of definite accusative, z5) on several members of prepositional phrases, as well as the phrase-final placement of the definite article, were signs of a similar tendency toward the inflectional marking of words as comembers of nominal groups in Armenian; the replacement of the genitive by the case of the head, especially the non-syncretic ablative and instrumental, was another manifestation of this tendency. Vogt had no typological explanation for the use of inflections for grouping purposes in both Old Georgian and Classical Armenian, nor could he account for the difference between the two languages in regard to the retention or dropping of the inner genitive. He did not really question his assumption that two nominals bound together by identical case inflections were a determinant and a determine in close-knit construction. Admittedly, he occasionally found it difficult to distinguish in Armenian between genuine attribution and the juxtaposition of nominals in asyndetic apposition; but this did not suggest to him that the syntax corresponding to the grouping type of inflection was somehow different. What explanations Vogt had to offer were historical, invoking an Ancient Near Eastern substratum (a little too ancient perhaps for Roma to benefit from it). Armenian-style case attraction purportedly had parallels in Hittite, and Suffixaufnahme, its functional analogue, was not limited to Old Georgian either, definitely occurring also in Mittani Hurrian and perhaps, although much was still obscure about this language, in Vannic Khaldean (i.e., Urartian). Judging by pairs of examples such as (19), the so-called genitivus genitivi in Etruscan belonged here as well.

22 (19)

Prologue a. lard cuclnie(s) velvur-us Larv Cuclnie Velvur-Gen 'Larv Cuclnie the son of Velvur' b. larv-al cuclnie-s vel ur-us-la Lard-Gen Cuclnie-Gen Velvur-Gen-Gen 'of Lard Cuclnie the son of Velvur'

The genitive on the last attribute seemed to appear only if the head was also in the genitive.17 The identical case marking on heads and nominal attributes was not the only feature that Vogt saw as stemming from an ancient substratum, possibly to be identified as South Caucasian.18 Others, shared at least by Georgian and Armenian, were the trend toward ordering the determinant before the determine (with the verb thus coming last), and the case-inflectability of genitives lacking an overt head (20). (20)

a. vmrt-isa-j god-Gen-Nom(Def)) 'that of/belonging to god' b. astuc-oy-k'-n god-GenSg-NomP1-Deff 'those of god'

What one might have gone on asking is whether there were any necessary structural links between these various features, or whether it was by accident that precisely these features rather than any others were found to co-occur in languages once or still neighbors.19

2.6. The Lingering of the Haupttypen: Lewy, Lohmann, Wagner Although the Haupttypen des Sprachbaus did not prove a popular success, in Norway or elsewhere, there were a few followers in Finck's footsteps. Apart from the later Bork, these were Ernst Lewy (1881-1966), a student at Berlin in Finck's time and an emigre to Ireland as Nazism held sway over many areas of whose languages Lewy had taken cognizance; Johannes Lohmann (18951983), a renegade Indo-Europeanist who latterly believed that an injection of Heideggerian philosophy would benefit linguistics (as a professor at Freiburg im Breisgau it was hard not to succumb to the genius loci); and Heinrich Wagner (1923-1988), a Swiss-born Celticist based at Belfast and then Dublin whose interests extended to the languages of all lands once perhaps roamed by the Celts, and beyond.20 With this fringe group, Finckian typology went areal. 2.6.1. While not uncritical of Finck's typology, Lewy wholeheartedly adopted the notion of group-inflection, in the straightforward sense of inflec-

(Re-)Introducing Suffixaufnahmeme

23

tions accruing to whole phrases rather than to words or stems. Intent on mapping the geographical distribution of structural types—a subject not broached in Finck's Haupttypen—he would go far beyond Bork's "Caucasian" and variously mention Ket and Kott (of the Yenisei-Ostyak group based in Siberia), Mordva (of the Volgaic branch of Finno-Ugric, purportedly Caucasian-influenced), Avar and indeed all of Northeast and also Northwest Caucasian, Lycian (an Indo-European language of ancient Anatolia, another favorite of Bork's), the modern Iranian languages, Basque at its earlier stages, Coptic (the last stage of Egyptian), Nama (Khoisan), Aranda (PamaNyungan) and other Australian languages as well as those of New Guinea, and Mayan Quiche alongside South Caucasian as showing significant features of group-inflection, and he would speculate that group-inflection was perhaps an old principle often superseded by the currently more widespread types.21 He saw, on the other hand, certain similarities between group-inflection and what he called "inflection-isolation," representing a possible development of stem-inflection and exemplified in English, for example, by the groupgenitive (his uncle and aunt's good graces) and prepositionally introduced subordinate clauses (the mystery of why she had invited them). Among the further correlates of group-inflection that he suggested were ergative alignment or the prevalence of clause constructions other than the accusative-type active transitive one, and, more enigmatic to him, the absence of nominal classification or gender. In regard to Suffixaufnahme, it seemed to Lewy an especially telling and not at all enigmatic characteristic of the group-inflecting type, because it proved suffixes to be appendages so loose that they could gain independence and be repeated on another constituent, bracketing it with the word from which the suffix was copied and thus establishing the boundaries of a complex syntactic group (Lewy 1951: §7). On another occasion (Lewy 1942: §272), taking his cue from Heinrich Winkler, he paraphrased the latter function as "explicative" or "pointing" and even "correcting" (deutend, berichtigend), somewhat comparable to that of cross-referencing pronominals on verbs in languages of the Balkans. Looseness of inflections was a theme borrowed from Finck, and it was pointed out above that the attribution of this trait specifically to group-inflection is unorthodox; it is at least as characteristic of the subordinating type (better known as agglutinating) and of Lewy's own inflection-isolating one (better known as analytic). Still, by emphasizing the separability of inflections Lewy arguably sought to distinguish Suffixaufnahme from plain agreement in the stem- or root-inflecting types, not normally involving the repetition of actual desinences but more abstract categorial correspondences. Thus the adjective theory evidently was not to Lewy's liking; his taking the function of Suffixaufnahme to be the bracketing of constituents was certainly closer in spirit to Finck and the later Bork than to Bopp and the early Bork. Nonetheless, he had once not seemed averse to regarding Suffixaufnahme or similar patterns as a source of adjective agreement (cf. Lewy 1920: §12). Considering its purported typological significance, Lewy must have been

24

Prologue

exasperated by the scarceness of Suffixaufnahme in the languages where he perceived markedly group-inflecting tendencies. In fact, he only knew it in Old Georgian, and therefore was glad early on to have found traces of what seemed a similar pattern in Erzya Mordva (Lewy 1920). In texts elicited from a man who some ten years before had left the area of Saratov on the Volga for that of Tomsk in Siberia and was now a prisoner of war, there were four instances of the postposition mara 'with' resuming the case and, if present, definiteness and number suffixes of a noun outside the postpositional phrase: (21) a. sure-n-t' kol't's'a mare-n-t' tapar'ik' pangkstsa finger-Gen-Def (Sg) ring with-Gen-Def wrap=up patches 'Wrap up the finger with the ring with patches!' b. ton s'e puz'iri-n-t' w'er' mare-n-t' putik kawalet alu you this (?) bladder-Gen-Def (Sg) blood with-Gen-Def put your=armpit under 'Put this bladder with blood under your armpit!' c. ki kantsindz'e w'edra-t-n'i-n' w'ed' mare-t-n'i-n' who carries bucket-Pl-Def-Gen water with-Pl-Def-Gen 'Who carries the buckets with water?' d. awanza putiz'e koika-s pruzhina mare-s his=mother he=laid bedstead-Ill (Sglndef) springs with-Ill 'He laid his mother into a bedstead with springs' This Erzya pattern differed from Old Georgian Suffixaufnahme in that the carrier of the suffix copied was not a case-inflected noun but a postposition; there was a genitive case in Erzya, but it did not do what its Old Georgian analogue was known to be doing. Moreover, mara was the only Erzya postposition on record in Lewy's texts as showing Suffixaufnahme. Lewy's interpretation of this peculiarly circumscribed pattern was that the copied suffixes strengthened the cohesion of the phrase consisting of a noun and the postposition and at the same time bracketed the postpositional phrase with the head noun that it complemented. And these were also the purposes served by Finck's Suffixaufnahme. As to bracketing, Lewy might have added that without Suffixaufnahme there could have been a danger of ambiguity in regard to the constituency of such postpositional phrases, or to their relational status as attributes or adverbials or secondary predicates. In an example such as 'Kuzma laid his mother into a bedstead with a pitchfork' the mam-phrase could conceivably be associated with the oblique object in the illative (with the bed's construction somehow including a pitchfork), the direct object (with the mother clutching her pitchfork), or the subject of the action (with Kuzma using a pitchfork for the transferral). Under such circumstances Suffixaufnahme is likely to disambiguate. I have been unable to confirm Lewy's find from other accounts of Mordva.22 He himself struck gold no more. Bork's pertinent discoveries in his

(Re-)lntroducing Suffixaufnahmmee

25

"Caucasian" continued to be unappreciated.23 In Strehlow's Aranda Phonetics and Grammar (1944) he saw enough to convince him that group-inflection also reigned in Australia. Aranda's genitives were a case in point: (22)

a. worra ingkata-kana-la son chief-Gen-Erg b. ingkata-ka worra-la chief-Gen son-Erg 'the son of the chief

The case that marked the external relation of the entire group of a head noun and its nominal attribute, ergative/locative -la, came at its very end, irrespective of whether the last constituent was the attribute in the genitive, hence carrying two cases, or the head. Why the genitive was -kana when followed by another case and -ka when alone eluded Lewy (1953b: 252); he might have found similar "lengthening syllables" after genitives with Suffixaufnahme in the Mittani language, had he paid heed to Bork. Group-inflecting though the first pattern looked, it was not Suffixaufnahme in Finck's sense since case marking was not repeated. It will be remembered that Bork had been less discriminating than Lewy was, lumping Sumerian "suffix-accumulation," analogous to (22a), with Suffixaufnahme a la Old Georgian and Hurrian. 2.6.2. Johannes Lohmann's aim was to replace Finck's psychological foundation of his eight cardinal types with an ontological one, based on how the relationship between "thing" and "attribution" (the suppositio of medieval logic) was conceived of on the one hand, and on how constituents were overtly interrelated on the other. As to the syntax of suppositio, Lohmann (1948, 1954) distinguished four types of clause construction: purely verbal, purely nominal, mixed verbo-nominal, and nominal and verbal clauses coexisting. As to morphology, he accepted Finck's eight modes of synthesis. The purely verbal clause construction was an "objective" rendering of predications, with all participants registered on the verb by means of pronominals (as schematically in 'the women they-give-him-it the beggar the bread') and without a distinction of active and passive voices and perhaps also of Aktionsarten (instead expressed in the encoding of the actants). This type of clause construction was, according to Lohmann, the necessary correlate of group-inflecting morphology. The areal domain of this twinning of traits was the lands around the Mediterranean, with Caucasian, Basque, and Sumerian specifically mentioned by Lohmann. That incorporation as practiced in America generated what seemed to be the same type of verbal construction did not bother Lohmann any more than it had Bork; he surmised that the group-inflecting type was perhaps only a refinement of the incorporating one. Suffixaufnahme as such appears not to have riveted Lohmann's attention. Not that he would have found its explanation difficult: it was just another instance of relating expressions (for such of course included cases) being

26

Prologue

loosely appended to notional expressions and thus bracketing them to single units, comparable to the pronominal copies clustering around the verb. Someone holding that group-inflection was a trait virtually immune to change (1948: 65) might have been perturbed by the recent scarcity of Suffixaufnahme in the circum-Mediterranean area; there are no indications that Lohmann was. But then, Lohmann's typology never got very far beyond the a priori. On the subject of group-inflection he might as well have reedited the mature Ferdinand Bork's writings on "Caucasian," with annotations from Heinrich Winkler and Hans Vogt. 2.6.3. The group-inflecting type, Finck's innovation and Lewy's object of desire, appears to have lost its appeal entirely for Lewy's friend, Heinrich Wagner. Outside Celtic and its alleged Near Eastern and African connections, Wagner's areal interests, none too circumscribed, were centered on the great Eurasian mountain belt extending from the Pyrenees via the Alps, Anatolia, the Caucasus, the Hindukush and the Karakorum range, the Himalayas and Tibet to the highlands of Yunnan Province and Indochina. For Wagner (cf. 1978, 1985) this vast area was one heartland of the subordinating (alias agglutinative) type; the three others were northern Eurasia and the Arctic region, central Asia including Korea and Japan, and New Guinea and Australia. The most salient structural traits of the subordinating type, apart from agglutinative morphology, supposedly were the ergative construction (missing in Altaic, though, and sometimes elsewhere too), the lack of personal or subjective verbs in the strict Indo-European sense, and SOV and allied word orders of Greenberg's (1963) Type III. The addition to verbs of pronominals cross-referencing subjects, objects, and perhaps further nominals was the preferred method of interrelating clausal constituents, with verbs lacking inherent relationality. The incorporation of entire nominals themselves, as in Paleosiberian and American languages and also occasionally in the Eurasian mountain Sprachbund, was but a radicalization of this method. Without much ado virtually all of Finck's, Bork's, and Lewy's candidates for group-inflection were thus reclaimed for the subordinating type by Wagner. As the group-inflecting type merged with the subordinating one, Suffixaufnahme lost its hard-gained prominence: once the hallmark of a separate type, it was tacitly reduced to an inconspicuous optional ornament of the subordinating supertype. Considering the contours of Wagner's areal typology, this was rather surprising, for eastern Asia Minor, northwestern Mesopotamia, northern Syria, and the northwestern and southern Caucasus represented the core area of his Eurasian mountain Sprachbund, and it was here that Suffixaufnahme had first been discovered. In fact, it is difficult to tell whether Wagner cold-shouldered Suffixaufnahme deliberately or unwittingly. When commenting on the typically subordinating ways of encoding attribution (1978: 54f; 1985: 54, 69-72), he suggested that the simple juxtaposition of dependent and governing noun was the basic or

(Re-)Introducing Suffixaufnahmeme

27

original strategy, often complemented by the addition of "pseudo-genitival" suffixes to dependents; furthermore, with increased morphological complexity attributive encoding would tend to become "pleonastic," with the dependent noun in the genitive and the governing noun agreeing with the preceding dependent in person, number, and perhaps gender—schematically, 'the king's his (-)palace'.24 Hattic (an extinct isolate of Asia Minor), Sumerian, Old Hittite, Ossetic (Northeastern Iranian, spoken in the Caucasus), Abkhaz (Northwest Caucasian), Mordva, Votyak (Permic branch of Finno-Ugric), and his native Alemannic (i.e., Alpine German) were languages where Wagner detected this pleonastic pattern. However, patterns which are formally quite different were also alluded to as characteristic of the subordinating type: the replacement of genitival nouns by derived adjectives as in the Anatolian group of IndoEuropean, and "inflected genitives." Under this latter rubric Wagner lumped three different things: suffix-accumulation a la Sumerian and Aranda (see (14) and (22a) above), Suffixaufnahme as in (Old) Georgian (getting only the most fleeting of mentions25), and genitives lacking overt governors but case-inflected like independent NPs as in the Anatolian group, extinct Akkadian (North Semitic), and Basque. Wagner may have had a reason not to overemphasize formal differences here: functionally, all these varieties of attributive encoding encountered in the subordinating type—with the exception of suffix accumulation, all in a sense pleonastic—could be seen as serving the single purpose of bracketing the dependent with its governing noun, just as nominals were bracketed with their verbs by means of verbal cross-reference markers. The less charitable interpretation is that Wagner—odd though it seems for someone out to salvage the heritage of Finck, Bork, and Lewy—did not realize that the three kinds of double case marking in particular were different, and thus he was in no position to gauge their respective areal or typological significance. It is fair to say, at any rate, that as Finck's structural types were areally anchored in the manner of Bork, Lewy, Lohmann, or Wagner, Suffixaufnahme was not the main beneficiary. At worst, it was the chief casualty, along with the group-inflecting type.

2.7. Encore After a brief interval the curtain rose again on a new cast reenacting the parts of Franz Bopp, Franz Nikolaus Finck, Ferdinand Bork, Henri Bourgeois, Hans Vogt, and Ernst Lewy et al., with the script slightly modernized to suit the taste of the 1980s. 2.7.1. Contemporary linguistics owes much to Igor Mel'cuk for the tidying up of its terminology. In the course of his effort to end all loose talk about cases, he surveyed various current subclassifications of cases ("cases 2," to be more precise), one of which distinguishes between governed cases and cases induced by agreement (1977: 20, 31f.; 1986: 36f., 68-70, 83f.). Properly speaking, the latter should not be known by the same name as the former; Mel'cuk

28

Prologue

suggested distinguishing the latter variety as casus concordatus as opposed to casus rectus. Casus concordati are typically found on adjectives, but according to Mel'cuk nouns may also agree in case when occurring as dependents in a syntactic relation amenable to agreement marking. As Old Georgian demonstrates, a governed case (genitive) and an agreement case (instrumental) may co-occur on a single noun: (23)

saxel-ita mam-isa-jta name-Ins father-Gen-Ins 'with father's name'

Asked by two colleagues, T. V. Bulygina and A. A. Zaliznjak, whether casus concordatus nominalis georgicus antiquus was not the same thing as the rather better-known casus concordatus adjectivaliss as found in Russian, Mel'cuk conceded that there is a functional similarity, but he also pointed out a crucial difference: Russian words agreeing in case and other relevant categories with head nouns, such as otc-ov-ym father-Adjct-InsSg, are adjectives, while their supposed Old Georgian analogues, such as mam-isa-jta, are nouns; suffixes such as Russian -ov are derivational, turning nouns into agreeing adjectives, while the Old Georgian genitive, providing the base for casus concordati, is inflectional. And Mel'cuk argued his case, if frugally. In Russian, words such as otcovym occupy the position of adjectives and cannot be modified by a further adjective, "etc." (1977: 31); in Old Georgian, words such as mamisajta occur after their heads, which (allegedly) is not the position of attributive adjectives, and can themselves be modified by attributive adjectives, "etc." (1977: 32). In another context Mel'cuk (1986: 68) said that casus concordati on nouns are "physically identical with (or similar to)" the casus recti of these nouns, which does not necessarily apply to case-agreeing desinences of adjectives. Admitting that the distinction between adjective and noun may not always be clearcut, Mel'cuk cited possessive adjectives in Slovak as permitting, in noun-like fashion, determination by a possessive pronoun in the genitive, where one might have expected the possessive to agree with the ultimate head, in adjective-like fashion and along with the derived adjective: (24) vasej materina dcera your=GenSg mother=Adjct=NomSgFem daughterFem (NomSg) 'the daughter of your mother'

But since possessive pronouns are the only kind of words to show such aberrant behavior in the company of words such as materin-, Mel'cuk categorized these as adjectives, and he was confident that a distinction between casus concordatus nominalis and adjectivalis would always be feasible. A perusal of Finck's grammar of German Romany might have served as an antidote. By 1986 Mel'cuk had become aware of casus concordati, supposedly likewise nominales, in languages other than Old Georgian, mentioning Basque

(Re-)Introducing Suffixaufnahmeme

29

(erroneously) and Ngarluma and Kayardild, both Australian. In Ngarluma the attributive constituents displaying casus concordati in Mel'cuk's example were actually relative clauses rather than simple nouns in the genitive; in addition to having its own case, each noun in a relative clause would here resume the case of the relative's head. For Kayardild Mel'cuk could report (courtesy of Nick Evans) a maximum of no fewer than four cases in a sequence. Thus in (25) the attribute (shown in boldface) first takes the genitive case, owing to its attributive function; then the instrumental, copied from its head; then the modal ablative (M.Ab1), signaling the sentence's modality (low reality status) and distributed over all its nominals other than the subject; and finally the purposive case (Purp), linking up a sentence with the previous discourse (its force here being that of a contradiction) and again distributed over all its parts, including the verb. (25))

ngijuwa yalawu-jarra-ntha yakuri-naa-ntha waytpala-karra-nguni-naa-ntha mijil-nguni-naa-ntha I=Nom=Purp catch-Past-Purp fish-M.Abl-Purp white=man-Gen-Ins-M.AblPurp net-Ins-M.Abl-Purp 'Yes, I did catch some fish with the white man's net'

Further categories of case or case-like notions had to be introduced to account for such distributed marking that was distinct from both casus rectus and casus concordatus. Daghestanian Tsakhur, mentioned en passant (1986: 82f. n. 12), forced Mel'cuk to recognize three separate genitives by his criteria of case distinction. The choice between these depends on the noun class (classes I to HI vs. class IV) and case (nominative vs. oblique) of the head, as exemplified in (26). (26)

a. hammaz-na dekh friend-Genl father 1 (Nom) 'friend's father' b. hammaz-m jik friend-Gen2 heartlv (Nom) 'friend's heart' c. hamaz-ni dek-is / jik-is friend-Gen3 father1-Dat / heartIV-Dat 'to friend's father / heart'

Ferdinand Bork would have been glad to know this; it might have fitted in with the information he had about Tsakhur genitives resuming the plural of their heads. For Mel'cuk the plurality of Tsakhur genitives, also found in the Tse/ group of Northeast Caucasian, appears to have borne no resemblance to casus concordatus nominalis. Mel'cuk's (perhaps unwitting) contributions to the Suffixaufnahme de-

30

Prologue

bate, then, were four: (a) he actually argued against, as opposed to merely rejected, Bopp's adjective theory for Old Georgian; (b) he accepted the possibility of nouns agreeing with other nouns in case; (c) he noted similarities between relative clauses and attributive nouns with respect to case agreement with their heads; and (d) he put Australian on the map. As to (b) and (c), he assumed nouns and relative clauses carrying a casus concordatus to be ordinary attributes, which deprived him of the motive that Finck, the later Bork, and Lewy could adduce as the raison d'etre and typological correlate of Suffixaufnahme. In dealing with the forms of case marking, Mel'cuk did touch on typological parameters such as agglutination (with non-cumulative exponents of case and number exemplified from Georgian) and group-inflection (with the principle of "one inflection per phrase" exemplified for phrases with coordinate or subordinate members from Tocharian, Gilyak, English, and Basque); but syntactic correlates of any such morphological differences were not part of his contribution. Among the instances of double case marking not involving a casus concordatus Mel'cuk mentioned Basque genitives, inflecting for case when functioning as independent NPs (lumped with Suffixaufnahme by Wagner): (27)

ggizon-aren-ari man-Genl-Dat 'to that of the man'

The question of a possible correlation between such hypostasis formations and casus concordatus nominalis once more remained unasked. And why should it have been asked if the topic were definitions? 2.7.2. The Australian connection was one rarely missed since. In Christian Lehmann's (1984) panorama of relative clauses, old Hurrian found itself in the immediate vicinity of newly famous (if dying) Dyirbal—of all languages— sharing with it not only embedded postnominal relative clauses with noninitial subordinator and (nominalized) relative clauses case-marked in agreement with their heads on their last constituent, but a host of further features as well, such as agglutinative and exclusively suffixing morphology, a rich supply of cases, unmarked verb-final and object-before-subject constituent order, some positional freedom of attributes, ergativity in syntax as well as morphology, and Suffixaufnahme (1984: 73-75): (28)

sen-iffu-ue-ne-ue asti-ue nigari brother-1SgPoss-Gen-ArtSg-Gen wife-Gen dowry 'dowry of the wife of my brother'

(29)

bagul wanal-gu banul-cjin-gu yara-nu-n din-gu ClassI=Dat boomerang-Dat ClassI=Gen-Lig-Dat man-Gen-Lig-Dat 'the man's boomerang (Dat)'

(Re-)Introducing Suffixaufnahmeme

31

In regard to Suffixaufnahme, there were even such identical details as the presence of a "catalytic" or "ligative" (Lig) suffix, in Hurrian an article distinguishing number, in between the genitive and the resumed case. Although interdependencies other than those to do with relative clauses were not Lehmann's topic, his sample might have suggested that Suffixaufnahme was not an isolated trait but implied some or even all the others that Hurrian and Dyirbal had in common. Some of them would also have been encountered in Old Georgian, had it been included. The implications clearly did not hold the other way round, since Suffixaufnahme was missing from numerous languages in the sample sharing much else with Hurrian and Dyirbal, such as Turkish, Quechua, Japanese, the Dravidian family (all non-ergative), Basque, Sumerian, and Greenlandic (all ergative). Recognizing a distinction between nouns and adjectives in his two Suffixaufnahme languages, Lehmann considered case-resuming words in the genitive to be (inflected) nouns rather than (derived) adjectives, although possessive adjectives could in other languages be used to perform partly similar functions (1983: 362). As to the syntactic relationship between the words or phrases linked by Suffixaufnahme, it was not conceived of as differing fundamentally from that of genitival dependents and their heads elsewhere. In principle, genitival dependents could be adnominal modifiers or governed complements of their heads; but since complements were not supposed to agree in case with their governors, Suffixaufnahme proved genitival dependents to be modifiers (cf. Lehmann 1982,1983). The more radical difference in Lehmann's taxonomy of relations was between dependency and "sociation," with coordination and apposition subsumed under the latter. For relative clauses non-dependency was a real option, with adjoined and appositional varieties not forming subordinate co-constituents with the nominals with which they referentially belong. But genitival attributes, and also adjectival ones, were supposedly different, being inherently so closely bound up with their heads as to preclude the head's anaphoric resumption (schematically 'the keys, those of heaven') and to resist extraposition ('the keys were lost of heaven') (Lehmann 1984: 205, 231). Finck, Bork II, and Lewy would have begged to differ. For them, even apposition as defined by Lehmann, requiring the appositum to form an immediate constituent with the nominal it is apposed to, might still have seemed too tight a relation for partners in Suffixaufnahme. 2.7.3. With notices of a certain exuberance of relational marking on nominals accumulating in descriptive accounts of individual Australian languages, comparatists down under had got something to answer for. 2.7.3.1. Although he conventionally listed the genitive among cases, Dixon (1980: 300, 321) was not quite sure whether it is really properly inflectional. Unlike other cases and like the "having" and "lacking" suffixes widespread in Australia (sometimes referred to as proprietive and privative cases), the genitive or the dative, also encoding possessors, can be followed—or according to

32

Prologue

Dixon (1972:12,105; 1980: 321), is "normally" followed in "all" or "almost all" or "most" Australian languages—by a further case inflection, with the attribute thus agreeing with the possessed noun that it modifies. Armed with the axiom that a nominal word consists of a root and a case inflection, with any affixes in between being derivational, this distributional evidence inclined Dixon (cf. also 1977:134) toward considering the genitive and its functional equivalents to be derivational, forming adjectival stems that could take case agreement.26 After Bork's defection, Franz Bopp had again recruited a disciple. Not all readers of The Dyirbal Language of North Queensland noticed, though. In a popular survey of inflectional morphology by Anderson (1985: 188), Dyirbal serves as the only example of that rarity, nouns (which was not what Dixon was convinced they were), those most reluctant agreers, agreeing in case, of all categories, upon inflecting (which was not what Dixon believed they did) for genitive. It was Dyirbal's ergative rather than its genitive that attracted the most attention at the time. 2.7.3.2. In Blake's overview of Australian Aboriginal grammar (1987: 31f., 77-99), however, genitives—and other markers of possessors which are followed by the cases that identify the relation of the possessed nominal within its clause—were beginning to catch up. They got a new label: "pre-cases." In phrase-marking (or group-inflecting) languages such as Alyawarra (30) or Aranda (22), such pre-cases, located on the final member of the possessor phrase, are followed by a case suffix doing duty for the entire phrase, whereas in word-marking languages such as Kalkatungu (31) or Dyirbal (29), the head nominals themselves are also case-marked, with the case suffix after the precase merely a copy. (30)

ayliyla artwa ampu-kinh-ila boomerang man old-Gen-Ins 'with the old man's boomerang'

(31)

kalpin-kuwa-thu yaun-kuwa-thu thuku-yu (ityayi-ngi) man-Gen-Erg big-Gen-Erg dog-Erg (bite-me) 'the big man's dog (bit me)'

While both patterns illustrate "Suffixhaufung" in Bork IIs sense, Suffixaufnahme in the narrower sense presupposes word-marking. Reconsidering the question of the inflectional or derivational status of precases, Blake's results were somewhat inconclusive. There are some properties that pre-cases share with bona fide inflections: they are very productive and semantically more or less regular; some are also found to mark complements of verbs. More on the derivational side are the possibilities of pre-case forms being lexicalized and undergoing bona fide derivation, changing their wordclass. Yet further properties, in particular their phrasal scope in phrasemarking languages, seem equally difficult to reconcile with either interpretation. Curiously, the fact that words carrying a pre-case agree in case in word-marking languages seemed to Blake (1987: 88f.) highly suggestive of

(Re-)Introducing Suffixaufnahmmee

33

their inflectional status; for Bopp/Dixon this was the clincher in favor of derivation. Overall, Blake's leanings were more toward derivation, but this was little more than a gut feeling. He might more profitably have asked what pre-cases implied for the feasibility of neatly compartmentalizing morphology, rather than taking the two compartments for granted and worrying about the proper allocation of pre-cases. What Blake was not very specific about was the word class of words derived or inflected by means of pre-cases. And this was perhaps not an inadvertent omission. How could one decide whether such words were nouns or adjectives if the two word classes were not really distinguished in the first place? For Australian languages the received wisdom was that adjectives are merely a subclass of nominals, inflecting like nouns and differing from them only in syntactic distribution.27 But it was precisely their syntactic distribution, acting as attributes, that had often earned nouns in the genitive (especially if they also agreed in case) the reputation of being adjectives, in Australia and elsewhere. However, informed by the analyses of individual Australian languages that had become increasingly popular since the late 1970s (including Heath 1978), Blake showed a way of accounting for the syntactic distribution and agreement behavior of nominals with pre-cases without assuming that they were genuine attributes. What seemed to be attributes from a Standard Average European perspective might really be nominals in their own right, contracting their own relationships with the predicate, as expressed by their pre-cases; what seemed complex NPs consisting of possessor and possession might not be phrases but appositive collocations of nominals in parallel, taking the same case in recognition of their parallelism (Blake 1987: 89; also 1983). A more accurate English rendering of a sentence such as (31) above might thus be something like 'the man-owned one, the one owned by a big one, the dog bit me'. Evidence for the lack of phrasal cohesion would be the freedom of linear order, the splitting up of possessor and possession by clitics and other material, and intermediate pauses. Also, the use of pre-cases followed by agreement cases extended beyond possessor phrases in some Australian languages such as Warlpiri (32), pointing to a parallel syntactic status of all nominals with pre-cases.28 (32)

ngarrka-ngku kurdu watyilipu-ngu ngurra-kurra-rlu man-Erg child (Nom) chase-Past camp-Allative-Erg 'The man chased the child to the camp'

Finck, Bork II, Lewy, and Vogt would have enthusiastically embraced this theory of a flat or non-configurational syntax corresponding to Suffixaufnahme morphology in terra australis no longer incognita. Now older and wiser, they might have sensed, though, that the notion of apposition was in need of further clarification, for it was hardly consistent with standard usage to say that two nominals in apposition with one another (hence agreeing in case) were independently contracting relationships with the predicate (hence their inner or pre-case marking).

34

Prologue

2.7.3.3. The first systematic attempt to survey multiple case marking in Australia in its entirety was made by Dench and Evans (1988). Distinguishing several functions that cases can fulfill in Australian languages—called relational, adnominal, referential, and complementizing—Dench and Evans showed that the potential for multiple case marking is considerable, with the agreement in case between a nominal in the genitive or another possessor case and the nominal denoting the possession being only one source of it. Dench and Evans were unimpressed by the arguments previously mustered in favor of the derivational nature of genitives and other pre-cases. Their relative word-internal order preceding other cases could not be decisive: a sequence of inflectional affixes outside derivational ones is not in principle objectionable; and there supposedly are clear instances in Australian languages of impeccable case suffixes being followed by unquestionably derivational ones. In Dench and Evans's opinion it was productivity, semantic regularity, and phrasal scope which really counted, and these criteria argued for inflection. The question of the word class of case-agreeing inflected words was largely irrelevant because most Australian languages arguably did not distinguish adjectives from nouns to begin with. In Dench and Evans's opinion, the drawback of Blake's notion of precases was that it failed to delimit a reasonably invariable subset of items across Australian languages. Often suffixes that would typically occur in pre-case position could also be unaccompanied by another case and mark subject or object functions; and sometimes, as in Warlpiri, virtually any case could occupy pre-case position. For Dench and Evans, then, the morphological fact of case-inflected nominals further inflecting for case, by virtue of agreement or otherwise, was not to be explained away. They adopted a cautious stance on the question of how agreement case marking could be motivated syntactically. While acknowledging the possibility of possessor and possession not forming a single NP but being in apposition, they were reluctant to accept this as the general rule for all Australian languages with Suffixaufnahme. What their examples of agreeing adnominal cases other than genitive, proprietive, and privative effectively showed, however, is how unclear the dividing line can be between attributes forming constituents with their heads on the one hand, and nominals related to one another referentially but not syntactically on the other. In (33), from Warlpiri, the nominal in the perlative case is supposed to be an attribute to the indirect object in the dative, agreeing with its head. (33)

ngarrka ka-rla marlu-ku jaarlparnka yuwurrku-wana-ku man (Nom) Nonpast-3SgIObj kangaroo-Dat intercept scrub-Per-Dat 'The man is intercepting the kangaroo on its way through the scrub'

Replacing the outer dative on the perlative nominal by zero, the marker of nominative, would suffice to relate this nominal to the subject instead ('the man is intercepting the kangaroo on his own way through the scrub'), as in

(Re-)Introducing Suffixaufnahme

35

(32) above—supposedly illustrating the referential function of case marking, where the nominals related by case agreement do not form a constituent; and the net semantic difference might be almost negligible under the circumstances. In (34), from Panyjima, the nominal in the locative is supposed to be an attribute of, hence to agree with, the nominal in the accusative: (34)

ngunha watharri-ku nyurna-yu warrapa-la-ku that look=for-Pres snake-Acc grass-Loc-Acc 'He is looking for the snake in the grass'

In (35), from Martuthunira, accusative agreement likewise relates two nominals, one denoting a part and the other the whole; although these are not supposed to form a constituent, it is only the lack of an inner non-agreement case on the nominal denoting the whole that would seem to distinguish (35) from (34). (35)

ngayu nhuwa-rninyji nyimi-i ngurnaa muyi-i I=Nom spear-Fut rib-Acc that=Acc dog-Acc I'll spear that dog in the ribs'

One is tempted to conclude from such examples that case agreement generally serves to establish the appropriate semantic connections, performing a referential function in Dench and Evans's terms, with syntactic constituency being largely immaterial. Nor would syntactic constituency seem crucial for "pre-cases." Their function might be to identify the semantic relationship of their nominals within their clauses or sentences (relational in Dench and Evans's terms); it would be immaterial, and sometimes perhaps indeterminable, whether such relationships are contracted with another nominal or with the predicate, with two nominals potentially contracting the same relationship to it. 2.7.3.4. It is unlikely that an awareness of its Old World equivalents would have sent those prospecting for Suffixaufnahme in Australia on a search for common historical origins or channels of diffusion. Clearly (at least for those with lingering doubts about proto-World), if there was an explanation for this transcontinental joint possession (provided the phenomenon indeed was the same) it had to be typological rather than genetic or areal. The two factors that competed in Australianist circles as the crucial conditions for the prospering of Suffixaufnahme were the derivational status of the genitive and appositive syntax; both had alternatively been invoked also for Greater "Caucasian." The predominantly agglutinative nature of Australian morphology, resembling that of Greater "Caucasian" or indeed the Great Eurasian Mountain Belt, might have been reckoned with too, as might ergativity; but neither were. On the other hand, as Dench and Evans (1988: 13 fn. 9) intimated, the difference between languages with and without Suffixaufnahme might as well be unrelated to almost anything else: it might

36

Prologue

simply be that double case marking is either permitted (as in Martuthunira, Kayardild, or Old Georgian) or it is not (as in Russian and almost everywhere else), for no particular reason either way. The only possible correlate of Suffixaufnahme on this agnostic view would be other kinds of double case marking. It will be remembered that Old World comparatists had sometimes mentioned case-inflecting genitives in hypostasis as an implicatum of Suffixaufnahme. Incidentally, what was never quite sorted out was whether all, almost all, most, or merely many Australian languages were practicing the custom at issue. 2.7.4. In modern times, the agglutinative type owed its most elaborate portrayal to Vladimir Skali ka. Among the mutually supportive traits of Skalicka's agglutinative "construct" were these: predominantly non-cumulative morphological markers; loose ties between stems or roots and affixes, manifesting themselves in easily recognizable word-internal morpheme boundaries, few morphologically conditioned word-internal alternations, the syllabic autonomy of affixes, and little word-internal bonding other than by vowel harmony; large inflectional paradigms and correspondingly less extensive use of function words (such as adpositions, articles, possessive and personal pronouns, conjunctions, auxiliaries); potentially long strings of affixes, especially suffixes, with the possibility of syntagmatically combining markers of the same paradigmatic category (such as case); uniform declension and conjugation, owing to the absence of synonymous inflections; little homonymy of inflections; systematic zero-expression of basic paradigmatic categories (such as nominative/absolutive and singular); a blurring of the distinction between derivation and inflection; no rigid lexical discrimination of wordclasses; an abundance of morphological modifiers of verbs; no nominal classes, and in particular no genders; phrasal marking rather than wordmarking, with each category only marked once per phrase, which precludes agreement; a coexistence of different kinds of clause construction, in particular a verbal and a nominal one; a predominance of nominalizations over finite subordinate clauses; and relatively rigid word order.29 In a characterological sketch of Hurrian, second to none in ergativity, Plank (1988) confronted Suffixaufnahme with this agglutinative construct. Although largely based on Turkish, Skali ka's admittedly idealized agglutinative profile bore a striking likeness to Hurrian, and could pass for a not-too-distorting outline of the structure of other languages with Suffixaufnahme. Even multiple case marking had been provided for, although Skali ka was thinking of inflected genitives in hypostasis formations and combinations of local cases rather than of Suffixaufnahme. And there was the rub. While overall more harmonious with the agglutinative type than with any other, granting that Skali ka's constructs were not entirely unfounded, Suffixaufnahme was in fact at odds with one agglutinative habit: the disinclination to have categories dis-

(Re-)Introducing Suffixaufnahme

37

tributed all over a phrase by virtue of agreement. This had also been the point of the long-forgotten Henri Bourgeois. Suffixaufnahme would be reconciled with the agglutinative construct if it turned out not to be ordinary NP-internal agreement like that of attributive adjectives with their heads, as Bourgeois (1909a, b) and Plank (1988) failed to notice. From this perspective too, then, the issue is seen to be one of syntax. If Hurrian, Old Georgian, Dyirbal, etc., have NPs in loose apposition rather than tight attribution, Suffixaufnahme might yet win accolades as being the ultimate in agglutination. But was not such syntax also the prerequisite of the group-inflection of old? The tacit subsuming of the group-inflecting type under the agglutinative one was not unprecedented. The factual basis of discrimination or lumping remained as unfirm as ever. 2.7.5. Australia, incidentally, was not the only continent where there were yet discoveries to be made. Thus, after some to-ing and fro-ing among the few experts, Wilhelm (1976) definitively established Suffixaufnahme also for Urartian, the shadowy close relative of Hurrian—itself meanwhile clarified considerably by Bush (1964) and Thiel (1975), among others. In IndoEuropean languages of ancient Anatolia, in particular Luwian (Hieroglyphic and Cuneiform) and Lycian (A and B), patterns were noticed, or perceived more distinctly, the extent of whose similarity to Suffixaufnahme in neighboring Hurrian-Urartian provoked some in-group discussion, notably by Mittelberger (1966), Stefanini (1969), and Neumann (1982). In eastern Africa the Ethiopian highlands, home to speakers of the Cushitic branch of Afroasiatic, also emerged as potentially fertile soil, judging by specialist accounts such as those of Palmer (1958), Hetzron (1976), and Hudson (1976). However, like similar tidings a little earlier from the Siberian tundra—inhabited, among other peoples, by speakers of Tungusic languages not averse to the occasional Suffixaufnahme, according to authorities such as Poppe (1927), Bouda (1950), and Benzing (1955a)—such news now rarely roused typological circles. Alas, the promising question No. 2.1.1.7 in Comrie and Smith's questionnaire for Croom Helm nee Lingua Descriptive Studies (1977: 33): Does the language display double case-marking? In other words, do nouns standing in a particular attributive relationship to another (head) noun exhibit, in addition to their own case-marking, case-agreement with the head noun? If this is so, describe: under what circumstances it occurs; which combinations of cases are possible; whether it is optional or obligatory to have the second case-marking.

failed to elicit a straightforwardly positive answer of the right sort for any of the languages covered before the regrettable demise of the series. But we have been straying from the past and the present perfect well into the present continuous.30 Before history continues, it is time to take stock and look at Suffixaufnahme from a systematic rather than historiographical angle.

38

Prologue

3. Nominals Interrelated: A Taxonomy

3.1. Functions and Patterns It was always against a backdrop of seemingly more ordinary forms of overt marking that Suffixaufnahme has sporadically attracted typological attention as something special—at least initially, with its apparent anomaly rendering it liable to be rationalized as not really so remarkable after all. Before passing judgment on its curiosity value or beginning to explain it away, however, it is appropriate to ask to what Suffixaufnahme actually is an alternative. Bearing in mind rationalizations like those pioneered by Bopp and Finck, it would be unwise to adopt too narrow a view on parts of speech and on the nature of syntactic relationships in surveying the range of options that is to include Suffixaufnahme. The following taxonomy of marking patterns is therefore intended to be neutral (a) as to whether the nominal to be related to another is a noun or something else (such as a derived adjective), and (b) as to whether its relationship is one of attribution or of some other kind (such as apposition)—and indeed, whether this relationship is that of an immediate adnominal constituent or not. The nominals in relation will be neutrally referred to as "primary" and "secondary." Attributes are prototypical secondaries vis-a-vis their heads, and they provide the bulk of the subsequent examples; but on referential and distributional grounds, secondary rank is also justified for the appositum in apposition or for a nominal indirectly related to another as a secondary predicate or the like. In encoding expressions containing primaries and secondaries there are two complementary tasks (as Heinrich Winkler had not failed to grasp): to identify the relations which the two nominals have in the expression in which they occur, and to indicate which are the nominals to be related to one another. The relations identified may be those of secondary (No. 1 in Table 1.1) or of primary (No. 2) or of both (No. 3), with the markers normally associated, morphologically or syntactically, with the respective nominals themselves. Relatednessindicators may occur on the secondary (No. 4), reflecting some property of the primary that it belongs with (such as its number, gender/class, person, or case); or they may be on the primary (No. 5), reflecting some property of the secondary that it belongs with, or on both (No. 6). If relatedness-indicators are specific to particular constructions, they may thus also serve to identify relations. (For example, in languages where attributes are indicated on nouns and subjects or objects on verbs, such person-number marking is sometimes formally differentiated, thus identifying its carriers as either a nominal head or a verbal predicate.) The two basic marking techniques may be used individually (Nos. 1-6) or in combination (Nos. 7-15). In combined relationshipidentification and relatedness-indication the marking may be separate, as shown schematically in Table 1.1, or cumulative, with single markers being simultaneously relationship-identifying and relatedness-indicating. There may alternatively be markers of the entire construction, linking primary and secondary without forming a morphological co-constituent of

(Re-)Introducing Suffixaufnahme

39

Table 1.1. Patterns of Marking

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.

(=l.+4.) (=l.+5.) (=l.+6.) (=2.+4.) (=2.+5.) (=2.+6.)

13.

(=3.+4.)

14. 15. 16. 17. 18. 19. 20. 21.

(=3. + 5.) (=3.+6.)

Primary

Secondary

X X-prim X-prim X X-y X-y X X-y X-y X-prim X-prim-y X-prim-y

Y-sec Y Y-sec Y-x Y Y-x Y-sec-x Y-sec Y-sec-x Y-x Y Y-x

X-prim

X-prim-y X-prim-y X X X X X X

Y-sec-x

Link Link-x Link-y Link-x-y # +

Y-sec Y-sec-X Y Y Y Y Y Y

either (No. 16); such links ("associative" markers or lexical items such as 'thing', 'possession/belong', 'place') may themselves be combined with indicators of the nominals thus related (Nos. 17-19), and the nominals may carry additional markers of their relation. Lastly, relationship-identification and relatedness-indication may be entrusted to the linear ordering of primary and secondary, joined in syntactic (No. 20) or morphological (No. 21) construction, either without any further segmental or suprasegmental marking or accompanied by any of the marking patterns above. Table 1.1 lists all twenty-one marking patterns that can be distinguished along such functional lines.31 Lower-case letters x and y symbolize any properties of the corresponding primaries or secondaries (represented by upper-case letters X and Y) that may be utilized for relatedness-indication; -prim and -sec symbolize any markers of the respective relations. The more plentiful the marking, the safer. Pattern No. 15 leaves little to be desired: either nominal is marked for its relation as well as for its partner, precluding any misinterpretation. While in general one would expect languages to strike a balance between clarity and economy and to avoid such formal overexpenditure, it is especially in attributive constructions, rather than in verbal ones, that relational non-ambiguity is at a premium.32 Sometimes the safety precautions of double or even triple marking, as in Nos. 3, 7, 8 and 9, 13, 14, are indeed being taken; nonetheless, it is still some of the formally more generous patterns that are cross-linguistically least favored,

40

Prologue

with Nos. 10 and 11 (both double marking), 12 (triple marking), and 15 (quadruple marking) being unrepresented in my current sample.33 In the following subsection all attested patterns except Nos. 20 and 21 are illustrated and sparsely annotated, with examples mostly culled from standard reference grammars (not specifically acknowledged here). Many, perhaps most, languages utilize more than one pattern; sometimes these alternatives are mentioned, though normally without dwelling on the factors—such as the alienability or inalienability of possession—that determine the choices between them.34

3.2. Illustration No. 1:

X Y-sec

(36)

Latin (Italic, Indo-European) domu-s/- r g-is house-NomSg/DatSg king-GenSg '(to) the house of the king' domu-s fr tr-is r g-is house-NomSg brother-GenSg king-GenSg 'the house of the brother of the king'

(37)

Mangarayi (ungrouped, Australian) na-muyg aya-Bagur iya NomMasc-dogMasc GenFem-Bagur iyaFem 'Bagur iya's dog' (alternatively No. 8 (71))

(38)

Igbo (Kwa, Niger-Congo) isi eyu head goat:Ass 'the head of a goat' (both words have high tone on all syllables, as in isolation, but there is a pitch-fall or 'downstep' on the attribute, functioning as a suprasegmental associative marker)

(39)

Israeli Hebrew (Northwest Semitic, Afroasiatic) talmid-im sel zamar-im studentMasc-PlMasc of singerMasc-PlMasc 'students of singers' (known as periphrastic genitive; alternatively Nos. 2 (45), 8 (69), and 14 (77))35

(40) English (Germanic, Indo-European) coast-al waters mother-ly kisses royal [< king-ly] palace (41)

Sumerian (isolate, Ancient Near East) e lugal-ak house king-Gen 'house of the king'

(Re-)Introducing Suffixaufnahmehhh

41

e lugal-ak-a house king-Gen-Loc 'in the house of the king' (cf. also (14) above; alternatively No. 8 (66)) e ses lugal-ak-ak-a house brother king-Gen-Gen-Loc 'in the house of the brother of the king' (42)

(Late) Elamite (isolate, Ancient Near East) siyan d.Insusinak-ni(-)ma temple Inshushinak-Gen(-)in 'in Inshushinak's temple' (alternatively Nos. 4 (50) and 5 (61))

(43)

Kanuri (East Saharan, Nilo-Saharan) tata kamu-ve-qa ruskma son woman-Gen-Acc I=have=seen 'I have seen the woman's son' (alternatively No. 8 (71))

Latin, Mangarayi, Igbo, and Israeli Hebrew illustrate the marking of secondaries by suffixes, prefixes, tone, and prepositions respectively, with the affixes happening to cumulate case and another inflectional category (number in Latin, gender/class in Mangarayi). English will do to show that the marking of secondaries may be derivational rather than inflectional, with basic nouns uncontroversially recategorized as adjectives. Sumerian, (Late) Elamite, and Kanuri—as well as Aranda (22) and Alyawarra (30), appearing in Section 1— differ only superficially from Latin, insofar as the markers of secondary status, like all other case marking, occur at the end of the whole phrase of which they are part rather than being bound to their nominals.36 If a secondary follows its primary, it will accordingly acquire the case which encodes the external relationship of the primary in addition to its own secondary marking; and since attribution is recursive, a batch of cases may accumulate on the last member of the entire phrase (three are exemplified for Sumerian).37 No. 2: X-prim Y (44)

East Aramaic (Northwest Semitic, Afroasiatic) bne mata sons=Const village 'villagers' (bne being the construct state of bnuni)

(45)

Israeli Hebrew (Northwest Semitic, Afroasiatic) talmid-ey zamar-im studentMasc-ConstPlMasc singerMasc-PlMasc 'students of singers' (known as synthetic genitive; alternatively Nos. 1 (39), 8 (69), 14 (77))

(46)

Arbore (Lowland East Cushitic, Afroasiatic) hikic-i hoggattu axe-Head Laborer

42

Prologue '(the) laborer's axe' (ef. (47) at No. 3; vowel-final nouns take none of the head suffixes, thus exemplifying No. 21, disregarding possible tonal marking of attribute alone or of both attribute and head)

Affixal or other markers identifying the relation of primary rather than secondary—prominent in, though not exclusive to, Afroasiatic—are traditionally referred to as "states" rather than "cases;" a nominal is in the construct state, contrasting with the absolute state, when accompanied by an attribute. In Israeli Hebrew, the suffix which identifies the noun as a primary also marks its number and gender. No. 3:

X-prim Y-sec

(47) Arbore (Lowland East Cushitic, Afroasiatic) hikic-i hoggattu-t axe-Head laborer-Gen '(the) laborer's axe' (genitive is optional, ef. (46) at No. 2) (48) Ndunga (Adamawa-Eastern, Niger-Congo) ma-l-a ta gulu head-ClassNumberIII-Const of pig 'head of a pig' (to may be a linking particle rather than a co-constituent of the secondary, in which case Ndunga would exemplify pattern No. 16, with additional marking of the primary function; alternatively No. 13 (75)) (49) Pitta-Pitta (Pama-Nyungan, Australian) kupakupa- a titi-wara-lu old=man-Dat brother-Possessed-Erg 'the old man's brother'

Markers identifying a nominal as a primary accompanied by a possession, as here illustrated from Pitta-Pitta and glossed as 'Possessed', are sometimes referred to as cases among Australianists; on this analysis, double case marking may ensue from primary nominals further attracting the case which encodes the external relation of the entire phrase, as illustrated in (30). No. 4:

X Y-x

(50) Elamite (isolate, Ancient Near East) pahi-r sunki-p-r(i) protector-3SgAnim king-3Pl-3SgAnim 'protector of kings' (alternatively Nos. 1 (42) and 5 (61)) (51) Pitjantjatjara (Pama-Nyungan, Australian) ngayu-la kata-ngka me-Loc head-Loc

(Re-)Introducing Suffixaufnahme

43

'on my head' (such case agreement is only used with body parts in part-whole relationships) (52)

Ritharngu (Pama-Nyungan, Australian) nu- u dawal you-Gen country (Abs) 'your country' nu:-kala-li? dawal-li? you-Lig-All country-All 'to your country' (such case agreement is only used with locative, allative, ablative, and pergressive as primary cases)

(53) Alawa (Maran, Australian) guyumu-ndu gudar-if nose-Loc hill-Loc 'along the point of the hill' (part obligatorily precedes whole; with alienable possession: X-Case Y-Gen, i.e. No. 1) (54)

(55)

Classical Armenian (Indo-European) bazmowt'-eamb zawr-awk'-n Hay-oc' crowd-InsSg force-InsPl-Def Armenian-GenPl 'with the crowd of the forces of the Armenians' (ef. (17); attested, as an alternative to No. 1, only with ablative or instrumental as cases of the primary; otherwise No. 1, as exemplified by the second attribute, is obligatory) Middle Hittite (Anatolian, Indo-European) Kumarbiy-aza E-ir-za Kumarbi-AblSg house-AblSg 'out of Kumarbi's house' D

(56)

Ancient Greek (Hellenic, Indo-European) Deiopite-n outasen omo-n Deiopites-AccSg he=wounded shoulder-AccSg 'He wounded Deiopites' shoulder; he wounded Deiopites at the shoulder' (an alternative to No. 1 in certain constructions, often involving nominals denoting parts and wholes, where it is unclear which is primary and which is secondary and whether they are co-constituents38)

(57)

Latin (Italic, Indo-European) sen-em dein dedolabo viscer-a old=man-AccSg then I=will=hew intestine-AccPl 'then I will hew away the intestines of the old man' (alternative to No. 1 under roughly the same circumstances as in Greek, (56)39)

Among the morphological categories through which secondaries may indicate which primaries they are to be related to, those illustrated from Elamite—person, number, and animacy—are fairly common. The case of the primary is utilized far less frequently for this purpose, and apparently

44

Prologue

never fully productively. There are probably few languages other than the Australian and ancient Indo-European ones listed at (51)-(55)—to which Martuthunira (35) must be added—in which the sole (segmental) marking of secondarihood consists in the secondary nominal, by a process that has been called case attraction or assimilation, being in the same case as the corresponding primary. (Part-whole constructions as in Ancient Greek (56) and Latin (57) are found more frequently.) As Pitjantjatjara, Alawa, and Latin clearly show, it is not necessarily the actual case marker of the primary that gets copied onto the secondary, but rather the case category, realized by the allomorph that is appropriate for the noun on which it ends up. As the Classical Armenian and Latin examples demonstrate, when case is expressed cumulatively with number, number need not be involved in the transfer of categories from primary to secondary. No. 5: X-y Y (58)

Turkish (Turkic, Altaic) oban kiz-i shepherd girl-3SgPoss 'the shepherd-girl' (alternatively No. 8 (67))

(59) Yagua (Peba-Yaguan, Ge-Pano-Carib) sa-rooriy Tomaasa 3SgClass-house Tom 'Tom's house' (60) Early Modern English (Germanic, Indo-European) the daulphin of France his power Juno hir bedde estrangers their ships (alternative to No. 1, possibly owing to the reanalysis of the genitive suffix -es as the possessive pronoun (h)is) (61)

Elamite (isolate, Ancient Near East) f.d.Nahhunte-untu par-e Nahhunte-untu descendants-3 'Nahhunte-untu's descendants' (-e being number- and gender-neutral; alternatively Nos. 1 (42) and 4 (50))

The categories of secondaries that are utilized by primaries—perhaps accompanied by an appropriate pronoun—in order to indicate their partners frequently include person, number, and gender or class. Theoretically the case of the secondary could also be used for this purpose, in analogy to No. 4; but a relationship-identifying case marker on the secondary would define a different pattern to begin with (see further at No. 8, with reference to Berta).

(Re-)Introducing Suffixaufnahme

45

No. 6: X-y Y-x

This pattern is probably found in Gola (West Atlantic, Niger-Congo), where according to Ultan (1978: 22) possessors may be marked for the noun class of the possessee and vice versa. In Mono (Uto-Aztecan; Langacker 1977: 25), marking is normally according to No. 5, with relatedness-indication on possessees in terms of person and number; but when possessor and possessee are discontinuous, being interrupted for instance by a subject pronoun gravitating to its preferred second position in the sentence, the possessor in addition copies the case suffix of the possessee: (62)

pahapi"ci-na nii a-"ki"ki-na a-na-"pu"ni-"ti bear-Acc I 3SgPoss-feet-Acc it-nearly-see-Tense 'I saw the bear's tracks'

No. 7: X Y-sec-x (63)

Luo (Chari-Nile, Nilo-Saharan) duong' ma-r piny greatness=Sg Prep-Sg land 'the greatness of the land' kinde ma-g yueyo times=PI Prep-Pi rest 'times of rest' (alternatively, especially with inalienable possession, No. 2, with primary noun in construct state and without preposition)

(64)

Swahili (Bantu, Niger-Congo) m-toto w-a Ali ClassSg-child ClassSg-of Ali wa-toto w[a]-a Ali ClassPl-child ClassPl-of Ali 'Ali's child/children' ki-tanda zh-a Ali ClassSg-bed ClassSg-of Ali 'Ali's bed'

The categories through which secondaries, already identified as such by adpositions (or associative markers), may point to their corresponding primaries include number (also utilized by Tsakhur (11)) and class or gender. In the most obvious instances of case being used for the same purpose, the marking of secondarihood appears to be derivational rather than inflectional: the denominal possessive adjectives of Latin, Russian, Finnish, and scores of other languages agree with their head nouns in, among other things, case: (65)

Latin (Italic, Indo-European) domu-s reg-i-us houseMasc-NomSg king-Adjct-NomSgMasc

46

Prologue domu-i reg-i-o houseMasc-DatSg king-Adjct-DatSgMasc '(to) the royal house'

Insofar as the case of the primary was among the relatedness-indicating categories on secondaries marked as such, the bulk of the illustrations in Section 1—from Human (8, 28), Old Georgian (1, 2, 9, 10, 15, 18, 23), Bats (12), Tsakhur (26), Ancient Greek (16), Erzya Mordva (21), Kayardild (25), Dyirbal (29), Kalkatungu (31), Warlpiri (32, 33), and Panyjima (34)— likewise fall under this pattern. To determine what is nevertheless special about these, is the aim of this volume. For the time being, a reference back to German Romany (5-7) or Slovak (24) ought to suffice as a reminder that differences may not always be clearcut. No. 8:

X-y Y-sec

(66)

Sumerian (isolate, Ancient Near East) lugal-ak e-ani king-Gen house-3SgAnim 'the king's house' (alternatively No. 1 (41))

(67)

Turkish (Turkic, Altaic) oban-in kiz-i shepherd-Gen daughter-3SgPoss 'the shepherd's daughter' (alternatively No. 5 (58))

(68)

Colloquial German (Germanic, Indo-European) dem Konig sein Haus the=DatSgMasc kingMasc=DatSg Poss=3SgMasc house 'the king's house'

(69)

Israeli Hebrew (Northwest Semitic, Afroasiatic) beit-o sel Dan house-3SgMascPoss of DanMasc beit-a sel Miriam house-3SgFemPoss of MiriamFem 'Dan's/Miriam's house' (alternatively Nos. 1 (39), 2 (45), 14 (77))

(70)

Ngandi (Gunwingguan, Australian) ma-warngurra'-gu gu-rerr-'nguthayi Class-bandicoot-Gen Class-camp-its 'the bandicoot's camp'

(71)

Mangarayi (ungrouped, Australian) na-bugbu -gu na-banam-nawu-yan GenMasc-old=manMasc-GenMasc LocNeut-campNeut-3SgPoss-LocNeut 'in the old man's camp' (alternatively No. 1 (37))

(Re-)Introducing Suffixaufnahme

47

(72) Kanuri (Saharan, Nilo-Saharan) ali-be far-nzS Ali-Gen horse-3Sg 'All's horse' (alternatively No. 1 (43), with alienable possession only) ali-be ya-nza Ali-Gen mother-3Sg ya-nza ali-be(-qa) mother-3Sg Ali-Gen(-Acc) 'All's mother' (inalienable possession lacks alternative No. 1)

Several formal varieties of secondarihood marking are here illustrated in combination with relatedness-marking on the primary, which in turn may be in terms of the secondary's person, number, animacy, gender or class (without full exemplification of relevant contrasts). Relatedness-marking in terms of the secondary's case—in a way the mirror image of No. 5—is probably an option in Berta, although the facts are unclear (to the extent that it must remain dubious whether there is any case marking in the first place). (73) Berta (Chari-Nile, Nilo-Saharan) sul(-a) ma:b-u house(-Gen) man-Gen 'house of the man' (exemplifying No. 1 when the genitive is omitted from the head)

No. 9:

X-y Y-sec-x

(74) Huallaga Quechua (Andean, Andean-Equatorial) hipash-nin-ta kuya-: Hwan-pa-ta daughter-3Poss-Acc love-1 Juan-Gen-Acc 'I love Juan's daughter' (only if possessor is separated from possessed; otherwise No. 8: Hwan-pa hipash-nin-ta kuya-: Juan-Gen daughter-3Poss-Acc love-1)

While the primary relates to its secondary in terms of person, the secondary, identified as such by its genitive, indicates its primary in terms of the case encoding the primary's external relation (accusative in this example). No. 10: No. 11: No. 12:

X-prim Y-x X-prim-y Y X-prim-y Y-x

All three are unattested in the present sample.

48

Prologue

No. 13:

X-prim Y-sec-x

(75)

Ndunga (Adamawa-Eastern, Niger-Congo) k-a ta kumu ko-k-a ClassNumberV-it of forest tree-ClassNumberV-Const 'the tree of the forest' (alternatively No. 3 (48))

(76)

Ge'ez (South Semitic, Afroasiatic) as'-a (za) hsywat treeMasc(Sg)-Const (Particle=MascSg) life 'the tree of life' (the particle optionally marking the secondary is also used as a relative pronoun; if omitted, we have pattern No. 2)

In addition to construct-state primarihood marking and the marking of secondarihood by an attributive preposition or particle, the secondaries indicate their primaries in terms of number and gender or class, marked either on a special carrier pronoun accompanying the secondary or on the particle itself. No. 14: (77)

X-prim-y Y-sec

Israeli Hebrew (Northwest Semitic, Afroasiatic) talmid-ey-hem sel zamar-im studentMasc-ConstPlMasc-3PlMasc of singerMasc-PlMasc 'students of singers' (known as double genitive; alternatively Nos. 1 (39), 2 (45), 8 (69))

Relatedness-indication on the primary is in terms of the secondary's person, number, and gender. No. 15:

X-prim-y Y-sec-x

Unattested in present sample. No. 16: (78)

Balanta (West Atlantic, Niger-Congo) bko ne esi head of cow 'the cow's head'

No. 17: (79)

X Link Y

X Link-x Y

Biafada (West Atlantic, Niger-Congo) boofa babe usa headclass of=Class man nnaga bbe usa cowclass of = Class man 'the man's head/cow'

(Re-)Introducing Suffixaufnahme No. 18: (80)

X Link-y Y

Nalu (West Atlantic, Niger-Congo) ki am nam-cel head of=Sg Sg-man ki ga b -cel head of=Pl Pi-man 'the man's/men's head'

No. 19: (81)

49

X Link-x-y Y

Laragia (Larakian, Australian) mangulmili-ma bilo-va bi -nagi-ma canoe-ClassIV man-ClassI ClassI-3SgPoss-ClassIV 'the man's canoe' (only if possessor is human and alienable possession does not belong to the same class but requires an independent possessive pronoun; otherwise No. 18: gwijarma -gwa madjir-a bi -nagi feather-ClassV bird-ClassI ClassI-3SgPoss 'the bird's feather')

(82) Maasai (Eastern Sudanic, Chari-Nile, Nilo-Saharan) l-c r 15 1-payyan MascSg-friend of=Mascprim=MascSgSec Masc-elder 'friend of the elder' il-c r -ta 1 1-payyan MascPl-friend-Pl of=MascFrim=MascSgSec Masc-elder 'friends of the elder' l-c r 1-payyan-i MascSg-friend of=Mascprim=MascPlSec Masc-elder-Pl 'friend of the elders' l-c r l -k rai MascSg-friend of=Mascprim=FemSgSec FemSg-child 'friend of the child' en-toki 5 1-payyan FemSg-thing of=Femprim=MascSgSec Masc-elder 'thing of the elder' en-toki 1-payyan-i FemSg-thing of=Femprim=MascPlSec Masc-elder-Pl 'thing of the elders' Relatedness-indication on a linking element, itself a pronoun agreeing in person and number with the primary in Laragia, is preferably in terms of class and/or number in patterns Nos. 17-19. In Maasai the linking particle relates to the primary solely in terms of gender and to the secondary in terms of both gender and number, consisting of 1-10- for masculine/feminine primary, and -51 - for masculine/feminine singular secondary and -.9.9 for plural secondary.

50

Prologue

4. Tema con variazioni

4.1. The Prototype Suffixaufnahme has traditionally been seen as a special instance of pattern No. 7, X Y-sec-x, although there has been the occasional temptation, not yet overcome completely,40 to lump it with "Suffixhaufung," i.e. that variety of pattern No. 1, X Y-sec, in which all case marking accumulates at the end of a phrase, with phrase-final secondaries thus carrying the case of their primary (not itself case-marked) in addition to their own. Now, in order to avoid underestimating or exaggerating its peculiarity, it is necessary to get some preliminary idea of (a) the relationship of this pattern to others, (b) its relationship to other instances of the same pattern, and (c) its range of formal variation. Reassuringly, there is some resemblance among the prospects set out in the preview to follow, Moravcsik's final retrospective, and what is sandwiched in between, although the viewpoints and highlights will often be found to differ. In what can be recognized as the prototype of Suffixaufnahme there is a nominal consisting of (what there are reasons to believe is) a noun or a personal pronoun in a relationship of (what there are reasons to believe is) attribution to another nominal, in (what there are reasons to believe is) the basic form that attributive constructions take in the language concerned, with the head nominal morphologically marked by a case suffix for its external syntactic relation, with the (bona fide) attribute carrying (what there are reasons to believe is) the inflectional marking of genitive case, and— crucially—with the attribute itself in addition separately marked for the same case, plus perhaps further categories expressed by suffixes, as the head. Words in boldface refer to parameters for possible variation; parentheses forebode controversy. Marking patterns in Chukchi and perhaps other Chukchi-Kamchatkan languages (whose "possessive" forms largely correspond to Indo-Europeanstyle genitives), Old Georgian and perhaps further Kartvelian languages and dialects, Bats (pending disconfirmation of the data in Schiefner 1856), Hurrian and Urartian, Awngi, and many Australian languages—and nowhere else—can reasonably, if not always uncontroversially, be supposed to answer to this description. The following samples exemplify what is covered extensively later in this volume by Koptjevskaja-Tamm for Chukchi, Boeder for Kartvel, Wilhelm and Wegner for the Ancient Near East, Hetzron and Aristar for Cushitic, and Schweiger, Austin, Dench, and Evans for Australian: (83)

Chukchi (Chukchi-Kamchatkan) Rult -n-ine-k tumg- k Rultyn-Sg-Poss-Loc friend-Loc 'at Rultyn's friend'

(Re-)Introducing Suffixaufnahme (84)

Old Georgian (Kartvel, Caucasian) perx-n-i [. . .] kac-isa-n-i foot-Pl-Nom [. . .] man-Gen-Pl-Nom '(the) feet of the man'

(85)

Bats (Northeast or North-Central Caucasian) bakhe-v thhe dad Daivth-e-v mouth-Instructive our father David-Gen-Instructive 'through the mouth of our father David' (=12a above)

(86)

Hurrian (Hurrian-Urartian) sen-iffu-ue-ne-z asti-z brother-lSgPoss-Gen-ArtSg-Erg wife-Erg 'my brother's wife'

(87)

Urartian (Hurrian-Urartian) Haldi-i-ne-ni alsuisi-ni Haldi-Gen-ArtSg-Ins greatness-Ins 'through the greatness of Haldi'

(88)

Awngi (Central Cushitic, Afroasiatic) woliji-kw-des aqi-kw-des an-ka-des old-Genpl-Abl man-Genpl-Abl house-Pl-Abl 'from the old man's houses'

(89)

Yidiny (Pama-Nyungan, Australian) wagal-ni- gu gudaga- gu wife-Gen-Erg dog-Erg '(my) wife's dog'

51

With the exception of Chukchi (and possibly Kerek, Koryak, Alutor, and Itelmen, its relatives) and Awngi, all these languages or families with prototypical Suffixaufnahme are old acquaintances from the historical narrative above.

4.2. More or Less Relational Marking 4.2.1. Case marking of nominals for secondarihood as well as for the case of the primary they belong with is to be found in more than one pattern of Table 1.1. No. 7, not always easily distinguished from No. 17, X Link-x Y (with the linking element generally tending toward the secondary—but there are presently no such examples with case being utilized for relatedness-indication), shows such double marking of secondaries without further adornments, while Nos. 9, 13, 15, and 19, containing Y-sec-x, add markers on the primary (or on the link) identifying it as such and/or indicating its partner. As long as their Y-sec-x part meets the specifications of Suffixaufnahme, such patterns with more lavish marking presumably warrant being lumped

52

Prologue

together with their more economical counterparts; in functional terms, at any rate, their similarity to Nos. 7/17 is evidently closer than that of the "suffixaccumulating" variety of No. 1, not practicing relatedness-indication at all. Empirically, though, such patterns turn out to be very unpopular. Representing the ultimate in formal marking, No. 15, X-prim-y Y-sec-x, is probably too cumbersome to be used anywhere. No specimen of No. 13, X-prim Y-secx, is on record where relatedness-indication is by case; in the examples given above the head is in the construct state and lacks case marking for its external syntactic relation. Nos. 9 and 19, X-y Y-sec-x and X Link-x-y Y, are in fact attested, but the kinds of Suffixaufnahme found in Quechua and Maasai deviate from the prototype, as will be seen presently.41 4.2.2. Comparing No. 7 (and No. 17) to patterns whose marking is minimally less extensive, No. 1, X Y-sec (and No. 16, X Link Y, in respect to No. 17) only differs from it by doing without relatedness-indication, while No. 4, X Y-x, instead forgoes relationship-identification on the secondary. Analogously, all further patterns containing Y-sec-x are more extensive by one degree than corresponding patterns containing either Y-sec or Y-x. However easily such hyperclasses can be defined by downplaying or ignoring individual markers present on secondaries, they need to prove their worth by figuring in typological generalizations and perhaps also in diachronic scenarios of transitions between classes. A priori, at any rate, pattern No. 4 would seem to be a good candidate for higher-level conflation with No. 7, with case apparently being the least common choice among the categories utilized for relatedness-indication in both patterns. Of the few languages of pattern No. 4 that opt for case, virtually all (only Ancient Greek and Latin excepted) are relatives and/or neighbors of languages of Australia, the Caucasus, or the Ancient Near East in which Suffixaufnahme occurs in its prototypical form, owing to additional relationship-identification on secondaries—which is absent in the relevant constructions of Pitjantjatjara, Ritharngu, Alawa, Classical Armenian, Middle Hittite, Ancient Greek, and Latin. These areal and genetic distributions suggest that the relatedness-indicating part of the pattern may have been borrowed, or alternatively that the relationship identifying part may have been lost. Insofar as the case marking on secondaries in pattern No. 4 is to be accounted for in terms of a process of case attraction or assimilation, one might argue that secondaries are indeed case-marked for their own relationship "underlyingly" and that this case marking is then exploited, following the genius loci, for the different purpose of relatedness-indication, undergoing the appropriate changes. The difference of such case attraction from Suffixaufnahme would thus be that nominals being assigned two cases—once by government, once by agreement—give the latter requirement precedence over the former (see further Moravcsik on this theme). When case is expressed on its own, secondaries may or may not accommodate two such cases; when case is cumulated with number or another category, they tend to

(Re-)Introducing Suffixaufnahme

53

tolerate no more than one. Doubly-case-marking Old Georgian would be exceptional in this respect, since Suffixaufnahme is not precluded by case being cumulated with number.42 There may in fact be transitions between the suffix-accumulating varieties of No. 1 and No. 7, the two patterns sometimes lumped together wholesale under the label of Suffixaufnahme understood broadly. Thus, basically, Oromo (or Galla) is purely relationship-identifying: it has secondaries in the genitive, and a case suffix (or perhaps rather an enclitic) marking an external relation of the entire NP other than those of subject, direct object, and attribute is added to the secondary in the genitive as the last member of the NP (Tucker and Bryan 1966: 515, Gragg 1976: 182-184): (90)

a. ilm hayy-uu son (Abs) chief-Gen 'the son of a chief (direct object) b. ilm hayy-uu-ti son (Abs) chief-Gen-Dat 'to the son of a chief

When they are subjects, nouns are in the nominative, and this relational marking—unlike dative, etc.—is not passed on to the last word of the NP: (90)

c. ilm-i hayy-uu son-Nom chief-Gen 'the son of a chief (subject)

By this unusual turn, which seems unparalleled in Lowland East Cushitic (in Somali, for instance, primaries are always in the absolutive/accusative and all external case marking is added to the genitive marking of secondaries) or indeed elsewhere, the external relation of NPs is thus marked twice: on the primary, where nominative contrasts with absolutive depending on whether the NP is subject (90c) or non-subject (90a/b); and on the secondary, whose own genitive suffix is followed, or not, by a further case marker depending on whether the NP is an indirect or oblique object on the one hand (90b) or a subject or direct object on the other (90a/c). In a way, such interdependent case markings—where a dative or other post-genitival case on the secondary implies absolutive on the primary, and nominative on the primary implies no further post-genitival case on the secondary—serve the purpose of relatednessindication. It is when secondaries carry two overt cases, as in (90b), that the similarity to Suffixaufnahme is closest.

4.3. Aufnahme of Case or Other Categories Owing to its functional rationale, the particular categories used for relatednessindication were disregarded in the typology of Table 1.1. Thus Luo and Swahili,

54

Prologue

for example, as well as Finck's Subya, join Chukchi, Old Georgian, Hurrian, Awngi, and Yidiny patternwise; and their application for membership in the exclusive circle of Suffixaufnahme languages—or indeed, a little less exclusively, of Aufnahme languages, considering that the elements resumed in Swahili are prefixes—might therefore be sympathetically considered. Occasionally such languages have indeed been admitted: recall Ferdinand Bork on Tsakhur and Elamite and Grande on Mishnaic Hebrew. Of course, subtypes can easily be distinguished depending on whether or not case is among the categories put to such use. Again, the question is whether this is a distinction that matters. A positive answer is suggested by the narrow view that has traditionally tended to be taken on what is to be regarded as Suffixaufnahme proper. However, as far as I can see, the point has never really been argued that Luo, Swahili, Subya, Mishnaic Hebrew, and their kind differ from the founder members of the Aufnahme club in anything other than the choice of relatedness-indicating categories—for instance, in the syntax of attribution. (Finck had surmised, though, that the relevant constructions in Swahili's relative Subya are appositive rather than attributive.) In fact, it does not assist this argument that this choice is virtually forced on a language, depending on the availability of case marking on primaries in the first place.43 On the other hand, the comparative infrequency of relatedness-indication on secondary nouns in terms of case, vis-a-vis the ordinariness and wider genetic and areal spread of the agreement categories of number, gender/class, or person, might bespeak some special status. Diverse structural correlates of relatedness-indication by means of case or non-case would also underline the significance of this distinction; however, previous suggestions to this effect, recorded in Section 2, have not been unambiguous and need thorough checking. What the present volume has to offer in this respect sheds little light on the typological unity or diversity of case and non-case Suffixaufnahme, since the latter is largeley disregarded. What all examples illustrating pattern No. 7 above have in common is that the categories concerned, case or non-case, are overtly marked on the primaries themselves. This is not necessarily so for all relatedness-indication: distinctions of person and of gender/class of the primary nominal may overtly appear only on the secondary (as in Early Modern English Jupiter/Juno hislhir bedde, although lacking relational identification on the secondary). An even wider view of Aufnahme phenomena might also take into account such categories lacking overt marking on primaries, suffixal or otherwise.

4.4. Aufnahme of Case Plus Other Categories If relatedness happens to be indicated on secondaries by case, it is conceivable that case may be the only category involved in Suffixaufnahme, or that it may be accompanied by other categories. While in most of the prototypical examples given above, case ostensibly was the only category of the primary to reappear on the secondary noun, plural also did so in the Old Georgian example (84).

(Re-)Introducing Suffixaufnahme

55

Obviously, as is shown by a further example from Chukchi (91), if case is expressed cumulatively with other categories, in particular number, such unsegmentable suffixes cannot possibly be torn apart, and any process pertaining to such cumulative markers is bound to involve both categories (or neither). (91)

inqej- rg-ine-t tumg-3t boy-Pl-Poss-AbsPl friend-AbsPl '(the) friends of (the) boys'

Conceivably, the category cumulated with case could simply be disregarded in indicating relatedness, with case mapped correctly onto the secondary and with number returning to unmarked singular value (with AbsSg instead of AbsPl as the final suffix of 'boy' in (91)); but this never actually happens.44 Of course, in the examples above where case is the only category involved, the primaries were not overtly marked for number, with zero forms functioning as singulars in typically agglutinative manner. When separate plural suffixes are added to those of case, both are seen to be subject to Suffixaufnahme in the Old Georgian example (84), and this indeed appears to be the general rule: there is no evidence from any of the languages with prototypical Suffixaufnahme to suggest that case may part company with overt number, as hypothetically illustrated in (84'). (84')

*perx-n-i kac-isa-j foot-Pl-Nom man-Gen-Nom '(the) feet of the man'

Note that in Awngi (88) the plural suffix of the primary, although not literally repeated on the secondary, is nevertheless reflected in the choice of its genitive alternant. In Hurrian and Urartian the elements between the secondary's genitive and the case copied from the secondary, glossed above as singular "articles," in fact reflect the primary's number rather than the secondary's. As to Tsakhur, where Suffixaufnahme is less prototypical (see Sections 4.12 and 4.15), the observations of Bork (11) and Mel'cuk (26) combine to show that here too both case and number are involved in relatedness-indication. In case attraction, by contrast, the number of the primary may be dissociated from its case. Recall that in Classical Armenian, when secondaries exchange their genitive so as to conform to the ablative or instrumental of the primary, exemplifying pattern No. 4, their number as such remains uninfluenced by the primary's number: (92) bazmowt'-eamb zawr-awk'-n . . . crowd-InsSg force-InsPl-Def . . . 'with the crowd of the forces . . .' (=54)

While it is not uncommon for nouns to inflect for number but not for case, case inflection without number inflection is rare. Number is thus very likely to

56

Prologue

be a second inflectional category whenever case is one. If Suffixaufnahme is appropriately characterized, then, as inflection-Aufnahme in languages where noun inflection includes case, case-Aufnahme is very likely to be accompanied by number-Aufnahme.45 When number escapes attraction along with case, as opposed to Aufnahme, it is presumably because referential information would otherwise be lost. If inflection comprises several subsystems of case marking, with cases on single nominals potentially coming in several layers, all of them ought to undergo Aufnahme on the principle that it applies generally to all inflections present on primaries—which is what they do in the most profusely casemarking language known: (93)

Kayardild (Tangkic, Australian) maku-ntha yalawu-jarra-ntha yakuri-naa-ntha dangka-karra-nguni-naa-ntha mijil-nguni-naa-nth woman-Purp catch-Past-Purp fish-M.Abl-Purp man-Gen-Ins-M.Abl-Purp netIns-M.Abl-Purp 'The woman must have caught fish with the man's net'

The secondary (in boldface) adds, after its own genitive, the relational case (instrumental) of its primary as well as the modal ablative and the oblique case (purposive) that the primary has acquired owing to its further syntactic associations in Kayardild (cf. (25) above, and see further Dench and Evans 1988 and Evans in this volume). In principle Suffixaufnahme might even extend beyond inflection, copying all morphological markings of primaries onto secondaries. It hardly ever does, though. For Hurrian, certain adverbial and derivational suffixes have been mentioned as accompanying the essive case in Suffixaufnahme; but the constituent parts of the relevant suffix conglomerates have so far resisted unequivocal identification.

4.5. Aufnahme of All or Some Cases It is tacitly understood that whenever case is subject to Suffixaufnahme, all terms realizing this category participate in the process; and analogously for number and perhaps further inflectional categories involved. Theoretically, however, it would be possible for Suffixaufnahme to be limited to any arbitrary or natural subset of cases (or numbers). Where case is used for relatedness-indication in the manner of pattern No. 4, X Y-x, several languages do indeed impose such limits. Ritharngu (52) copies only the locative, allative, ablative, and pergressive from primaries onto secondaries; Classical Armenian (54) copies only the ablative and instrumental, and Ancient Greek (56) and Latin (57) preferably only the accusative. In languages where secondaries are more elaborately marked Y-sec-x,

(Re-)Introducing Suffixaufnahme

57

analogous restrictions do not seem to obtain. What may happen, however— preferably on the condition that the marking of secondarihood and of relatedness-indication in terms of case are fused, as illustrated presently in Section 4.12—is that the case contrasts on secondaries are fewer than the corresponding ones on primaries. Thus in the Tsez family of Northeast Caucasian, cases come in two subsets, direct (nominative) and oblique (comprising all other cases), with the latter often adding a stem formative; the contrast of genitives on secondaries is only binary, reflecting that between direct and oblique (rather than between all individual cases) on primaries (see Kibrik in this volume). In the Cushitic languages Darasa and Burji, where relatednessindicators on secondaries are less easily segmented than in Awngi, the genitival alternation is likewise a two-way, i.e., minimal one; but here the case paradigm itself has no more than three members in the first place. On Boeder's interpretation (in this volume) of the alternation between short and long genitive (-is vs. -isa) in Old Georgian as determined by the head being either in the absolutive (the only case without overt suffix) or in any of the other cases, this would be an analogous instance of the reduction of a case paradigm to a two-way contrast for purposes of relatedness-indication; however, relatedness-indicating -a would here be segmentable from relationshipidentifying -is.

4.6. Aufnahme of Suffix, Prefix, or Tone 4.6.1. The very term "Suffixaufnahme" implies a constraint on the expression of case (plus perhaps further inflectional categories) amenable to Aufnahme. Nonetheless, as illustrated above at (36)-(38), morphologically bound relationship-identification of the kind traditionally regarded as case marking is formally diverse, creating a potential for variation. Considering the overwhelming predominance of case suffixation, it is not in fact surprising that Aufnahme should prototypically involve suffixes rather than prefixes. Contrary to occasional claims, however, case prefixes are not nonexistent. Languages where at least some case marking on nouns or pronouns is arguably prefixal, as opposed to prepositional, are to be found in Australian (Mangarayi, Nungali, Burarra, Mara, and Alawa—five non-PamaNyungan languages of the Northern Territory that are geographically very close to one another, but with only Mara and Alawa as co-members of a low-level genetic group), Philippine Austronesian (Murut), Indo-European (Classical Armenian), Ugric (Hungarian46), South Semitic (Amharic), Kwa (Idoma), South Eastern Bantu (Zulu), Khoisan (Eastern Bushman), Penutian (Coos, Siuslaw, Tsimshian), Hokan (Huamelultec Chontal), and Salish (Squamish).47 There are prefixes, then, that are in principle available for Aufnahme. When searching for actual instances of Prafixaufnahme, one is encouraged by the observation that the habitat of some languages with case prefixes is in Suffixaufnahme territory. It is surely in Australia that the expectations of such

58

Prologue

discoveries are highest, although none is made in Schweiger's 50-language search later in this volume. As seen earlier, Mangarayi encodes attributive constructions according to patterns Nos. 1 (37) and 8 (71) and thus does not practice Aufnahme at all (Merlan 1982). On the information available to me (Glasgow 1984), Burarra has "descriptives" (attributive adjectives) which agree in case and class with their heads, but its possessive pronouns lack such agreement, and there does not seem to be a genitive case or its equivalent. Alawa's cases are suffixal, but gender-marked nominals (most feminines and agreeing adjectives, and a few masculines, mainly kin terms) have a prefix distinguishing gender as well as case (nominative an-lna-, all other cases af-la-, feminine and masculine respectively), and these lack a separate ergative suffix. Alienable possession is expressed according to pattern No. 1, and inalienable possession according to pattern No. 4, with the secondary (possessor) copying the case of the primary (possessed) but lacking genitival marking (see above, (53)). Unfortunately, Sharpe (1970, 1976), my source for Alawa, does not include examples analogous to (53) with the primaries and possibly the secondaries carrying such gender-cum-case prefixes—schematically, ar-PRIMARYLoc (ar-?)SECONDARY-Loc. Closely related Mara (Heath 1981) has five cases likewise marked by suffixes, including a genitive or rather a more general purposive, plus a zeromarked nominative. Class and number prefixes, which distinguish masculine singular, feminine singular, neuter, dual, and plural, come in two sets, one for absolutive (intransitive subject and direct object, where case is nominative) and another for oblique relations (including transitive subject and instrument, also marked by nominative case), and thus they participate in relational marking. Such prefixes of primaries reappear on secondaries in the genitive just as their case suffixes do, in constructions which thus conform to pattern No. 7 and exhibit both Prafixaufnahme and Suffixaufnahme, both without ligative: (94) a. n-rja-radbur n-jawuru

Class-NeutAbs-campNeul (Nom) Class-3Sg=Gen 'his camp'48

b. na-radbur-yu(r) na-yawuru-yu(r) Neutobl-campNeut-All Neutobl-3Sg=Gen-All 'to his camp'

No example with a noun instead of a pronoun as secondary could be found. Nungali (a member of the Djamindjungan group, now probably extinct; cf. Hoddinott and Kofod 1976 and Bolt, Cleverly, and Hoddinott 1970) has five to seven cases, depending on the part of speech inflected, and four noun classes (below identified by roman numerals). These classes are distinguished by prefixes that simultaneously distinguish the cases, with elative and allative in addition taking suffixes. Pronouns have suffixes for dative and possessive.

(Re-)Introducing Suffixaufnahme

59

Some nouns, mainly borrowings from neighboring Ngaliwuru, do not take these case-and-class prefixes and use plain case suffixes instead. The case-andclass prefixes of primaries are copied, inter alia (indeed all inflectable words related to the primary), onto secondaries marked by the dative/possessive (a prefix) if nominal, or by the distinct possessive (a suffix) if pronominal, again without ligative:49 (95)

a. gi- arg-ina gi-ya-mad Datl-me-Poss DatI-Male-siblingj 'to/of my brother' b. gi-ganji-narurj gi-ya-mad DatI-DatH-womann Datl-Male-sibling, 'to the woman's brother'

The example with a pronoun as secondary (95a) is attested in Hoddinott et al.; that with a noun (95b) is constructed and must be taken with a pinch of salt, according to my Australianist consultants. In view of the scarcity of case prefixes, the chances are that Mara and Nungali are the only two languages in the world with Prafixaufnahme. Since the prefixes undergoing Aufnahme here happen to cumulate case with gender/class and possibly with number, there is no analogue to Suffixaufnahme involving pure case markers. Aufnahme of case infixes or circumfixes, if there are any, has not been recorded, nor has Aufnahme of non-additive marking, including subtraction (such as the omission of class prefixes from nouns in certain adverbial, especially locative, relations in Eastern Nilotic) and segmental modification (such as metathesis or gradation in Estonian—e.g., teder vs. tedre 'black grouse', tuba vs. toa 'room', nominative vs. genitive singular respectively—or initial mutations in Celtic—e.g., Irish an bad vs. an bhdid 'the boat', nominative/ accusative vs. genitive). 4.6.2. A further rare form of case marking, common only in Eastern Nilotic (or Paranilotic) and attested also in Cushitic (and perhaps Omotic) and Bantu, Bantoid, and Kwa (cf. Igbo "downstep" at (38)), is accomplished exclusively by tone.50 Tone-Aufnahme is thus a theoretical possibility; in fact, it is not at all a figment of the imagination. Paranilotic Maasai (Tucker and Tompo ole Mpaayei 1955, Tucker and Bryan 1966: 443-494) has two principal cases: accusative (or absolute), used among other things for the possessor after the particle linking it to the possession; and nominative, used among other things for subjects. The two cases are distinguished solely by tone. The tonal contrast differs depending on the tone class of the noun; compare the accusative vs. nominative contrasts in en-tito vs. en-tito 'girl' (tone class I), in-toki-tin vs. in-toki-tin 'things' (tone class II), and en-toki vs. en-toki 'thing' (tone class III). (Tone marks: a high level, a low level, a high-low falling, a mid level.) Nouns are marked for gender by pre-

60

Prologue

fixes, and plural is expressed mainly by suffixes, with plurals usually being in a different tone class than corresponding singulars. As exemplified above at (82), the attributive construction of Maasai conforms to pattern No. 19, X Link-x-y Y, with the linking particle relating to the primary in terms of gender and to the secondary in terms of gender and number. What was disregarded earlier is the fact that the secondary noun is always in the accusative case, marked tonally; but the primary may be in the accusative or nominative, also marked tonally, depending on its external syntactic relation. When the secondary noun is singular, the case of the primary is immaterial to the linking particle. However, with plural secondaries the linking particle consists of two vocalic segments, and their tones differ depending on whether the primary is accusative (mid level, then high level) or nominative (high level, then mid level): (96) a. en-toki oo 1-payyan-i

FemSg-thing:Acc of=Femprim=MascPlSec:Acc Masc-elder-Pl:Acc

b. en-toki oo 1-payyan-i FemSg-thing:Nom of=FemFrim=MascPlSec:Nom 'the thing of the elders' (object/subject)

Masc-elder-Pl:Acc

c. ol-core loo 1-ayio-k MascSg-friend:Acc of=Mascprim=MascPlSec:Acc Masc-boy-Pl:Acc d. ol-core 15o 1-ayio-k MascSg-friend:Nom of=Mascprim=MascPlSec:Nom Masc-boy-Pl:Acc 'the friend of the boys' (object/subject)

(A colon instead of a hyphen in the glosses indicates that the marking is suprasegmental.) The linking particle that appears with plural secondaries can thus be considered an element itself amenable to case marking by tone. The relational encoding of attribution in Maasai exceeds the minimum required for it to qualify as an instance of Aufnahme, viz. Y-sec-x or Link-x Y; illustrating pattern No. 19, with Link-x-y, rather than No. 17, it nevertheless resembles Suffixaufnahme patterns by virtue of the tonal case of the linking particle corresponding to the case of the primary. In the (ki-)Ntaandu dialect of Kongo (central western area of Bantu) nouns fall into four tone classes and are inflected by changes in their respective patterns of high and low tones. According to Daeleman (1983), four tonal cases can be distinguished (here identified by roman numerals): a joint case for subject and object (I); another for predicative and attributive function (II); a third for the head constituent of a phrase in subject or object function (III), thus simultaneously marking headhood and external relation; and lastly one for negated predicate nominals (IV). In terms of segmental marking, attributive constructions conform to pattern No. 17 in typically Bantu manner, with a pronominal copy of the head's class-and-number prefix serving as a link between head and attribute.51 Tonal marking, however, also contributes to

(Re-)Introducing Suffixaufnahme

61

identifying both the head and the attributive relation, and that of the attribute varies with the tonal case of the head. If the head is in case II, marking predicative function, attributes (themselves also in case II) undergo and induce further tonal alternations: they form a "tone-bridge" with their head by raising all tones between the first high tone of the head and their own last high tone; when they are of tone class 4 (such as ma-bulukutu, ma-bulukutu 'Lippia (Verbena)', case I and II forms respectively) they additionally advance their high tone to the preceding syllable, raising it to extra-high. (97) a. ma-zina m& bi-menina ClassNumber-name:II ClassNumber ClassNumber-plant:IIa '[these are] the names of the plants' b. tu-kaya tu ma-bulukutu ClassNumber-leaverll ClassNumber ClassNumber-Lippia:IIa '[these are] the leaves of the Lippia'

(Acute indicates high tone, double acute extra-high; low tone is unmarked. Ha indicates the tonal alternations accompanying case II here.) Without the tone-bridge and high-advancement adjustments, the case II forms of the attributive nouns in (97) would be bi-menina and ma-bulukutu; without tonebridge, the case II forms of their heads would be ma-zina and tu-kaya. When the heads of attributes occur as subject or object, they are in case III, with low tones throughout. Their attributes are in case II, as in the company of heads whose external function is predicative, but with no further tonal alternation except a shift of the high tone onto the prefix of nouns of tone classes 3 and 4 (here glossed as lib): (98)

a. ma-zina ma bi-menina . . . ClassNumber-name:II ClassNumber ClassNumber-plant:IIb . . . 'the names of the plants . . .' b. tu-kaya tu ma-bulukutu tuyumini ClassNumber-leave:III ClassNumber ClassNumber-Lippia:IIb are=dry 'The leaves of the Lippia are dry'

Ntaandu thus represents the tonal equivalent of fused Suffixaufnahme a la Tsez and perhaps Darasa and Burji, with no (easily) segmentable markers for relationship-identification and relatedness-indication on the secondary (see Section 4.12). It is tonal case II that, inter alia, identifies secondaries, and it is the different tonal alternations associated with case II—tone-bridge and possibly high-advancement on the one hand and no further tone change or shift of high tone to the prefix on the other—that relate secondaries to primaries in tonal cases I and III respectively. Taking into account both segmental and suprasegmental marking, the encoding of attributive constructions in Ntaandu combines patterns No. 17 (X Link-x Y, segmentally) and No. 13 (X-prim Y-sec-x, tonally). Insofar as the

62

Prologue

overall pattern, X-prim Link-x Y-sec-x, contains Y-sec-x and this variable x represents case, the criteria for Suffixaufnahme are satisfied—except that case is expressed by tones rather than by suffixes. It has not been determined so far whether further Bantu languages with tonal case marking, including in particular Ntaandu's central western relatives such as Hungu (or ki-Holo), (ki-)Mbundu, (ki-)Pende, (yi-)Yaka, (ci-)Cookwe, and other dialects of (ki-)Kongo, show similar tonal alternations of attributive cases (co-)conditioned by the case of the head. Cushitic and the rest of Paranilotic apparently have no analogues of either the Maasai or the Ntaandu tonal pattern at issue. Case is sporadically distinguished by stress alternations, alone or in combination with additive marking. In the Highland East Cushitic language Kambata, for example, the relation of indirect object is encoded by retracting the main stress to the last syllable of nouns (Hudson 1976: 253). Stress-Aufnahme is not on record, however. 4.6.3. Case, then, lends itself to Aufnahme whether expressed by suffixes or prefixes or tones. With the scope of Aufnahme encompassing all major formal varieties of bound case marking, the traditional term is evidently too narrow. It has nonetheless been generally retained in the present volume, including the title, because Aufnahme of suffixes, however infrequent cross-linguistically, is still considerably more common than Aufnahme of prefixes or tones. Those with strong opinions about terminological precision should replace it everywhere by "case-and-other-inflection-Aufnahme."

4.7. Aufnahme of Bound or Free Markers Grammatical markers of the external relation of primaries may be morphologically bound to stems, taking the form of affixes or of segmental or suprasegmental modifications, including tones; or they may be independent words such as adpositions, relational nouns, or serial verbs—often inclining to cliticize on their hosts, however. This distinction is a gradual one and possibly admits of intermediate species (such as phrase-bound affixes).52 Conceivably, the scope of Aufnahme might thus extend to any looser appendages of nominals that are functional analogues of case affixes. However, relational markers that are not morphologically bound have occasionally been claimed not to be available at all for relatedness-indication; or, to put it the other way round, if categories figure in agreement within NPs, their expression is normally taken to be affixal rather than adpositional on this evidence alone. On this assumption, the external relation of NPs is free to be marked on both nouns and attributive adjectives in languages such as Latin where such marking, cumulating case with number and gender, is by affixation (domu-i regi-o houseMasc-DatSg kingly-DatSgMasc '(to) the royal house'); but it is perforce limited to a single marking per NP in languages such as English that employ adpositions (*to the royal to (the) house). Contrary to this assumption, however, there are languages—Slavonic

(Re-)Introducing Suffixaufnahme

63

and Baltic ones, including Old and later non-literary Russian, Old Czech, Old Serbian, and Lithuanian (Worth 1982), and also Hungarian (Tompa 1972: 180f.)—where relation-identifying elements that qualify as adpositions rather than affixes on all other counts are optionally (Slavonic, Baltic) or obligatorily (contemporary Hungarian) distributed over more than one NP constituent. In Old Russian, for example, prepositions may be repeated with attributive adjectives (99a/b) and appositive nouns (99c), but normally only when these are in postnominal position; in Hungarian, postpositions not governing an affixal case must (lOOa)—and some postpositions governing a non-nominative case marginally may (lOOb)—be repeated after a few demonstrative pronouns, which are the only NP constituents also to agree in case and number with nouns when these are overtly marked for these categories.53 (99)

(100)

a. ko knjaz-ju k velik-omu to dukeMasc-DatSg to grand-DatSgMasc 'to the grand(-)duke' b. na sydn-o na posl-ov-o to boatNeut-AccSg to envoy-Adjct-AccSgNeut 'to the envoy's boat' c. s knjaz-em s Ivan-om with duke-InsSg with Ivan-InsSg 'with Duke Ivan' a. e(-)f616tt a hajo folott this(-)above the ship above 'above this ship' b. ez-en (??tul) a hid-on tul this-Superessive (??beyond) the bridge-Superessive beyond 'beyond this bridge'

Adposition-Aufnahme must thus be reckoned with as a theoretical possibility. In fact, in afterthoughts and similarly loose-knit constructions this very pattern enjoys worldwide popularity. Even English has it, with or without a pronominal copy of the primary, depending on the kind of attributive marker (preposition o/vs. suffix or postposition 's): (101)

a. in the palace, in that of the duke b. in the palace, in the duke's

Occurring in what are clearly not basic attributive constructions, such instances of Aufnahme, however, hardly qualify as prototypical. Interestingly, in the languages where adpositions are repeated on attributes in less marginal constructions, relationship-identified nouns are not normally among the elements prompting such relatedness-indication. In Old and later non-literary Russian, where preposition-repetition has plausibly been claimed to signal a high degree of syntactic integration of the attribute or

64

Prologue

appositum with its head (rather than special emphasis), and likewise in other relevant Slavonic and Baltic languages, attributive nouns are as a rule eligible only upon their conversion into possessive adjectives, thus failing another prototypicality criterion of Aufnahme (see Section 4.15).54 Examples straightforwardly analogous to (99b), except with the attributive noun in the genitive rather than transformed into an adjective as in (102a), do not seem to be on record. What are found are a very few instances of preposition-repetition with straight genitives that precede their heads—rather than follow, as is the norm for this method of signaling joint membership in an NP; an example is (102b), where the inflection of the attributive noun and its accompanying possessive pronoun is moreover ambiguous (at least orthographically) between genitive and prepositive (Pre), the case governed by the preposition. (102)

a. *na sydn-o na posl-a to boatNeut-AccSg to envoy-GenSg 'to the envoy's boat' b. v tvo-ej brat'-i v votcin-ax on your-GenSg brother-GenSg on seat-PrePl 'on your brother's ancestral seats'

It is only on the strength of exceedingly rare and unusual examples such as (102b) that Old Russian could be credited with adposition-Aufnahme. An earlier example of postposition-Aufnahme from Old Georgian, repeated as (103), is not fully convincing either, because gan, apart from retaining properties of a postposition (e.g., governing a case, viz. genitive), shows signs of increasing fusion; moreover, its copying is not perfect, with the dative taking its place on the secondary (see next section). (103)

pir-isa-gan uymrto-ta-sa face-Gen-in infidel-OblPl-Dat 'from the face of the infidels' (=7)

The elements subject to Aufnahme in Chukchi, Old Georgian, Bats, Hurrian, Urartian, Awngi, Yidiny, et al., exemplifying the prototype, are not so tightly fused with stems as the case inflections in, say, Latin or Russian. They are usually, and not inappropriately, analysed as suffixes rather than postpositions, although one has to bear in mind that in languages with predominantly agglutinative morphology the difference tends to be far less tangible than under flective-type circumstances. It is in this border area of rather weak morphological cohesion that Aufnahme appears to thrive. 4.8. Formal or (Actual or Virtual) Categorial Copying The definitional requirement that the attribute be marked for the same case as the head may be construed as pertaining either to the mere form of the

(Re-)Introducing Suffixaufnahme

65

marking (hence automatically to category as well) or to its category, with "category" in turn permitting either a concrete or a more abstract reading. In the examples above from Old Georgian (84), Bats (85), Hurrian (86), Urartian (87), Awngi (88), Yidiny (89), Kayardild (93), Mara (94), and Nungali (95), the exponents of the primary's case inflection were copied onto the attribute without formal change. By contrast, in Chukchi (83, 90), and similarly in the Australian representatives of pattern No. 4, X Y-x, mentioned earlier (including Pit j ant j at jar a (51) and Alawa (53)), the copies were not quite true to the originals: they were instead the allomorphs of the primary's case that the phonological environment of the secondary demanded (e.g., locative -k vs. -ak, absolutive plural -/ vs. -3t in Chukchi, locative -la vs. -ngka in Pitjantjatjara). Presumably the inference is legitimate that whenever there are phonologically conditioned allomorphs, those appropriate for the host environment will be chosen. In Old Georgian et al., those appearing in the examples just happened to be phonologically appropriate in both primary and host environments. In accordance with the superiority of category over form, tone-Aufnahme does not consist in the mere copying of the primary's tone. Making allowances for low-level phonological adjustments does not suffice to account for variation along this parameter: morphological conditioning may also be involved. In itself, this would not be unusual for attributive agreement; when there are different inflection classes of adjectives, for instance, it is the adjective's own inflection class that selects the allomorph of the categories in which it agrees with its head (cf. Latin vir-lsapient-es manMascNomPl wise-NomPlMasc 'wise men', not vir-l sapient-i). Similarly, in Latin examples of pattern No. 4, the exponent of case and number on the secondary is that of its own declension (accusative singular -em in (57), where the primary has accusative plural -a). However, the morphological alternations possible in Aufnahme patterns are less straightforward. In Gugu-Yalanji of Queensland, Australia, the factors determining the allomorphs of case suffixes include phonological (trisyllabic vs. other word bases) and morphological ones (animate vs. inanimate word bases); whether the primary takes the animate, inanimate, or trisyllabic allomorph, the copy of this case on the secondary (regardless of its own animacy or inanimacy) is invariably the inanimate allomorph (Hershberger 1964): (104)

Dicki-ndamun-du kaya-ngka Dick-Gen-Erglnanim dog-ErgAnim 'Dick's dog'

Although in Gugu-Yalanji the divergence between primary and secondary marking does not transcend the limits of allomorphy, there are other instances where the markings on primary and secondary arguably do not represent the same category.55 First, in Old Georgian, the interrogative and indefinite pronoun vin is among the items that are exempt from overt ergative marking when used as transitive subject in clause types where other nominals would take ergative case; nonetheless, this pronoun's secondaries are in the ergative,

66

Prologue

as with any other primaries overtly marked ergative (Boeder 1987: 42, with an alternative interpretation offered in the present volume): (105)

vin-me . . . mocape-ta-gan-man who-Indef . . . disciple-OblPl-of-Erg 'one of the disciples'

Other NPs have no particularly good reasons for not carrying an ergative suffix; when irregularly occurring in oblique plural form in transitive subject function, their secondaries nevertheless likewise have an ergative suffix -man (Boeder 1987: 48): (106)

mitxres mydeltmosyuar-ta da xuces-ta ma-t Huria-ta-man said=to=me high=priest-OblPl and elder-OblPl Art-OblPl Jew-OblPl-Erg 'The high priests and elders of the Jews said to me'

In a variation on this theme, also occurring in Old Georgian, a relational marker such as ablatival gan, halfway between a postposition and a case suffix, is not itself resumed on secondaries but is here substituted for by the dative, a general local case but not in fact governed by the erstwhile postposition; see (103) above.56 And in yet another variation, in Jiwarli (Mantharta group) and perhaps other Australian languages where transitive subjects and objects are each case-marked differently depending on the position the nominals hold on the animacy hierarchy, secondaries to such primaries take the subject or object cases that are required by their own animacy rather than by that of their primaries. Thus inanimate objects (such as 'camp') themselves remain unmarked in Jiwarli, while animate ones take the accusative suffix, but their secondaries in the dative case (suppletive in the 1st person pronoun) take an overt accusative suffix when animate, as alienable possessors usually are (Austin, this volume): (107)

warri nganaju-nha ngurra panyi-ma not I=Dat-Acc camp (Acc/Abs) disturb-Imp 'Don't disturb my camp!'

In order to subsume such patterns under the general heading of Aufnahme, as seems advisable, the notion of copying must be conceived of more abstractly. It is not only the actual relationship-identifying categories on primaries (or their phonologically or morphologically conditioned allomorphs) that may be used again on secondaries for purposes of relatedness-indication, but also virtual categories, occurring with alternative lexical realizations of primaries in the same external relationships. These may include animate, inanimate, or trisyllabic case allomorphs, as in Gugu-Yalanji; ergative or absolutive case or plural oblique for transitive subjects as in Old Georgian; genuine case or incipient case/erstwhile postposition for ablatival nominals as in Old Georgian; or unmarked absolutive/accusative or marked accusative case for transitive ob-

(Re-)Introducing Suffixaufnahme

67

jects as in Jiwarli. If there is a choice, a single alternative may be prescribed for all secondaries (inanimate allomorphs in Gugu-Yalanji, ergative in case of transitive subject primaries and dative for ablatival primaries in Old Georgian); or secondaries may determine the choice in terms of their own semantics or morphological features (unmarked absolutive/accusative or marked accusative in the case of transitive object primaries in Jiwarli). If the choice is determined by the secondaries themselves, one may wonder—as Austin (in this volume) comes close to doing—whether there is any copying occurring in the first place; primary and secondary might instead be assigned their respective cases, matching only in the abstract, independently.

4.9. Relative Order of Multiple Marking The order adopted in our schematic representations of relevant marking patterns has relationship-identifying markers before relatedness-indicating ones; thus, Y-sec-x rather than Y-x-sec in No. 7. Although purely conventional, this relative order (or its mirror image x-sec-Y to account for prefixation) is in fact the only one attested: no languages are on record in which a relatednessindicating case affix is closer to the stem of the secondary nominal than is a relationship-identifying case affix, whenever forms fulfilling these two functions are segmentable. Instances of a relationship-identifying case being suffixed and a relatedness-indicating one prefixed, as in Mara (94) and Nungali (95), do not invalidate this generalization. When relatedness-indicating case and perhaps other inflectional marking is repeated in response to hierarchical syntactic structures, with secondaries accompanying nominals that are in turn secondaries, syntactic immediacy is mirrored by linear morphological precedence (cf. Dench and Evans 1988: 6f., Plank 1990), as in this example from Old Georgian: (108)

sasxdomel-eb-i igi msqidel-ta-j ma-t tred-isa-ta-j stall-Pl-Nom Art (Norn) seller-OblPl-Nom Art-OblPl pigeon-Gen-OblPl-Nom 'the stalls of the sellers of pigeons'

Both secondaries here have their own relationship-identifying inflections first (-ta OblPl, -isa Gen); the second secondary then has the case-number of its immediate primary (-ta OblPl) before the case of the ultimate primary (-il-j Nom). Kayardild, already noted for its several layers of case marking, is the only known language sometimes violating this mirror-image correspondence between morphology and syntax; to get around the prohibition against an oblique case (formerly also known as purposive) being followed by any other case, the two may simply be reordered (Dench and Evans 1988: 41, and Evans in this volume): (109)

ngada kurri-jarra dangka-na yalawu-n-kina yakuri-naa-ntha mijil-nguni-naanth / *mijil-nguni-nja-na

68

Prologue lSg=Nom see-Past man-M.Abl catch-Nomin-M.Abl fish-M.Abl-Obl net-InsM.Abl-Obl / *net-Ins-Obl-M.Abl 'I saw the man catching fish with the net'

Notwithstanding virtually unexceptional regulations elsewhere, Kayardild proves the relative order of affixation to be a potential variable in Aufnahme patterns.

4.10. Double Marking Retained, Reduced, or Replaced Sequences of relationship-identifying and relatedness-indicating case markers on secondaries, where in principle permissible,57 may undergo reduction or replacement. Though limited to certain well-defined circumstances, such special surface treatment evidently detracts from the full flavor of otherwise exemplary Aufnahme. Not altogether unexpectedly in view of the havoc known occasionally to be wrought by morphological haplology, when adjacent relationship-identifying and relatedness-indicating inflections are overtly identical, one of them may obligatorily get omitted—as in Jiwarli and Old Georgian (Austin in this volume, Boeder 1987: 47 and in this volume, and Dench and Evans 1988: 36-39): (110)

thuthu-wu purrarti-yi(*-yi) dog-Dat woman-Dat(*-Dat) '(of/to) the woman's dog'

(111)

kar-ta kalak-ta(*-ta) gate-OblPl city-OblPl(*-OblPl) 'the gates of the cities'

Haplology being notoriously capricious, it is not surprising either that potential victims may also survive—as does double dative in Thalanyji (obligatorily, again see Austin) or double genitive in Old Georgian (optionally): (112)

kaparla-ku wartirra-ku-ku dog-Dat woman-Dat-Dat 'the woman's dog'

(113)

3-isa kac-isa(-jsa) son-Gen man-Gen(-Gen) 'of the son of man'

Apart from haplology, a further possible reason for relatedness-indicating case suffixes to be dropped in Suffixaufnahme languages is the proximity of particular relationship-identifying cases. Thus in Djapu of Arnhem Land, relatedness-indicating ergative, instrumental, causal, and local cases (the last only if the primary is non-human) are omitted after a relationship-identifying oblique case (Dench and Evans 1988: 41, Schweiger in this volume):

(Re-)Introducing Suffixaufnahme (114)

69

waanga-ngur yapa-'mirringu-wal(*-ngur) ngarra-kalangu-wal(*-ngur) camp-Loc sister-KinProp-Obl(*-Loc) !Sg-OblStem-Obl(*-Loc)58 'at my sister's camp'

This may also be the fate of relationship-identifying cases. However, the example which Dench and Evans (1988: 40) provide from Kungarakany of Arnhem Land, where the dative is omitted from pronominal possessors when it would be followed by a case marking the primary's external relation, is one of suffix-accumulation rather than Suffixaufnahme, since the primary itself is not case-marked: (115)

lok ngirrpa(*-kini)-wu place lSg(*-Dat)-Loc 'to my place'

Finally, instead of being omitted, a relationship-identifying case may be replaced by another case in particular morphological environments. Thus, whenever a relatedness-indicating (or other) case other than oblique, such as the modal proprietive (M.Prop), would follow the locative in Kayardild, the associative, expressing temporary possession and thus semantically not altogether inappropriate, substitutes for the locative (Dench and Evans 1988: 38): (116)

dangka-wu yubuyubu(*-ya -H»)-nurru-wuru man-M.Prop road(*-Loc —»)-Ass-M.Prop 'the man on the road'

There apparently are no straightforwardly analogous replacements of relatedness-indicating cases; all adaptations here involve virtual categorial copying in the above sense (Section 4.8). It would seem that some languages, such as those of the Ancient Near East, are less prone than others—especially those of Australia—to any such superficial distortion of the prototypical Aufnahme pattern.

4.11. Direct or Mediated Connection of Double Marking The relatedness-indicating marker of case, possibly plus number or other inflection, may be added to the relationship-identifying marker of the secondary either directly or after an additional "ligative" affix (also known as a "catalytic" or "linking" affix). In the examples of prototypical Suffixaufnahme above, such ligatives occurred only in Hurrian and Urartian. However, Yidiny (89), Kayardild (93), and Thalanyji (112) are not representative for all of Australia; other languages of the continent, including Dyirbal (117), require ligatives as well. (117)

nay-gu-djn-du yabu-nu-jidjn-du barjgun guda-rjgu ISg-Gen-Lig-Erg mother-Gen-Lig-Erg ClassII=Erg dogn-Erg 'my mother's dog'

70

Prologue

While such ligatives, on the face of it, appear to be mere buffers in those Australian languages that have them, in Hurrian and Urartian they also reflect number distinctions of the primary.59 However, even ligatives that are now invariable (or whose allomorphs are phonologically conditioned, as in Dyirbal) may have had some kind of phoric or pronominal function in the past; at least this would provide a rationale for the use of such additional elements, as is argued by Aristar in this volume. In languages that employ ligatives these generally seem to be obligatory rather than optional, although occasional omissions or limitations to certain case environments have been reported. Regardless of the past or present functions of ligatives, their presence or absence is an obvious parameter of variation, but one that does not obviously affect the pro to typicality of Suffixaufnahme. To generalize from two instances, Prafixaufnahme shuns ligatives.

4.12. Separation or Fusion of Double Marking In the Aufnahme prototype the relationship-identifying marker (genitive) and the relatedness-indicating marker (copying the primary's case plus perhaps other inflections) on the secondary nominal are separate suffixes. In a less prototypical form of Aufnahme, though one that would still seem to merit this title, the two markers are systematically fused in a single suffix not amenable to further internal segmentation. The few pertinent languages hail from areas and families also boasting prototypical Suffixaufnahme. Recall that in Awngi (88) the primary's plural suffix -kd was not actually repeated on the secondary but was co-expressed with the genitive, with -kw constrasting with the genitive alternants -w and -t, which are used in association with masculine and feminine singular primaries. In two other Cushitic languages (see Hetzron in this volume and Hudson 1976) such genitive alternants also reflect—among other categories of the secondary (number and gender, the latter possibly polarity-reversed) and also of the primary (proper vs. common noun)—the case of their primaries, with accusative (or absolute, the citation form) and nominative (the form that also serves as the base for genitival suffixing) being the only cases available. (118)

Burji (Highland East Cushitic, Afroasiatic) a. samee-nta amaa SameeProper-Gen wifeFem (Ace) b. samee-cci [a]-lSgAbsPro-Connect wife-3SgPoss-Gen dowry-3SgPoss-Dir 'to the dowry of the wife of my brother' (Mit. IV 46f.) Here too the genitive attribute of lower rank agrees with the genitive attribute of higher rank, but neither is related to the head by Suffixaufnahme.

130

The Ancient Near East 5. Suffixaufnahme in Urartian

The patterns of Suffixaufnahme in Urartian are basically identical with those in Hurrian. Suffixaufnahme with genitive attributes has been obscured by the fact that the Urartian genitive morpheme in the singular (-i) is sometimes difficult to identify.

5.7. Genitive Attributes Absolutive, head singular / attribute singular:10 (70) Ispuini=se . . . Minua=se . . . Halde=i susi sid=ist=u=ni Ispuini-Erg . . . Minua-Erg . . . Halde-Gen temple.tower build-RC-Trans3SgErg/3Abs 'Ispuini . . . (and) Minua . . . built a temple-tower of (the god) Haldi' (HChI 8 I = UKN 25:1-27/6-7) (71) Minua=i pili tini Minua-Gen canal name ' "Minua's canal" is (his) name' (HchI 29a-d I = UKN 43:3) Genitive, head singular / attribute singular: (72) Minua=i=ne=i sila=ie Tariria=i ini uldi Minua-Gen-RltSg-Gen daughter-Gen Tariria-Gen this vineyard 'this vineyard (is that of) of Tariria, the daughter of Minua' (HchI 40 = UKN lll:l-2//4-5) (73) Umesini tini . . . Rusa=i=ne=e hubi=ii Umesini name . . . Rusa-Gen-RltSg-Gen valley-Gen 'Umesini is the name of Rusa's valley' (HchI 126 III = UKN 281:15-17) (74) Minua=i=ne=i urishusi=ne=i Minua-Gen-RltSg-Gen storehouse-RltSg-Gen '(object) of the storehouse of Minua' (HchI 40 A = UKN 112-117) Dative, head singular 7 attribute plural: (75) Haldi=i=ne=i patari=e pulusi Haldi-Gen-RltSg-Dat city-Dat stela 'a stela for the city of Haldi' (HchI 79 = UKN 301:1) —, head plural 7 attribute singular: (76) Haldi=i=na=ue sesti=na=ue Haldi-Gen-RltPl-DatPlgate-RltPl-DatPl 'for the gates of Haldi' (HchI 10 II16 // X 58 = UKN 27:16//58)n

Suffixaufnahme

in Human and Urartian

131

Instrumental, head singular / attribute singular: (77)

Haldi=i=ne=ni alsuisi=ni Haldi-Gen-RltSg-Ins greatness-Ins 'through the greatness of Haldi' (HchI, UKN passim)

5.2. Relational Adjectives in -he as Attributes Such adjectives are ubiquitous in Urartian because they are used to form patronyms. Absolutive: (78)

Rusa=ni Argisti=he Rusa-3SgPro Argisti-Adjct 'Rusa, the son of Argisti'

Ergative: (79)

Rusa=se Sardure=he=ne=se Rusa-Erg Sardure-Adjct-RltSg-Erg 'Rusa, the son of Sardure'

Genitive: (80)

Rusa=i Argisti=he=ne=i Rusa-Gen Argisti-Adjct-RltSg-Gen

Dative: (81)

Ispuini=e Sardure=he=ne=e Ispuini-Dat Sardure-Adjct-RltSg-Dat

The Urartian city-names of the type Rusa=he=ne=li follow the same pattern, although their plural head, understood as 'settlements', 'achievements' or the like, is regularly omitted (Salvini 1979). Thus, Rusa=he=na—ue genitive/ dative, Rusa=he=na=idi directional (originally dative with postposition), Rusa=he=na=a locative. Notes 1. Morpheme boundaries in Hurrian examples are conventionally marked by an equality sign ( = ) rather than by a hyphen (-), which is used in the "broad transcription" (transliteration) of syllabic cuneiform. 2. 'Connect' glosses a connective enclitic.

132

The Ancient Near East

3. The phrase is paralleled by en=n(a)=a=az=uz tevi=n(a) =

az=uz 'the words of the gods' in 1. 18'. katki has been correctly linked with the verbal root had- 'to speak' by Salvini (1975: 236). A parallel formation with suffix -klgi is itki 'mortar' from id- 'to beat', on which see Neu (1988a: 29; 1988b: 108) and earlier Thiel and Wegner (1984: 201ff.). 4. The ending here attributed to a separate "e-case" has been described as an allomorph of the genitive suffix -ve after the 3rd person singular possessive suffix (Speiser 1941: 56; Bush 1964: 91, 127f.; Xacikjan 1985: 115f.). The e-ending appears with the words edi 'person' and avi 'face' used as postpositions. Although this ending always occurs in final position, its vowel in most instances is written plene (-e-e), which is never the case with the genitive suffix -ve. Since there are clear attestations of the sequence of the 3rd singular possessive suffix -/- and the genitive suffix -ve that do not show the assumed shift -i=ve > *i-ie, the ending -e should be kept apart from the genitive suffix. If this is accepted, it will no longer be necessary to postulate an exception to the rules of Suffixaufnahme and an unusual plene-writing of the singular article in final position, as Diakonoff (1971: 153 with n. 162) was forced to do when commenting on example (15): "Hierbei wird -ne (Artikel) nach dem Kennzeichen des Genitivs gesetzt, und eine Suffixaufnahme findet nicht statt." (Because of the scantiness of attestations, however, Diakonoff himself called this explanation into question.) 5. Girbal (1988: 131f.) has argued for a case ending in -nna, to which he attributes the two functions of an equative and an adverbial. Since the recently discovered forms ending in -oz (see below) have appropriately been labeled "equative" as well, I suggest the rather vague term "associative" for the forms at issue here, which covers both functions described by Girbal. Very much like a case ending, -nna requires the oblique stem form of 1st person singular pronouns -iffu- 'my' and so 'me'. However, there is a suffix -nni that in certain contexts (Mil. 115, IV 37) seems to fulfill the same function as -nna (cf. Bush 1973: 47f.; 1978: 222). It might well be that -nna is a combination of -nni and the essive case ending -a. The suffix -nni may also be combined with the instrumental suffix that also serves as an adverbial (cf. sena=nn(i)=ae 'in the manner of, acting as, a brother', Mit. IV 12). It is evident from example (21) that -nni is a derivational rather than a case suffix, because its position is to the left of another derivational suffix. 6. The distribution of -ge and -hhe is not yet well understood. Frequently -ge appears in topographical adjectives, such as hurrlhurv=6=ge 'Hurrian' (from Hurri, *Hurvi), hlbg (Ugaritic alphabetic script) = halb=a=ge 'Aleppian' (from Halab, Halba), alzyg = *alaziia=ge 'Cypriot' (from Alaiiid). The suffix appears frequently in cultic texts from Bogazkoy to denote local forms of deities: Ankuua=ge, Hattarina = ge, KiZZuuadna=ge, Kuliusna=ge, SamuHa=ge, Sabinuua=ge, Tameninga-ge, hatt= o=ge (from Hattf), manuZ=o=ge (from Manuzzf). With consonantal stems, the suffix may be attached directly to the stem: tugris = he (from Tugriz). If the ge-adjective is based on a verbal root nominalized with -(', this vowel remains unchanged; cf. pa=i=ge 'building plot' (from pa- 'to build'), pass=i=ge 'consignment' (from pass- 'to send'), na=i=ge 'pasture' (from nav- 'to graze'), haval=ge 'enclosure' (from halv- 'to fence', with syncope, anaptyxis, and metathesis, following well-known patterns). The variant -hhe is attested in words like hiiar=o=hhe 'gold, golden' (from hiiari), siniber=o=hhe 'of ivory' (from *siniberi 'ivory' < Akkadian sinni piri), asth = a$t=o=hhe 'female' (from asti 'wife, woman'), trh = tur=o=hhe 'male' (from *turi 'man'), tiz=n = o = hhe 'heart-shaped' (from tiz = ni 'heart'), simig = o = hhe 'pertaining to the sun-god' (name of a city gate).

Suffixaufnahme

in Human and Urartian

133

7. CV is short for "connecting vowel." 8. The fact that the name Mittani is frequently written with final -nni in Akkadian and Hittite contexts may be explained as a result of stress, which certainly was on the paenultima because of the shift mai- > mi in the first syllable. 9. Cf. ibid. 11 I If.; da-a-la-rwuu^-u-s[i\ re-eb^-ri-in (ChS I/I No. 11 Vs. 15). 10. RC is short for "root-complement." Urartian fuses verbal subject and object markers, here indicated by a slash. 11. Urartian is a little less agglutinative than Hurrian, fusing, for example, dative and plural.

Sources of Examples ChS I/I

ChS 1/5

HchI HSS 5 KBo XXXII Mit.

RS Ugaritica 5

Haas, Volkert. 1984. Die Serien itkahi und itkalzi des AZU-Priesters, Rituale fur Tasmisarri und Tatuhepa sowie weitere Texte mit Bezug auf Tasmisarri. (Corpus der hurritischen Sprachdenkmaler, I. Abteilung: Die Texte aus Bogazkoy, vol. 1.) Rome: Multigrafica Editrice. Haas, Volkert, and Use Wegner. 1988. Die Rituale der Beschworerinnen SAL SU.GI. (Corpus der hurritischen Sprachdenkmaler, I. Abteilung: Die Texte aus Bogazkoy, vol. 5.) Rome: Multigrafica Editrice. Konig, Friedrich Wilhelm. 1955-57. Handbuch der chaldischen Inschriften. (Archiv fur Orientforschung, Beiheft 8.) Graz 19551957. (Osnabriick: Biblio-Verlag, 1967, 2d ed.) Chiera, Edward. 1929. Texts of Varied Contents. (Excavations at Nuzi, vol. 1; Harvard Semitic Series, vol. 5.) Cambridge, Mass.: Harvard University Press. Otten, Heinrich, and Christel Riister. 1990 [1991]. Die hurritischhethitische Bilingue und weitere Texte aus der Oberstadt. (Keilschrifttexte aus Boghazkoi, Heft 32.) Berlin: Gebr. Mann Verlag. The Mittani Letter (letter of King Tusratta of Mittani to Pharao Amenophis III written a few years before 1352 BCE). Last full transliteration: Friedrich, Johannes. 1932. Kleinasiatische Sprachdenkmaler, 8-32. (Kleine Texte fur Vorlesungen und Ubungen, ed. by H. Lietzmann, 163.) Berlin: de Gruyter. Cuneiform edition: Schroeder, Otto. 1915. Die Tontafeln von ElAmarna, zweiter Teil. (Vorderasiatische Schriftdenkmaler der koniglichen Museen zu Berlin, Heft 12.) Leipzig: J.C. Hinrich'sche Buchhandlung, No. 200. Latest translation (into French): Les lettres d'El Amarna, traduction de William L. Moran. (Litteratures anciennes du Proche-Orient, 13.) Paris: Les Editions du Cerf, 1987 (139-151: EA 24 [French translation of an unpublished translation into German by Gernot Wilhelm]). Tafelsignatur Ras Samra (=Ugarit); cf. Ugaritica 5. Nougayrol, Jean, Emmanuel Laroche, Charles Virolleaud, and Claude F. A. Schaeffer. 1968. Ugaritica 5. (Mission de Ras

134

UKN

The Ancient Near East Shamra, tome 16.) Paris: Imprimerie Nationale and Paul Geuthner. Melikisvili, Georgij A. 1960. Urartskie klinoobraznye nadpisi. Moscow: Nauka.

References Benedict, W. C. 1958. Urartian Phonology and Morphology. Ph.D. dissertation, University of Michigan. Bork, Ferdinand. 1909. Die Mitannisprache. Mitteilungen der Vorderasiatischen Gesellschaft 14, 1/2. Berlin: Wolf Peiser. Bush, Frederic William. 1964. A Grammar of the Human Language. Ph.D. dissertation, Brandeis University (Ann Arbor: University Microfilms No. 64-12,852). Bush, Frederic William. 1973. The relationship between the Hurrian suffixes -nel-na and -nnilel-nna. In Orient and Occident: Essays Presented to Cyrus H. Gordon, ed. by Harry A. Hoffner, Jr., 39-52. (Alter Orient und Altes Testament 22.) Kevelaer and Neukirchen-Vluyn: Butzon & Bercker, Neukirchener Verlag. Bush, Frederic William. 1978. The suffixes -ne and -na in Hurrian. In Biblical and Near Eastern Studies: Essays in Honor of William Sanford LaSor, ed. by Gary A. Tuttle, 220-230. Grand Rapids, Mich.: Eerdmans. Diakonoff, Igor M. 1971. Hurrisch und Urartiiisch. (Miinchener Studien zur Sprachwissenschaft, Beiheft 6 N.F.) Munich: Kitzinger. Diakonoff, Igor M., and S. M. Kashkai. 1981. Geographical Names According to Urartian Texts. (Repertoire geographique des textes cuneiformes 9; Tiibinger Atlas des Vorderen Orients, Beihefte, Reihe B, No. 7/9.) Wiesbaden: Reichert. Diakonoff, Igor M., and S. A. Starostin. 1986. Hurro-Urartian as an Eastern Caucasian Language. Munich: Kitzinger. Friedrich, Johannes. 1935. Zum Subaraischen und Urartaischen. In Festschrift A. Deimel, 122-135. (Analecta Orientalia 12.) Roma: Pontificium Institutum Biblicum. Friedrich, Johannes. 1939. Kleine Beitrage zur churritischen Grammatik. Mitteilungen

der Vorderasiatisch-Agyptischen Gesellschaft 42, 2. Leipzig: J. C. Hinrichs.

Girbal, Christian. 1988. Der Paragraph 24 des Mittani-Briefes. Zeitschrift fur Assyriologie 78, 122-136. Jensen, Peter. 1899. ZurErklarungdesMitanni. Zeitschrift fur Assyriologie 14,173-181. Laroche, Emmanuel. 1980. Glossaire de la langue hourrite. Paris: Editions Klincksieck. Melikisvili, G. A. 1971. Die urartdische Sprache. (Studia Pohl 7.) Rome: Biblical Institute Press. Messerschmidt, Leopold. 1899. Mitanni-Studien. Mitteilungen der Vorderasiatischen Gesellschaft 4/4. Berlin: Wolf Peiser. Neu, Erich. 1988a. Varia Hurritica: Sprachliche Beobachtungen an der hurritischhethitischen Bilingue aus Hattusa. In Documentum Asiae Minoris Antiquae: Festschrift fur Heinrich Otten zum 75. Geburtstag, ed. by Erich Neu and Christel Ruster, 235-254. Wiesbaden: Harrassowitz. Neu, Erich. 1988b. Hurritische Verbalformen auf -ai aus der hurritisch-hethitischen Bilingue. In Studia indogermanica et slavica: Festgabe fur Werner Thomas zum 65. Geburtstag, ed. by P. Kosta, 503-513. (Specimina philologiae slavicae Suppl. 26.) Munich: Otto Sagner.

Suffixaufnahme

in Human and Urartian

135

Plank, Frans. 1988. Das Hurritische und die Sprachwissenschaft. In Hurriter und Hurritisch, ed. by Volkert Haas, 69-93. (Xenia 21.) Konstanz: Universitatsverlag. Plank, Frans. 1990. Suffix copying as a mirror-image phenomenon. Linguistics 28, 1039-1045. Salvini, Mirjo. 1975. Note sulle 'sentenze' hurriche nei rituali ittiti di KBo XXI e XXII. Oriens Antiquus 14, 227-241. Salvini, Mirjo. 1978. Hourrite et Urarteen. Revue hittite et asianique 36, 157-172. Salvini, Mirjo. 1979. Problemi della morfologia nominale in Urarteo. AI N 1, 97115. Smeets, Rieks. 1989. On Hurro-Urartian as an Eastern Caucasian language. Bibliotheca Orientalis 46, 259-279. Speiser, Ephraim A. 1941. Introduction to Human. (Annual of the American Schools of Oriental Research 20.) New Haven: American Schools of Oriental Research. Thiel, Hans-Jochen, and Ilse Wegner. 1984. Eine Anrufung an den Gott Tessup von Halab in hurritischer Sprache. Studi micenei ed egeo-anatolici 24, 187-213. Thureau-Dangin, Francois. 1931. Vocabulaires de Ras-Shamra. Syria 12, 225-266. van Loon, Maurits. 1983. Teksten uit Urartu uit de 9e-8e eeuw v. Chr. In Schrijvend verleden, documenten uit oude Nabije Oosten, ed. by K. R. Veenhof, 32-45. Leiden: Ex Oriente Lux. Wilhelm, Gernot. 1976a. Zur urartaischen Nominalflexion. Zeitschrift fur Assyriologie 66, 105-119. Wilhelm, Gernot. 1976b. Parrattarna, Saustatar und die absolute Datierung der NuziTafeln. Acta Antiqua Academiae Scientiarum Hungaricae 24, 149-161. Wilhelm, Gernot. 1983. Der hurritische Ablativ-Instrumentalis /ne/. Zeitschrift fur Assyriologie 73, 96-113. Wilhelm, Gernot. 1984. Die Inschrift auf der Statue der Tatuhepa und die hurritischen deiktischen Pronomina. Studi micenei ed egeo-anatolici 24, 215-222. Wilhelm, Gernot. 1989. The Hurrians. Warminister: Aris and Phillips. Wilhelm, Gernot. 1992. Zum hurritischen Verbalsystem. In Texte, Satze, Worter und Moneme: Festschrift fur Klaus Heger zum 65. Geburtstag, ed. by Susanne R. Anschutz, 659-671. Heidelberg: Heidelberger Orientverlag. Wilhelm, Gernot. 1993. Zur Grammatik und zum Lexikon des Hurritischen. Zeitschrift fur Assyriologie 83, 99-118. Winckler, Hugo, and Ludwig Abel. 1889. Der Thontafelfund von El Amarna. (Mittheilungen aus den orientalischen Sammlungen der Museen zu Berlin, Hefte 1— 3.) Berlin: Spemann. Xacikjan, Margarita L. 1985. Xurritskij i urartskij jazyki. Yerevan: Akademija Nauk Armjanskoj SSR.

3 Suffixaufnahme in Hurrian: Normal Cases and Special Cases Use Wegner 1. Instant Insight The phenomenon of Suffixaufnahme—the repetition of the number and case marking of the head noun on the attributive noun in the genitive case—has been on record since Hurrian first began to be investigated. It is commonly regarded as a peculiarity of Hurrian and as one of the most characteristic features of this agglutinative and exclusively suffixing language of the Ancient Near East (cf. Speiser 1941: 201). Suffixaufnahme or "Suffixanreihung" (suffix-stringing) in Hurrian genitival constructions was first recognized by the Assyriologist Leopold Messerschmidt, who described it in his first comprehensive interpretation of the Mittani Letter in 1899, about ten years after its discovery. Taking as his point of departure this passage of the Mittani Letter, (1) Immoria=ve KUR Mizirre=ve=NE=ve evri=ve 'of Immoria, of the land of Egypt's lord' (Mit. III 104-105),

Messerschmidt (1899: 4f.) interpreted the ending -ve-NE-ve (which he actually read as -pi-ni-e-pi) thus: Apparently in order to render the connection between the two words in such cases as tight as possible, the genitive receives the ending of the noun governing it [i.e., -ve of evri=ve lord-Gen] in addition to its own ending [Mizirre=ve Egypt-Gen]. . . . A lengthening-syliable -ni-e [i.e. -NE] is further added to -pi [i.e. -ve in Mizirre=ve], and it is almost always found when the genitive is followed by the head noun. (My translation.)

Noting a further example of this kind in the Mittani Letter, (2) sen = iffu = ve = NE = ve asti=ve 'of my brother's wife' (Mit. III 76),

Suffixaufnahme

in Hurrian: Normal Cases and Special Cases

137

Messerschmidt commented: "Again the rectum [sen-iffu brother-my], in addition to the genitive ending [-ve], carries the ending of the regens [-ve of asti=ve wife-Gen], which is added by means of the lengthening-syllable [-NE]." This sequence of suffixes had in fact already been observed by Peter Jensen (1894: 438), but Jensen initially mistook -ni-e-pi for a gentilitial ending. When Jensen dealt with the Mittani Letter once more a few years later, his revised interpretation was essentially the same as Messerschmidt's: "The case ending of the regens may, apparently randomly, be attached to the preceding governed word, after its own genitive ending and the linking element -ne" (Jensen 1899: 179, my translation). Jensen now also pointed to a similar pattern in nominalizations, as in example (13) below. In two studies on Hurrian, Ferdinand Bork (1905, 1913) also emphasized that the ending of governing nouns may here be added to the genitive of governed nouns, and he recognized what he saw as comparable patterns especially in Elamite (where, however, the added ending is not one of case) and Old Georgian. This purported peculiarity of Hurrian is thus one of the very few grammatical phenomena whose very earliest interpretations have proved essentially correct; this is all the more remarkable considering that progress in Hurritology over the past century has not otherwise been steady and smooth. Concerning Suffixaufnahme, it is mainly terminology that has tended to vacillate since the pioneering researches of Messerschmidt and Jensen, and to some extent their interpretation of the "lengthening" or "linking" element has undergone revision as well. As an alternative to Messerschmidt's "Suffixanreihung" Johannes Friedrich (1935:124) adopted the term "Suffixaufnahme" (suffix-resumption) coined by Franz Nikolaus Finck (1910: 141). Later Friedrich (1939: 3) came to prefer "Suffixuibertragung" (suffix-transference), which seemed to him more neutral with regard to directionality, as was appropriate for Hurrian, where the suffix"resuming" genitive could precede as well as follow its head noun. Covering Hurrian for the Handbuch der Orientalistik, Friedrich (1969: 21) returned to "Suffixaufnahme," the term that by now enjoyed the widest currency in Hurritology. "Suffixanreihung," however, has also remained popular (cf. Kammenhuber 1968a: 248; Diakonoff 1971: 88), and "suffix-duplication" (Speiser 1941: 95; Bush 1964: 150) and "suffix-copying" (Plank 1990) are more recent additions to the terminological repertoire. Messerschmidt's "lengthening-syllable" and Jensen's "linking element" was renamed "suffixanreihend" (suffix-stringing) by Friedrich (1935: 123ff.), a term he had been employing in his Urartian studies; the interpretation of this particle, read as -ni by Friedrich, remained problematic, however. Terminological alternatives of more recent origin are "suffix-stringing article" and "carrier suffix" (Girbal 1988: 130; 1990: 99f.). The confusion about this suffix—and this is not merely a terminological matter—is due to its ability to occur in two different positions in the suffixal chain: (a) immediately after the stem (i.e. after root and root extension), functioning as what has been called a singular "article," and (b) after the

138

The Ancient Near East

genitive suffix, forming the base for Suffixaufnahme. Apart from correctly distinguishing these two functions of -ne, viz. those of "article" and of "carrier" (in the present paper the carrier function is indicated by capitalization), Bush (1964: 153f.) pointed out the anaphoric or cataphoric force of the carrier suffix (thus (2), for example, is to be rendered more accurately as 'of the wife [of] that of my brother'). It is largely owing to Bush's (1964: 148ff.) fundamental contribution that there is now a general consensus about the range of occurrence of Suffixaufnahme in Hurrian. Apart from straightforward genitival constructions, Suffixaufnahme occurs with attributes formed by the suffixes -hhe and -sse, which derive relational adjectives and nominalize verbs, respectively. In addition, Suffixaufnahme is attested with some further, albeit fairly rare, suffixes, including "adjectivalizing" -n(n)i (Bush 1973: 50; Wilhelm 1976: 149). The suffixal conglomerate -nnohha (Kammenhuber 1968a: 250f.; Wilhelm 1985: 495; Xacikjan 1985: 128), possibly to be analysed as -nni-hhe-a, is special insofar as it may occur on both members of an attributive construction, as is seen in (3). (3) sen=iffu=ve=NE=nnohha tiza=nnohha 'according to what is in my brother's heart (tiza)' (Mit. II 10)

Thus in the language of the Mittani Letter Suffixaufnahme essentially consists in the repetition on the attribute of the case and number suffixes of the head noun, including the suffixed "article" capable of distinguishing number (-nel-nd), separated by the anaphoric suffix -NE/-NA (Sg/Pl) from the attribute's own genitive (-ve), adjectivalizing (-hhe), or nominalizing (-sse) suffixes. In Bogazkoy Hurrian it occasionally seems that further elements could be resumed on attributes; but the relevant contexts are presently too unclear to be certain about this. Suffixaufnahme serves to indicate which two words are to be related to each other as the members of an attributive construction. This particular technique of aggreement-marking is not among the traits that Skalicka (1979) considers characteristic of the agglutinative type. Since Hurrian otherwise conforms almost perfectly to the holistic "construct" as conceived by Skalicka (cf. Plank 1988: 80f.), one would not be astonished if its application in this prototypically agglutinative language were highly irregular. This, however, is not the case. Although Suffixaufnahme is not absolutely obligatory in Hurrian, deviations from the norm as sketched above are exceedingly rare. To judge from Hurrian, Suffixaufnahme and agglutinative morphology would thus seem to be typological traits showing a closer affinity to each other than was anticipated by Skalicka. After a selection of standard instances of Suffixaufnahme is presented in Section 2, complementing that in Wilhelm's contribution to this volume, its more unusual varieties will be documented in Section 3. (Suffixaufnahme with derived adjectives and nominalizations is covered in greater detail by Wilhelm.)

Suffixaufnahme

in Hurrian: Normal Cases and Special Cases

139

The examples given here are drawn from the Mittani Letter (Mit.) as well as from the Hurrian texts of the archives of Bogazkoy (Bog.; publications are identified by abbreviations, for which see Sources of Examples below). The Mittani Letter formed part of the voluminous correspondence between Tusratta, king of Mittani, and pharaoh Amenophis III of Egypt (c. 13901352 BCE) concerning the prospective marriage between Tusratta's daughter and Amenophis III. Found in Amarna in Egypt in 1887, this is the only letter of this correspondence to be written in Hurrian; all the others are in Akkadian. With its approximately 500 lines, it is the longest coherent Hurrian text available, and despite occasional obscure passages, it has formed the basis of all grammars of Hurrian to date. Its orthography is as consistent as its syntax is complex. The texts preserved at Bogazkoy, the site of the Hittite capital Hattusa, are linguistically far more heterogeneous, although the content of most of them is uniformly religious, and their investigation lags behind that of the Mittani Letter. To a considerable extent the Bogazkoy texts have so far defied translation, and some are corrupted owing to the long period over which they were copied again and again by scribes with a limited, or perhaps no, command of Hurrian.

2. Normal Cases Examples (4)-(12), sparsely annotated, illustrate what must be considered the normal pattern of Suffixaufnahme in Mittani and Bogazkoy Hurrian. If the head noun is in the absolutive singular, which is zero-marked, there are no overt number and case endings that might be transferred to the preceding genitival attribute: (4) sen=iffu=ve asti brother-1SgPoss-Gen wife (AbsSg) 'my brother's wife' (Mit. III 21; cf. Bush 1964: 150)

The singular (5) and plural (6) forms of the definite "article" of the head noun, with zero marking for case (absolutive) in these examples, may be copied onto the genitival attribute like ordinary markers of number and case. (5) segurni=ve=ne tuppi=ne fate(?)-Gen-ArtSg tablet-ArtSg (Abs) 'the tablet of fate (or life)(?)' (Bog.; ChS I/1 No. 9 III 42) (6)

D

erve = na D NIN. GAL=ve=na dog-ArtPl (Abs) Ningal-Gen-ArtPl 'the dogs of (the goddess) Ningal' (Bog.; KUB XLV 47 + Bo 4186 III 19')

There is, however, no carrier suffix -NE/-NA here, which would be homonymous with the resumed "articles." Suffixaufnahme of the singular "article" is

140

The Ancient Near East

probably not obligatory and is not attested at all in the Mittani Letter, so far as unambiguous examples are concerned. The plural article, however, is copied regularly. In the following examples the head noun has an overt case suffix (genitive (7), directional (8), ergative (9, 10), equative (11)), and the attribute always adds the appropriate carrier suffix (singular or plural depending on the number marking of the head) after its own case marking and before the copy of the head's case and possibly its pluralizer (8, 10, 11). Recursive attribution, as in (7), where the topmost head lacks overt number and case marking, will be dealt with in Section 3. (7) sen=iffu=ve=NE=ve asti=ve nigari brother-1SgPoss-Gen-CarrSg-Gen wife-Gen (Sg) dowry (AbsSg) 'the dowry of the wife of my brother' (Mit. IV 48; cf. Bush 1964: 150) (8) en(i)=na=as=ta attan(i)=ne=ve=NA=as=ta god-ArtPl-Pl-Dir father-ArtSg-Gen-CarrPl-Pl-Dir 'to the gods of the father' (Bog.; ChS 1/2 No. 43 Rs. 19) (9) sen=iffu=ve=NE=z asti=z brother-1SgPoss-Gen-CarrSg-Erg wife-Erg (Sg) 'my brother's wife' (Mit. III 7; cf. Bush 1964: 151) (10)

itkalzi=ne=ve=NA=azu=z itkitanni=m(ma) edi=v siye=na=azu=z . . . D Hebatte ( D Hebat-ve)=NA=azu=z itkitanni=m(ma) edi=v siye=na=azu=z purification-ArtSg-Gen-CarrPl-Pl-Erg VERB-2SgAbsPro body-2SgPoss waterArtPl-Pl-Erg . . . Hepat-Gen-CarrPl-Pl-Erg VERB-2SgAbsPro body2SgPoss water-ArtPl-Pl-Erg The waters of purification may purify you, your body . . . the waters of (the goddess) Hepat may purify you, your body' (Bog.; Chs I/1 No. 9 Vs. II 2931) [The pluralizer -azu consists of -as plus a linking vowel u [a], appearing only in front of an overt case marker. The analysis of the non-indicative verb is unclear; -m is the short form of the enclitic 2nd person singular absolutive pronoun.]

(11)

puttukki=azo=nna=lla=an Sove=NA=azo=nna im=o=kko achievement-Pl-Equ-3PlAbsPro-and 1SgGenPro-CarrPl-Pl-Equ equal-IntransNeg 'and they do not equal my achievements' (Mit. III 60; cf. Girbal 1988: 130f.; 1992) [-lla is the enclitic 3rd person absolutive pronoun identifying the subject; sove is the genitive of the independent 1st person singular pronoun, behaving like a genitive noun with respect to Suffixaufnahme.]

Finally, here are two example of a nominalized verb, functionally analogous to a relative clause, acting as the target of Suffixaufnahme:

Suffixaufnahme

in Human: Normal Cases and Special Cases

141

(12)

satti=dil kad=i=u=sse=na [tive=na] 1PlAbsPro-1PlAbsPro tell-Trans-2Sg-Nomin-ArtPl [word-ArtPl] '[the words] which you tell us' (Bog.; ChS I/1 No. 41 Vs. II 29f.; cf. Wegner 1993)

(13)

sove=man tuppi nigar(i)=ne=ve ar=oz=av=sse=NE=ve 1SgGenPro-Particle tablet (AbsSg) dowry-ArtSg-Gen give-Past-Trans1SgNomin-CarrSg-Gen 'my tablet of the dowry which I have given' (Mit. III 40f.; cf. Bush 1964: 152; Girbal 1993)

The verb kad- 'tell' in (12), nominalized by means of -sse, is (bi-)transitive and hence takes the appropriate transitive marker -i that is obligatory in the present tense with 2nd or 3rd person agents; -u [o] is a marker of 2nd person singular subject. The recipient of the information, here consisting of the independent 1st person plural pronoun (satti) combined with the corresponding enclitic pronoun (-dil), is in the absolutive case. What the recipient is told is to be inferred contextually; its definiteness and plurality is, however, reflected in the final suffix -na of the nominalized verb. In (13), the verb ar 'give', also nominalized by means of -sse, adds the singular carrier suffix -NE before resuming the genitive suffix of its head, itself an attribute. (If one were to relate the genitive possessive pronoun sove to 'dowry' rather than to 'tablet', it would be expected to copy the genitive of this attribute; possibly tuppi nigar(i) =ne=ve is such a close-knit unit as to preclude Suffixaufnahme.) 3. Special Cases Though in general remarkably regular, Suffixaufnahme in Hurrian is not absolutely dependable. The present section surveys circumstances under which it is sometimes or perhaps always found not to apply, or found to apply less exuberantly than might be expected, or found to apply where one might not necessarily expect it to apply.

3.1. Close-knit Genitives Suffixaufnahme is missing in what Wilhelm (1983: 102) has called "close-knit genitive constructions," as illustrated most clearly in (14). (14)

URU

ninua=ve NE=va DSavuska=va Ninive-Gen Sawuska-Dat 'to the goddess Sawuska of the city of Ninive' (Mit. III 98)

Although he was unaware of the possible influence of the closeness of the ties between head and attribute, with the two even capable of forming a kind of

142

The Ancient Near East

lexical unit, Messerschmidt (1899) had already noted that what one would expect here is (14'), with the dative also on the attribute. (14')

URU

ninua=ve=NE=va DSavuska=va

3.2. Recursive Attribution While generally unexceptional as long as there is only a single attribute, Suffixaufnahme is usually limited in its application in recursive attribution. Like (7) above, (15) shows two levels of attribution, but unlike in (7), the topmost head in (15) is overtly marked for case (viz. directional). (15)

sen=iffu=ve=NE=va=t(ta)=an asti=j=ve nigari=j=da brother-1SgPoss-Gen-CarrSg-Gen-1SgAbs-and wife-3SgPoss-Gen dowry3SgPoss-Dir 'and to the dowry of the wife of my brother' (Mit. IV 46; cf. Wilhelm 1985: 491;Girbal 1990: 99f.)

The hierarchically lowest genitival attribute, sen=iffu=ve, copies the case of its immediate head, which is genitive because this head itself functions as attribute. (The actual form of the genitive copy is -va rather than -ve, with its vowel assimilated to that of the following enclitic pronoun.) The genitival attribute at the first level of embedding, however, does not copy the case of its head (directional); and this topmost case does not trickle down to the lowest attribute either, thus appearing only once in the entire construction. (15') shows what would have been the result if Suffixaufnahme had applied mechanically (cf. Girbal 1990: 100), with suffix-order at the lowest level obeying the mirror-image principle (Plank 1990). (15') sen=iffu=ve=NE=ve=NE=da=t(ta)=an asti=j=ve=NE=da nigari=j=da Suffixaufnahme may also pattern rather differently in recursive attribution, with the relevant examples coming from Bogazkoy Human: (16)

en(i)=na attan(i)=ne=ve=na D Savuska=ve=na god-ArtPl (Abs) father-ArtSg-Gen-ArtPl Sawuska-Gen-ArtPl 'the gods of the father of the goddess Sawuska' (Bog.; KUB XXVII1171; cf. Wegner 1981: 45 note 179)

The topmost head here is without overt case marking, but the first genitival attribute from above, attan(i)=ne=ve, resumes its plural number marking; and this plural suffix is also passed on to the lowest genitive, DSavuska=ve, which, differing from its counterpart in (15), fails to resume the genitive of its immediate head. On the model of (15) one would have expected this: (16') en(i)=na attan(i)=ne=ve D Savuska=ve=NE=ve

Suffixaufnahme

in Hurrian: Normal Cases and Special Cases

143

To account for the form actually attested, one might take en(i)=na attan(i) = ne=ve=na for another close-knit construction (despite the repeated occurrence of plural number), with DSavuska=ve then being a genitival attribute at the highest level of embedding, regularly entitled to receive the number suffix of the head, en(i)=na. (The alternative translation 'the gods of the father and of the goddess Sawuska', not implausible on structural grounds, is ruled out by the context.) With another level of embedding added, (17) resembles (16) in respect to the topmost head and the first two attributes: (17)

en(i)=na attan(i)=ne=ve=na avar(i)=ne=ve=NE=ve=na D Savuska=[ve=na] god-ArtPl (Abs) father-ArtSg-Gen-ArtPl field-ArtSg-Gen-CarrSg-Gen-ArtPl Sawuska[-Gen-ArtPl] 'the gods of the father of the goddess Sawuska of the field' (Bog.; KUB XXVII 6 I 26-27)

However, in what seems to be a further variation, the lowest genitival attribute, avar(i)=ne=ve, copies the genitive of its immediate head (on the Mittani model of (15)) as well as the plural marker that must ultimately be traced to the topmost head. What is not on record so far are examples of recursive embedding where Suffixaufnahme would be consistently practiced at each successive level. The general pattern of Suffixaufnahme in recursive attributions for Mittani Hurrian can thus be represented as in Figure 3.1, (i), with pattern (ii) accommodating what seems to be a specialty of Bogazkoy Hurrian. 3.3. Missing Heads A third deviation from the norm is Suffixaufnahme in spite of the overt absence of heads, which need to be inferred from the context. Straightforward examples of hypostasis, with attributes carrying the case marking, including zero, that would be appropriate for their heads if they were present, can be found in the Mittani Letter as well as in Bogazkoy Hurrian. (i) HEAD-Case [ATTRIBUTE1-Genitive [ATTRIBUTE2-Genitive-Genitive]]

(ii) HEAD-Suffix [ATTR-Gen-Suffix [ATTR2-Gen-Suffix [ATTR3-Gen-Gen-Suffix]]]

Figure 3.1. Recursive Attribution in Hurrian.

144

The Ancient Near East

(18)

fez=nna=an atta=iffu=va uaduranna keban=oz=o=sse tea attai=v=(v)e=NE=dan 2SgErgPro-3SgPro-and father-1SgPoss-Dat NOUN? [in the essive case: uadurann(i)=a?] send-Past-2Sg-Nomin ADVERB father-2SgPoss-GenCarrSg-Abl 'and what you have sent my father as . . . , was more than your father's' (Mit. III 68--69; cf. Bush 1964: 153)

(19)

adi=nin D Simige=ne=ve=NE=mmaman amm=oz=a thus-Particle Simige-ArtSg-Gen-CarrSg-Particle reach-Past-3SgTrans 'thus he/she/it reached that [sc. the city] of the [sun] god Simige' (Mit. I 94)

(20)

. . . aga=ve am=ud=o=m eza=ve=ma fur=ud=o=m . . . yonder-Gen (AbsSg) reach-Neg-3SgTrans there-Gen-Particle (AbsSg) see-Neg-3SgTrans '. . . the one [sc. shore] yonder he does not reach, the one [sc. shore] here he does not see' (Bog.; KBo XXXII 14 Vs. I 28-29)

The genitival attribute in (18) lacks an overt head but nonetheless is in the ablative, as would be appropriate for noun phrases expressing the standard of comparison in such sentences. In (19) the carrier suffix after the attribute's genitive is not followed by number or case suffixes because the unexpressed head is singular and absolutive (in the relation of direct object), both unmarked. In (20) it is demonstrative stems that are in the genitive and thus imply a head element, contextually reconstructable as 'shore'. The resulting hypostases are also singular and absolutive and hence lack further number and case suffixes; unlike DSimige=ne=ve in (19), they also lack a carrier suffix. In view of the possibility of such absolute, headless occurrences of genitival attributes with carrier suffixes and Suffixaufnahme, the following instances of recursive attribution, both from Bogazkoy texts, can be considered perfectly regular. (21) huessa DTessub=va egli=ve=NE=da subri=ve=NE=da en(i)=na=az=(v)e=NE=da sarri=ne=da en(i)=na=az=(v)a=1(1)a? egli=ve=NE=ve=NA=az=(v)a=1(1)a? subri=ve=NE= ve=NA=az=(v)a=1(1)a ? turohhe=na=az=(v)a=1(1)a ? furul(i)=ne= ve=NA=az=(v)a=1(1)a? D E.A=da mad=o=nne=NE=da furul(i)=ne=ve=NE=da agri . . . VERB? Tessub-Dat salvation-Gen-CarrSg-Dir ,swbri?-Gen-CarrSg-Dir godArtPl-Pl-Gen-CarrSg-Dirking-ArtSg-Dirgod-ArtPl-Pl-Dat-3PlAbsPro? salvation-Gen-CarrSg-Gen-CarrPl-Pl-Dat-3PlAbsPro?.subri?-Gen-CarrSgGen-CarrPl-Pl-Dat-3PlAbsPro7male-ArtPl-Pl-Dat-3PlAbsPro? house-ArtSgGen-CarrPl-Pl-Dat-3PlAbsPro? EA-Dirwise-Derivation?-Adjct-CarrSg-Dir house-ArtSg-Gen-CarrSg-Dir incense (AbsSg) 'VERB?, to the [weather] god Tessub, to the one of salvation, to the one of subri, to the king of the gods; to the gods of the one of salvation, of the one

Suffixaufnahme

in Hurrian: Normal Cases and Special Cases

145

of the subri; to the male [gods] of the house; to the wise god EA of the house, the incense . . .' (Bog.; ChS I/1 No.43 Vs. II 17'-21'; cf. Kammenhuber 1968a: 138f.) Dative and directional being equivalent cases (as also in Mittani Hurrian), it is clear that the two genitives egli=ve and subri=ve must be syntactically parallel to DTessub=va, and therefore they can only be hypostases, with overt heads missing but with the case marking appropriate for headed noun phrases in this environment (directional/dative). Such hypostases now have the same syntactic potential as ordinary genitives, and this includes functioning again as attributes, requiring a second genitive suffix plus whatever suffixes need to be copied from heads. Such triple case marking is exemplified here by the double genitives egli=ve=NE=ve and subri=ve=NE=ve, resuming plural, case, and another suffix of their joint head, en(i)=na=az=(v)a=l(l)a?. This sentence in addition shows three further regular examples of simple Suffixaufnahme with overt heads, but it is not fully understood because the meaning of what appears to be the verb, huessa (with stem hu(e)-?), is unknown, and the function of enclitic -l(l)a is unclear. The same words with a double genitive, one owing to Suffixaufnahme from a non-expressed head and a further case suffix copied from an overt head (here again dative), occur in (22). (22) unni=ma DTessub=va sarri=ne=va evre=n(i)=ne=va . . . en(i)=na=az=(v)a egli=ve=NE=ve=NA=az=(v)a subri=ve=NE=ve=NA=az=(v)a . . . un=a now-Particle Tessub-Dat king-ArtSg-Dat lord-Indiv-ArtSg-Dat . . . god-ArtPl-Pl-Datsalvation-Gen-CarrSg-Gen-CarrPl-Pl-Dat fuftn'-Gen-CarrSg-Gen-CarrPl-Pl-Dat . . . come-Intrans 'Now he comes to Tessub, to the king, to the lord . . . , to the gods of the savior [lit. of the one of salvation], of the one of the Subri . . .' (Bog.; ChS I/ 1 No. 2 Vs. 14'-15') (The suffix -ni in evre=n(i)=ne=ve has an individualizing function.) Triple case marking as exemplified in (21) and (22) is rare. Significantly, its only known occurrences are due to the ellipsis of heads—which enables a genitival attribute to take another genitive (reflecting its own syntactic function) and further suffixes, including one for case (accruing from Suffixaufnahme)—rather than to straightforward recursive attribution as such, where Hurrian tends to utilize Suffixaufnahme economically (see Section 3.2). Sources of Examples ChS

I/1

Corpus der hurritischen Sprachdenkmaler, ed. by Volkert Haas, Mirjo Salvini, Ilse Wegner, and Gernot Wilhelm. I. Abteilung: Die Texte aus Bogazkoy. Band 1: Die Serien itkahi und itkalzi des AZU-Priesters,

146

ChS 1/2

KBo KUB Mit.

The Ancient Near East Rituale fur Tasmisarri und Tatuhepa sowie weitere Texte mit Bezug auf Tasmisarri, by Volkert Haas. Rome: Multigrafica Editrice, 1984. Corpus der hurritischen Sprachdenkmaler, ed. by Volkert Haas, Mirjo Salvini, Ilse Wegner, Gernot Wilhelm. I. Abteilung: Die Texte aus Bogazkoy. Band 2: Die Rituale des AZU-Priesters, by Mirjo Salvini and Ilse Wegner. Rome: Multigrafica Editrice, 1986. Keilschrifttexte aus Boghazkoi, Heft 1-33. Leipzig, later Berlin, 19201990 (continuing), Keilschrifturkunden aus Boghazkoi, Heft 1-60. Berlin, 1921-1990 (continuing). The so-called Mittani Letter, text No. 20 in Die Tontafeln von El-Amarna (Vorderasiatische Schriftdenkmaler 12), Leipzig: J. C. Hinrichs, 1915; used in conjunction with Johannes Friedrich's transliteration in Kleinasiatische Sprachdenkmaler, Berlin: de Gruyter, 1932.

References Bork, Ferdinand. 1905. Kaukasisches. Orientalistische Litteratur-Zeitung 8, 184-187. Bork, Ferdinand. 1909. Die Mitannisprache. Mitteilungen der Vorderasiatischen Gesellschaft 14, 1/2. Berlin: Wolf Peiser. Bork, Ferdinand. 1913. Kaukasisches (Nachtrag zu OLZ 1905 Sp. 184 ff.). Orientalistische Literaturzeitung 16, 385-387. Bush, Frederic W. 1964. A Grammar of the Human Language. Ph.D. dissertation, Brandeis University (Ann Arbor: University Microfilms). Bush, Frederic W. 1973. The relationship between the Hurrian suffixes -nel-na and -nni/e/-nna. In Orient and Occident: Essays Presented to Cyrus H. Gordon, ed. by Harry A. Hoffner, Jr., 39-52. (Alter Orient und Altes Testament 22.) Kevelaer and Neukirchen-Vluyn: Butzon & Bercker, Neukirchener Verlag. Diakonoff, Igor M. 1971. Hurrisch und Urartaisch. (Munchener Studien zur Sprachwissenschaft, Beiheft 6 N.F.) Munich: Kitzinger. Finck, Franz Nikolaus. 1910. Die Haupttypen des Sprachbaus. Leipzig: Teubner. Friedrich, Johannes. 1935. Zum Subaraischen und Urartaischen. In Miscellanea Orientalia, dedicata Antonio Deimel annos LXX complenti, 122-135. (Analecta Orientalia 12.) Rome: Pontificium Institutum Biblicum. Friedrich, Johannes. 1939. Kleine Beitrage zur churritischen Grammatik. Mitteilungen der Vorderasiatisch-Agyptischen Gesellschaft 42, 2. Leipzig: J. C. Hinrichs. Friedrich, Johannes. 1969. Churritisch. Handbuch der Orientalistik II; 1/2, 1-30. Leiden: Brill. Girbal, Christian. 1988. Der Paragraph 24 des Mittani-Briefes. Zeitschrift fur Assyriologie 78, 122-136. Girbal, Christian. 1990. Zur Grammatik des Mittani-Hurritischen. Zeitschrift fur Assyriologie 80, 92-101. Girbal, Christian. 1992. Zum hurritischen Vokabular. Studi micenei ed egeo-anatolici 29, 159-169. Girbal, Christian. 1993. Selbstandige Personalpronomina im Urartaischen und Hurritischen. In The Anatolian Connexion: Memorial Offerings for Charles Carter in Anatolian and Other Asia Minor Languages, ed. by Yoel L. Arbeitman. Jensen, Peter. 1894. Grundlagen fur eine Entzifferung der (hatischen oder) cilicischen

Suffixaufnahme

in Human: Normal Cases and Special Cases

147

(?) Inschriften. Zeitschrift der Deutschen Morgenlandischen Gesellschaft 48, 235-352, 429-485. Jensen, Peter. 1899. Zur Erklarung des Mitanni. Zeitschrift fur Assyriologie 14, 173181. Kammenhuber, Annelies. 1968a. Hurrische Nomina. In Studien zur Sprachwissenschaft und Kulturkunde: Gedenkschrift W. Brandenstein, ed. by Manfred Mayrhofer, 247-258. Innsbruck: Innsbrucker Beitrage zur Kulturwissenschaft 14. Kammenhuber, Annelies. 1968b. Die Arier im Vorderen Orient. Heidelberg: Carl Winter. Messerschmidt, Leopold. 1899. Mitanni-Studien. Mitteilungen der Vorderasiatischen Gesellschaft 4/4. Berlin: Wolf Peiser. Plank, Frans. 1988. Das Hurritische und die Sprachwissenschaft. In Hurriter und Hurritisch, ed. by Volkert Haas, 63-93. (Xenia 21.) Konstanz: Universitatsverlag. Plank, Frans. 1990. Suffix copying as a mirror-image phenomenon. Linguistics 28, 1039-1045. Skalicka, Vladimir. 1979. Typologische Studien. Braunschweig: Vieweg. Speiser, Ephraim A. 1941. Introduction to Human. (Annual of the American Schools of Oriental Research 20.) New Haven: American Schools of Oriental Research. Wegner, Ilse. 1981. Gestalt und Kult der Istar-Sawuska in Kleinasien. (Alter Orient und Altes Testament 36; Hurritologische Studien 3.) Kevelaer and NeukirchenVluyn: Butzon & Bercker, Neukirchener Verlag. Wegner, Ilse. 1993. Hurritische Verba dicendi mit einfacher und doppelter AbsolutivRektion. In The Anatolian Connexion: Memorial Offerings for Charles Carter in Anatolian and Other Asia Minor Languages, ed. by Yoel L. Arbeitman. Wilhelm, Gernot. 1976. Parrattarna, Saustatar und die absolute Datierung der NuziTafeln. Acta Antiqua Academiae Scientiarum Hungaricae 24, 149-161. Wilhelm, Gernot. 1983. Der hurritische Ablativ-Instrumentalis /ne/. Zeitschrift fur Assyriologie 73, 96-113. Wilhelm, Gernot. 1985. Hurritische Lexikographie (Rezensionsartikel zu Emmanuel La Roche, Glossaire de la langue hourrite). Orientalia 54, 487-496. Xacikjan, Margarita L. 1985. Xurritskij i urartskij jazyki. Yerevan: Akademija Nauk Armjanskogo SSR.

This page intentionally left blank

III THE CAUCASUS

This page intentionally left blank

4

Suffixaufnahme in Kartvelian Winfried Boeder ids natesavobitman bmnuamancueulebaj priad sakebeli, e.i. tan-qmobaj, romeltaca saxelta, ganmatwsebelad sexuedebis sitquasa sina. 'The genitive case knows a very laudable habit, namely agreement with nouns with which it meets as determiner in a sentence.' (Antoni I, royal prince and patriarch of the Georgian church, in his Georgian grammar of 1767 [1885: 136])

1. Introduction

1.1. On Kartvelian The Kartvelian or South Caucasian languages form a family of genetically closely related languages which, as far as we know, have always been spoken in approximately the same area: the western and middle regions of Transcaucasia, including parts of present-day northeastern Turkey. Kartvelian comprises Georgian in the eastern and middle regions, Svan in the northwestern mountain region, Mingrelian in the northwest, and Laz in the southwest of the area. Whereas the latter two languages have been attested mainly since the nineteenth century, Georgian has a written tradition from the fifth century onward. It is the language of a rich literature beginning with translations of the Gospel. Other religious texts were translated from Greek, Armenian, Syriac, or Arabic and contain occasional caiques, but there are long original works, such as Giorgi Merchule's tenth-century biography of Gregory of Khandzta, that afford an impression of elaborate Old Georgian prose. For the medieval period we have an even larger body of texts, including chronicles and epics such as Shota Rustaveli's Man in the Panther's Skin from around 1200. Standard Modern Georgian developed in the nineteenth century and is the official language of the Republic of Georgia and the written language of about 3,900,000 people (according to the census of 1989), Mingrelians and Svans included.

152

The Caucasus

Georgian has been unusually conservative during the last 1,500 years; to a large extent old texts can still be understood by educated readers. Nonetheless, some grammatical rules have changed, and one of them concerns Suffixaufnahme. 1.2. Plan The principal aim of this chapter is to show that Suffixaufnahme is ordinary agreement, as undergone also, inter alia, by adjectives, and that genitive NPs can undergo ordinary agreement only after the basic constituent structure of NPs has been restructured. Since Suffixaufnahme in Old Georgian is inextricably connected with various aspects of NP structure, an array of phonological, morphological, syntactic, and even orthographic issues—some of which have also been addressed by previous researchers, notably Hans Vogt and the Georgian scholars Ivane Imnaisvili and Anton Kiziria—need to be raised in the course of this argument. The behavior of clitics in NPs and its implications for the boundaries and constituent structure of NPs is the first of these issues (Section 2). Some basic questions will then be asked about Suffixaufnahme itself, concerning its scope (3.2), the position of the NPs carrying it (3.3), and the internal constituent structure of these NPs, as revealed by constituent movement, discontinuity, and punctuation (3.4). Section 4 will look at recursive attribution. The focus of Section 5 will be on morphological and syntactic details such as heads without overt case marking (5.1), head marking not straightforwardly copied (5.2), the difference between genitives and adjectives (5.3), the actual host of Suffixaufnahme (5.4), and the relationship of Suffixaufnahme to multiple case-number marking in elliptic expressions (5.5). Section 6 surveys the modern Kartvelian languages, where Suffixaufnahme is partly retained. Finally, the possible functions and origins of Suffixaufnahme, as well as parallels in languages with which Kartvelian has been in contact, will be explored in Section 7.

1.3. Structural Sketch With some reservations, Georgian can be called a free-word-order and nonconfigurational language (Boeder 1989a). There is much evidence for operator-operand (adjunct-head) as the underlying order on both the clause and the phrase level, with final ordering in particular of the verb, of the head noun, and of adpositions. In spite of the enormous variation in Old Georgian word order (cf. Kalase 1961, Sar vela e 1984: 510-542), this assumption allows most surface structures to be derived by rules that refer to phonological, semantic, and communicative features of the constituents involved. On the phrase level, modifier-head and NP-postposition is the exclusive order in Modern Georgian and in most other Caucasian languages (see, however, 7.2). Transitive verb constructions show a split in relational typology, with an ergative case-marking pattern limited mainly to aorist forms of the verb

Suffixaufnahme

in Kartvelian

153

Table 4.1. Old Georgian Nominal Inflection. Common Nouns Absolutive Vocative Nominative Ergative Genitive Dative Instrumental Adverbial

Singular kackac-o kac-i kac-man kac-is(-a) kac-s(-a) kac-it(-a) kac-ad

Proper Names Plural — kac-n-o kac-n-i

1 kac-ta

Singular GrigolGrigol-is Grigol-s Grigol-it Grigol-ad

(Boeder 1979). Some scholars, however, prefer to assign Georgian to the "active" language type instead. Kartvelian has a rich inflectional morphology. In particular, the verb is polypersonal, containing prefixed exponents of all core arguments, in addition to preverbs (whose function in Modern Georgian is a perfectivizing one) and suffixes for mood, tense, and number. While verbal morphology displays much formal variation, the inflection of nouns, adjectives, and participles is— with a few minor exceptions—completely uniform. Table 4.1 gives the Old Georgian paradigms of the common noun kac- 'man' and of the proper noun Grigol 'Gregory'. Bearing on the topic at issue here, there is a kind of syncretism between the oblique cases in the plural: -ta indicates the plural but not the difference between ergative, genitive, and dative. In a sense this is in accordance with a universal tendency for case syncretism to take place in forms marked for number (Greenberg 1966: 27, Boeder 1976). In addition, number tends to be neutralized in Georgian, with the singular often used instead of the plural, in the instrumental and adverbial cases, which are the most highly marked cases in terms of feature composition (Sara eni e 1942, Boeder 1987a: 50). 2. The Delimitation of Noun Phrases

2.1. The Location of Clitics Although it is doubtful whether Old Georgian had verb phrases, it certainly had well-defined NPs, for there are rules that clearly refer to them. As the delimitation of NPs is of some relevance here, these rules need to be examined. One rule refers to the location of clitics such as the so-called articles and some particles in the NP. Articles are unstressed variants of demonstrative pronouns whose stressed forms occur in initial position. There is thus a contrast between (1), where igi is a demonstrative, and (2), where it is an article and an enclitic of the noun mteri, its "host" (to use a term coined by Arnold Zwicky).

154 (1)

The Caucasus igi mter-i that=Nom enemy=Nom 'that enemy'

(2) mter-i igi enemy-Nom Art=Nom 'the enemy'

Similarly, (3) contrasts with the article use of 'that' in (4), where ma-s cannot be the demonstrative because the context does not allow for textual deixis. (3) ma-s zam-sa (Mt 26,55) that-Dat time-Dat 'at that time' (4) zam-sa ma-s sakumevel-isa kumev-isa-sa (L 1,10DE) time-Dat Art-Dat incense-Gen burning-Gen-Dat 'at the time of incense'

Taking this contrast for granted, we see that the clitic occurs in non-final position if the NP consists of more than just the head noun: (5) qovel-sa ma-s kueqana-sa (Mt 9,31DE) all-Dat Art-Dat land-Dat 'the whole country'

More complex examples demonstrate that the clitic actually occupies second position in the phrase, abiding by Wackernagel's Law within phrases: (6) cmida-j igi mcire-j eklesia-j (Gm V 24) holy-Norn Art=Nom little-Norn church-Norn 'the holy little church'

Particles such as -ca 'and, too', comparable to Latin -que, can likewise move to second position; if there is also a clitic article, the clitic particle precedes it: (7) a. ama-s-ca cel-sa (L 13,8) this-Dat-too year-Dat 'this year also' b. mdidar-i-ca igi (L 16,22DE; Xanm) rich-Nom-too Art=Nom 'the rich man too' Like the article, the particle -ca cannot cross phrasal boundaries without changing its scope; compare (8a) with (8b).

Suffixaufnahme

in Kartvelian

155

(8) a. m-is tana mosrul-n-i-ca igi Huria-n-i (J 11,33) she-Gen with having=come-Pl-Nom-too Art=Nom Jew-Pl=Norn 'the Jews also who had come with her' b. m-is tana-ca mosrul-n-i igi Huria-n-i 'the Jews who had come with her, too' While clitic movement is regular in most texts and has even been considered obligatory (Kala e 1961: 105), there are a few examples without it. Thus the Adish manuscript of the Gospel, apparently reflecting a different dialect than the versions of the fifth- or sixth-century Khanmeti palimpsests and the Dshruchi and Parkhali manuscripts, is particularly reluctant to employ clitic movement:1 (9) a. qovel-sa kueqana-sa ma-s (Mt 9,31C; cf. (5)) all-Dat land-Dat Art-Dat 'the whole country' b. mdidar-i igi-ca (L 16,22C; cf. (7b)) rich-Nom Art=Nom-too 'the rich man too' Assuming a phrase-initial position of clitics in basic structure, a clitic movement rule is formulated in (10), with Y representing the first non-clitic constituent and X a clitic dominated by the same phrasal node as Y and Z. (10)

[ X - Y - Z ] 1 2 3

213

Clitic movement can now be used to determine the boundary of NPs:2 the constituent to the left of a clitic must be the first constituent of the NP. While such an observation might look trivial, determining the boundaries of NPs will soon be seen to be a central issue.

2.2. Ordinary Agreement A second rule that refers to NPs is ordinary agreement. Modifiers such as adjectives, numerals, pronouns, and articles agree with their nominal head in case and number. Examples are (6), where two adjectives and the article visibly agree with a nominative head, and (11), also showing number agreement. (11) a. or-n-i brma-n-i (Mt 9,27) two-Pl-Nom blind-Pl-Nom 'two blind men' b. or-ta mocape-ta (L 7,19) two-Pl(Dat) disciple-Pl(Dat) '[calling] two of his disciples'

156

The Caucasus c. netar-o Grigolblessed-Voc Gregorynetar-i Grigolblessed-Norn Gregorynetar-man Grigolblessed-Erg Gregorynetar-is Grigol-is blessed-Gen Gregory-Gen, etc.

Agreement occurs between immediate constituents of the NP, and only between them. For reasons that will become clear later (3.3.1), we need at least one additional, intermediate level between lexical categories on the lowest level of NP structure—noun, adjective, article, postposition, etc.—and NP on the highest. This intermediate level consists of adjective phrases, determiners, etc., and in particular of "small noun phrases" (NP') containing the head noun. Postpositional phrases also behave like adjective phrases (3.2.2), and in apparently later texts even non-nominal adverbial phrases have ordinary agreement: (12) aka-ta kac-ta ena-j (Sin. 43, p.28v apud Kiz 253) here-Pl(Gen) man-Pl(Gen) language-Norn 'the language of the people here'

These phrases are immediate constituents of NP that agree in case. But notice that agreement in case does not require an "agreeing" case marker on all immediate constituents. Some constitutents (proper nouns, 1st and 2nd person pronouns, etc.) refuse a case marker for the vocative, nominative, and ergative. Accordingly, Grigol in (11c) exhibits no case marker for these cases, while its modifier does. Case features assigned to the NP as a whole are spelt out selectively on those immediate constituents that do not refuse the corresponding case features. Case markers are attached on the right in each of these constituents.3 Number is copied from the head noun onto all coconstituents and is again spelt out on their right. In order for case-number markers to be attached correctly, heads (adjectives, nouns, articles, or postpositions) must be in final position in their own phrases. In (6') the operation of agreement is exemplified for NPs such as (6), enriched by the plural. (6') NP

Suffixaufnahme

in Kartvelian

157

2.3. Agreement in [case] Unlike these kinds of modifiers, dependent NPs in the genitive do not seem to agree with their heads in case and number: (13)

sikudil-isa gemo-j (J 8,52) death-Gen taste-Norn 'the taste of death'

There is a different type of agreement at work here, however. The paradigms in Table 4.1 offer two variants of the genitive, and also of the dative and instrumental: one long (-isa, -sa, -ita) and the other short (-is, -s, -it). The textual evidence suggests that this variation is not random (Dondua 1930 [1975: 32]). The adnominal genitive has the short form if its head noun has the short form (14a). From this point of view, the absolutive is simply the short form corresponding to the long nominative form. Thus absolutive heads, which are used for instance as predicates, tend to occur with the short form of the genitive (14b). (14) a. ded-is mucl-it-gan (Mt 19.12C) mother-Gen womb-Ins-from 'from the mother's womb' b. iqos ubralo- ert-is- col-is- kmar- (Tt 1,6 apud Imn 185) he=should=be blameless-Abs one-Gen-Abs wife-Gen-Abs husband-Abs 'He should be blameless, the husband of one wife'4

As seen in Table 4.2, nominative -i and the -a of the other case forms are in complementary distribution: -a occurs after case markers, -i after the pure stem (Boeder 1979: 448). If morphological case is to be represented in terms of features, then we may say that the feature [ + case] is spelt out as -i/-a on the head noun and that it triggers the long form of the genitive.5 Non-absolutive examples like (14a) are extremely rare (because short forms of the genitive and the dative occur only in a few archaisms), but we may safely posit a rule according to which genitives agree with their nominal head co-constituent in the feature [case]. What matters here is that adnominal adjectives and genitives share this type of agreement. Agreement in [case] has an interesting parallel in some Daghestanian languages (Bork 1913: 186, Boeder 1987a: 56f.), in particular in the Tsez group and in Tsakhur (Mel'cuk 1986: 82f., Boguslavskaja and Kibrik in this volume). In these languages, spoken in the mountains adjacent to a Georgian-speaking valley area, there are two types of genitive, depending on whether the head noun is in the direct case ("nominative") or in an oblique case. Functionally, there is a slight difference between Old Georgian and Daghestanian insofar as the Daghestanian direct case corresponds to both absolutive and nominative in Old Georgian, where the absolutive but not the nominative triggers the short genitive form. Historically, however,

158

The Caucasus Table 4.2. Inflection of Head Noun and Preceding Genitive in Old Georgian. Absolutive Nominative Genitive Dative Instrumental

col-is- kmar-$ col-is-a kmar-i col-is-a kmr-is-a col-is-a kmar-s-a col-is-a kmr-it-a 'wife's husband'

this differentiation between absolutive (pure stem) and nominative (ending in -z) may have been preceded by a stage in which both functions were covered by a formally unmarked "absolutive" (Boeder 1979: 461), triggering the direct case type of genitive as in Tsez.

2.4. Non-agreement There is perhaps one further regularity making reference to NPs. In appositive constructions like 'King Herod', the proper name can optionally be left uninflected when preceding the head noun (15a/b), whereas it must be inflected when following it (15c; Imnaisvili 1957: 575-580, Kiziria 1963: 227). (15)

a. Herode- mepe-sa (Mt 2,3) Herod- king-Dat b. Herode-s mepe-sa (Mk 6,14DE) Herod-Dat king-Dat c. mepe-sa Herode-s (Mk 6,14C) king-Dat Herod-Dat 'to King Herod'

These variants may differ semantically—'King Herod' (15a), 'Herod, the king' (15b), 'the king, Herod' (15c)—but the main point here is the formal opposition between possible non-agreement of preposed proper names (15a) and agreement otherwise (15b/c). (15a) could possibly be interpreted as the underlying structure; Herode mepe-sa would be a "small noun phrase" (2.2) with (phrasal) case marking occurring only once. In both (15b) and (15c) one of the terms has become a co-constituent of the small noun phrase, and the terms must agree in case: (15')

a. [[Herode mepe-sa]] b. [[Herode-s] [mepe-sa]] c. [[mepe-sa] [Herode-s]]

Suffixaufnahme

in Kartvelian

159

Notice, first, that Herode- has the short form of the dative in (15b/c). Proper names are inherently marked for "refusing" [+case] and therefore have short forms of case markers: dative Herode-s, genitive Herode-js, instrumental Herode-jt (see Table 4.1). Nominative and ergative have zero marking (Herode) because they require the feature [+case].6 The absence of [+case] in proper names, the prototypical referential expressions, on the one hand, and in non-referential predicate nominals on the other ("absolutive," 2.3) points to a referential meaning of [+case]. Second, agreement shows a double contrast: agreement in the dative case (15b/c) contrasts with its absence (15a) on the one hand, and with agreement in [case] on the other, with mepe- but not Herode- marked for [+case] in (15a/b/c). Suffixaufnahme is agreement in case (in the usual sense), in number, and in [case] (see 5.1). 3. Suffixaufnahme Essentials

3.1. Suffixaufnahme

Introduced

With NPs delimited, we can now turn to Suffixaufnahme. One of the oldest Georgian inscriptions, dated to 494 CE, begins with the words: (16)

secevn-ita cmid-isa sameb-isa-jta (Sos 1980 no. 2) help-Ins holy-Gen trinity-Gen-Ins 'with the help of the Holy Trinity'

Unlike all examples adduced so far, the genitive NP is postposed here and has two case suffixes: the genitive suffix signaling its adnominal relation, and the instrumental suffix agreeing with the suffix of its head noun in case. Similarly, the adnominal genitive adds plural nominative marking in (17a) and plural vocative marking in (17b). (17)

a. srosan-n-i vel-isa-n-i (Mt 6,28) lily-Pl-Nom field-Gen-Pl-Nom 'the lilies of the field' b. asul-n-o lerusalem-isa-n-o (L 23,28) daughter-PI-Voc Jerusalem-Gen-Pi-Voc 'daughters of Jerusalem!'

This is a phenomenon that Franz Bopp (1846: 275 [1972: 413]) found remarkable (cf. Boeder 1987a: 42). It was called "Suffixaufnahme" by Franz Nikolaus Finck (1910: 141)7 and "genitive with relationship" (roditel'nyy s" otnoseniem) by Marr (1925: 64); and the Georgian tradition knows it as "agreement of the degenitival determinans" (nanatesaobitari msazyvrelis setanxmeba) or "agreeing governed determination" (martul-setanxmebuli gansazyvreba).

160

The Caucasus

3.2. Phrasal Scope A first question raised by such constructions concerns the identity of the constituent that agrees with the head noun by Suffixaufnahme. With Dench and Evans (1988: 2f.) we may call the relevant constituent the "scope" of Suffixaufnahme and define it as the "host argument" of the case marker. The scope of Suffixaufnahme is a genitive NP whose final constituent agrees with its agreement controller in case and number (3.2.1). Other phrases also exhibit multiple case-number marking, but not Suffixaufnahme in the narrow sense (3.2.2). 3.2.1. Suffixaufnahme marks only one constituent of its scope. Extending the definition of Dench and Evans (1988: 3), we may refer to the range of distribution of case and number within the constituent over which the case has scope as the "domain" of case and number marking. Suffixaufnahme is coded by final case-number marking; the last constituent is thus the domain of Suffixaufnahme (see however 5.4). This is best seen in complex examples. In (18) the final suffix -j of the last noun of the whole coordinated group takes up the nominative suffix of keba-j 'praise', which is separated from it by punctuation (see further 3.4.4). (18)

aymovtkua keba-j . mam-isa da 3-isa da cmid-isa sul-isa-j (Phys XXXIV) I=may=pronounce praise-Nom . father-Gen and son-Gen and holy-Gen soulGen-Nom 'that I may pronounce the praise of the Father, the Son, and the Holy Ghost'

This suggests that the whole group 'the Father, the Son, and the Holy Ghost' forms one single NP consisting in turn of three different NPs:

By contrast, the standard wording of the often-repeated prayer formula is as in (19), with multiple case marking occurring with every single genitive noun. (19)

saxel-ita mam-isa-jta da 3-isa-jta da sul-isa cmid-isa-jta (Mt 28,19) name-Ins father-Gen-Ins and son-Gen-Ins and soul-Gen holy-Gen-Ins 'in the name of the Father, and of the Son, and of the Holy Ghost'

Suffixaufnahme

in Kartvelian

161

Arguably this pattern presupposes conjunction reduction, with (19) to be derived from 'in the name of the Father, and in the name of the Son, and in the name of the Holy Ghost'. Thus, while mam-isa-jta has Suffixaufnahme, 3isa-jta and sul-isa cmid-isa-jta presuppose ellipsis (see 5.5), and (19), thus analysed, does not contradict our interpretation of Suffixaufnahme as phrasal case-number marking. Australian-style "complete concord" (Dench and Evans 1988: 3-5), with the domain of Suffixaufnahme extending to every nominal constituent of the genitive NP, would also be conceivable here; but examples such as the following, constructed accordingly, are ungrammatical in Old Georgian. (15')

*secevn-ita cmid-isa-jta sameb-isa-jta help-Ins holy-Gen-Ins trinity-Gen-Ins

(18")

*keba-j [. . .] cmid-isa-j sul-isa-j praise-Nom holy-Gen-Nom soul-Gen-Nom

3.2.2. Besides genitive NPs, postpositional phrases also take multiple case marking: (20) a. hrkua vinme ma-s mocape-ta-gan-man (L 11,1) said somebody (Erg) he-Dat disciple-Pl(Gen)-from-Erg 'one of his disciples said unto him' b. rajta carcqmdes ert-i [sc. aso-j] aso-ta sen-ta-gan-i (Mt 5,29) in=order=that it=may=perish one-Norn [sc. limb] limb-Pl(Gen) yourPl(Gen)-from-Nom 'that one of thy members should perish' c. ert-man lesu-js-tana-man (Mt 26,51E) one-Erg Jesus-Gen-at-Erg 'one of them which were with Jesus' d. viet-n-i-me ma-t-gan-n-i itqodes (J 11,37) some-Pl-Nom-Particle he-Pl(Gen)-from-Nom said 'some of them said'

On the basis of examples like (20c), Finck (1910: 141, 144) introduced the term "Gruppenflexion" (group-inflection). However, while the modern Kartvelian languages sometimes do mark one member of a syntactic group for the relation borne by the whole group, in accordance with this notion of group-inflection (6.2), Old Georgian in actual fact does not, and in this respect it resembles most Indo-European languages (Mel'cuk 1986: 49). Constructions as in (21), where the final ergative suffix marks the whole NP, are ungrammatical in Old Georgian. (21)

*ert cmida kac-man one holy man-Erg

162

The Caucasus

There is one peculiarity of Old Georgian postpositional phrases that distinguishes them from genitives with Suffixaufnahme: their case marker almost always lacks a counterpart on an overt head noun. The indefinite pronoun vinme in (20a) exhibits no case marker that could have been copied onto mocapeta-gan-man (cf. Boeder 1987a: 42, Plank 1990: 1040f.).8 This pronoun vin-me may be the determiner of a non-overt head noun (5.5), as is erti in (20b). Accordingly, (20a) would be the result of deleting mocape-man disciple-Erg: (20')

a. [sc. mocape-man] vinme mocape-ta-gan-man [disciple-Erg] some (Erg) disciple-Pl(Gen)-from-Erg

Similarly, ert-man in (20c) is not a head noun: the head noun is abstract ('person'). Examples like (22) are extremely rare, and an expansion of (20c) as in (20'c) probably deviates from the norm. (22)

(20')

Mose-s mier-ta ma-t cign-ta (Hieronymus apud Sanie 1976: 51 §67) Moses-Gen by-Pl(Gen) Art-Pl(Gen) book-Pl(Gen) 'of the books by Moses' c. ?ert-man lesu-js-tana-man kac-man one-Erg Jesus-Gen-at-Erg man-Erg

But as we saw above (2.2), agreement in case is not the copying of overt case markers but rather derives from the case marking of the NP as a whole. Therefore the head status of vin-me in (20a), or even of viet-n-i-me in (20d)— which does have a case marker—need not concern us here. Thus, although we may have to posit an underlying non-overt head noun in examples like (20a), postpositional phrases behave like genitive NPs with Suffixaufnahme in exhibiting ordinary agreement with their head nouns. But are (20a-d) really instances of Suffixaufnahme? Postpositional phrases do occur without case-number markers as verbal or participial adjuncts, as in (23). (23) esre ceril ars cinacarmetquel-isa mier (Mt 2,5) so written is prophet-Gen by 'thus it is written by the prophet'

As NP modifiers, they must undergo ordinary agreement, however. In this they concur with adjectives but contrast with genitive NPs, where casenumber marking may be present or absent inside NPs. Although in principle every kind of ordinary agreement could of course be called "Suffixaufnahme," there appear to be no rules jointly referring to genitive NPs and postpositional phrases to the exclusion, for instance, of adjective phrases. Nothing is gained, then, by extending the notion of Suffixaufnahme to postpositional phrases. And although genitive NPs, postpositional phrases, and adjective phrases partially overlap in their distribution, I prefer to reserve the term "Suffix-

Suffixaufnahme

in Kartvelian

163

aufnahme" for the prototypical case with genitive NPs, where the contrast between presence and absence of Suffixaufnahme corresponds to a difference in structural relation, as will be argued below. Postpositional phrases will thus be said to have multiple case-number marking rather than Suffixaufnahme in the narrow sense.

3.3. Suffixaufnahme and the Position of the Genitive Suffixaufnahme normally occurs with postposed NPs as in (16) and (17), but even in the earlier texts there are rare examples of agreement with preposed NPs: (24)

a. gunebavs lesu-js-i xilva-j (J 12,21E; C has Iesu-s xilva-j) we=want Jesus-Gen-Nom seeing-Nom 'we would like to see Jesus' b. nic-isa ketil-isa-j micema-j (Mt 7,11C; Xanm has ketil-isa) gift-Gen good-Gen-Nom giving-Nom 'to give good gifts'

In another passage, two manuscripts of around 1000 CE (Tbeti, 995; Mestia, 1033) have Suffixaufnahme on preceding genitives (25a) from which it is missing in the earlier manuscripts (25b). (25)

a. sameupo-jsa-man ma-n kac-man (J 4,49DE) kingdom-Gen-Erg Art-Erg man-Erg b. sameupo-jsa kac-man kingdom-Gen man-Erg 'the nobleman'

Similarly, one manuscript (C) has Suffixaufnahme where it is lacking in the version of the oldest manuscripts.9 (26)

a. anu ara ese ars xuro-jsa-j igi 3e-j (Mt 13,55DE; Xanm) or not this is carpenter-Gen-Nom Art=Nom son-Norn b. anu ara ese ars-a xuro-jsa m-is 3e-] or not this=Nom is-Qu carpenter-Gen Art-Gen son-Norn '(Is not this) the carpenter's son?'

In fact, in the earliest dated Georgian manuscript of 864 (many earlier manuscripts are undated), the Sinuri Mravaltavi, examples with Suffixaufnahme by preposed genitives abound; Ivane Imnaisvili (1957: 584) actually contends that this is an archaism and that preposed genitives without Suffixaufnahme are a secondary development.10 In view of their increasing frequency in later texts, however, it is more probable that preposed genitives with Suffixaufnahme became popular between the seventh and the ninth century, and that this popu-

164

The Caucasus

larity was somehow connected with rhetorical practices. At any rate, what matters here is that the construction was possible in the earliest texts and that the position before or after the head noun cannot be the factor that triggers Suffixaufnahme. A change in linear order can, but need not, be connected with Suffixaufnahme: in examples like (24/25) genitives with Suffixaufnahme occur in the same prenominal position as genitives without Suffixaufnahme, to be discussed in the next section.

3.4. Constituent Structure A third central question concerns constituency in expressions with genitive NPs. After some considerations about the basic structure of genitive NPs (3.4.1), three kinds of evidence will be examined to determine constituent structures: clitic and non-clitic movement (3.4.2), discontinuity (3.4.3), and punctuation (3.4.4). 3.4.1. In their basic structure, genitive NPs are assumed to be simple in the sense of lacking Suffixaufnahme and to precede their head nouns, as in (27). (27) Davit-is galob-isa muql-ta ama-t cartkuma-j (GM VII 23-24) David-Gen singing-Gen verse-Pl(Gen) Art-Pl(Gen) recitation-Norn '[He ordered] the recitation of the verses of the song of David'

Structures like (27) are admittedly less frequent than their reversed counterparts with Suffixaufnahme as in (16) and (17); in the Khanmeti texts the ratio is 1:2 (Sar vela e 1984:527). Nonetheless, modifier-head has become the obligatory order in Modern Georgian, and, as observed above, this also harmonizes with postposed adpositions and with the final position of verbs, which are arguably the unmarked orders in Old Georgian. Furthermore, genitives without Suffixaufnahme appear never to be separated from their heads by any modifiers: (28)

a. venaq-is mokmed-sa ma-s (L 13,7) vineyard-Gen laborer-Dat Art-Dat 'the dresser of his vineyard' b. cmid-isa m-is 3ma-n-i igi (GM XX 33) holy-Gen Art-Gen brother-Pl-Nom Art=Nom 'the brothers of the saint' c. xut-ta ma-t krtil-isa pur-ta-gan (J 6,13C) nve-Pl(Gen) Art-Pl(Gen) barley-Gen bread-Pl(Gen)-from 'of the five barley loaves'

This seems to indicate that genitive NPs and their heads form one single constituent, a "small noun phrase" (NP'), inside the larger NP: (28') c. [xut-ta ma-t [krtil-isa pur-ta] NP .] NP

Suffixaufnahme

in Kartvelian

165

Consider now the location of the article in (28a/b). In (28a) ma-s should move to the right of venaq-is if this noun were its co-constituent, and similarly igi in (28b). Clitic movement can be saved by assuming a higher node dominating the article on the one hand and the rest on the other:

(Here and subsequently, elements or constituents moved are underlined in structural representations.) Since the clitic cannot "enter" its co-constituent NP' (by "lowering"), it moves behind it. The same applies to (28b). In (28c), however, ma-t can move to the second position after xut-ta: (28')

c. [ma-t xut-ta [ [krtil-isa]NP2 pur-ta]NP,]NP1

Comparing Suffixaufnahme and its absence with agreement and nonagreement in the appositional structures mentioned above (2.4), we see that (15a) corresponds to preposed genitive NPs without Suffixaufnahme as in (27); (15b) to preposed genitives with Suffixaufnahme as in (24), (25a), and (26a); and (15c) to postposed Suffixaufnahme as in (16) and (17). Arguably, appositive nouns behave like genitive NPs in occurring either as immediate sister constituents of the head noun or as higher-level co-constituents on a par with determiners and adjective modifiers; only in the latter case do they exhibit ordinary agreement. But the syntax of apposition in Georgian is not understood well enough to pursue such considerations. As a result, we get two levels of constituent structure, with a lower-node NP' dominating the genitive NP and its head noun with which it agrees in [case] but not in other features, and with a higher-node NP dominating all kinds of determiners and adjectival or postpositional modifiers undergoing ordinary agreement. Raising a constituent X to the higher-node NP implies that ordinary agreement must apply to X. On this analysis, Suffixaufnahme is ordinary agreement, not an additional kind of agreement. It must remain an open question here whether all genitive NPs are generated as sister constituents of the head noun, as will subsequently be assumed for the sake of simplicity, or whether some of them are base-generated at higher levels of the NP as sister constituents of adjective phrases and therefore exhibit Suffixaufnahme.

166

The Caucasus

Recursiveness demonstrates that at least some instances of Suffixaufnahme must be the result of raising (4.2). 3.4.2. On the present analysis, Suffixaufnahme presupposes hierarchical restructuring. A change in linear order is not necessarily involved. As to Suffixaufnahme on prenominal genitives originating from the raising of such genitives, note that although modifiers do not intervene between a genitive NP and its head noun, there are examples where the article does so intervene: in (25a) and (26a) the articles man and igi occur to the left of their head nouns. What, then, is the difference between (25a/26a) and (28a/b)? The preposed genitive NP has Suffixaufnahme, and the article has moved to its right. Clitic movement presupposes a different constituent status for the genitive NP, which at the same time explains its Suffixaufnahme. With (25'a) as the basic structure of (25a), sameupo-jsa (NP2) is extracted from NP' and raised to the next higher node (NP1). (25')

b. [man [ [sameupo-jsa]NP2 kac-man]NP.]NP1

This in turn allows the article man to move after NP2, which has by now become its leftmost co-constituent: (25")

b. [man [sameupo-jsa]NP2 kac-man]NF]NP1

At the same time, clitic movement proves NP2 to be still a constituent of NP1: in this sense, NP2 is not dissociated from the rest of NP1. Similarly, (24), (25a), and (26a) should be interpreted as the result of leftward raising, with subsequent clitic movement of man and igi where such elements are present. In addition, NP2, being now a co-constituent of NP' under NP, meets the conditions of ordinary agreement and hence has Suffixaufnahme. However, "complete concord" cannot occur because agreement does not extend further "downward" beyond the level of immediate constituents of the NP. If clitic movement of the article is a valid assumption, the following example poses a problem because, by analogy with (26a), one would expect Suffixaufnahme on such prenominal genitives (mter-isa-j enemy-Gen-Nom). (29) mter-isa igi saqopel-i (Cartq lerus 32,8 apud Imnaisvili 1955: 262) enemy-Gen Art=Nom residence-Norn 'the residence of the enemy'

Such examples indeed seem to be rare. Probably (29) results from phonetic simplification, with the nominative exponent being, as it were, swallowed up by following igi (mter-isa-j igi > mter-isa igi). The non-occurrence and perhaps ungrammaticality of analogous examples with forms of the article that do not begin with i (29') is in line with this interpretation.

Suffixaufnahme (29')

in Kartvelian

167

?

mter-isa ma-s saqopel-sa enemy-Gen Art-Dat residence-Dat

It is doubtful, at any rate, that the absence of Suffixaufnahme in (29) is a morphological phenomenon of the same kind as its absence in other examples of postposed Suffixaufnahme (see 5.2 below). Besides clitic movement, there is additional evidence for raising with preposed Suffixaufnahme.11 The structure underlying (30) must be (30'): (30)

Pavle-js-i tav-isa mokueta-j (PO 105,12) Paul-Gen-Nom head-Gen cutting-Nom 'the decapitation of the head of Paul'

(30')

[ [[Pavle-js] tav-isa] mokueta-j]

As tav-isa mokueta-j is not a constituent, it cannot have been dissociated from Pavle-js by rightward movement. Thus, if we want to account for the difference between (30) and basic (30') in terms of constituent structure, Pavle-js must be assumed to have been raised to the next higher node and to have become a co-constituent of mokueta-j. Taking prenominal genitives as basic, Suffixaufnahme in postposed genitive NPs as in (16) and (17) involves rightward movement in addition to raising. In order to understand this kind of movement, let us first consider some evidence for rightward movement in postpositional phrases. Compare unmoved (31) with (32) and (33), resembling the products of clitic movement. (31)

ma-s dye-sa sina (Mt 7,22) that-Dat day-Dat in 'on that day'

(32)

ert-sa sina dye-sa (L 8,22) one-Dat in day-Dat 'on one day'

(33)

qovel-ta gan cmida-ta adgil-ta (GM XVI,4) all-Pl(Gen) from holy-Pl(Gen) place-Pl(Gen) 'from all the holy places'

If postpositions were clitics, (32) and (33) would not require rightward movement at all. But there are three problems with this interpretation. First, postpositions always occur after the article. This suggests different levels of constituent structure, with the postposition staying in its final position and governing an NP which in turn contains the article—cf. (34): (34)

qovel-ta ma-t zeda didebuleba-ta (L 13,17) all-Pl(Dat) Art-Pl(Dat) on glory-Pl(Dat) '[All the people rejoiced] for all the glorious things'

168

The Caucasus

Second, postpositions often occur in final position even in complex phrases, as is seen in (35), where sina could have occurred after the article: (35)

a. qovel-ta ma-t cmida-ta udabno-ta sina (GM X 42) all-Pl(Dat) Art-Pl(Dat) holy-Pl(Dat) wilderness-Pl(Dat) in 'in all the holy wildernesses' b. or-ta ma-t mis-ta senebul-ta monaster sina (GM XXXIII 1) two-Pl(Dat) Art-Pl(Dat) his-Pl(Dat) built-Pl(Dat) monastery-Pl(Dat) in 'in the two monasteries built by him'

Third, the NP governed by the postposition can easily be pronominalized, which is a usual test of constituency; ma-t sina that-Pl(Dat) in 'in them' could refer to (35b). The simplest explanation of these facts is that postpositions are indeed phrase-final in basic structure and that they cannot intrude into their coconstituent NPs. If the postpositions in (32)-(34) have not been moved by clitic movement, the occurrence of material on their right side must be the result of rightward movement:12 (32')

ert-sa dye-sa] p sina

Although (32') illustrates the extraction of constituents from phrases and the rightward movement of these constituents, it does not show the landing-site of the moved constituents inside or outside the postpositional phrase. Evidence for the resulting structure comes from clitic movement. Postposed genitive NPs may themselves contain an article, and its behavior can be used as a diagnostic test of constituent structure. Compare the following examples, assuming that the relevant forms are indeed articles:13 (36)

a. aysdarcues sartul-i saxl-isa-j m-is (Mk 2,4) they=uncovered roof-Norn house-Gen-Nom Art-Gen 'They uncovered the roof of the house' b. upal-i igi venaq-isa-j m-is (Mt 21,40DE) lord-Norn Art=Nom vineyard-Gen-Nom Art-Gen c. upal-i igi m-is saqursen-isa-j (Mt 21,40C) lord-Norn Art=Nom Art-Gen vineyard-Gen-Nom 'the lord of the vineyard' d. kar-sa ma-s m-is saplav-isa-sa (Mt 27,60C; DE omit ma-s) door-Dat Art-Dat Art-Gen tomb-Gen-Dat 'to the door of the sepulchre' e. mama-j m-is qrm-isa-j da deda-j (Mk 5,40DE) father-Norn Art-Gen youth-Gen-Nom and mother-Norn 'the father and the mother of the damsel' f. godol-sa zeda m-is tasr-isa-sa (Mt 4,5C; DE omit m-is) pinnacle-Dat on Art-Gen temple-Gen '[Then the devil taketh him up into the holy city, and setteth him] on a ninnacle of tho temnlo'

Suffixaufnahme

in Kartvelian

169

In (36a/b) the genitive article moves to second position, which is the final one in this example, whereas it stays in its initial position in (36c-f), as depicted in (36'c).

Such behavior of m-is shows, first, that clitics move only within their phrase, effecting a change of precedence but not of dominance. Contrary to igi, m-is does not move after upal-i but stays in its initial position inside NP2, and it does not move after saqur sen-isa-j because it has NP' as its host. Second, the variable Y in rule (10) refers to a possible host that may already host a clitic. Notice that clitic movement simply implies phonological attachment to a preceding constituent (represented by a dotted line in (36'c)) rather than syntactic attachment to a host (Klavans 1985). Third, and most important for present purposes, the data in (36) show that some NPs with Suffixaufnahme belong to the same phrase as their head noun, but others do not. Clitic movement of the genitive article is a diagnostic criterion: in (36c-f) the article m-is stays in its initial position because it has a non-clitic host (NP' in (36'c)), suggesting that the genitive NP belongs to the same phrase (NP1 in (36'c)) as its host (NP'). In (36a/b), by contrast, the genitive article m-is moves after its head noun, suggesting that this counts as a first constituent in its phrase.14 Thus, NPs of this type must have been dissociated from their head nouns; it is only after extraction from their source NPs that they meet the condition for clitic movement, which requires the absence of a non-clitic constituent before the clitic. On this analysis (36a/b) consist of two separate NPs each. The same line of argument for raising applies to preposed genitives with Suffixaufnahme. In examples such as (37), starting from an underlying structure with the genitive NP contiguous with its head noun, we may assume that it is extracted from its NP and raised to the higher NP node, as indicated in (37'). (37)

(37')

mqec-ta-j ma-t U3yeb-i igi mswnvareba-j (Parx 1206b) beast-Pl-Nom Art-Pi insatiable-Norn Art=Nom raging-Nom 'the insatiable raging of the beasts' [ iai u yeb-i [[m ec-ta ma-t]

170

The Caucasus

But this cannot be the surface structure of (37), because igi occurs after u3yeb-i rather than after mqec-ta ma-t (NP2). From the point of view of clitic movement, usyeb-i must be the first constituent of its NP. Again, this means that mqec-ta-j ma-t and u3yeb-i igi m3wnvareba-j are two separate NPs in (37). Instances of hyperbaton (or discontinuity) justify such assumptions of separate NPs. 3.4.3. In all examples considered so far, genitive NPs were adjacent to the constituents of their head NP, and there was evidence that in some cases adnominal phrase and head noun belong to one single NP. But there are examples of discontinuity that imply that adnominal phrase and head NP form separate constituents. This extraposition often results in structures where the genitive NP is separated from its head noun by intervening material that cannot belong to the source NP. Thus the verb 'we are' in (38a), with Suffixaufnahme, separates the prenominal genitive from its head noun, extraposed to the right; (38b), without Sufnxaufnahme, is the version of another manuscript. (38)

a. cuen Mose-s-n-i vart mocape-n-i (J 9,28DE) we Moses-Gen-Pl-Nom we=are disciple-Pl-Nom b. cuen Mose-s mocape-n-i vart (J 9,28C) we Moses-Gen disciple-Pl-Nom we=are 'We are Moses' disciples'

Using the terminology of ancient rhetorics, this phenomenon will be referred to as "hyperbaton." Hyperbaton makes Suffixaufnahme obligatory in (38a). (For an additional example see (52).) On general grounds, rightward extraposition of head nouns as in (38a) might seem a problematic move; however, analogous separations of head nouns from adjectival modifiers (39) and, less sparsely and earlier, from quantifiers (40) are attested. (39)

tkbil-n-i mesmodes me sitqua-n-i sen-gan (Kim I 54,22 apud Kal 118) sweet-Pl-Nom may=1=hear I word-Pl-Nom thou-from 'May I hear sweet words from thee'

(40) ese picar-i mraval-ta aymoikitxes Huria-ta (J 19,20DE; C: Huria-ta mraval-ta) this=Nom title-Norn many-Pl(Erg) read Jew-Pl(Erg) 'This title then read many of the Jews'

Examples such as the following show that it is indeed the head noun, rather than a phrase, that is extraposed:15 (41)

a. madl-isa-j igi gamoacina saun3e-j (SinMr 24,9) grace-Gen-Nom Art = Nom showed treasure-Nom 'He showed the treasure of (his) grace' b. rajta [. . .] cxoreb-isa-j igi mohmadlos natel-i (SinMr 37,18-20)

Suffixaufnahme

in Kartvelian

171

so=that [. . .] life-Gen-Nom Art=Nom he=may=bestow=on=them lightNorn 'so that he may bestow them the light of life' In (41 a) madl-isa-j igi cannot be a constituent and therefore could not have been moved to the left after being extracted from the NP whose head is saun3e. Postposed genitive NPs can also be separated from their heads: (42)

a. ukuetu 36-j xar ymrt-isa-j (L 4,9C) if son-Norn you=are god-Gen-Nom 'if thou be the Son of God' b. qma-j isma mam-isa-j maylit (Cil-etr 27 p. 22) voice-Norn was=heard father-Gen-Nom from=high 'The voice of his father was to be heard from above' c. mdinare-n-i mucl-isa mis-isa-gan gamodiodian cqal-ta cxovel-ta-n-i (J 7,38C) river-Pl-Nom belly-Gen his-Gen-from they=come=out water-Pl(Gen) living-Pl(Gen)-Pl-Nom 'Out of his belly shall flow rivers of living water' d. dapu3neba-sa aucqebda eklesia-jsa-sa (SinMr 195,24) foundation-Dat he=laid=it church-Gen-Dat 'He laid the foundation of the church' e. mcignobroba-j-ca iscava mraval-ta ena-ta-j (GM II 42) literacy-Nom-too he=learnt=it many-Pl(Gen) language-Pl(Gen)-Nom 'He learned to read and write in many languages' f. mklav-ta zeda miikua Swmeon-is-ta da Ana-js-ta (SinMr 49,2-3) arm-Pl(Dat) on was=taken=up Simeon-Gen-Pl(Dat) and Anna-GenPl(Dat) 'He was taken up on the arms of Simeon and of Anna'

The last example shows that adnominal phrases can be extracted from NPs themselves governed by a postposition. Extraposition with hyperbaton is not confined to genitive NPs. It also occurs with adjective modifiers (43) and with participial phrases (44). (43)

qelmcipe-sa vi-s-me did-sa svil-i moukuda pirmso-j (Kim I 51,21 apud Kiz 250) king-Dat some-Dat-Particle great-Dat child-Nom died=to=him first=born=Nom 'There was a king whose first-born son died'

(44) kac-sa ma-s hxedvides ma-t tana mdgomare-sa (Act 4,14 apud Kiz 250) man-Dat Art-Dat they=saw=him that-Pl(Dat) near standing-Dat 'beholding the man [which was healed] standing with him' The implications of hyperbaton for the basic and derived constituent structure of Suffixaufnahme constructions are as follows. Assuming that Old Geor-

172

The Caucasus

gian had no verb phrase in the sense that English, for example, has one, (42a) has an underlying structure like (42'a); NP2 is first moved to the right and raised to the node dominating NP1 (42"a); then NP2 is extracted from its source, NP1, and attached to the node S, which dominates both its source phrase and the intervening material (42"'a). Finally, NP2 is moved across the verb resulting in (42a). (42')

a. [ukuetu [[ymrt- isa] 3e-j] xarv]

(42")

a. [ukuetu [[[ymrt-isa] NP2 3e-j]NP',]xarv]s

(42'") a. [ukuetu [[3e-j]NP, [ymrt-isa-j]NP2]NP1 xar v ]

Notice that hyperbaton cannot operate in one move: only constituents immediately dominated by a node that is itself dominated by S can be raised to S. Interestingly, the intrusive constituent in (42)-(44) is always a verb, removed from its final position by hyperbaton. Imnaisvili's (1975: 101f.) data include only some more complex examples of verb intrusion after adjectival modifiers: (45)

ubrcqinvales-n-i cmida-man kalcul-man aymotkuna sitqua-n-i (SinMr 13,29) most=brilliant-Pl-Nom holy-Erg virgin-Erg pronounced word-Pl-Nom 'The Holy Virgin pronounced most brilliant words'

Intrusive nouns appear unattested and may be ruled out by general restrictions of cross-over phenomena. Hyperbaton is probably favored, among other factors, by the rhythmic tendencies of natural speech. An intrusive verb is normally shorter than the last constituent with Suffixaufnahme and was probably unstressed. The cursus of Medieval Latin rhetorics, the successor of the ancient clausula, recommended final words with more than two syllables (Lausberg 1963: §462), and Byzantine rhetorics may have found its way into Georgian prose composition, strengthening natural inclinations. 3.4.4. The third type of relevant data comes from punctuated Old Georgian texts. Some manuscripts use punctuation to mark units below the sentence level, and such punctuation marks are likely to have indicated breaks for potential, though not obligatory, pausing. Most old manuscripts were meant to be read aloud, and it was important for the reader to divide the text into appropriate segments. Although pauses do not necessarily coincide with syntactic boundaries, there is a strong correlation between syntax and tone groups in planned speech. Here are some examples from Marr's edition of the Physiologus: (46)

a. da moixuna ornive prteni . sul-ita sulneleb-isa-jta . simqn-ita da nic-ita . cata-jta (X)

Suffixaufnahme

in Kartvelian

173

and he=brought both wings . breath-Ins fragrance-Gen-Ins . valiance-Ins and gift-Ins . heaven-Pl(Gen)-Ins 'and he [sc. the phoenix] brought with him both wings with the smell of the fragrance, with valiance and the gift of the heavens' b. moiqvanis sesakrebel-sa . saravanded-isa-sa . (XXV) she=leads=it meeting=place-Dat . crowned-Gen-Dat . 'She [sc. the virgin] leads it [sc. the unicorn] to the audience-chamber of the king' c. saxl-sa . Davit-is mon-isa tws-isa-sa (XXV) house-Dat . David-Gen servant-Gen his(Reflexive)-Gen-Dat '[and has raised up a horn of salvation for us] in the house of his servant David' (=L 1,69) d. seipqrenit mel-n-i mcbier-n-i . ganmrqunel-n-i . saqur n-isa cem-isa-n-i (XVIII) catch fox-Pl-Nom sly-Pl-Nom . devastator-Pl-Nom . vineyard-Gen my-GenPl-Nom 'Catch the sly foxes, the devastators of my vineyard' e. da dahqaris naclev-i . mprinvel-ta-j ma-t (XVIII) and he=sheds=down intestine-Nom . bird-Pl(Gen)-Nom Art-Pl(Gen) 'and he sheds down the intestines of the birds' This phrasing is all the more remarkable because such pauses are not expected when reading the English translations. Appositions are always separated by punctuation even in the Life of Gregory of Khandzta, a text that otherwise has very little punctuation below the clause level: (47)

masenebel-i Xan3t-isa da Satberd-isa-j . or-ta ma-t didebul-ta monaster-ta-j . (GM II 24-25) builder-Nom Khandzta-Gen and Shatberd-Gen-Nom . two-Pl(Gen) ArtPl(Gen) magnificent-Pl(Gen) monastery-Pl(Gen)-Nom. 'the builder of Khandzta and Shatberd, the two magnificent monasteries'

Arguably, such punctuation bears on constituent structure, with pauses thus marked occurring after phrases immediately dominated by S (cf. Boeder 1987b). Although there is no material intervening between head noun and genitive NP, we may again stipulate structures with raising and extraposition. Taking (46e) as an example, raising would apply to an underlying structure like (46'e); then NP2 would be extraposed (46"e); and finally the article ma-t would be moved into the second position of NP2 (46"'e), resulting in (46e).

174

The Caucasus

(46')

e. da dahqaris [[[mg-t mprinvel-ta] NP2 naclev-i]NP, ]NP1]S

(46")

e. [da dahqaris [[naclev-i]NP. [ma-t mprinvel-ta-j] NP2 ] Np1 ]s

(46'") e. [da dahqaris [[naclev-i]NP.]NP1 . [ma-t mprinvel-ta-j]NP2]s |

The evidence from punctuation converges with that from constituent structure: the movement of the article in (46" 'e) points to the separation of the extraposed NP2. Thus punctuation and clitic movement require the same constituent structure of the sentence as that suggested by hyperbaton. But most importantly, punctuation as in (46) occurs between head noun and postposed modifier, whereas ordinary preposed modifiers without Suffixaufnahme are never separated from their head in this way. Whatever the structural basis of punctuation may be, it must signal some kind of separation that is not found between simple preposed genitives or other modifiers and their heads. The separation effected by extraction, as diagnosed by clitic movement, hyperbaton, and punctuation, has a particular function and is, diachronically, at the root of Suffixaufnahme, as will be argued in Section 7. 4. Recursiveness As pointed out by Plank (1990), the recursiveness of Suffixaufnahme reflects interestingly on the interplay between morphology and syntax. Two kinds of hierarchical restructuring, along the lines of the previous section, appear to be required to account for recursive genitives in Old Georgian. In the first variant, which is by far the most frequent, the highest genitive NP moves to the right of its head noun and takes up its case marking, with the same procedure then applying to the next lower NP. Thus, assuming (48'), with recursive left-branching, as the basic structure of (48), the complex NP2 moves to the right of its head noun; then its constituent, NP3, moves to the right of its head (48"); then the article igi shifts after saidumlo-j as an enclitic while m-is stays in its phrase-initial position (48'"). NP2 and NP3 take up the case-number marking of their respective head nouns, nominative singular stemming from saidumlo-j and genitive singular from sasupevel-isa (48""). (48)

tkuenda micemul ars cnob-ad saidumlo-j igi sasupevel-isa m-is ymrt-isa-jsa-j (Mk4,11C) to=you given is knowing-Adv mystery-Norn Art=Nom kingdom-Gen ArtGen god-Gen-Gen-Nom 'Unto you it is given to know the mystery of the kingdom of God'

(48f)

[igi [[m-is [[ymrt-isa]NP sasupevel-isa] NP .] NP2 saidumlo-j]Np.]NP1

(48 (48

Suffixaufnahme

in Kartvelian

175

Thus comes about multiple Suffixaufnahme on one genitive NP of two present, with the order of successive layers of case markers mirroring syntactic constituency. The evidence for the other type of recursive extraposition, though somewhat scanty,16 appears to be unambiguous; it should be noted, however, that only nominative Suffixaufnahme is attested so far. In (49), derived from basic structures like (49') as sketched in (49")(49""), and analogously in (50), we find extraposition from extraposed genitive NPs, with separate, unrepeated Suffixaufnahme on the two genitives. (49)

da sul-i sulneleb-isa-j . sakumevel-ta-j . gamoqdis pir-isa-gan mis-isa (Phys XX) and breath-Nom fragrance-Gen-Nom . incense-Pl(Gen)-Nom . it=comes= out mouth-Gen-from his-Gen 'and a whiff of fragrance of incenses comes out of its [sc. the dragon-fish's] mouth'

(49 (49 (49

176 (50)

The Caucasus xilva-j cmid-isa m-is ber-isa-j da adgil-isa-j m-is ketil-isa saqopl-isa ymrtivmocemul-isa-j (GM VI 68) seeing-Nom holy-Gen Art-Gen monk-Gen-Nom and place-Gen-Nom Art-Gen good-Gen dwelling-Gen god-given-Gen-Nom 'the seeing of the holy monk and of the place of good dwelling given by God'

The entire NP2 is first moved to the right of its head noun (49'); then NP3, originally included within NP2, is moved to the right of its head (49"); sharing a head, sul-i, NP2 and NP3 take up the same case-number marking, nominative singular (49'"); finally, to account for punctuation and pausing, NP3 is raised so as to be immediately dominated by S (49""). Crucially, in the course of this derivation, NP3 changes its agreement controller: after sulneleb-isa in NP2, sul-i finally becomes its agreement controller in NP1; Suffixaufnahme again applies after raising. A structure like (49"") raises further problems. In structures like (48"") and (49'") Suffixaufnahme could apply on the basis of the configuration of their final constituent structure: NPs agree with their sister constituents dominated by the same node. But if (49"") is correct, the nominative in NP3 cannot be explained on the basis of a sister constituent NP1; the domain of ordinary agreement is a phrase, and there is no indication of agreement between NPs dominated by the same node S. An alternative interpretation is that Suffixaufnahme applies after each cycle, with the case-number marking of earlier steps of the derivation being retained. Thus, NPS in (49"") retains the nominative marker it got under NP1 even after being raised to S. Support for a derivational interpretation of (49"") comes from the following examples, which are parallel to both (48) and (49/50): like (48) they have multiple Suffixaufnahme on the genitive NP in rightmost position, and like (49/50) they have simple Suffixaufnahme also on the first genitive. (51)

a. qovel-i igi sisxl-i saxl-isa-j m-is Saul-is-isa-j (2 Rg 16,8 apud Kiziria 1946: 330) all-Norn Art=Nom blood-Norn house-Gen-Nom Art-Gen Saul-Gen-GenNom 'all the blood of the house of Saul' b. sasxdomel-eb-i igi msqidel-ta-j ma-t tred-isa-ta-j (Mk 11,15DE) seat-Pl-Nom Art=Nom seller-Pl(Gen)-Nom Art-Pl(Gen) dove-GenPl(Gen)-Nom '[and overthrew] the seats of them that sold doves'

Consider (5la): starting from an underlying structure (51'a), NP2 is moved after NP' and, being dominated by NP1, undergoes agreement with the nominative sisxl-i (51 "a). Next NP3 is raised to the higher node NP2 and, being a co-constituent of the genitive saxl-isa, it is marked for the genitive: saxl-isa

Suffixaufnahme

in Kartvelian

177

Saul-is-isa-j (51 '"a). Finally NP3 Saul-is-isa is again raised to the next higher node and, being now sister constituents of sisxl-i, both saxl-isa and Saul-is-isa get the nominative (51""a). The articles undergo clitic movement. The crucial fact here is that Saul-is-isa retains its Suffixaufnahme triggered by saxl-isa on the earlier cycle, although saxl-isa is no longer the head noun of Saul-is under the domination of NP1. (51')

a. [igi qovel-i [[m-is [[SauI-is]NP3 saxl-isa]NP.]NP2 sisxl-i]NP,]Npl

(51")

a. [igi qovel-i [sisxl-i]NF [m-is [[Saul-is]NP3 saxl-isa]NP,-j]NP2]NP1

(51'")

a. [igi qovel-i [sisxl-i]NP, [m-is [saxl-isa]NP, [[Saul-is]NP, -isa]NP3 -j]NP2]NP1

(51"")

a. [igi qovel-i [sisxl-i]Np, [m-is [saxl-isa]NP, -j ]NP2 [[Saul-is]Np. -isa -j]NP3]NP1

Similarly, (51b) is to be derived from (51'b) by multiple extraposition: (51') b. [[[tred-isa]NP3 ma-t msqidel-ta]NP2 igi sasxdomel-eb-i]Npl Vogt (1947: 130) considered msqidel-ta-j in (51b) a scribal mistake; but while one manuscript (Urbnisi) indeed omits the nominative suffix, the revised Athonite manuscripts (OtxBol) have it as well. The difference between the two kinds of recursive extraposition, then, is not one of principle; the second type simply adds a further application of raising. Under this interpretation, (49) should be sul-i sulneleb-isa-j sakumevel-taisa-j. But -ta-j is a simplification of -ta-isa-j owing to a morpheme sequence constraint (see 5.2). Therefore, the cyclic nature of Suffixaufnahme is unrecognizable in (49/50). Genitive NPs preceding their heads, as in basic structure, do not seem to show multiple Suffixaufnahme. (52), where the pre-head genitive takes up the plural and nominative of the head, dye-n-i, is only an apparent example. (52)

Elisabed-is-n-i ayivsnes dye-n-i sob-isa-n-i (L 1,57C) Elizabeth-Gen-Pl-Nom were=fulfilled day-Pl-Nom bearing-Gen-Pl-Nom 'Elizabeth's full time came that she should be delivered'

The derivation of (52) resembles that of (49/50), as seen in (52'a/b), if (52'a) is its correct underlying structure. Its only peculiarity is rightward movement of the phrase dye-n-i sob-isa-n-i in (52'c), thus ensuring precedence for the subjective genitive Elisabed-is (see 7.1). (52')

a. [[[Elisabed-is] sob-isa] dye-n-i] ayivsnes b. [[dye-n-i] [[Elisabed-is] sob-isa-n-i]] ayivsnes

c.

178

The Caucasus

These considerations suggest some formal properties of Suffixaufnahme. First, restructuring of recursive attribution works upside down, proceeding from more to less complex NP constituents. Raising and postposing the least complex constituents first, while possible, would yield results that are unattested. Second, while the restructuring of dominance relations by itself allows raising both to the left and to the right of the current head, recursive genitive raising is strictly unidirectional: only rightward recursive raising occurs. Hence (49) cannot mean 'the whiff of incenses of fragrance', regardless of how plausible or otherwise such meanings are. Third, preposed genitive NPs cannot have recursive Suffixaufnahme at all. A unifying principle seems to underlie these restrictions. This becomes particularly clear if we consider postpositional phrases. In addition to multiple raising in recursive genitive structures, these seem to offer one further possibility: taking a simplified version of (42f) as an example (cf. (42'a)), NP2 can be raised to NP1 (42'", cf. 42"a), then to PP— i.e., the node that dominates the postposition and the NP that it governs (42 "'f) — and finally to S (42" "f, cf. 42" 'a). (42')

f. [[[[[Swmeon-is]Np2 mklav-ta]NP,]NP1 zeda]PP miikuav]s

(42")

f. [[[[[Swmeo«-«]NP2 mklav-ta]NP, ]NP1 zeda]PP miikuav]s

(42'")

f. [[[[mklav-ta]NP, {Swmeon -is - to] MP2] zeda ]PP miikuav]s

(42"")

f. [[[[mklav-ta]NP.]MP1 zeda {Swmeon-is-ta\ NP7 ] pp miikua v ]s

But a structure like (42 " ' ) , where the genitive NP2 intervenes between its head noun and the postposition, does not seem to occur at all (Kala e 1961: 99). 17 If this is correct, we may speculate about a general constraint on raising in Old Georgian postpositional phrases and NPs. In a traditional sense, both postpositions and nouns are heads that govern the head of their co-constituent NPs. In addition to dominance, government also seems to be defined in terms of precedence in Old Georgian: both head nouns and postpositions govern only those co-constituents that occur to their left (as in underlying structure) . Probably the following constraint holds for Old Georgian: raising cannot create new government relations. In most cases both dominance and precedence relations are altered by raising without bringing new constituents into the scope of the government of a head. In (42"'f), however, mklav-ta and Swmeon-is(-ta) would both be governed by zeda. If precedence is altered as in (42""), zeda governs only mklav-ta, as in (42'f). The formal properties of Suffixaufnahme can now be derived from this constraint. It predicts that raising must be upside down, because raising a deeper NP first would bring it into the government scope of a different (higher) head noun. Taking (48) as an example, both ymrt-isa-jsa sasupevel-isa saidumlo-j (God-Gen-Gen kingdom-Gen mystery) and sasupevel-isa ymrt-isa-isa saidumlo-j (kingdomGen God-Gen-Gen mystery) are in all probability ungrammatical:

Suffixaufnahme

in Kartvelian

179

Second, recursive rightward raising may occur because it moves NPs outside the scope of any head noun. Third, leftward raising is possible once, because it does not alter the original government relation. If the constraint on raising is correct, it lends some support to the assumption that phrases are underlyingly head-final (1.3). 5. Adding Morphological and Syntactic Detail 5.1. Suffixaufnahme

as Agreement in [case]

Suffixaufnahme repeats not only case and number but also the exponents of [+case]; it thus reflects the contrast between short and long forms (2.3-4): (53)

a. pirmso-jt-gan kac-ta-jt (Exodus 13,15 apud Imn 597) firstborn-Ins-from man-Pl(Gen)-Ins 'from the firstborn of men' b. mucl-it-gan ded-isa tws-isa-jt (Mt 19,12DE) womb-Ins-from mother-Gen her=own-Gen-Ins 'from the womb of his mother' c. rajta xiqvnet svil- mam-isa tkuen-isa- (Mt 5,45 Xanm) so=that you=may=be child-Abs father-Gen your-Gen-Abs 'that ye may be the children of your father' d. kma- ars mocap-isa- (Mt 10,25 Xanm) sufficient-Abs it=is disciple-Gen-Abs 'It is enough for the disciple [that he be as his master]'

The long form of the instrumental case suffix -ita is required in (16), its short form in (53a). But compare (53b) with its variant (14a): while Suffixaufnahme itself always exhibits the "regular pattern," the genitive forms themselves have the short form (ded-is) in (14a) as required by the short form of the head noun mud-it, but the long form in (53b) (ded-is-a tws-is-d). Similarly, Suffixaufnahme is vacuous in (53c/d): both the absolutive and the genitive have [ — case], i.e. zero. But again, the genitives themselves exhibit the long form: the underlying forms mam-is tkuen-is svil and mocap-is kma contrast with svil mam-is-a tkuen-is-a and kma mocap-is-a respectively. Imnaisvili (1957: 588) cites svil natl-isa child- light-Gen as a later source's remodeling of the older form svil natl-is (J 12,36) child-0 light-Gen with the short form of the genitive. Similarly, genitives of proper names sometimes have the long form of the genitive, although proper names normally refuse the feature [+case] (see 2.4): (54)

kueqana-j Zabilon-isa-j (Mt 4,15D; Lortkipani e 1959: 389) land-Norn Zabulon-Gen-Nom 'the land of Zabulon'

180

The Caucasus

Here kueqana-j Zabilon-is-i (as attested in C) would conform to the norm. There is a tendency, then, to mark postposed modifiers for [+case], even if Suffixaufnahme sticks to strict agreement in [case]. But even Suffixaufnahme may fail to repeat zero as an agreement marker of the absolutive: in suffixaufnahme the short form absolutive may be replaced by its corresponding long form, the nominative. The following implication seems to be valid, however: if the predicative noun is in the nominative instead of the absolutive, there is nominative Suffixaufnahme, but the reverse is not true. This is exemplified in (55) and, with ellipsis of the head, (56). (55)

a. sabrke- ars sul-isa cem-isa-j (GM XII 18-19) snare-Abs is soul-Gen my-Gen-Nom 'He is the snare of my soul' b. msgavs-i iqo cinajscarmetquel-isa Davit-is-i da netar-isa Bagrat- mocikulisa da mydel-mocam-isa . mocap-isa cmid-isa Petre- mocikul-isa-j (GM XXIV 9-11) similar-Nom was prophet-Gen David-Gen-Nom and blessed-Gen BagratAbs disciple-Gen and priest-witness-Gen . pupil-Gen holy-Gen PeterAbs disciple-Gen-Nom 'He was similar to the prophet David and the blessed disciple and hieromartyr Bagrat, the disciple of the holy apostle Peter'

(56)

sen kueqan-isa qelmcipe xar . xolo Kriste-j [sc. qelmcipe aris] zec-isa da kueqan-isa da kuesknel-ta-j : sen natesav-ta ama-t mepe xar . xolo Kriste-j [sc. mepe] qovel-ta dabadebul-ta-j (GM XXVI 18-19) you earth-Gen ruler- you=are . but Christ-Norn [sc. ruler- is] heavenGen and earth-Gen and nether=world-Pl(Gen)-Nom . you kinsmanPl(Gen) this-Pl(Gen) king- you=are . but Christ-Norn [sc. king- ] allPl(Gen) creature-Pl(Gen)-Nom 'You are the ruler of the earth, but Christ [sc. is the ruler] of heaven and earth and the nether world; you are the king of this nation, but Christ [sc. is the king] of all creatures'

In (55a) the head noun is in the absolutive and thus conforms to the rule, but the extraposed (hyperbaton) genitive phrase is in the nominative, committing a double offense. First, the genitive should agree with its head noun in [case]: the expected form is sabrke . . . sul-is cem-is with the short form of the genitive, as explained above. Second, nominative Suffixaufnahme is not justified by the head noun in (55a), which is in the absolutive—i.e., not marked for [+case]. But both long form and Suffixaufnahme presuppose the same feature, namely [+case], which should be absent in the postposed genitive noun because it is absent from its head noun. In (55b) both the head, an adjective, and the extraposed genitives are in the nominative, although the head should be in the absolutive according to the older norm. In general, then, case marking on the head noun implies case marking in postposed modifiers, but not vice versa. There seems to be a general tendency

Suffixaufnahme

in Kartvellan

181

to mark every noun for [ + case] along an implicational scale: first postposed genitives, which get the long instead of the short form (-is > -is-a) (53/56); second, Suffixaufnahme, where the nominative -i is used instead of the absolutive zero (55/56); and last, the head noun itself, where the absolutive is again replaced by the nominative. This variational statement in all probability has its counterpart in the history of Georgian. It is part of a long-term development leading to the loss of the absolutive and of the [±case] contrast in general. The implicational scale also conforms to a further tendency: if case marking is not assigned to all nominals in an NP, then the preference in Kartvelian is to mark the final one, perhaps preparing the ground for the later development toward group-inflection.

5.2. Deviations from Straightforward Copying To begin with, it should be kept in mind that Suffixaufnahme does not necessarily require corresponding marking on head nouns (see 2.2, 3.2.2): NPs such as Simon- lona-js-o (J 21,15 Xanm) Simon-0 Jonas-Gen-Voc 'Simon, son of Jonas!' are perfectly regular, although the vocative on the genitive NP has no counterpart on its head noun Simon (which refuses [+case], see 2.4); the vocative characterizes the NP as a whole, but it cannot be spelt out on the proper name. On the other hand, every case-number suffix of the head noun normally requires corresponding Suffixaufnahme marking. But there are exceptions and complications, too—or rather, simplifications. First, not every plural marker can be taken up. There are three plural markers in Georgian: -n- in the nominative and vocative, -ta- in the oblique cases, and -eb- in all cases, -eb- is rather rare in Old Georgian, but it is beginning to expand in medieval texts and is the normal nominal plural suffix in Modern Georgian. Originally it had collective meaning, which explains its syntactic behavior: as is seen in (51b) above, -eb- does not take part in Suffixaufnahme (cf. Vogt 1947: 130). Second, Suffixaufnahme does not necessarily produce straight morphemic copies of the cases of the head. Owing to the possibility of— in principle unlimited—recursion, multiple case-number marking can become rather complex (see Imnaisvili 1957: 623-631 for examples of triple marking): (57)

a. qovel-man krebul-man 3e-ta Israel-isa-ta-man (Exodus 12,6 apud Imn 629) all-Erg assembly-Erg son-Pl(Gen) Israel-Gen-Pl(Gen)-Erg 'the whole assembly of the sons of Israel' b. moyeb-ita madl-isa m-is Kriste-js-isa-jta (Abo 199,24 apud Imn 630) receiving-Ins grace-Gen Art-Gen Christ-Gen-Gen-Ins 'by the grace of Christ' c. qrma-n-i igi mtavar-ta ma-t sopl-eb-isa-ta-n-i (3(1) Rg 20,19 apud Imn 622) youth-Pl-Nom Art=Nom chief-Pl(Gen) Art-Pl(Gen) country-Pl-GenPl(Gen)-Pl-Nom 'the young [warriors] of the governors of the provinces'

182

The Caucasus

In principle, Old Georgian tolerates such complexities (Imnaisvili 1957: 625, 631); and as all suffixes in examples as complex as (57c) are overtly different, they may not be too difficult to process. Yet two sentences earlier (20,17), the same text offers a simplified version of (57c) that omits the genitive plural suffix -ta, the first to be copied by Suffixaufnahme, attesting to the operation of "morphological sequence constraints" as known from Australia (Dench and Evans 1988: 35-43): (57)

d. qrma-n-i mtavar-ta ma-t sopl-eb-isa-n-i (Imn 632)

Often such simplifications are haplological. Thus, repetitions of the same suffixes as in (58a) are commonly simplified as in (58b), and in the case of the repeated oblique plural marker -ta haplology is in fact obligatory (59).18 (58)

a. 3-isa kac-isa-jsa son-Gen man-Gen-Gen 'the son of man' b. 3-isa kac-isa (Mt 10,23 Xanm)

(59)

a. *kar-ta kalak-ta-ta door-Pl(Obl)city-Pl(Gen)-Pl(Gen) 'the gates of the cities' b. kar-ta kalak-ta

Note that the sequence -ta-ta is phonotactically well-formed: mta-ta mountainPl(Obl). The obligatoriness of morphological haplology in this case may have to do with the syncretic nature of this suffix (on which see below). The surface structure of (60) would lead one to expect Huria-ta, haplologically simplified from Huria-ta-ta, rather than Huria-ta-man: (60)

mitxres mydelt-mogyuar-ta da xuces-ta ma-t Huria-ta-man (Sakme mocikultaj 186,15 apud Kal 102) said=to=me priest-teacher-Pl(Erg) and elder-Pl(Erg) Art-Pl(Erg) JewPl(Gen)-Erg 'the high priests and elders of the Jews said to me'

The suffix -man that Suffixaufnahme unexpectedly produces here is ergative, as is indeed appropriate for a subject of a transitive verb such as mitxres. The non-appearance of -man also on the head nouns of the subject NP in (60) is traditionally considered to reflect a syncretism, with the contrast among ergative, genitive, and dative being neutralized in favor of plural marking by -ta (see Table 4.1). In accordance with the essentially agglutinative character of Kartvelian noun inflection, -ta might also, and perhaps more plausibly, be analysed as being reduced from -ta-man, -ta-jsa, -ta-sa, rather than as a portmanteau fusing case and number (Dondua 1931, Boeder 1976). There is a remarkable difference between morphological sequence condi-

Suffixaufnahme

in Kartvelian

183

tions in "normal" noun inflection and in Suffixaufnahme: while a one-suffixonly constraint is valid for most of the usual Old Georgian noun paradigm,19 Suffixaufnahme is less restrictive. Consider again (60): the sequence -ta-man in Huria-ta-man, with the "underlying" case marker -man making it to the surface, would not be allowed outside Suffixaufnahme. The derivational history of Huria-ta-man is sketched in (60'), with its plural -ta and ergative -man stemming from the "underlying" forms of the two conjuncts of the head, *mydelt-mo3yuar-ta-man and *xuces-ta-man. (60')

a. *Huria-ta-jsa-ta-man b. *Huria-ta-ta-man c. Huria-ta-man

(by "inner case deletion," Dench and Evans 1988: 39) (by haplology)

Similarly, the relevant NPs in (61) presuppose the underlying morphological forms qel-ta-sa kac-ta-jsa-ta-sa (hand-Pl-Dat man-Pl-Gen-Pl-Dat) and pir-tajsa gan mtavar-ta-jsa-ta-jsa (face-Pl-Gen from chief-Pl-Gen-Pl-Gen).20 (61)

a. 3er-ars 3e-j kac-isa-j micem-ad qel-ta kac-ta-sa (Mt 17,22DE apud Kiz 276) it=fit=is son-Norn man-Gen-Nom giving-Adv hand-Pi man-Pl(Gen)-Dat 'The son of man shall be betrayed into the hands of man' b. ara sesineboda pir-ta gan mtavar-ta-jsa (Kim II 2-3 apud Dondua 1931) not he=was=afraid face-Pl(Gen) from chief-Pi-Gen 'He was not afraid before the rulers'

There is a further possibility that resembles the Indo-European type of syncretism. An adverbial case marker, copied in (62a), may either be suppressed (62b/c) or, albeit rarely, be replaced by a dative (62d/e) or even a genitive (62f) (Imnaisvili 1957: 590-595). (62)

a. mica-dsikvdil-isa-d(Ps21(22),16) earth-Adv death-Gen-Adv 'into the dust of death' b. 3egl-ad maril-isa (Sap 10,7) pillar-Adv salt-Gen 'as a pillar of salt' c. 3e-d ymrt-isa (J 19,7) son-Adv God-Gen '[He made himself] the son of God'21 d. kalak-ad Nakor-is-sa (Genesis 24,10 apud Kiziria 1963: 266) city-Adv Nahor-Gen-Dat 'the city of Nahor' e. moklva-d mam-isa Grigol-is-sa (GM 68,17 apud Deeters 1927: 20) killing-Adv father-Gen Gregory-Gen-Dat 'to kill father Gregory'

184

The Caucasus f. zyude-d-mde urakparak-ta-jsa (Neh 3,8 apud Imn 594) wall-Adv-up=to place-Pl(Gen)-Gen 'up to the wall of the places'

Replacement of the adverbial by a dative or a genitive is another strategy to reduce morphological complexity, because dative and genitive are less complex than the adverbial in terms of feature composition. The same development is attested in adjective agreement from medieval times up to the early nineteenth century (cf. citel-s vasl-ad red-Dat apple-Adv 'as a red apple').22 Thus we have three forms of "syncretism" that account for deviations from simple copying. One constrains sequences of different case-number markers (-ta-man > -ta etc.); it is obligatory in simple noun inflection but optional in sequences produced by Suffixaufnahme. The second form is haplology, which is obligatory for -ta-ta -Pl(Obl)-Pl(Obl) but optional for -isa-jsa -Gen-Gen, probably because the first is more complex than the second in "underlying structure." The third form of syncretism is feature simplification, which occurs in Suffixaufnahme but not in simple noun inflection. The differences between simple noun inflection and Suffixaufnahme in the first and third type of syncretism point to a difference in cohesion: noun stems and their own case and number suffixes form phonological as well as morphological words, permitting interactions between the parts, such as a syncope of the stem vowel conditioned by suffixes (e.g. kmar-i vs. kmr-isa, kmr-ita in Table 4.2); by contrast, the suffixes attached by virtue of Suffixaufnahme are looser, cliticlike appendages forming phonological but not morphological words with their hosts and hence can be omitted or added in manners uncharacteristic of word morphology.

5.3. The Adjective Theory As was seen above (2.3), adjectives and genitives share agreement in [case], and genitives without Suffixaufnahme could therefore be considered a separate class of "adjectives." The different layers of case-number marking would accordingly have different morphological status: non-final case-number would be derivational, producing adjectives from nouns, and case-number due to Suffixaufnahme would be inflectional—no very remarkable state of affairs cross-linguistically. The diagnostics speak against this version of the adjective theory. Genitive NPs have the same privileges of occurrence as non-genitive NPs, which differ from those of adjectives: unlike adjectives, genitive NPs also occur as objects of verbs such as 'to fear' or 'to be ashamed of (Imnaisvili 1957: 715), as adverbials ('(to buy/sell) at the price of, 'in (winter)', etc.), and as objects of some postpositions. But could genitives with Suffixaufnahme be "adjectives"? This was the theory espoused by Franz Bopp (1846: 275 [1972: 413])23 and later also by Deeters (1926/1927: 36); and in view of the well-known affinities between genitive and adjective in many languages, its appeal is undeniable. However, even though both ordinary agreement and Suffisaufnahme have phrases

Suffixaufnahme

in Kartvelian

185

(headed by adjectives, nouns, determiners, or postpositions) as their scope (see 2.2 and 3.2), syntactic position and agreement as its corollary are the only features that adjectives share with genitives. In other respects they differ. First, unlike genitive NPs, adjectives have no non-agreeing counterparts in attributive positions. Second, the morphological constraints on adjectives, as well as on ordinary nouns, are different from those applying to genitive NPs with Suffixaufnahme (5.2). Third, adjectives are typically formed from lexical bases, while genitives have internal syntactic structure. In the absence of any other rule applying to both typical adjectives and genitives with Suffixaufnahme, little would be gained by subsuming them under one label. The cross-linguistic tendency for attributive adjectives rather than genitives to agree with their heads hardly justifies the identification of genitives with adjectives in Georgian. This is not to deny the possibility of case markers deriving adjectives from nouns. In fact, some were used productively for this purpose in the history of Georgian.24 5.4. Suffixa5.4. Suffixaufnahme's Hostufnahme's Host

As demonstrated above (3.2), the scope of Suffixaufnahme is the phrase, and its domain is one—and only one—constituent. The question that remains is, what is its phonological host? In the usual examples the case-number markers transferred by Suffixaufnahme attach to genitive head nouns. But they may also be hosted by other phrase-final elements, such as possessive pronouns (46c, 63a) or adjectival modifiers (19, 63b/c). (63)

a. mam-ta cuen-ta-j ganbneva-j (Kim II 216,14 apud Imn 584) father-P1(Gen) our-P1(Gen)-Nom scatter(ing)-Nom 'the dispersion of our fathers' b. sisxl-ita kac-ta ma-t martal-ta-jta (Kim II 106,13 apud Kiz 272) blood-Ins man-P1(Gen) Art-P1(Gen) just-P1(Gen)-Ins 'with the blood of the just men' c. sul-ta qrma-ta ccwl-ta-jsa (Job 24,12 apud Imn 601) soul-Pl(Gen) youth-P1(Gen) tender-P1(Gen)-Gen 'the souls of tender youths'

In fact, the host of Suffixaufnahme, though usually phrase-final, need not be in this position; in (36a/b) and (46e) the host is followed by the article. The appropriate generalization appears to be that the transferred markers are attached to the last non-clitic word of the phrasal scope of Suffixaufnahme. As the article is a clitic, Suffixaufnahme cannot be hosted by the article. This restriction actually follows from the fact that Suffixaufnahme, as a syntactic process, precedes clitic movement. Being phrasal rather than word-inflectional, Suffixaufnahme markers are not sensitive to any morphosyntactic properties of their phonological hosts,

186

The Caucasus

such as word-class or morphological classifications. Conceivably they might be sensitive to such properties of the head nouns, but they are not. In fact, Georgian nominal inflection is remarkably uniform in the first place, with only a few marginal deviations—such as proper nouns remaining uninflected for nominative, vocative, and ergative; the 1st and 2nd person pronouns me, sen, even and tkven being syntactically nominative, genitive, and dative; or the ergative form of the demonstrative and article man failing to be segmentable into stem and ergative suffix (being identical with this case suffix instead). It is precisely these deviantly-inflected words that cannot be heads of genitives, and the regularity of case-number marking in Suffixaufnahme is thus undisturbed.25 While perhaps not necessarily a precondition, uniformity of inflection, a characteristic of the agglutinative type of morphology, would then at least seem to be favorable to Suffixaufnahme.

5.5. Ellipsis Surface structures often do not contain an overt noun in the same sentence supplying the case and number suffixes for Suffixaufnahme (Imnaisvili 1957: 617-619, Boeder 1987a: 42). Thus the head noun mokmed- 'laborer' has been omitted in the second sentence of (64), being identical to a noun in the preceding sentence. (64)

"moucode mokmed-ta ma-t" [. . .] mo=raj=ukue=vides [sc. mokmed-n-i] meatertmet-isa m-is zam-isa-n-i igi miiyes tito-j satir-i (Mt 20,8-9C) "call worker-P1(Dat) Art-P1(Dat)" [. . .] now=when=they=came [sc. laborer-P1-Nom] eleventh-Gen Art-Gen timeGen-P1-Nom Art=Nom they=got one(each)-Nom denary-Nom '[The lord of the vineyard saith unto his steward:] "Call the laborers!" [. . .] when they came that were hired about the eleventh hour, they received every man a penny' (translated from Greek: elthontes de hoi peri ten hendekaten horan [sc. ergatai] elabon ana denarion)

Similarly, head nouns are deleted in conjunction reduction (3.2.1). Sometimes it is not the linguistic context that sanctions ellipsis; general nouns such as 'man, people, matter' may also lack overt expression: (65)

a. masin Huriastan-isa-n-i ivltoded mta-d (Mt 24,16C) then Judaea-Gen-P1-Nom they=should=flee mountain-Adv 'Let them which be in Judaea flee into the mountains' b. romel-i kueqan-isa am-is-gan-i ars, kueqan-isa-gan ars da kueqan-isa-sa itqws (J 3,31C) who-Nom earth-Gen Art-Gen-from-Nom is earth-Gen-from is and earthGen-Dat says 'He that is of the earth is earthly, and speaketh of the earth'

Suffixaufnahme

in Kartvelian

187

c. xolo iqo netar-i Mikel kueqan-isa SavSet-isa-j sopl-isa Norgial-isa-j (GM XXXII 25) but was blessed-Nom Michael earth-Gen Shavshet-Gen-Nom village-Gen Norgial-Gen-Nom 'but the blessed Michael was from the land Shavsheti from the village of Norieli' d. area moiqsenet xut-ta ma-t pur-ta-j xut atas-ta-j ma-t (Mt 16,9E) and=not you=remembered five-P1(Gen) Art-P1(Gen) bread-P1(Gen)-Nom five thousand-P1(Gen)-Nom Art-P1(Gen) 'Do you [. . .] neither remember the five loaves of the five thousand?' Such case-number marking stemming from abstract, elliptic sources should not be identified with Suffixaufnahme in the narrow sense. Reconsidering (64), where the head noun is assumed to precede the genitive NP, an interpretation involving both Suffixaufnahme and ellipsis is required. Obviously, however, non-overt head nouns are hard to locate without additional evidence. The underlying representation of the second sentence of (64) might as well be as in (64'), with the suffixal sequence -isa-n-i resulting from the contextually sanctioned deletion of the stem mokmed-, present in the adjacent sentence. (64')

meatertmet-isa m-is zam-isa [mokmed]-n-i igi eleventh-Gen Art-Gen hour-Gen [worker]-P1-Nom Art=Nom

Multiple case-number marking would thus not result in the Suffixaufnahme manner; instead of encoding a syntagmatic relationship between two NPs, it would be the inflectional trace of a contextually recoverable noun stem.26 Instances where the nominal that sanctions the ellipsis of a noun stem immediately precedes the deletion site in the same sentence would be ambiguous between the Suffixaufname and the ellipsis interpretations; and such ambiguities arguably play a role in the origin of genuine Suffixaufnahme (see 7.2 below). The histories of Suffixaufnahme and ellipsis as such, however, have been very different in modern Kartvelian, with the former showing signs of receding as early as the twelfth century—owing, inter alia, to the disinclination of modifiers to be postposed—and with ellipsis continuing to thrive. Notice that a language like Basque has productive multiple case-number marking with ellipsis and afterthought, but no Suffixaufnahme (Lafitte 1979: 61 §145): (66)

a. aita-ren father-Gen 'father's' b. aita-ren-a father-Gen-'Art' 'the one of the father'

188

The Caucasus c. ait-en father-GenP1 'of the fathers' d. ait-en-a father-GenP1-'Art' 'the one of the fathers' e. ait-en-a-ren father-Gen-P1-'Art'-Gen 'of the one of the fathers', etc. 6. Suffixaufnahme in Modern Kartvelian

6.1. Modern Georgian Double case-number marking due to ellipsis is fully preserved both in the written language (67a) and in dialogue (67b). (67)

a. romelic sveli sistem-is porm-eb-tan ertad axl-isa-c blomad gvicvenebs (Patsch 1964: 143) which old system-Gen form-P1(Dat)-with together new-Gen-too abundantly shows=us ( < axali [sistem-]is [porm-eb-sa-]c blomad gvicvenebs)27 'which together with forms of the old system show us those of the new one' b. vi-s saxl-s esebs? — cemi amxanag-isa-s (Dondua 1930 [1975: 24, note 14]) who-Gen house-Dat he=looks=for — my comrade-Gen-Dat 'Whose house is he looking for? — The one of my comrade'

In Old Georgian, enriched by the word amxanag-, (67b) would have been cem-isa amxanag-isa [saxl-]sa. 28 Note that the answer preserves the old form of the genitive -isa. Even two-year old children master this pattern: (68)

visi sal-it moxvedi? — bebia-s-it (Kaxase 1969: 19) whose shawl-Ins have-you-come? — grandma-Gen-Ins ( < bebia-s sal-it) 'With whose shawl have you come? — With grandma's' The difference between ellipsis and Suffixaufnahme becomes apparent in

(69): (69)

kakl-is xe-eb-i ezo-eb-si da xil-isa-n-i bayc-eb-sa da bay-eb-si (I. Cavcavase apud Kvacase 1977: 167) walnut-Gen tree-P1-Nom yard-P1(Dat)-in and fruit-Gen-P1-Nom little=gardenP1-Dat and garden-P1(Dat)-in 'walnut-trees in the yards and (sc. trees) of fruit in the little gardens and orchards'

Suffixaufnahme

in Kartvelian

189

Suffixaufnahme here cannot take up the plural marker -eb- (5.2), but ellipsis requires a plural marker even in Modern Georgian. The old plural suffix -ncontinues to be fully alive with ellipsis: it is used on genitives and other modifiers in the absence of an immediately following head noun, as in (70), to be contrasted with (71). (70)

a. sxva sm-eb-i other brother-P1-Nom 'other brothers' b. or-i 3ma two-Nom brother (Nom) 'two brothers'

(71)

a. sxva-n-i ar mosulan other-P1-Nom not come=Perfect 'The others did not come' b. or-n-i vart two-P1-Nom we=are 'We are two'

Owing to their tendency toward group-inflection, Modern Kartvelian languages have developed a distinction between contact forms and non-contact forms, used respectively in the presence or absence of a following head noun. Non-contact forms are partially different from forms with Suffixaufnahme. The contact form of adjectival modifiers in Modern Georgian has a partially reduced form of agreement: (72)

a. 3vel-ma cign-ma old-Erg book-Erg b. svel cign-s old book-Dat (cf. Old Georgian 3vel-sa cign-sa)

In some dialects all prenominal adjectives can be left uninflected. This "demorphologization" (Klimov 1962: 127-129) of preposed modifiers, already attested in medieval Georgian texts (Kartozia 1968: 162), is characteristic of all Kartvelian languages. As a result, modifier-head constructions can look like compounds (Cikobava 1936: I, 71). Postposed adjectives, occurring only in literary or even poetic language, must be fully inflected in non-contact position (cign-s 3vel-s(a)). Preposed genitives have a reduced "short" contact form, and postposed genitives—not occurring frequently in the first place—have "long" noncontact as well as "short" contact forms; the "long" non-contact forms optionally undergo Suffixaufnahme. Table 4.3 gives the three paradigms of postnominal genitives as recognized by Topuria (1936), paradigm I being the long non-contact form with Suffixaufnahme, paradigm II the long non-contact form without Suffixaufnahme, and paradigm III the short contact form.

190

The Caucasus

Table 4.3. Inflection of Head Noun and Following Genitive ('the book of a/the woman') in Modern Georgian. I

II

III

Nominative cign-i kal-isa-0 cign-i kal-is-a fign-i kal-is Ergative Dative etc.

cign-ma kal-isa-m cign-s kal-isa-s

cign-ma kal-is-a cign-s kal-is-a

cign-ma kal-is cign-s kal-is

Paradigm III is very common in poetry, and it has been infiltrating prose, in particular the "rhythmic prose" of some authors (Gigineisvili 1984: 62). It is attested since the early nineteenth century at the latest: (73)

zrunva-sa mep-is (Nikoloz Baratasvili) care-Dat king-Gen 'the care of the king'

The simple non-contact form, as in paradigm II of Table 4.3, sometimes occurs in everyday speech, in frequent formulas (74a) and elsewhere (74b). (74) a. dila msvidob-isa morning (Nom) peace-Gen 'good morning!' b. esa-a norma axal-i Kartul-i saliteraturo en-isa (from Gigineisvili 1984) this=Nom-is norm (Nom) Modern-Nom Georgian-Nom literary (Nom) language-Gen 'This is the norm of the Modern Georgian literary language'29

In (74b) the prenominal position of the genitive, taking the short contact form (axali Kartuli saliteraturo en-is norma), is probably dispreferred because 'the Modern Georgian literary language' is rhematic and is being contrasted with the Old Georgian literary language.30 It is unknown when paradigm II developed. Gigineisvili (1984: 57) seems to think that it already existed in the twelfth century, a period when many other changes pertaining to nominal inflection got under way.31 The literary language, on the other hand, preserved the old norms of paradigm I for many centuries, with Sulkhan-Saba Orbeliani (1658-1725), for example, regularly using Suffixaufnahme even in the plain prose of his Wisdom of the Lie. Suffixaufnahme on the non-contact form, as in paradigm I, is the recommended norm in Modern Georgian and as a rule occurs in literary prose. Nonetheless, it seems to have a bookish or solemn flavor for some authors, and "stylistic reasons" for its use are sometimes invoked (Topuria 1936). Cikobava (1950: 049) calls it "bookish and artificial." According to Gigineisvili, it is rarely used and gives the expression an "archaic air," but it must be used in some cases where the preposed construction "cannot replace it," for reasons unspecified

Suffixaufnahme

in Kartvelian

191

(1984: 43). Dondua (1930 [1973: 24]) adduces a typical example from a newspaper (75a) with postposition of a "heavy" constituent, and even hyperbaton occurs as in Old Georgian (75b). (75)

a. mizez-s nacionalizm-is gam3lavreb-isa-s reason-Dat nationalism-Gen strengthening-Gen-Dat '[In the same articles we explain] the reason for the strengthening of nationalism' b. porm-eb-i im 3gup-s ganekutvneba saxel-eb-isa-s (Patsch 1964: 142) form-P1-Nom that group-Dat will=belong noun-P1-Gen-Dat 'The forms will belong to that group of nouns'

Suffixaufnahme often fails to apply even in the modern literary language, and it is deficient in the sense that it never occurs with the genitive and the instrumental: (76)

a. *cign-isa kal-isa-sa ( < kal-isa-isa) book-Gen woman-Gen-Gen b. *cign-it(a) kal-isa-ta ( < kal-isa-ita) book-Ins woman-Gen-Ins32

Cikobava (1950: 049) observes that Suffixaufnahme is rendered more improbable as the distance between the word in the genitive and its head increases. The nominative of the non-contact genitives deserves particular attention: that of paradigm I has early become indistinguishable from that of II owing to a phonetic rule of Modern Georgian. Before a word boundary or a consonant, /i/ is elided (cf. okro-is gold-Gen > okros), which erases the nominative suffix -i after genitive -isa that one expects with Suffixaufnahme (kal-isa-i > kalisa), and enables such genitives to be reanalysed as not having undergone Suffixaufnahme in the first place (like prenominal genitives). Since the nominative is the most frequent case, and since Suffixaufnahme was generally "bled" by the increasing preference for preposed genitives anyway, this reanalysis of -isa could be extended to all cases, potentially eroding paradigm I. The reanalysed genitives in -isa could then be interpreted either as a purely positional variant of the prenominal contact form, eventually developing into the "short" contact form of paradigm III, or as instances of non-contact forms (paradigm II) being used for that conjunct in coordinate constructions that is distant from the head (77). (77)

kal-isa da bavsv-is cign-s woman-Gen and child-Gen book-Dat 'the book of the woman and the child'

Two factors that may well have delayed the replacement of paradigm I by II— hence the demise of Suffixaufnahme—presumably are the prevalence of the Old Georgian literary norm until the beginning of the nineteenth century, and the continuing vitality of double case marking in ellipsis. The influence of

192

The Caucasus

literary norms is confirmed by developments in non-literary Kartvelian languages, to be examined next.

6.2. Mingrelian and Laz Mingrelian and Laz are sometimes considered dialects of the same language, and they can be treated together for present purposes. Ellipsis gives rise to double case marking in both Mingrelian (78a/b) and Laz (78c). (78)

a. e xam-ks te bosi-si-k i?uu (Qipsise 1914: 0135; Chrest 24,27) this knife-Erg this youth-Gen-Erg was 'The knife was the youth's [sc. knife]'33 b. ma mejckvank ziri cxen-s — arti tavadi-si-s do mazira qazaqi-si-s (Xub 54,23 [?] apud Klimov 1962: 129) I send=you two horse-Dat — one prince-Gen-Dat and second peasantGen-Dat 'I will send you two horses: one of a prince and a second one of a peasant' c. joyori-si mota-s lakoti vucumelt, katu-si-s — ckintali, puji-si-s — geni, txa-si-s — tikani (Cikobava apud Klimov 1962: 125) dog-Gen young-Dat lakoti we=call=it, cat-Gen-Dat — ckintali, cow-GenDat — geni, goat-Gen-Dat — tikani 'The young of a dog we call 'lakoti', that of a cat 'ckintali', that of a cow 'geni', that of a goat 'tikani' '

Mingrelian NPs typically show group-inflection; most frequently it is the head nouns that are in final position and hence carry case and number marking, with adjectives and genitives in pre-head position remaining uninflected: (79)

(80)

a. skvami cira-k / cira-si beautiful girl-Erg / girl-Gen b. skvami ciren-k / cir-ep-is(i) beautiful girl(P1)-Erg / girl-P1-Gen 'beautiful girl(s)' 3a-si oxori wood-Gen house = Nom 'house of wood'

Often, however, the marker of the preposed genitive is reduced: -(i)si > -(i)s > j > 0 (81); and only the last conjunct in coordinate constructions is inflected (82).34 (81) a. koci-sguri man-Gen heart = Nom 'the man's heart'

Suffixaufnahme

in Kartvelian

193

b. xencipe-j skua king-Gen son=Nom 'the king's son' c. cai-si suma- dro-s tea-Gen drinking (Gen) time-Dat 'at the time of tea-drinking' (Lomtase 1954) (82)

kinosk d ate sapule-s xencspe do musi cili do skua do vezir-en-k (Kartozia 1968: 154 = Xub 206,11-13) stayed this cemetery-Dat king and his wife and child and vezir-P1-Erg 'The king and his wife and his child and the vezirs stayed on this cemetery'

Sometimes modifiers, in particular possessive pronouns, follow the head noun, then acting as carriers of case inflection, while number continues to be marked on the nouns themselves: (83)

a. 3uma ckimi-k brother my-Erg b. 3uma-lepe ckimi-k brother-Pi my-Erg 'my brother(s)'

(84)

?ude didi-k house big-Erg 'big house'

(85)

skani skvami tol-ep xantir-s your beautiful eye-Pi painted-Dat '[I want to look for ever at] your beautiful, painted eyes'

As pointed out by Kartozia (1968: 160f.), all these phenomena can—to a lesser degree and more rarely—also be found in Georgian, even in the literary sources. But there are additional possibilities in Mingrelian and Laz. Both adjectival modifiers and genitives can be postposed after the inflected group and receive their own case-number markers. Without prejudging the question of constituent structure, there being no relevant diagnostics such as clitic movement in Modern Kartvelian, this pattern will be referred to as "extraposition." The contrast between group-inflection and repeated marking on head nouns and extraposed possessives is illustrated for Mingrelian in (86).35 (86)

a. ucu vezir musi-s (Xub 69,8) he=said=to=him vezir his-Dat 'he said to his vezir' b. ucu osur-k musi-k mazia ima-s (Xub 276,9) she=said=to=him wife-Erg her-Erg second brother-Dat 'the wife said to her second brother'

194

The Caucasus

In (87), likewise from Mingrelian, the repeated marking on an extraposed modifier also involves number: (87)

art xemcape-s Puns breli cxen-epi cak-epi (Kiziria 1982: 223; Xub 137,2) one king-Dat has many horse-P1=Nom fill-P1=Nom 'One king has many fillies'

There is a corresponding contrast in Mingrelian between genitive NPs with group-inflection (88a) and those with repeated case marking (88b-e), the latter thus conforming to the Suffixaufnahme paradigm I of Modern Georgian; in Laz postnominal genitives, though acceptable (cf. (90c) below), are unusual (Kartozia 1968: 166, Klimov 1962: 125). (88)

a. burjua-s-en-s usaxana uca teri musa-si-sa (Kartozia 1968: 162f.; Xub 309,3-8) bourgeois-son-Pl-Dat they=call=them black enemy worker-Gen-Dat 'They call the sons of the bourgeois the enemy of the worker' b. kimibvanki-a xorcobua-s nadiri-si-s (ib.; Xub 305,13-14) that=I=bring-Quot meats-Dat game-Gen-Dat 'that I bring the meats of the game' c. giantxuudesa kata-s te xencspe-si mala-si-s (Qipsise 1914: 0135; Chrest 55,10) They=attacked people-Dat this king-Gen dominion-Gen-Dat 'they attacked people of this king's dominion' d. gverd tqebi-s kimercanti-a si-a te tura-si-si-a (Kartozia 1968: 163; Xub 262,23) half skin-Dat we=will=give=you-Quot you-Quot this jackal-Gen-DatQuot 'We will give you half of the skin, he said, you, he said, of this jackal, he said' e. seudirt ostat-en-k keteba-s ?ude-si-s (Xub 21,20 apud Klimov 1962: 125) they=set=to=work master-P1-Erg making-Dat house-Gen-Dat 'The masters began to construct the house'

Group-inflection (88a) develops from repeated case marking structures (88be) by deletion of internal case marking. This development conforms to a general preference for final case marking (5.1). Hyperbaton with case marking repeated analogously also occurs in Mingrelian (89) and Laz (see (91c) below). (89)

imendi-ks meudina cira-si xvamilapa-si-ka (Qipsise 1914: 0135; Chrest 63,3334) hope-Erg he=lost girl-Gen elope-Gen-Erg 'He lost any hope of eloping with this girl'

Suffixaufnahme

in Kartvelian

195

Unlike adjectival modifiers, extraposed genitives have yet another option in Mingrelian (90a/b) and Laz (90c), corresponding to paradigm II of Table 4.3 in Georgian (Kartozia 1968: 164). (90)

a. iPu ak cxupi-k otxolo-si (Xub 264,31) it=came=about here quarrel-Erg four-Gen 'A quarrel began between the four' b. mela-k demi molee?one oskvidapu-sa badidi-si (Qipsise 1939: 42,28) fox-Erg Dev=Nom brought=with=him strangling-Dir grandfather-Gen 'The fox brought with him the Dev (i.e. a giant) to strangle the grandfather' c. tito canas ar-ar tane muqonasunonan oxori-sa bic-epe-si (Kartozia 1968: 166) every year one-one specimen they=must=bring=it house-Dir boy-P1-Gen 'Every year they must bring one into the young men's house'

While (90b) can be explained as the result of a morphological sequence condition, (90a) and (90c) cannot: Suffixaufnahme simply does not apply in this example. What (90c) shows is that if postposed genitive NPs occur at all in Laz, they lack Suffixaufnahme. Even hyperbaton allows for non-agreement: (91)

a. dro-k komortu bayana-si dabadeba-si (Qipsi3e 1939: 83,8-9 apud Kartozia 1968: 164) time-Erg came child-Gen bearing-Gen 'the time of bearing a child' (cf. (52)) b. oze-sa miniles arabi-si-ni, arabi-k dou3ax Ale-s (Kartozia ib.; Xub 172,19— 20) courtyard-to they=went Arab-Gen-when, Arab-Erg he=called Ale-Dat 'When they went into the courtyard, the Arab called for Ale' c. dezgini mazdu cxeni-si (Cikobava 1936: II,39,1 apud Klimov 1962: 125) bridle=Nom he=took=it=to=me horse-Gen 'He took the bridle of my horse'

Suffixaufnahme on extraposed genitives is thus not obligatory in Mingrelian. In fact, the pattern referred to as Suffixaufnahme sometimes permits an ellipsis interpretation. Interestingly, Qipsise (1914: 032) puts commas around extraposed modifiers, which look like appositions with ellipsis of the head noun: (92)

a. arti Jima-k, uncasi-k, doyuru (Chrest 52,22) one brother-Erg, older-Erg, died 'One brother, the older one, died' b. ckimi nacinebu cira-k, skvam-k, kamorta my acquaintance girl-Erg, beautiful-Erg, came 'A girl whom I am acquainted with, a beautiful one, came'

196

The Caucasus

Similarly, the editor of the text from which (87b) is taken (Xub) puts a comma after xorcobuas (deleted by Kartozia 1968: 163 note 55), which would no doubt be fully justified in (87d) because the quotative particle always marks a phrasal boundary. On the other hand, extraposed possessive pronouns as in (86b) can hardly be interpreted as having their head elided.

6.3. Svan Like the other Modern Kartvelian languages, Svan distinguishes between non-reduced non-contact forms of the genitive (93a) and a reduced contact form (93b). A non-reduced contact form occurs only in some dialects (Lakhamula, Lower Bal dialect; Kaldani 1956: 167) and in archaic poetry (93c) side-by-side with a reduced form (93d) (Abesa3e 1984: 13f.). (93)

a. katlas hen=Gen (nominative: katal) b. katla pas hen=Gen price=Nom c. Dalas kvaba-s Dal=Gen grotto-Dat 'in the grotto of the Dal (a female mythological being protecting the game)' d. Dala lasgar Dal=Gen host=Nom 'the host of the Dal'

The non-contact form is also the form used in ellipsis. It is less clear, however, whether ellipsis requires multiple case-number marking in Svan. Since a nominative marker after the genitive marker -is would be zero in Svan, examples like (94a) (with a non-reduced genitive cincl-is of cincil) do not tell us if multiple case-number markers occur with ellipsis, and I have been unable to find a non-nominative example; on the other hand, I can offer at least one Lower Bal example (94b) where multiple case marking occurs with an adjective modifier (isgnem-sv < isgnem [si-ar]-sv). (94) a. dadv katla pas semi abaz li, cinclis semi saur (Gabliani 1925: 176) female hen=Gen price three=Nom abaz=Nom is, chicken=Gen three=Nom shaur=Nom 'The price of a hen is three abzis, that of chicken three shauris' b. jar eser mica si-ar-sv imsai, jar — isgnem-sv (Dav 119,6) sorne=Nom Quot his=own hand-P1-Ins works, some=Nom — other=OblIns 'Some work with their own hands, others with other people's (sc. hands)'

Thus multiple case-number marking cannot be ruled out with genitive modifiers in ellipsis. Preposed modifiers show partial agreement, as they do in Georgian:

Suffixaufnahme

in Kartvelian

197

(95) a. xoca 3 mil good=Nom brother=Nom b. xocam 3 mil-d good=Obi brother-Erg 'good brother' There is group-inflection as in Mingrelian: (96)

a. sarsgan ankid kutxva misgva-d (Abesage 1984: 12; SvP 140: 37) roar=Norn stopped Kutkha my-Erg 'My Kutkha (name of a rifle) stopped (its) roaring' b. ale Bucu Goginos-d amsdebe (Dondua 1930 [1975: 24])36 this=Nom Buchu Gogino=Gen-Erg he=lent=it=to=me 'Buchu's son lent this to me'

Extraposed modifiers do occur with case marking, as in Mingrelian (97a); but they may also lack case marking—showing, however, partial agreement (97b). (97)

a. idreqiv gvisgvej paqva m gem-s cl-s may=he=protect our hat-Gen wearing-Dat all-Dat 'May he (i.e. Saint George) protect all our hat-wearers (i.e. men)' (a traditional toast communicated to me by Ciuri Gabliani) b. samun-d misgva cisx lajpx kne (Abesa3e 1984: 12; SvP 146,34) brother=in=law-Erg my=Obl foot=Nom stretched 'My brother-in-law stretched out his feet'

In (97b) misgva is the form that would also occur in pre-head position. As we saw above, genitives are reduced when preposed, i.e. in contact position (93b/d). However, the full, non-contact, form is obligatory in coordinate constructions for the conjunct that is distant from the head (98a) or that occurs after the head, separated from the first conjunct (98b). (98)

a. Rostom-is i Zurabi saplav-te-zi (Abesase 1984: 15; SvPr 234,18) Rostom-Gen and Zurab = Obl grave-to-upon 'to the grave of Rostom and Zurab' b. simindi liqde i manas-is (SvPr 302,24-25) maize=Obl collecting=Nom and rye-Gen 'the collecting of maize and rye'37

In prose as well as in poetry, the full form is normal with all types of genitives not followed by their head noun (99), including hyperbaton constructions (100).38 (99)

a. ecanyo kanjesx bajrax-s, xas Tarisel-is-s (Onian 25,25, Lashkh dialect apud Dondua 1930: 24 note 14)

198

The Caucasus thereafter they=bring=out banner-Dat icon-Dat Taridzel-Gen-Dat 'Then they bring out the banner and the icon of the archangel Taridzel' b. al gim-s Domentls-s des wodi (Dondua's note, ib.) that soil-Dat Domenti=Gen-Dat not=possible I=give=away 'I cannot give away Domenti's share of land' c. mewar xokwibx sukwisa lixwje-s ej-kalib-ar-es-s (Onian 6,33 apud Deeters 1927: 37) very they=fear on=their=way meeting-Dat such-form-P1-Gen-Dat They are afraid to meet such people (?) on the way' d. merma laday acad 3a-j latxel-te xosam laxvbas (SvPr 391,34) second day he=went himself-Nom look=for-to elder=Obl brother=P1=Gen 'On the second day he (i.e. the younger brother) went himself to look for his elder brothers' e. mag des acad asxvzi sukv-s maqal-sv zah-is (SvPr 34,5) all not=possible we=went together path-Dat fear-Ins avalanche-Gen 'We could not go up the path together for fear of an avalanche' f. esxu kalmax vokrvres lasv (SvPr 380:38) One trout=Nom gold=Gen was 'one trout was of gold' g. hangir oxqad exer laskad, vokvres-vercxl-is (SvPr 70,15) saddle=Nom came=out well forged, gold=Gen-silver-Gen 'The saddle came out well-forged, of gold and silver' h. ka maCkvre lusdgvar-s vokvras (Abesase 1975: 11; SvPr 88,10) Particle he=cuts=off tress-Dat gold=Gen 'He cuts off the tresses of gold'

(100)

a. ezer licvdi xoxal nasur-is (SvPr 51,9) good spell=Nom he=knows=it surfeited-Gen 'He knows a good spell for a surfeited man' b. val xagdanda davas (Abesase 1975: 10; SvPr 65,28) debt=Nom was=on=him Dev=Gen 'He was indebted to the Dev' c. diiva kor lag ceis (Abesase 1975: 11; SvPr 307,30) Dev=Gen house=Nom stands copper=Gen 'There stands the Dev's house of copper'

While there are clear instances with (99a-c) and without (99e/h) repeated case marking on genitive NPs, (99f/g) and (l00a-c) are ambiguous in this respect, because a separate nominative suffix is impossible after -is. In light of (99a-c) it is curious that Abesase (1975: 11) should simply deny the existence of Suffixaufnahme in Svan. As these examples are from the Lashkh dialect, spoken in the neighborhood of Georgian, it is possible, however, that Suf-

Suffixaufnahme

in Kartvelian

199

fixaufnahme is now dialectally limited, having been abandoned in Upper Svan, the dialect of the other examples, as it was in Laz as opposed to Mingrelian. Alternatively, genitives with repeated case marking in Lashkh may have to be interpreted as afterthoughts with ellipsis ('the icon, that of Taridzel'). In Upper Bal, all postposed modifiers are positional variants of the preposed modifiers; (97a) may be an exception, perhaps being an archaism. On the other hand, the common Kartvelian distinction between non-reduced non-contact forms and reduced contact forms is preserved. Similarly, multiple case marking seems to be preserved with ellipsis, again as in Laz. Whereas nominal inflection in Mingrelian is uniform—mentioned earlier as a factor presumably encouraging Suffixaufnahme—there are several declensions in Svan. The dative marker occurring in (99a-c), however, happens to be shared by all declensions in the Lashkh dialect. Considering a non-uniform dative such as mara man=Dat from the Upper Bal dialect (cf. mare man==Nom), the question is whether 'the man of Domenti' (cf. (99b)) would in this dialect be rendered as in (l0la) or as in (l0lb). (101)

a. mara Domentis-s b. mara Domentis

On the analogy of (99h), the second alternative, without Suffixaufnahme but with the non-contact form, appears to be the more probable one. 7. Functions, Origins, Parallels

7.1. Functions Although in a general way Suffixaufnahme can be made sense of as one solution of the problem of encoding multiple relations (see Moravcsik, this volume), one is entitled to wonder whether it might not fulfill more particular functions in the languages going for this option. For Old Georgian, where genitives with Suffixaufnahme co-occur with genitives without it, one would have to ask, then, what advantages accrue from the restructuring of NPs and Suffixaufnahme, its formal concomitant. It is not immediately obvious why this habit should be so laudable as it seemed to Prince Antoni I. Some possible leads will be explored in this section. As for Suffixaufnahme on prenominal genitives, an inspection of Imnaisvili's (1975: l0lf.) material suggests some connection with the so-called infinitive construction: in most examples, the head is a verbal noun (masdar) and the genitive NP with Suffixaufnahme is its subject or object. In Old Georgian as in other languages,39 non-finite clauses may undergo several forms of incorporation into the main clause. Consider for instance the "impersonal verb" mnebav-s 'I want', which takes a nominative complement in (102a) (as in it pleases me (to . . .)); in (102b) the verbal noun is demoted and takes the adverbial case, while its object ('his resurrection') is raised and promoted to

200

The Caucasus

the surface subject position formerly held by the verbal noun, and consequently takes the nominative. (102)

a. m-nebav-s xilva-j lSgObject-want-3Sg see(ing)-Nom 'I want to see' b. mnebavs xilva-d adgoma-j-ca m-is-i (SinMr 172,16 apud Cxubianisvili 1972: 87) I=want=it see-Adv stand=up-Nom-too he-Gen-Nom 'I want to see his resurrection, too'

In an example like (24a), on the other hand, repeated here for convenience, the verbal noun retains its syntactic position as in (102a), but the object ('Jesus'), inflected for its adnominal relation (genitive), is raised and consequently undergoes Suffixaufnahme. (24)

a. gunebavs lesu-js-i xilva-j (J 12,21E) we=want Jesus-Gen-Nom seeing-Nom 'We would like to see Jesus'

Thus Suffixaufnahme with the objective genitive in (24a) and promotion with the object in (102b) serve the same purpose of raising that is characteristic of different non-finite constructions. But this function can hardly be generalized to all relevant occurrences. Sometimes a contrastive or focusing function has also been assumed for Suffixaufnahme on prenominal genitives (Dondua 1930 [1975: 23]), which is not implausible in instances such as (26b) ('isn't this the carpenter's son?') or (103), Dondua's own example: (103)

[da ara sad gansisuldis da area ibanis sada da hrcxuenia visame igi sisueli da] ar-ca tws-ta xorc-ta-j hnebavn xilva-j (Kim 16,20-22) [. . .] not-and his=own-P1(Gen) flesh-Pl(Gen)-Nom he=would=want seeingNom '[and he used never to bare himself nor ever to bathe himself and was ashamed to see anybody naked] nor would he look at the parts of his own body'

The only hint so far at a possible function of Suffixaufnahme on postposed genitives—its most frequent occurrence—is the idea, first aired by Dondua (1930 [1975: 31]), that it marks the "closedness and completeness" of its scope (zakoncennosf'; zamknutaja i cel'naja sintaksiceskaja edinica).40 One might add that Suffixaufnahme, like multiple case-number marking in ellipsis and like the non-contact forms of Modern Kartvelian, here also signals the absence of a following head noun. While it may indeed be advantageous to mark the end of postposed genitive NPs, especially of long ones, and to indicate a deletion site, the actual functional load on Suffixaufnahme in these respects is

Suffixaufnahme

in Kartvelian

201

fairly low. In eleven pages (of 37 lines each) of elaborate Old Georgian prose (GM), I counted some 110 instances of Suffixaufnahme (about ten per page on average): ca. 70 were clause-final; 30 preceded words unable to govern genitives; and only ten were followed by a noun, thus creating a situation where syntactic constituency is potentially ambiguous. However, on semantic grounds these following nouns could almost never be mistaken for heads of the genitive, rendering Suffixaufnahme essentially redundant as a bracketing device. A case in point is (104), where the construal of Kristeane-ta, without Suffixaufnahme, with mydeloba-sa would yield the unorthodox meaning 'to the priesthood of the Christians'. (104)

cxoreb-isa sasoeba-j Kristeane-ta-j mydeloba-sa sehvedra (GM III 30) life-Gen hope-Norn Christian-P1(Gen)-Nom priesthood-Dat he=entrusted=it=to=it 'He (i.e. God) entrusted the Christians' hope for salvation to the priesthood'

The positional versatility of the genitive permits a distinction between subjective and objective genitives when the two co-occur. They occur on opposite sides of their joint head: the object precedes it and the subject follows it in (104), but (105) shows the reverse order. (105)

Pilate-js-i igi dabana-j qel-ta-j (SinMr 92,34) Pilatus-Gen-Nom Art=Nom washing-Nom hand-P1(Gen)-Nom 'Pilatus's washing of his hands'

There is no doubt that Suffixaufnahme helps to bracket recursive genitive NPs, because final case-number markers unambiguously code the relation between successive NPs: (106) cannot mean 'the keys of the heavens of the kingdom' (Plank 1992). (106)

klite-n-i sasupevel-isa ca-ta-jsa-n-i (Mt 16,19) key-Pl-Nom kingdom-Gen heaven-P1(Gen)-Gen-P1-Nom 'the keys of the kingdom of the heavens'

But even if Old Georgian were a language without Suffixaufnahme, other things being equal, a sequence like (106') would be unambiguous because of the raising constraint (Section 4): we would know that the sequence of genitives mirrors its derivational history, which produces unidirectional dependencies. (106')

*klite-n-i sasupevel-isa ca-ta

There is a possibility that simple, prenominal genitives and postposed genitives with Suffixaufnahme differ semantically—the former tending to be "descriptive" (as in English a fool's paradise, without definiteness of the head induced by the genitive41) and the latter "determinative" (Quirk et al. 1985: 327). Previous examples were largely of the determinative sort; here are others to illustrate the descriptive use of the genitive:

202 (107)

The Caucasus a. tevzta simravle-j did-i (L 5,6 Xanm) fish-P1(Gen) multitude-Nom great-Nom 'a great multitude of fish' (with fish being contextually expected: Jesus tells the unsuccessful fishermen on Lake Gennasaret to try again and they do catch fish in plenty) b. mam-isa sen-isa sakme-n-i (Sus IV) father-Gen your-Gen deed-P1-Nom '[Your father raised up sepulchres for the martyrs and built churches, and you have ruined] the deeds of your father' (translation by D. M. Lang; instead of 'father', again contextually given, one might translate 'paternal') c. tkuen-i tav-isa tma-n-i (Mt 10,30 Xanm) your-Nom head-Gen hair-P1-Nom 'the very hairs of your head [are all numbered]' (with hair of the head not opposed to other hair; the possessive belongs with 'hairs' rather than 'head') d. deda-ta buneba-j (Sus IV) women-P1(Gen) nature-Nom 'women's nature [is narrow]' (a cliche, translated by D. M. Lang as 'women are always liable to be unreasonable') e. pur-isa upal-man (J 2,9) bread-Gen lord-Erg 'lord of the bread' (cf. Greek arkhitrlklinos, German Speisemeister) f. mcuxr-isa zam-i (Sus II) eventide-Gen hour-Norn 'evening service' g. Somex-ta spajpet-i (Sus I) Armenian-P1(Gen) commander=in=chief-Nom 'generalissimo of the Armenians'

In addition to descriptive genitives Georgian also employs compounds, which differ from simple prenominal genitives in having the short form of the genitive suffix (-is/-t rather than -is/-td) and in being semantically sometimes noncompositional. (108)

a. 3m-is-col-i (Sus V) brother-Gen-wife-Nom 'sister-in-law' b. mydel-t-mo3yuar-i priest-P1(Gen)-teacher-Nom 'chief priest' (cf. (60))

Suffixaufnahme

in Kartvelian

203

The three degrees of formal "closeness" between the members of genitival constructions, mirroring the extent of their semantic independence, are brought out nicely by the following minimal pair, or rather triplet:42 (109)

a. kac-i [. . .] er-isa-j man-Norn [. . .] people-Gen-Nom 'man of the people' (cf. mcignobar-n-i er-isa-n-i (Mt 2,4) scribe-P1-Nom people-Gen-P1-Nom 'the scribes of the people' (in contrast to the priests, who, also literate, form their own caste)) b. er-isa kac-i (Sus V) people-Gen man-Norn 'layman' (contrasting with the priests) c. er-is-kac-i (Act 21,32) people-Gen-man-Nom 'soldier'

Yet another possible motive for the use of Suffixaufnahme (hinted at by Sani3e 1976: §276) is suggested by minimal pairs as found in the Martyrdom of St. Eusthatius. The title in (110a) is repeated a few lines later as (110b); similarly, (111a) is immediately repeated as (111b). (110)

a. Kartl-isa marzapan-sa (II) Kartl-Gen marzapan-Dat b. marzapan-sa Kartl-isa-sa marzapan-Dat Kartl-Gen-Dat 'the (Persian) governor-general of Kartli (Georgia)'

(111)

a. Kartl-isa mtavar-n-i (III) Kartl-Gen chief-P1-Nom b. mtavar-n-i Kartl-isa-n-i chief-P1-Nom Kartl-Gen-P1-Nom 'the princes of Georgia'

This use of Suffixaufnahme is anaphoric; it seems to derive from the employment of extraposed genitive NPs to add afterthoughts. As argued in Section 3, the majority of postposed genitive NPs are separated from their heads, ending up as immediate constituents of the clause. In such "flat" structures Suffixaufnahme enhances the recoverability of syntactic-semantic interdependencies. It is these separated genitive NPs in flat structures that suggest a possible point of departure for the development of Suffixaufnahme.

204

The Caucasus

7.2. Origins Multiple case marking ensuing from the ellipsis of heads, whose retained casenumber marking acts as a trace, occurs in all Kartvelian languages and may well be a common heritage. It is a phenomenon that is found in many languages without Suffixaufnahme, such as Basque (5.5), while Suffixaufnahme, on the other hand, in all probability implies the possibility of ellipsis and afterthought with multiple case marking. Therefore, if we find cases of overlap or ambiguity between ellipsis and Suffixaufnahme, then the latter can be the result of reanalysed ellipsis. As seen above, genuine Suffixaufnahme is not always easily distinguished from ellipsis. In fact, genitive NPs analysed above as separated from their heads, but not as involving ellipsis, sometimes give the impression of being added as afterthoughts and as such would imply ellipsis. Thus (46e) might also be rendered as 'and he sheds down the intestines, those of the birds', and (18) as 'that I may pronounce a praise, the one of the Father, and the Son, and the Holy Ghost'. Often, however, interpretations as afterthoughts are very awkward, especially with instances of multiple Suffixaufnahme (e.g., 'unto you it is given to know the mystery, that of the kingdom, (of) that of God' (48)); and Suffixaufnahme on prenominal genitives can hardly have been cataphoric (e.g., 'is this not the one of the carpenter, the son?' (26b)). Thus, synchronically, Old Georgian (and other early Kartvelian) Suffixaufnahme cannot be explained away completely as afterthought constructions including ellipsis. The overlap between them may point to a diachronic link, however. It is postposed genitives adjacent to their head nouns that appear to hold the clue to the development of genuine Suffixaufnahme from afterthought constructions. Not uncommonly, units anaphorically related to adjacent constituents are integrated into them as their co-constituents, with the boundary between them disappearing. In this manner conjunctions such as English that, originally cataphoric demonstrative pronouns within the main clause (I believe that: Georgian is a world language'), became part of the following subordinate clause they used to anticipate.43 Along similar lines, afterthought NPs as in ' ... the intestines, those of the birds', hardly separated by obligatory long pauses, are likely to have been reanalysed as attaching less loosely to the preceding NP and indeed its entire clause, ' . . . the intestines of the birds'. This reanalysis that gives rise to genuine Suffixaufnahme may have been further encouraged by a particular feature of Old Georgian, possessive pronouns. 3rd person possessive pronouns in Old Georgian are morphologically genitives with Suffixaufnahme, and they follow their heads unless they are contrastive or otherwise marked; differing from nouns in the genitive, however, they lack a counterpart without Suffixaufnahme (Kartozia 1968: 155):44 (112)

a. deda-j m-is-i; deda-j ma-t-i mother-Norn (s)he-Gen-Nom; mother-Nom (s)he-P1(Gen)-Nom 'his/her mother'; 'their mother'

Suffixaufnahme

in Kartvelian

205

b. m-is-i deda-j c. *m-is deda-j d. * deda-j m-is Arguably, the syntactic status of possessive pronouns is like that of adjectives and in particular of determiners such as articles and demonstratives, and like these they are always immediate constituents of their NPs and accordingly always agree with their head nouns. Since prenominai position of possessive pronouns is rare in Old Georgian,45 possessive pronouns and nominal genitives with Suffixaufnahme resemble NPs with elliptic heads in respect to order, in addition to the fact that all of them take up their head's case-number marking. Now, considering only possessive pronouns and ellipsis, on the assumption that Suffixaufnahme is more recent than both, the following developmental sequence seems plausible: (a) Frequent elliptical genitive NPs with an afterthought function could become a normal alternative to simple preposed genitives.46 (b) Afterthought NPs contiguous with the NPs they related to were reanalysed by analogy with postnominal third person possessive pronouns as co-constituents of these NPs; their case-number marking, which originally represented a trace of their elided head, was reinterpreted as being due to agreement inside the antecedent NP. (c) As agreeing immediate constituents of an NP, such former afterthoughts could be interpreted as the postnominal counterparts of prenominal genitive NPs. (d) When Suffixaufnahme in postposed NPs was interpreted as the concomitant of a hierarchical restructuring of prenominai genitives, creating such structures as would permit ordinary agreement, preposed counterparts were also possible, although they were less frequent. (e) Eventually, reversing the former process of integration, genitives located at a distance from the NPs whose case-number marking they reflected came to be reinterpreted as extracted from these NPs. That possessive pronouns should have been instrumental in getting these developments started is in line with the crucial role pronouns and other deictics are known to play in linguistic evolution, exerting a constant pressure in particular on the constitution of nominal categories (cf. Kurylowicz 1972). All the same, and despite the anaphoric function that Suffixaufnahme shares with ellipsis, Suffixaufnahme on genitive nouns as such never involved real anaphoric pronouns in the known or reconstructible history of Kartvelian. (In this sense Kartvelian differs from languages described by Aristar in this volume.)

7.3. Parallels and Analogues In the Caucasus the only close parallels to genuine Suffixaufnahme appear to be caiques. Thus Bats (or Tsova-Tush), whose speakers form a small enclave of North Caucasian Vainakh (Chechen-Ingush) in a Georgian-speaking area and

206

The Caucasus

have been bilingual for a very long time, borrowing much from Georgian morphology and syntax, has been credited with Suffixaufnahme by Schiefner (1859: 68 §235). Schiefner's examples of Suffixaufnahme are all taken from a translation of the Bible, and their naturalness remains to be examined. One may speculate that Suffixaufnahme is acceptable to Bats speakers in the same sense as Old Georgian syntax is acceptable to speakers of Modern Georgian: both may be accustomed to Suffixaufnahme as an archaism appropriate in religious language. No hint at the pattern has yet been found in other descriptions of Bats. Concerning Tsakhur, a member of Daghestanian, Bork (1913: 186) observed that sporadic and possibly archaic examples of Suffixaufnahme, cited from an earlier and less reliable source (R. von Erckert), might likewise be due to Georgian influence. Again, there is no hint at this phenomenon in later descriptions of Tsakhur. Widening the areal perspective, however, formal and functional parallels or analogues of Suffixaufnahme and its possible antecedent can be found in the ancient languages of adjacent areas. Suffixaufnahme is attested in its prototypical form in Hurrian and Urartian (see Wilhelm and Wegner in this volume). In post-Homeric Greek (113) and Iranian (114/115) the article and the relative pronoun respectively are reminiscent of Georgian Suffixaufnahme insofar as they repeat the case and number marking of the head.47 (113)

ho demos ho ton Athenaion Art=Nom people=Nom Art=Nom Art=PI=Gen Athenian=P1=Nom 'the people of the Athenians'

(114)

Yauna tyaiy uskahya uta tyaiy drayahya (DPe 12-14) Ionian=P1=Nom who=P1=Nom mainland=Sg=Gen and who=PI=Norn sea=Sg=Locative 'the lonians who are of the mainland and those who are by the sea' (Old Persian inscription of Darius in Persepolis; Fraenkel 1950: 83)

(115)

ma ram yim haurvatato (Yasna 31,6) spell=Sg=Ace which=Sg=Ace wholeness=Sg=Gen '[he will tell me] the spell of health' (from the most archaic Iranian text, the Gathas; Fraenkel 1950: 83)

In view of further typological isoglosses (cf. Boeder 1983), such resemblances are hardly coincidental, despite the temporal distances. Losely connected kinds of genitive NPs, like presumably the antecedents of Old Georgian Suffixaufnahme, are fairly common. From the point of view of processing, it is convenient to divide the information to be transmitted in a sentence into small packages that are phonetically and syntactically as selfcontained as possible. This is best done by starting with a core clause, in particular with a verb encapsulating the relational frame and the bare arguments thus interrelated or their pronominal representatives, and by successively adding further specifications and afterthoughts. The old Indo-European languages commonly employ this "appositional" mode of syntax, relying heavily on such

Suffixaufnahme

in Kartvelian

207

amplifications or "trains" (German Schleppe)—here illustrated from Homer's Iliad (116), a Greek mercenary inscription found in seventh-century BCE Egypt (117), an Anglo-Saxon formula (118), and an Old Norse narrative (119).48 (116)

(117)

(118)

(119)

to [sc. pedion] de pan hydatos plet' ekkhymenoio ( 300) this [sc. plain] then all water=Gen was-filled spilt=out=Gen 'The plain was covered with spilt-out water' basileos elthontos Elephantman Psamatikho king=Gen coming=Gen Elephantine=Ace Psammetikhos=Gen 'when Psammetikhos, the king, came to Elephantine' in usses dryhtnes noman haelendes cristes in our=Gen Lord=Gen name=Dat savior=Gen Christ=Gen 'in the name of our Lord, the savior Christ' en at o ru vari fara peir vestr pangat, Pornsteinssynir, tolf saman, til Valseyrarpings 'but next spring, they travel thither, the sons of Thorstein, twelve altogether, to the Valseyrar-Thing'

The results of such packaging are flatter syntactic structures whose components are easier to arrange and rearrange, as required by rhythm and shortterm planning. The avoidance of embedding is one further feature of flat syntax (Boeder 1987a, 1989b). As to Old Georgian, the oldest texts indeed have almost no embedded clauses other than occasional relative clauses, on which incipient embedding constructions were later modeled. Arguably, Kartvelian Suffixaufnahme too began with the loose syntax of afterthought NPs. The integration of postposed genitive NPs containing their heads relinquishes their syntactic independence and is a step toward the creation of more complex NP structures. Notes I am indebted to Lorraine McNeish for correcting and improving the English of an earlier version of this chapter, and to the editor for his extremely helpful "radical reform" that has improved the form and clarified the content of this investigation. All mistakes are of course mine. 1. Cf. Imnaisvili (1955: 262) for the article. 2. See also Taylor (1988), which came to my attention after I finished this paper. 3. Technically speaking, case features are assigned to the dominating NP node and percolate down to the nodes immediately dominated by NP but not further. 4. For more examples see Imnaisvili (1957: 183-190, 588). 5. [+case] is the necessary and sufficient feature of the nominative, while other cases require additional features (Boeder 1987a: 31); the nominative is the "minimal case" in this sense. The ergative with -man also counts as a long form triggering the long form of the genitive; i.e., it has the feature [+case] as a necessary but not a sufficient component. Ergative -man can only be spelt out if [ + case] plus some additional feature have been assigned to the noun. In other words, -man is the only real

208

The Caucasus

portmanteau morph in Old Georgian noun inflection, fusing the referential feature [+case] (see 2.4) and a case feature proper. Incidentally, if the short form 3e-0 kac-is-o son-Voc man-Gen-Voc 'the son of man' (Dondua 1930 [1975: 28]) is trustworthy (the texts usually have the long form kac-isa-o), it shows that the vocative has [-case], triggering the short form of the genitive according to the old norm (see 5.1). 6. Admittedly, however, anthroponyms with long forms of the case marker are on the increase in the history of Georgian. See Sarjvelase (1984: 364-372). 7. Oddly enough, Finck's example is not prototypical, probably because Suffixaufnahme in the narrow sense did not happen to occur in his sample text. What he offers is an example of multiple case-number marking in postpositional phrases (see 3.2.2): camoi axa ert-ma bavsv-ta-gan-ma exclaimed one-Erg child-P1(Gen)-from-Erg 'one of the children exclaimed'. 8. The pronoun vinme is to be analysed as vin 'who?' followed by the indefiniteness marker -me. It is true that vi-n is the nominative or the ergative form of 'who?' and contrasts with the dative form vi-s (cf. vi-s-me in (41)). But synchronically, -n is a case marker like zero in the nominative/ergative form of Grigol in Table 4.1, and it does not occur as such elsewhere. Vin(me) 'somebody', personal pronouns, and proper names like Grigol are at the top end of the animacy hierarchy (Boeder 1979: 438, 450, 457). 9. For more examples see Imnaisvili (1957: 383-385). 10. Similarly Lortkipani e (1959). Kiziria (1963: 255f.) and Patsch (1964: 133) hold the opposite view. Marr (1907: 297) saw preposed Suifixauinahme as a "pseudoarchaism." See also Deeters (1926/1927: 36). 11. See Imnaisvili (1957: 583f.) for further examples. 12. In (33) this possibly presupposes one single constituent cmida-ta adgil-ta, which would mean that NPs have more structure than is assumed in this paper. 13. The position of the genitive article for instance in (36c) is ambiguous: m-is could be a non-clitic, meaning 'that'. The validity of the argument rests on unambiguous examples such as (36f), where 'the temple' is not anaphoric in any sense but has a "unique denotation." 14. More examples of this kind are listed in Imnaisvili (1955: 265). 15. See further data in Imnaisvili (1975: 102-104). 16. Other examples are: Sur-i saxl-isa-j m-is mona-ta sen-ta-jsa-j (Parx apud Imn 629) envy-Nom house-Gen-Nom Art-Gen servant-P1(Gen) your-P1(Gen)-Gen-Nom 'the envy for the house of your servants'; sen xar Daniel je-ta-gan natijuenav-isa-ta-j Huriastan-isa-ta-j (Daniel 5,13 apud Imn 628) you are Daniel son-P1(Gen)-from bootyGen-Pl(Gen)-Nom Judaea-Gen-P1(Gen)-Nom 'you are Daniel, one of the sons of those who were carried away from Judaea'; qovel-i-ve cmel-i qbo-jsa-j m-is codva-tajsa-j (Leviticus 4,8 apud Imn 629) all-Nom-yet fat-Nom calf-Gen-Nom Art-Gen sinPl(Gen)-Gen-Nom 'all the fat of the (expiatory) calf of sins'. 17. Only one exception is available so far: 3-isa kac-isa-tws (Mt 12,32 Xanm, DE) son-Gen man-Gen-for 'against the son of man'. Imnaisvili (1957: 589) stresses the double deviance of this example: not only does the preposition occur at the "wrong" place (where 3-isa-tws kac-isa is expected), it also lacks Suffixaufnahme (only the Adysh manuscript (C) has kac-isa-jsa, but the editor points out that -jsa has been "almost deleted" to correct it into the form given by the other manuscripts). Possibly it was felt to be too much for 3-isa kac-isa, a frozen expression, to be separated by a preposition. 18. See Imnaisvili (1957: 632), Kiziria (1963: 266f., 274), Boeder (1987a: 49). In examples such as svil-isa Mariam-isa child-Gen Mary-Gen 'of the child of Mary', -isa on the second constituent is not necessarily an irregular long form for proper names (Lortkipanise 1959: 389) but may be due to the haplological simplification oiMariam-

Suffixaufnahme

in Kartvelian

209

is-isa, where -is is the expected short form with proper names (see 2.4) and -isa is due to Suffixaufnahme. 19. Seeming exceptions are nominative and vocative plural (see (17)), and the "modern" paradigm with the plural marker -eb (5.2). 20. See Imnaisvili (1957: 597-615), Kiziria (1963: 266f., 275-277, 280f.), and Dondua (1931) for further examples. 21. According to Imnaisvili (1957: 591), -isa in svil-ad Abraham-isa (Mt 3,9) childAdv Abraham-Gen '[as] children to Abraham' is to be interpreted as -is Gen + a Adv, and Abraham-is is indeed the correct genitive form for a proper name. Admittedly, there is a variant -a of -(a)d (Imnaisvili 1957: 41f.), and most examples are of the genitive plus adverbial type in proper names and pronouns (Imnaisvili 1957: 409-411, Kiziria 1963: 292f.); but the long form -isa has a tendency to occur in Suffixaufnahme where -is is expected (5.1). Thus, Abraham-isa and ymrt-isa (62b/c) are ambiguous, and -isa could be Suffixaufnahme with zero instead of -(a)d. 22. Cf. Boeder (1987a: 33-34, 38) and Kartozia (1968: 162). Unfortunately, the evidence for the type of agreement in (62d-f) is very scanty so far; see Imnaisvili (1957: 590-595). Compare also the refusal of the directional case to undergo Suffixaufnahme in Mingrelian (Kartozia 1968: 164); see (90b) below. For irregularities of Suffixaufnahme in general see Imnaisvili (1957: 594, 600-615) and Kiziria (1963: 265-276). 23. See Plank's introduction to this volume. 24. Thus the Modern Georgian adjective yr-it-ad-i 'fundamental' is to be analysed as root-Ins-Adv-Nom, illustrating a kind of word formation that occurs rather early (gulit-adikmodet (Kol 3,23) heart-Ins-Adv you=should=do=it 'do it heartily', Sam'se 1976: 50 §65); this type seems to have flourished later in the philosophical literature of the high Middle Ages and abounds in contemporary scientific and technical terminology. The genitive itself is not used derivationally. On the other hand, genitive -is, -ta and their counterparts in the other Kartvelian languages do occur in thousands of toponyms, such as the city name Tbil-is-i warm-Gen(?)-Nom or the village name Cxikv-ta jayPl(Gen?)(Nom) (Sam'se 1973: §155), and the possibility that such uses reflect a prehistoric stage where the genitive was derivational, having collective or similar meaning, has been much debated. Alternatively, such toponyms may result from ellipses of heads, implying original Suffixaufnahme: Ru-is-i < daba-j ru-is-i village-Nom creek-Gen-Nom. Notice that the latter interpretation implies Suffixaufnahme, while it is perfectly possible to posit a reversed order ru-is daba-j with simple ellipsis of the stem. Also, as Dondua (1930 [1975: 30]) rightly points out, the expected form is not the short genitive ru-is but ru-isa, agreeing with daba-j in [case] (see 2.3, 5.1). 25. Recall also the "deviant" pronoun vinme (3.2.2). 26. Deleted or abstract noun stems also play an important role in the development of several forms, productive in literary Modern Georgian, that are described concisely by Jedlicka (1963) under the rubric of "derivational declension"—such as semosvl-isa-s entering-Gen-Dat 'when entering', somehow related to semosvl-is dro-s entering-Gen time-Dat 'at the time of entering' (see also Patsch 1964: 143). Georgian dialects offer further examples of multiple case marking that might have to do with ellipsis, possibly involving a non-Georgian (Vainakh?) substratum influence; cf. for instance Khevsurian 3ut-isa-s Djuti-Gen-Dat 'in(to) (the village of) Djuti', ak-isa-s here-Gen-Dat 'here, hither', cixe-eb-si-it fortress-P1-in-Ins(=Ablative) 'from out of the fortresses' (Cincarauli 1960: 48-51). 27. Note the morphological difference between axal(i) and axl-: since the latter has syncope of -a- before -isa, ellipsis applies before the appropriate morphological form is chosen.

210

The Caucasus

28. Or more likely amxanag-isa cem-isa saxl-sa, with postposed possessive pronoun; see 7.2. 29. As will be seen presently, examples like these are ambiguous: en-isa could be interpreted as an example of Suffixaufnahme with zero nominative suffix. 30. Patsch (1964: 135) also mentions a contrastive meaning in another example of this kind: saSual zmna-s porma mokmedebit-isa akvs, punkcia — vnebit-isa middle verb-Dat form (Nom) active-Gen has, function (Nom) passive-Gen 'Middle verbs have an active form but a passive function'

31. Gigineisvili's examples from Rustaveli are not very convincing, though. 32. Notice that kal-isa-sa is homonymous with the dative form (woman-Gen-Dat) and kal-isa-ta with the plural oblique (woman-Gen-Pl). 33. In Mingrelian the ergative is used with both transitive and intransitive verbs in the aorist and thus also encodes tense. 34. See Cikobava (1936: I, 71) for Laz analogues. 35. Cf. Kiziria (1982: 220), and see Kartozia (1968: 161) for more Mingrelian and Laz examples with adjective modifiers. 36. This example is probably from the Lashkh dialect. Dondua actually writes Goginos-d-d, but his explanation of this form is opaque. 37. According to Aleksandre Oniani, to whom I owe the Georgian translation of this example, this is the most natural form in Svan, more natural than the one with the head noun repeated with the second conjunct. 38. I owe the Georgian translations of (94d-g) and (95a) to Ciuri Gabliani, Mestia. 39. For a close parallel in Vedic Indian see Boeder (1980). 40. Similarly, Patsch (1964: 134, 144) suggests that Suffixaufnahme, representing a "bracketing principle," effects a "closure of the nominal phrase;" and Schmidt (1959: 19) sees it as a "means to achieve unity of the nominal complex." 41. I owe this characterization of "descriptiveness" to Frans Plank. 42. Wilhelm (1983: 102, and in this volume) recognizes a similar distinction between feste Genitivverbindungen without Suffixaufnahme and less close genitive constructions with Suffixaufnahme in Hurrian. 43. See, for instance, Paul (1920: §211) on Gliederungsverschiebung. 44. This is probably why Sanise (1973: §157) considers the etymologically genitive suffixes -is and -t to have become derivational in possessive pronouns. 45. The modern Kartvelian languages have to some extent retained postnominal possessives; see (83) and (86). 46. This is a phenomenon known from other cases of word order change. See Hyman (1975). 47. See further Schmidt (1959) and Boeder (1987a: 56). 48. See further, also for these and other examples, Schulze (1892: 440 note), Krause (1924: 247), Gonda (1959: 38), Liberman (1990).

Sources of Examples Abo

lovane Sabanis3e, Camebaj cmidisa mocamisa Habojsi [The martyrdom of St. Abo of Tbilisi]. Kartuli enis istoriuli krestpmatia I, ed. by Ivane Imnaisvili. Tbilisi 1953.

Suffixaufnahme Act C Cartqlerus Chrest Ciletr D Dav

E GM Imn J Kal Kim Kiz Kol L Mk Mt Neh Onian OtxBol Otxsvel Parx Phys PO Rg Sap SinMr Sos Sus SvP SvPr Tt

in Kartvelian

211

Acts of the Apostles. Adish manuscript of the Gospels (897 CE) in Otxsvel. Cartquenvaj lerusalemisaj. Chrestomathy of Mingrelian texts in QipsiSe 1914. Cil-etratis iadgari, ed. by Akaki Sani3e et al. Tbilisi 1977. 3ruci manuscript of the Gospels (936 CE) in Otxgvel. Aleksi Davitiani, Svanuri andazebi [Svan proverbs], ed. by Akaki Sani3e and Maksime Kaldani. (Masalebi Kartvelur enata sescavlisatvis 5.) Tbilisi: Mecniereba 1977 (Sakartvelos SSR Mecnierebata akademia, Enatmecnierebis instituti). Parxali manuscript of the Gospels (973 CE) in Otxsvel. Georgij Mercul' [Giorgi Mercule], Zitie sv. Grigorija Chandzt'ijskago. Gruzinskij tekst. Vvedenie, izdanie, perevod Nikolaj Marra [Giorgi Merchuli: Life of Grigol of Khandzta]. St. Petersburg 1911. Imnaisvili 1957. Gospel according to St. John. KalaSe 1961. Kimeni, ed. by Korneli KekeliSe. Tbilisi 1918/1946, vols. I/II. Kiziria 1963. Epistle to the Colossians. Gospel according to St. Luke. Gospel according to St. Mark. Gospel according to St. Matthew. Nehemia. Arsen" Onian, Lusnu ambwar L lasxu sumi ninsw [Svan texts in the Lashkh dialect]. Petrograd 1917. Kartuli otxtavis on bolo redakcia [The last (Athonite) recensions of the Georgian Gospels], ed. by Ivane Imnaisvili, Tbilisi 1979. Kartuli otxtavis ori 3veli redakciis sami Satberduli xelnaceris mixedvit [Two old recensions of the Georgian Gospels according to three Shatberd manuscripts], ed. by Akaki Sani3e. Tbilisi 1945. Parxlis mravaltavi. Fiziolog. Armjano-gruzinskij izvod. Gruzinskij i armjanskij teksty, issledoval' izdal' i pereveP Nikolaj Marr [Old Georgian and Armenian translation of the Physiologus]. St. Petersburg 1904. Patrologia Orientalis 19, ed. by Nikolaj Marr, 1926. Kings. The Wisdom of Solomon. Sinuri mravaltavi 864 clisa [Old Georgian polykephalaion from Mount Sinai], ed. by Akaki Sani3e et al. Tbilisi 1959. Kartuli carcerebis korpusi I: Lapidaruli carcerebi I: Aymosavlet da Samxret Sakartvelo (V-X ss.) [Corpus of Georgian inscriptions on stone: East and South Georgia], ed. by N. Sosiasvili. Tbilisi 1980. Martvilobaj Susanikisi [The Martydom of Shushanik], ed. by I. Abula3e. Tbilisi 1938. Svanuripoezia I: Simyerebi [Poesie svane I: Chansons], ed. by Akaki Sani3e et al. Tbilisi 1939. Svanuri prozauli tekstebi I: Balszemouri kilo [Svan texts in the Upper Bal dialect], ed. by Akaki Sani3e and Varlam Topuria. Tbilisi 1939. Epistle to Titus.

212 Xanm Xub

The Caucasus Xanmeti tekstebi I [Khanmeti texts], ed. by KaSaia. Tbilisi 1984. Makar Xubua, Megruli tekstebi [Textes megreliens]. Tbilisi 1937.

References Abesase, Nia. 1975. Brunvata sintaksuri punkciebi Svanursi Kartultan mimartebit [The syntactic functions of cases in Svan in comparison with Georgian]. Tbilisis universitetis sromebi 164, 5-13. Abesase, Nia. 1984. Gansazyvreba Svanursi [Modifiers in Svan]. Tbilisis universitetis sromebi 245, 5-21. [Antoni I Katolikosi.] 1885. Kartuli yrammatika sedgenili Anton I-is mier [Georgian grammar by Anton I]. Tbilisi: Ekvtime Xelasis stamba. Boeder, Winfried. 1976. Morphologische Kategorien. In Grammatik. Akten des 10. Linguistischen Kolloquiums Tubingen 1975, vol. 2, ed. by Kurt Braunmiiller and Wilfried Kiirschner, 117-126. Tubingen: Niemeyer. Boeder, Winfried. 1979. Ergative syntax in language change: The South Caucasian languages. In Ergativity: Towards a Theory of Grammatical Relations, ed. by Frans Plank, 435-480. London: Academic Press. Boeder, Winfried. 1980. Zur Rekonstruktion von Infinitivkonstruktionen im Indogermanischen. In Linguistic Reconstruction and Indo-European Syntax, ed. by Paolo Ramat et al., 207-224. Amsterdam: Benjamins. Boeder, Winfried. 1983. 'Und' in den siidkaukasischen Sprachen I: Einige Verwendungen von da 'und' in der altgeorgischen Literatursprache. Folia Linguistica 17, 287-326. Boeder, Winfried. 1987a. Einfachheit und Komplexitat in der Geschichte der Kartvelsprachen. Iberiul-kavkasiuri enatmecnierebis celicdeuli (Annual of IberoCaucasian Linguistics, Tbilisi) 14, 23-64. Boeder, Winfried. 1987b. Punctuation, segmentation and syntactic structure in Old Georgian. Paper read in the Linguistic Institute and in the Institute of Manuscripts of the Academy of Sciences of the Georgian SSR. Boeder, Winfried. 1989a. Verbal person marking, noun phrase and word order in Georgian. In Configurationality: The Typology of Asymmetries, ed. by Laszlo Maracz and Pieter Muysken, 159-184. Dordrecht: Foris. Boeder, Winfried. 1989b. Zur Typologie der Satzverkniipfung in den kaukasischen Sprachen. Iberiul-kavkasiuri enatmecnierebis celicdeuli 16, 67-87. Bopp, Franz. 1846. Uber das Georgische in sprachverwandtschaftlicher Beziehung. Abhandlungen der Koniglich-PreufSischen Akademie der Wissenschaften zu Berlin. [Reprinted in Bopp, Kleine Schriften zur vergleichenden Sprachwissenschaft: Gesammelte Berliner Akademieabhandlungen 1824-1854, 397-477. Leipzig 1972.] Bork, Ferdinand. 1913. Kaukasisches (Nachtrag zu OLZ 1905 Sp. 184ff.). Orientalistische Literaturzeitung 16, 385-387. Cxubianisvili, Darejan. 1972. Inpinitivis sakitxisatvis 3vel Kartulsi [On the question of the infinitive in Old Georgian]. Tbilisi: Mecniereba (Sakartvelos SSR Mecnierebata akademia, Enatmecnierebis instituti). Cikobava, Arnold. 1936. Canurisgramatikulianalizi, tekstebiturt [Grammatical analysis of Tchan (Laz), with texts]. Tbilisi: SSRK Mecnierebata akademiis Sakartvelos pilialis gamocema (Caucasus Polyglottus II; SSRK, Mecnierebata akademiis Sakartvelos piliali).

Suffixaufnahme

in Kartvelian

213

Cikobava, Arnold. 1950. Kartuli enis zogadi daxasiateba [General characterization of Georgian]. In Kartuli enis ganmartebiti leksikoni I, 018-053. Tbilisi: Sakartvelos SSR Mecnierebata akademiis gamomcemloba. Cincarauli, Aleksi. 1960. Xevsurulis taviseburebani, tekstebita da indeksit [Characteristics of Khevsurian with texts and an index]. Tbilisi: Tbilisis universitetis gamomcemloba (Sakartvelos SSR Mecnierebata akademia, Enatmecnierebis instituti). Deeters, Gerhard. 1926/1927. Armenisch und Siidkaukasisch: Bin Beitrag zur Frage der Sprachmischung. Caucasica 3, 37-82; 4, 1-64. Dench, Alan, and Nicholas Evans. 1988. Multiple case-marking in Australian languages. Australian Journal of Lingusitics 8, 1-47. Dondua, Karpez. 1930. K voprosu o roditel'nom emfaticeskom v drevneliteraturnom gruzinskom jazyke [On the question of the emphatic genitive in the Old Georgian literary language]. Izvestija AN SSSR, Otdelenie gumanitarnix nauk (Leningrad) 1930, 3, 195-209. [=Dondua 1975, 21-33.] Dondua, Karpez. 1931. Ob agljutinativnom xaraktere gruzinskogo sklonenija. DANSSSR B 1931, 4, 63-68. [=Dondua 1975, 34-38.] Dondua, Karpez. 1975. Stat'i po obscemu i kavkazskomu jazykoznaniju. Leningrad: Nauka. (ANSSSR, Naucnyj sovet po teorii sovetskogo jazykoznanija pri otdelenii literatury i jazyka). Finck, Franz Nikolaus. 1910. Die Haupttypen des Sprachbaus. Leipzig: Teubner. Fraenkel, Ernst. 1950. Die baltischen Sprachen, ihre Beziehungen zu einander und zu den indogermanischen Schwesteridiomen als Einfuhrung in die baltische Sprachwissenschaft. Heidelberg: Winter. Gabliani, Egnate. 1925. 3vel da axali Svaneti [Old and Modern Svanetia]. Tbilisi: SSSR saxelmcipo gamomcemloba. Gigineisvili, Ivane. 1984. Postpoziciuri martuli msazyvreli Axali Kartuli leksis enasi [The postposed governed determinans in Modern Georgian poetry]. Kartuli enis kulturis sakitxebi 6, 40-63. Gonda, Jan. 1959. On amplified sentences and similar structures in the Veda. In Four Studies in the Language of the Veda, by Jan Gonda, 7-70. (Disputationes Rheno-Trajectanae 2.) The Hague: Mouton. Greenberg, Joseph. 1966. Language Universals, with Special Reference to Feature Composition. The Hague: Mouton. Hyman, Larry. 1975. The change from SOV to SVO. In Word Order and Word Order Change, ed. by Charles N. Li, 113-147. Austin: University of Texas Press. Imnaisvili, Ivane. 1955. Nacevari 3vel Kartulsi [The article in Old Georgian]. Tbilisis universitetis sromebi 61, 249-276. Imnaisvili, Ivane. 1957. Saxelta bruneba da brunvata punkciebi 3vel Kartulsi [Noun inflection and functions of cases in Old Georgian]. (3veli Kartuli enis katedris sromebi 4.) Tbilisi: Tbilisis saxelmcipo universitetis gamomcemloba. Imnaisvili, Ivane. 1975. Sinuri mravaltavi: Gamokvleva da leksikoni [The Mravaltavi manuscript from Mount Sinai: Investigation and lexicon]. (3veli Kartuli enis katedris sromebi 17.) Tbilisi: Tbilisis saxelmcipo universitetis gamomcemloba. Jedlicka, Jaromir. 1963. Derivative Deklination im Georgischen. Bedi Kartlisa 15-16, 103-106. Kaxase, Vaxtang. 1969. 2-dan 5-clamde bavsvis metqveleba da misi ganvitarebis gzebi [The speech of a child from 2 to 5 and the lines of its development], Tbilisi: Ganatleba. Kalase, Cira. 1961. Gansazyvreba 3vel Kartulsi [Determination (attribution) in Old Georgian]. Kartveluri enata strukturis sakitxebi 2, 73-122.

214

The Caucasus

Kaldani, Maksime. 1956. Svanuri enis Laxamuluri kilokavis gramatikuli taviseburebani [Grammatical peculiarities of the Svan dialect of Lakhamula]. Iberiul-kavkasiuri enatmecniereba 8, 161-177. Kartozia, Guram. 1968. Saxeluri sesitqveba Megrul-Lazursi [The nominal group in Mingrelian and Laz]. Macne 1968, 6,147-166. Kiziria, Anton. 1946. Nanatesaobitari mzazyvrelis urtiertobisatvis sazvyrultan (3vel Kartulsi) [The relationship between degenitive determinans and determinatum]. Iberiul-kavkasiuri enatmecniereba 1, 329-336. Kiziria, Anton. 1963. Martivi cinadadebis sedgeniloba 3vel Kartulsi [The composition of simple sentences in Old Georgian]. Tbilisi: Mecniereba (Sakartvelos SSR Mecnierebata akademia, Enatmecnierebis instituti). Kiziria, Anton. 1982. Martivi cinadadebis sedgeniloba Kartvelur enebsi [The composition of simple sentences in the Kartvelian languages]. Tbilisi: Mecniereba (Sakartvelos SSR Mecnierebata akademia, Enatmecnierebis instituti). Klavans, Judith L. 1985. The independence of syntax and phonology in cliticization. Language 61, 95--120. Klimov, Georgij A. 1962. Sklonenie v kartvel'skix jazykax v sravnitel'no-istoriceskom aspekte [Noun inflection in the Kartvelian languages under a comparative and historical aspect]. Moscow: Izdatel'stvo Akademii nauk SSSR. Krause, Wolfgang. 1924. Die Entwickelung einer alien elliptischen Konstruktion in den indogermanischen Sprachen. Zeitschrift fur vergleichende Sprachforschung 52, 223-249. Kurylowicz, Jerzy. 1972. The role of deictic elements in linguistic evolution. Semiotica 5, 174-183. Kvacase, Leo. 1977. Tanamedrove Kartuli enis sintaksi [Syntax of Modern Georgian]. Tbilisi: Ganatleba, 2d ed. Lafitte, Pierre. 1979. Grammaire basque (Navarro-Labourdin litteraire). Donostia (San Sebastian): Elkar, 3d ed. Lausberg, Heinrich. 1963. Elemente der literarischen Rhetorik. Munich: Hueber, 2d ed. Liberman, Anatoly. 1990. 'Afterthought' as a feature of Old Icelandic syntax. In Syntax gesprochener Sprache, ed. by Brigitte K. Halford and Herbert Pilch, 4560. Tubingen: Narr. Lomtase, Elizbar. 1954. Msazyvrel-sazyvrulis urtiertoba Megrulsi [The relationship between determinans and determinatum in Mingrelian]. Iberiul-kavkasiuri enatmecniereba 6, 207-242. Lortkipanise, Ketevan. 1959. Msazyvrel-sazyvrulis urtiertobis istoriidan [A chapter from the history of the relationship between determinans and determinatum]. Tbilisis Saxelmcipo pedagogiuri institutis sromebi 13, 389-395. Marr, Nikolaj Jakovlevic. 1907. Dejanija trex svjatyx bliznecov mucenikov Spevsipa, Elasipa i Melasipa [The deeds of the triplet martyrs Speusippus, Elasippus and Melasippus]. Zapiski Vostocnogo Otdelenija Imperatorskogo Russkogo Arxeologiceskogo Obscestva 17, 284-344. Marr, Nikolaj Jakovlevic. 1925. Grammatika drevneliteraturnogo gruzinskogo jazyka. (Materialy po jafeticeskomu jazykoznaniju 12.) Leningrad. Mel'cuk, Igor. 1986. Toward a definition of case. In Case in Slavic, ed. by Richard D. Brecht and James S. Levine, 35-85. Columbus, Ohio: Slavica. Patsch, Gertrud. 1964. Zur Frage der doppelten Relation im Georgischen. Bedi Kartlisa 17-18, 132-145. Paul, Hermann. 1920. Prinzipien der Sprachgeschichte. Halle: Niemeyer.

Suffixaufnahme

in Kartvelian

215

Plank, Frans. 1990. Suffix copying as a mirror-image phenomenon. Linguistics 28, 1039-1045. Plank, Frans. 1992. Advantage Albanian: Grouping in multiple attribution. EUROTYP Working Papers VII/17. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik. 1985. A Comprehensive Grammar of the English Language. London: Longman. [Qipsise, loseb] I. Kipsidze. 1914. Grammatika mingrel'skogo (iverskogo) jazyka s xrestomatieju i slovaren [Grammar of the Mingrelian (Iberian) language with a chrestomathy and a dictionary]. (Materialy po jafeticeskomu jazykoznaniju 7.) St. Petersburg. Qipsise, loseb. 1939. Canuri tekstebi [Laz texts], Arnold Cikobavas redakciit. Tbilisi: SSRK Mecnierebata akademiis Sakartvelos pilialis gamomcemloba. Sanise, Akaki. 1973. Kartuli enis grama(ikis sapuzvlebi I: Morpologia [Fundamentals of Georgian grammar I: Morphology]. (3veli Kartuli enis katedris sromebi 15.) Tbilisi: Tbilisis universitetis gamomcemloba. Sanise, Akaki. 1976. 3veli Kartuli enis gramatika [Old Georgian grammar]. (3veli Kartuli enis katedris sromebi 18.) Tbilisi: Tbilisis universitetis gamomcemloba. Sarasenise, Tinatin. 1942. -ta supiksiani mravlobiti mokmedebitisa da vitarebitis brunvebsi [Plurals with -ta in the instrumental and adverbial case]. Tbilisis universitetis sromebi 24, 129-144. Sarjvelase, Zurab. 1984. Kartuli saliteraturo enis istoriis sesavali [Introduction to the history of literary Georgian]. Tbilisi: Ganatleba. Schiefner, Franz Anton von. 1859. Versuch uber die Thusch-Sprache oder khistische Mundart in Tuschetien. Memoires de I'Academie imperiale des sciences de SaintPetersbourg VI. Sciences politiques 9 et dernier, 1-160. Schmidt, Karl Horst. 1959. Zur Komplexion attributiv bestimmter Nomina und zur Frage der 'bestimmten Adjektive'. Munchener Studien zur Sprachwissenschaft 14, 13-22. Schulze, Wilhelm. 1982. Quaestiones epicae. Giitersloh: Bertelsmann. Topuria, Varlam. 1936. Nanatesaobiti msazyvrelis setanxmeba sazyvrultan (cigns amxanagisas tu cigns amxanagisa?) [The agreement of genitive determinants with their determinatum]. Saliteraturo Kartulis normebil. Moxsenebata tezisebi, 22-24. Tbilisi. [= Tanamedrove Kartuli saliteraturo enis normebi. Pirveli krebuli, ed. by Ivane Gigineisvili, 135-138. Tbilisi: Mecniereba (Sakartvelos SSR Mecnierebata akademia, Enatmecnierebis instituti), 1970.] Taylor, Ann. 1988. The use of clitics as a diagnostic of phrase structure. Proceedings of the Eastern States Conference on Linguistics (ESCOL) 5, 465-476. Vogt, Hans. 1947. Le systeme des cas en georgien ancien. Norsk Tidsskrift for Sprogvidenskap 14, 98-140. Wilhelm, Gernot. 1983. Der hurritische Ablativ-Instrumental /ne/. Zeitschrift fur Assyriologie und Vorderasiatische Archaologie 73, 96-113.

5 Direct-Oblique Agreement of Attributes in Daghestanian AJeksandr E. Kibrik

1. Introduction Although Daghestanian languages are not normally included among those showing Suffixaufnahme, the encoding of attributive constructions in some of them bears a striking resemblance to this pattern insofar as markers of nominal attributes are found to vary with the case of the head (as was noted in Plank 1990: 1040 for Tsakhur). The comparative examination of these peculiarities across relevant Daghestanian languages is intended to bring out their similarities to, and differences from, prototypical Suffixaufnahme, and thus to contribute to a more comprehensive understanding of agreement possibilities in the NP and of multiple case marking. The thirty or so Daghestanian (or North-East Caucasian) languages, genetically related to the Nakh (or North-Central Caucasian) family, are spoken by about 1.7 million speakers in the relatively small area of the Daghestanian highlands and also in Georgia and northern Azerbaijan. Their rich morphology is predominantly agglutinative. Among their morphological categories are case (forming unusually elaborate systems), noun classes (also involved in agreement), and, in the verbal domain, aspect and tense. As to relational clause patterns, they are well-known representatives of the ergative type. Word order is relatively free, although there is a preference for dependent elements to precede their heads. The consonant systems of Daghestanian languages are typically characterized by pharyngealization (here indicated by the symbol /) and labialization (indicated by a subscript circle after the consonant), by an intensiveness opposition (intensiveness being indicated by a macron over the consonant), by ejectives (e.g. t', k') and laterals (affricate L and fricative f). Among the features of their vowel systems are pharyngealization, nasalization, and a length opposition (length being indicated by a colon).

Direct-Oblique Agreement of Attributes in Daghestanian

217

2. Kinds of Attributes: Genitive vs. Adjective To begin with, it is necessary to have criteria for determining, in each individual language, whether attributes are to be categorized as nominals in the genitive case or as derived adjectives. This distinction can be drawn on the basis of criteria such as the following. A. Are there different markers for nominal and adjectival attributes? There are, for example, in Archi (on which see Kibrik 1977): (1) a. ac'i —> ac'i-li-n (sob) disease (Nom) disease-Obi-Gen (end) 'the end of the disease' b. mu beauty

mu-tu-t (sob) beautiful-Adjct-IV (end) 'beautiful end'

The suffix -n is a member of the nominal case paradigm (cf. ac'i-li-s with final dative suffix); oblique case markers are usually preceded by the marker of an oblique stem (Obi). If the attributive position is filled by words for states, qualities, or actions, the attributive marker is -tu (glossed as 'Adjct'), followed by a class-number agreement marker. Such data thus demonstrate that Archi distinguishes two kinds of attributes—nominals (in the genitive case) and adjectives (or participles). B. May genitive and adjectivalizing markers co-occur in a single word form? They may in Archi: (2) ac'i-li-n-nu-t lo disease-Obi-Gen-Adjct-IV child 'unhealthy child' The possibility of deriving adjectives from nouns in the genitive (with -tu assimilated to -nu) is additional evidence for the distinction of two kinds of attributes in languages such as Archi. C. Are there different sets of grammatical categories for nominal and adjectival attributes? There are such different sets, for example, in Bezhta (on which see Madieva 1965, Kibrik and Testelec 1994) insofar as adjectives agree in number with their head nouns, whereas nominal attributes do not:

218

The Caucasus

(3) a. Haldij-obiLo white-Adj house (NomSg) 'white house' b. Haldij-ar-6 biLo-da white-P1-Adj house-Pi 'white houses' (4) a. abo-s is father-Gen brother (NomSg) 'father's brother' b. abo-s is-na father-Gen brother-NomP1 'father's brothers'

The function of the suffix -o glossed as 'Adj' will be explained presently. D. May nominal and adjectival attributes have different sets of syntactic functions? For Lak, for example, the answer is positive, because nominals in the genitive may occur as clause-actants in the ergative, but adjectives may not (cf. Murkelinskij 1967: 492). (5) a. buta-1 qata father-Gen/Erg house 'father's house' b. butu-1 ninu duruclaj dur father-Gen/Erg mother (Nom) take care 'Father takes care of mother'

There are Daghestanian languages where it is difficult to distinguish genitives and adjectives on these criteria. Rutul, exemplified in (6), is one of them (pace Ibragimov 1978). (6) bilax-a-d xad t'amir-di i spring-Obi-Adj ct/Gen water (Nom) clean-Adj ct be 'The water of spring is clean'

As to criterion A, the marker -dl-di (postvocalic/postconsonantal) is the same for derived adjectives (regardless of whether they are used attributively or predicatively, as is seen in (6)) and for the attributive form of nouns. As to criterion B, it is impossible to add adjectivalizing -dl-di to nouns that already have the suffix -dl-di (*bilax-a-d-di). As to criteria C and D, there are no different sets of categories and functions for adjectives and attributive nouns. The interpretation of the attributive form of nouns in Rutul as a derived adjective rather than as a noun in the genitive case, based on this evidence, is

Direct-Oblique Agreement of Attributes in Daghestanian

219

supported indirectly by the ability of the attributive marker in question, -dl-di, to be added to a dative case: (7) Xidi-ma-Si-s-di Xal relative-Pl-OblPl-Dat-Adjct room 'The room (designed) for relatives'

It would hardly be plausible to analyse such a post-dative marker as a genitive. Subsequently the term "genitive" will only be used for forms distinguishable by the above criteria and qualifying as members of case paradigms rather than as derived adjectives. The following chapter by Boguslavskaja offers a more extensive discussion of the distinction between nominal and adjectival attributes in Daghestanian. 3. Direct vs. Oblique in Tsez Languages The allomorphy of the genitive is among the features usually mentioned in descriptive grammars of Tsez (or Dido) languages (such as Bokarev 1959 and Madieva 1965), and attention has also been drawn to it in Dirr's (1928: 318ff.) comparative survey of Caucasian. It is this distinction of so-called genitive I and genitive II that is reminiscent of the Suffixaufnahme pattern, insofar as the marking of the attributive noun reflects a case contrast of the head: one genitive form is used when the head noun is nominative, the other when the head noun is in any oblique case. These two genitives, as found in Bezhta, Tsez, Khvarsh, and Hinugh, will be referred to as "direct" and "oblique" respectively (GenDct and Gen°bl). Bezhta, Tsez, and Khvarsh also distinguish "direct" and "oblique" adjective forms (AdjDct and Adj°bl) along the same lines. Hunzib is the only member of Tsez to lack the direct-oblique contrast of attributes entirely. Bezhta and Tsez will be examined in more detail in Sections 4 and 5, and similar patterns elsewhere in Daghestanian will be surveyed in Section 6. Two genitives are also documented for the Andi subgroup of Daghestanian (a co-member with Tsez of the Avar-Andi-Tsez branch), but their distribution is different. It depends crucially on the class membership of the attributive nouns themselves, with masculine, or class I, nouns having genitive alternants that agree with the head noun in class, and with nouns of other classes having an invariable genitive. 4. Bezhta 4.1. Two Genitives As is exemplified in (8) and (9), Bezhta has two declensions, one for one-stem nouns such as abo, whose stem is invariable, and the other for two-stem nouns

220

The Caucasus

such as is, which use the basic stem for the nominative (direct) case and a derived (oblique) stem for the oblique cases (cf. Kibrik 1991). (8)

a. abo-s is father-GenDct brother (Nom) 'father's brother' b. abo-la is-t'i-1 father-Genobl brother-Obl-Dat 'to father's brother'

(9)

a. is-t'i-s biLo brother-Obi-Gen0" house (Nom) '(the) brother's house' b. is-t'i-la biLo-? brother-Obl-Genobl house-In (Ess) 'in (the) brother's house'

(In (9b) the head carries only an orientational suffix since the essive lacks an overt marker.) However, declension type evidently does not influence the attributive marking of nouns: -s appears if heads are nominative, and -la if heads are in an oblique case (such as dative or inessive), regardless of the declension of the genitive noun or its head. The genitive alternants are formally identical to locative case suffixes, with -s expressing both direct genitive and ablative ('movement from') and -la expressing both oblique genitive and translative ('movement along/through'). In Bezhta, as in other Daghestanian languages, these locative cases obligatorily co-occur with markers of orientation, such as -?o 'inside' or -L'a 'on': (10) a. do biLo-?o-s eL'ec I house-In-Abl go (Pres) 'I am going from the house' b. do honol-L'a-la eL'ec I road-Super-Transl go (Pres) 'I am going along the road'

There are contexts where the identity of genitive and locative markers may actually cause indeterminacies: (11)

a. li-?i-s L'alo water-In-GenDct/Ab1 stone (Nom) 'the stone from the water' b. li-?i-la L'al-L'a water-In-Genob1/Transl stone-Super (Ess) 'on the stone from the water'

Direct-Oblique Agreement of Attributes in Daghestanian

221

Owing to the formal identity of direct genitive and ablative, li-? i-s in (11 a) simultaneously permits attributive and locative interpretations. Example (lib), where K-?i-la likewise has attributive as well as locative meaning, is more problematic in regard to the morphological identity of -la. On the attributive analysis, -la is oblique genitive, as it should be since the head is in an oblique case. On the locative analysis, however, there should be an ablative rather than a translative on semantic grounds; and one would have to assume that, in attributive position, the ablative has a direct (-s) and an oblique allomorph (-la, coinciding with translative), just as the genitive does. Apart from being simpler, the genitive interpretation would here seem to be preferable also on the evidence of the genitive's ability to co-occur with other oblique cases too (as shown presently).

4.2. Genitive from Comitative The two genitives can be added to nouns that are already case-marked. It is especially the comitative that permits such double case marking: (12)

a. kid adam-li-Ror-so jeL'a-?a girl (Norn) man-Obl-Com-GenDct leave-Neg The girl (who is) with the man does not leave' b. wahago kid wahalo Yadam-li-Ror-la suk'o-1 jegama'j this girl (Nom) that man-Obl-Com-Genobl person-Dat see (Pres) 'The person (who is) with that man sees this girl'

The genitive is direct in (12a), being the attribute to a nominative head, and oblique in (12b), where the head is dative, the oblique case governed by the verb. (The vowels of the genitive suffixes, appearing only after a consonant in the case of the direct alternant, harmonize with root vowels.)

4.3. Independence of Attribute If its head is missing, the genitive behaves like an independent nominal and may, among other cases, acquire a further genitive marker: (13)

usup-i-la abo-s k'obala biXalo gaxi, ibrahim-i-la-s basijo gaxi Jusup-Obl-Genobl father-GenDct stick (Nom) long be, Ibrahim-Obl-GenoblGenDct short be 'The stick of Jusup's father is long, that of Ibrahim's is short'

As the non-elliptic version of the subject NP of the second part of this sentence (13') shows, the first genitive on Ibrahim in (13) is oblique because its (elided) immediate head, being itself an attribute, would be in the genitive (i.e., in an oblique case); and the second genitive is direct because the (elided) topmost head would be nominative (i.e., in the direct case).

222 (13')

The Caucasus ibrahim-i-la abo-s k'obola Ibrahim-Ob1-Genobl father-GenDct stick (Norn)

Because there is some variation in languages with prototypical Suffixaufnahme on this point (cf. Plank 1990, and various contributions in this volume), it should be noted that in recursive attribution in Bezhta, as in (13) and (13'), it is always only the immediate head that is reflected in the direct or oblique forms of genitives.

4.4. Adjectives As noted above, adjectives in Bezhta also distinguish direct and oblique forms, although the actual markers differ from the nominal ones, with -ol-o as the direct and -al-d (harmonizing with the root vowel) as the oblique adjectival marker: (14)

a. Haldij-obiLo white-AdjDct house (Nom) 'white house' b. Haldij-a biLo-? white-Adjobl house-In (Ess) 'in the white house'

(15)

a. Haldij-ar-o biLo-da white-Pi-AdjDct house-Pi (Nom) 'white houses' b. Haldij-ar-a biLo-da-? white-Pl-Adjobl house-Pl-In (Ess) 'in the white houses'

5. Tsez

5.1. Two Genitives Like Bezhta, Tsez distinguishes a direct (-s) and an oblique (-z) genitive: (16)

a. obi-s esij idu-r ajsi father-GenDct brother (Nom) house-In (Ess) come (Past) 'Father's brother came home' b. obi-z esi-s jol Rlutku father-Genobl brother-GenDct be house (Nom) 'Father's brother owns a house'

Formally the oblique genitives of Tsez and Bezhta have different sources, while the direct ones are evidently identical. Tsez genitives do not coincide

Direct-Oblique Agreement of Attributes in Daghestanian

223

with locatives either, the ablative being -aj, the translative -aza, and the lative -or (see Bokarev 1967: 408f.). As is seen in the possessive (or existential) construction in (16b), the direct genitive functions also at clause-level in Tsez. The oblique genitive likewise extends beyond the attributive sphere, being governed by certain verbs as an oblique object case: (17)

a. obij-a: gulu-z curet' bok'si father-Erg horse-Genobl whip (Nom) hit (Past) 'Father hit the horse with the whip' b. aHmad-a: pat'imat-ez uba bojsi Ahmed-Erg Fatima-Geriobl kiss make (Past) 'Ahmed kissed Fatima'

5.2. Independence of Attribute If the heads of genitival attributes are elided, these are able to represent the entire NP and, in addition to their own genitive, they can attract the appropriate case (and orientation) marking, just as in Bezhta. (18)

a. gere-z tung-a: li riXer Xosi, c'edo-z-a: ril kek Xosi iron-Genobl jug-In (Ess) water (Nom) carry be, earth-Genobl-In (Ess) butter (Nom) churn be 'Water is carried in a jug of iron, butter is churned in one of earth' b. gere-z tung-os hut' jol, c'edo-z-os anu iron-Genobl jug-GenDct spout (Nom) be, earth-Genobl-GenDct not 'The jug of iron has a spout, the one of earth does not'

The oblique genitives in the elliptical NPs in the second conjuncts of (18a) and (18b) add the orientational suffix and the direct genitive, respectively, carried by the heads in their non-elliptical counterparts in the first conjuncts. 5.3.

Adjective

As in Bezhta, adjectives in Tsez also distinguish direct and oblique forms: (19)

a. HonL'o L'iri ec'no Rlutku jol mountain on new (AdjDct) house (Nom) be 'There is a new house on the mountain' b. ze iciX ec'no-zo Rlutk-a: he live new-Adjobl house-In (Ess) 'He lives in a new house'

With head nouns in the direct case, adjectives remain unmarked, while the oblique adjectival suffix -zo resembles that of oblique genitive.

224

The Caucasus

5.4. Participle and Predicate Nominal Attributes of a verbal kind likewise distinguish direct and oblique forms: (20)

a. li HaLuXosi zek'u dej esij jol water (Nom) drink-PrtcplDct man (Nom) my brother (Nom) be 'The man drinking water is my brother' b. li HaLuXo-zo zek'-a: dow-L'-or ozuri kursi water (Nom) drink-Prtcplobl man-Erg you-Super-Lat glance (Nom) cast 'The man drinking water cast a glance at you'

With heads in the direct case such participles take the suffix -si, resembling the direct genitive; with heads in an oblique case they take -zo, coinciding with the oblique alternant of both adjectives and genitives. Predicate nominals in attributive function share the direct and oblique markers of such participles: (21)

a. bikori gamac' joL' jol snake (Nom) stone (Nom) on is 'The snake is on the stone' b. gamac' joL'-si bikori dar Hit bikoada:-si zowsi stone (Nom) on-PrtcplDct snake (Nom) I (Dat) yesterday see-Prtcpl be (Past) 'Yesterday I saw the snake (that was) on the stone'

The predicate nominal of (21a), consisting of a noun in the nominative (gamac) followed by a postposition (JoL'), occurs as an attribute of a noun in the nominative (bikori) in (21b) and hence takes the direct participial marker (-si), suffixed to the postposition. Like genitives, such attributive predicate nominals may function as full NPs when their head is elided, adding the appropriate case markers: (22) azom-L'o-zo bikor-a-s jol zahru, gamac' joL'-zo-s anu tree-Super(Ess)-Prtcplobl snake-Ob1-GenDct be poison (Nom), stone (Nom) onPrtcplobl-GenDct not 'The snake on the tree is poisonous, the one on the stone is not'

The oblique participial suffix on the predicate nominals here is due to their head, elided in the second conjunct, being in the genitive. 6. Other Daghestanian Languages

6.1. Alternants of Attributive Marker Tsakhur, a member of the Lezgian group, resembles closely-related Rutul (see (6) above) in that it lacks a genitive and expresses nominal and adjectival

Direct-Oblique Agreement of Attributes in Daghestanian

225

attributes identically (cf. Ibragimov 1990). What is peculiar to Tsakhur, however, are the factors controlling the choice between the alternants of the attributive marker (glossed as 'AttrDct' and 'AttrObl'). (23) a. jiz-da / xunase-na coz I-AttrDct / woman-AttrDct brother, (Nom) 'my / the woman's brother' b. jiz-de / xunase-ne coz-iqa I-AttrObl / woman-Attr Obl brotherrCom 'with my / the woman's brother' (24) a. jiz-in / xunase-n kitab I-AttrDct / woman-AttrDct bookIV (Nom) 'my / the woman's book' b. jiz-de / xunase-ne kitab-e I-AttrObl / woman-Attr Obl bookIV-In (Ess) 'in my / the woman's book'

The three determining factors, exemplified in these examples, are the final segment of the attribute and the case and the class (indicated by the inferior roman numeral in the gloss) of the head noun. Table 5.1 sets out how these interact. It is the dependence of the attributive marking on the direct or oblique case of the corresponding head noun that Tsakhur thus shares with the Tsez languages. An essentially similar pattern, although only within a limited lexical domain, emerges in Khinalug, remotely related to the Tsez languages and Tsakhur. Here it is only the demonstrative pronouns that, apart from class, distinguish direct and oblique forms (cf. Kibrik et al. 1972), as exemplified in (25) and catalogued in Table 5.2. (25)

a. 3i 3ina this (DemDct, IV) worldlv (Nom) 'this world' b. sa 3in-e dahg-irdir this (DemObl, IV) worldIV-Gen affair-Pi (Nom) 'this world's affairs'

Table 5.1. Choice of Attributive Markers in Tsalchur. Case of head

Nominative Oblique

Class of head

IV I-III

Final segment of attribute sibilant

other

, -da -de

-na -ne

-n

226

The Caucasus Table 5.2. Paradigm of Khinalug Demonstrative Pronouns. Proximal

Direct Oblique

I du

RolRii

II da

III

Distal Si sa

IV

I

hu

II

III

hind

ha

IV

6.2. Attribute from Oblique Case Archi, another member of the Lezgian group, resembles the Tsez languages in that attributes can be formed from nouns in an oblique case: (26)

haltsr-ce-qla-s-du-b c'ele river-Obl-Inter-Abl-Adjct-III stoneIII (Nom) 'the stone from the river'

The adjectivalizing suffix -t u (realized as -du after a consonant) is here added to an oblique-stem noun in a locative case, with the entire attribute agreeing in class with the head.

6.3. Independence of Attribute In Archi, attributes with adjectivalizing suffix may occur independently, and they then inflect like the nominal head of an NP. (27)

haltar-ce-qla-s-du-m-mi-n rang river-Ob1-Inter-Ab1-Adjct-III-Obl-Gen color (Nom) 'the color of the one from the river'

In (27), the attribute of (26) is used as an NP in its own right, with its head elided, and it functions as a nominal attribute, requiring the oblique stem suffix -mi (which assimilates the preceding class marker) and the genitive (not distinguishing direct and oblique in Archi). Kubachi, a remote Dargva dialect, is among the languages where two genitives may come to co-occur within a single word in this manner (cf. Magometov 1963: 109): (28)

a. a:s-la milq'a silver-Gen spoon (Nom) 'silver spoon' b. a:s-la-l-la qulluq silver-Gen-Obi-Gen business (Nom) 'the business of the one of silver'

It is only the oblique stem marker, required by nouns taking an oblique case in Kubachi, that intervenes between the two genitives on the attribute in (28b).

Direct-Oblique Agreement of Attributes in Daghestanian

227

In a special type of such multiple case marking, attested in Archi, locatives and an orientation marker are crucial components. Consider (29), where old, the suppletive genitive of the 1st person plural pronoun nen, acts like an independent nominal: (29)

c'eba olo-ma-si go (Imp) we(Gen)-Orient-All 'Come to our house!'

It is the orientation marker -ma, preceding a locative case suffix, including allative, that adds the meaning component 'typical place for X', where X is the referent of the relevant pronoun or human noun. Whereas locative forms are normally based on the oblique stem, the basic genitive here would seem to suggest an implicit head, 'the one of X'. Considering that the genitive prototypically expresses possessive relations, and that among one's most typical possessions is the place one habitually occupies, the meaning 'place where X lives' is the most obvious one to infer. As to the historical origin of such forms with -ma, it should be noted (cf. Kibrik 1977: 105) that Archi has similar forms of locative gerunds with -ma as a localization marker ('there where') followed by standard locative case markers, as in (30). (30)

was L'am-ma-si uqla you (Dat) like-Orient-All go (Imp) 'Go where you like!'

Nouns usually have no such marker. The word biqj 'place' is one of the few exceptions, having the regular oblique marker -II (biqj-li Erg), but showing a more archaic form in the inessive, biqj-m-a 'in the place', where -m is the oblique marker, -a is a marker of orientation for 'in', and essive remains unmarked. Oblique -mi is regular for nominalized attributes such as demonstrative pronouns (e.g. to-r 'that', i.e. woman, belonging to class II, marked by -r), which, when occurring independently, can be used in oblique cases (e.g. to-r-mi-n that-II-Ob1-Gen 'her'). Thus -ma is not originally simple but consists of an oblique and an orientation marker. Accordingly, gerunds as in (30) can be historically analysed as nominalizations of verbs (L'an 'to love, like'), declined with the help of the oblique marker -m(i), the orientation marker -a, and case endings. The history of the oblique, orientation, and case markers of personal pronouns, as in (29), would seem to be the same. What remains to be explained here is the genitive of the pronoun. Lacking a verb 'to have', Archi puts the possessor in the genitive in clauses with an existential verb, from which may be derived a locative gerund by means of -ma. Inflected genitives such as olomasi would then be explicable as historical reductions of full clauses such as (31), with the noun redundantly identifying the possession and the existential verb omitted.

228 (31)

The Caucasus olo noL' i-ma-si we (Gen) house be-Orient-All 'toward where we have a house, where our house is'

7. Conclusions It has been shown here that some Daghestanian languages satisfy the usual criterion of Suffixaufnahme insofar as nominal attributes mark both their own attributive relation (by virtue of being in the genitive or of carrying some more general attributive marker) and the external syntactic relation of their head (by virtue of selecting the direct or oblique alternant of the genitive or other attributive marker). However, whereas in the prototypical instances of Suffixaufnahme this dual marking on nominal attributes is distributed over two co-occurring affixes, the Daghestanian variation on this theme cumulates the case-marking and agreement-marking functions. Although nominal attributes can in principle be distinguished from adjectival/participial ones in at least some of the relevant Daghestanian languages, the frequent extension of the direct-oblique alternation beyond what can strictly be called nominal genitives suggests that a special affinity between nominal and adjectival attributes might be a general prerequisite of Suffix aufnahme and similar patterns. What the comparative Daghestanian evidence demonstrates, finally, is that multiple case marking as such may occur without anything resembling Suffixaufnahme being used in a language. The genitive is especially prone to be the first case in doubly case-marked expressions, owing to the potential of attributes to act as independent nominals, with heads being usually recoverable from the context. Although double case marking thus does not imply Suffixaufnahme, Suffixaufnahme does imply the possibility of genitives acting as independent nominals and being case-marked again, so far as Daghestanian is concerned. The languages showing Suffixaufnahme of a kind are a minority within the thirty or so members of Daghestanian. It is natural to assume that this particular technique of marking was developed independently in the relevant subset of this genetic group. Nonetheless, its appearance would not seem to have been accidental, but rather to have been motivated by the fundamental opposition between direct case (nominative) and oblique cases, an opposition reasserting itself in the direct-oblique alternation of attributive markers. The two-stem declension, distinguishing a direct and an oblique stem, as now found in most Daghestanian languages (cf. Kibrik 1991), is an archaic trait of noun paradigms, and it may in fact be the trace of a case system with merely two members in the proto-language. On the assumption of such a twocase system, the genitive alternation may have come about through genuine repetition of the head's class (known to be an old agreement category) and case marking on the attribute, or perhaps also through "free marking" (in Moravcsik's sense, this volume) of the whole NP's external function on any of

Direct-Oblique Agreement of Attributes in Daghestanian

229

its members. Tsakhur (Section 6.1) is especially telling in this respect, since its attributive marker still reflects both the class and the case of the head.

Acknowledgment I am grateful to Frans Plank for editorial help and advice, and for having suggested the topic in the first place.

References Bokarev, Evgenij A. 1959. Cezskie (didojskie) jazyki Dagestana. Moscow: Nauka. Bokarev, Evgenij A. 1967. Cezskij jazyk. In Jazyki narodov SSSR, ed. by V. V. Vinogradov et al., vol. 4, 404-420. Moscow and Leningrad: Nauka. Dirr, Adolf. 1928. Einfuhrung in das Studium der kaukasischen Sprachen. Leipzig: Asia Major. Ibragimov, Garun X. 1978. Rutul'skij jazyk. Moscow: Nauka. Ibragimov, Garun X. 1990. Caxurskij jazyk. Moscow: Nauka. Kibrik, Aleksandr E. 1977. Opyt strukturnogo opisanija arcinskogo jazyka, vol. 2, Taksonomiceskaja grammatika. Moscow: Nauka. Kibrik, Aleksandr E. 1991. Organising principles for nominal paradigms in Daghestanian languages: Comparative and typological observations. In Paradigms: The Economy of Inflection, ed. by Frans Plank, 255-274. Berlin: Mouton de Gruyter. Kibrik, Aleksandr E., Sandro V. Kodzasov, and Irina P. Olov'annikova. 1972. Fragmenty grammatiki xinalugskogo jazyka. Moscow: Izdatel'stvo Moskovskogo universiteta. Kibrik, Aleksandr E., and Jakov G. Testelec. 1994. The Bezhta language. In Indigenous Languages of the Caucasus, ed. by Rick Smeets. Delmar, N.Y.: Caravan. Madieva, Gjul'zaxan I. 1965. Grammaticeskij ocerk beztinskogo jazyka. Maxackala: Dagestanskoe ucebno-pedagogiceskoe izdatel'stvo. Magometov, Aleksandr A. 1963. Kubacinskij jazyk. Tbilisi: Izdatel'stvo Akademii Nauk Gruzinskoj SSR. Murkelinskij, Gadzi B. 1967. Lakskij jazyk. In Jazyki narodov SSSR, ed. by V. V. Vinogradov et al., vol. 4, 488-507. Moscow and Leningrad: Nauka. Plank, Frans. 1990. Suffix copying as a mirror-image phenomenon. Linguistics 28, 1039-1045.

6 Genitives and Adjectives as Attributes in Daghestanian Ol'ga/u. Boguslavskaja

1. Introduction The phenomenon of Suffixaufnahme as such (Finck 1910, Plank 1990) is not characteristic of Daghestanian languages. Even adjectival modifiers rarely agree with their head nouns in case, and for Daghestanian at least this appears to be a precondition of Suffixaufnahme. The closest approximation to Suffixaufnahme occurs in Tsakhur, and similar patterns are found in the Tsez group, where a modifier in the genitive agrees in case with its head— although, unlike in prototypical Suffixaufnahme, the case marker of the head is not literally repeated on the modifier. It is precisely in these languages that the case of the noun is marked in some way on adjectival modifiers as well, thus pointing to a strong affinity between adjectives and nouns in the genitive. The aim of this chapter, then, is to examine systematically, in a representative sample of Daghestanian languages, what genitival nouns and adjectives have in common and what distinguishes them in the first place.

2. Comparing Genitives All descriptions of Daghestanian languages postulate at least one case whose main function is to mark nouns governed by other nouns. This is the genitive; however, it is by no means clear how this Daghestanian genitive compares to genitives elsewhere. For Indo-European two views of the genitive can be distinguished. Kurylowicz (1949) and Benveniste (1962) see the major function of the IndoEuropean genitive as being determined by the transformation of the verbal syntagm into a nominal one, i.e. by nominalization. Adnominal uses of the genitive that cannot be reduced to relations of subject and object are considered secondary. For Tesniere (1959), on the other hand, the genitive is mainly 230

Genitives and Adjectives as Attributes in Daghestanian

231

a means of providing a noun with adjectival functions, with its nominalizing ability a secondary and optional property. On Kurylowicz's and Benveniste's approach no genitive can be postulated for Daghestanian with any degree of certainty, for there are clear differences from Indo-European with respect to nominalization. In most Daghestanian languages the actants of masdars (adverbal nouns) retain the case of prototypical verbal constructions. Chamalal is such a language: (1) ima yi?-e-c' dl viXideda father (Nom) arrival-Masd-Loc I (Nom) was=delighted 'I was delighted at father's arrival'

In (1) two elementary predications are combined, 'father has arrived' and 'I was delighted at something'. In Chamalal the second actant of the verb 'to delight' is in the locative, and the single actant of 'to arrive' is in the nominative. As seen in (1), the masdar is in the locative governed by the main verb, and 'father' retains the nominative rather than acquiring the genitive (which would be imi). There are also two elementary predications combined in (2), 'father beats sister' and 'I am not glad at something': (2) im-ud jac jiL'al-e-c' di razi iko' father-Erg sister (Nom) beat-Masd-Loc I (Nom) glad not 'I am not glad at father's beating [my] sister'

The first actant of the verb 'to beat' is in the ergative, and the second in the nominative. The nouns associated with the masdar in (2) retain these cases, with the masdar itself taking locative, as required by the main verb. Neither ergative nor nominative can be replaced by the genitive, unless a different meaning is intended: (2')

im-i jac jiL'al-e-c' di razi iko' father-Gen sister (Nom) beat-Masd-Loc I (Nom) glad not 'I am not glad at father's sister being beaten'

As is typical of all Daghestanian languages, there are two nominatives in one clause in (1) and (2). It has been noted by Klimov and Alekseev (1980: 233-235) that in Tabasaran, Rutul, Tsakhur, and Lezgian the genitive can occur in nominalizations serving functions regularly performed by the ergative and nominative in verbal constructions. This is also the case in Aghul (of the Lezgian group) and Bezhta-Tl'adal (of the Tsez group). However, the genitive in nominalizations is always subject to a number of conditions, and prototypically verbal case marking is almost always possible as an alternative (see further Boguslavskaja 1989: 115-124). Nominalizations thus hardly justify pigeonholing the Daghestanian genitive

232

The Caucasus

with its Indo-European namesake. Nonetheless, there are traits they do have in common. They are to be found in adnominal constructions not representing nominalizations of verbal ones. In Daghestanian, as in Indo-European, the genitive is the main case marking actants and circumstantial subordinates of nouns. Typologically, then, this would seem to be the most important function of this protean case. 3. Comparing Genitives and Adjectives Following Tesniere (1959), the genitive will here be considered a means of imparting adjectival function to nouns. Nouns in the genitive and adjectives may accordingly be expected to share certain properties. The extent of the affinity between genitival nouns and adjectives in Daghestanian languages can be gauged by five criteria: referentiality (Section 3.1), the opposition of direct and oblique genitive (3.2), agreement (3.3), restrictivity (3.4), and identity of exponents (3.5).

3.1. Referentiality In Russian possessive adjectives can only be used non-referentially, while nouns in the genitive can be used both referentially and non-referentially. Demonstrative or possessive pronouns accompanying genitival nouns ensure the referential interpretation (3); otherwise both interpretations coexist (4a/b). (3)

platok etoj zensciny; golova moej sobaki kerchief this=Gen woman=Gen; head my=Gen dog=Gen 'this woman's kerchief; 'my dog's head'

(4) a. golova sobaki byla ispackana kraskoj head=Nom dog=Gen was stained paint=Ins 'The dog's head was stained with paint' b. bozestvo s golovoj sobaki deity=Nom with head=Ins dog=Gen 'a deity with a dog's head'

The most natural interpretation of (4b) does not presuppose any referent for sobaki; on this interpretation the genitive is synonymous with a possessive adjective: (5) bozestvo s soba6'ej golovoj deity=Nom with dog=Adjct=Ins head=Ins

In Daghestanian a noun in the genitive can also be either referential or non-referential, but there is no formal distinction between attributes of these kinds. Genitives as in (6), from Akhvakh, may be translated into Russian or English either as genitives or as adjectives.

Genitives and Adjectives as Attributes in Daghestanian

233

(6) ako'a-Li k'azi women-Gen kerchief (Nom) 'kerchief of the women, feminine kerchief As long as such genitives are interpreted referentially, their relation is that of owner or user, filling the corresponding valency of their head. On the nonreferential interpretation the relation is not one of possession; the noun in the genitive here denotes some property of the noun it modifies and is circumstantial rather than valency-bound. Circumstantial genitive constructions are much more widespread in Daghestanian than, for example, in Slavic. Thus in (7) and (8), from Chamalal and Aghul respectively, the genitival noun denotes a material, and in (9) and (10), from Akhvakh, a property. (7) hinc'a-L haq' stone-Gen house 'stone house' (8) ruq'-an Hajag iron-Gen kettle 'iron kettle' (9) c'ada-Li zo rain-Gen day 'rainy day' (10) b-asi-d-a mina-Li eko'a white head-Gen man 'grey-headed man' (with the adjective agreeing with 'head' in class and number) Russian can use only adjectives to translate such genitives (kamennyj doml *dom kamnja 'stone house', etc.).

3.2. Direct vs. Oblique Genitive The opposition of a direct and an oblique genitive, reflecting a corresponding case contrast of governing nouns, is primarily characteristic of the Tsez group and of Tsakhur. Compare the following examples from Bezhta: (11) a. uq'-o ozo biLo-? oq'ojo big-AdjDct boy (Nom) house-Loc come=Past 'A big boy came into the house' b. uq'-a oz-di hudo buca big-Adjobl boy-Erg wood (Nom) chop=Pres 'A big boy chops wood'

234 (12)

The Caucasus a. kiba-s biLo girl-GenDct house (Nom) 'the girl's house' b. t'ek huxlona kiba-la biLo-Ra book (Nom) remain girl-Genobl house-Loc 'The book remained in the girl's house'

The attributes are adjectival in (11) and genitival in (12). The choice of the adjectival case suffix depends on whether the head noun is in the nominative (-0) or in an oblique case (-a). The same principle determines the choice of the alternative genitival suffixes -s (direct) and -la (oblique). The same contrasts obtain in Khvarsh: (13)

a. koku-s lo logu goli spring-GenDct water (Nom) good is 'The water of the spring is good' b. kok-la ieho spring-GenObl water=Loc 'near the water of the spring'

There is a similarity to Suffixaufnahme insofar as the case of the head noun is reflected by its genitival attribute in Bezhta, Khvarsh, and other Tsez languages, as examined by Kibrik in the preceding chapter. Differing from genuine Suffixaufnahme, although only superficially, the case marker of the head is not literally repeated on the genitival noun. Also, attributes have merely a two-way contrast of direct and oblique, whereas head nouns distinguish several oblique cases. In accordance with Zaliznjak's (1967) and Uspenskij's (1957) criteria for case distinction, two different case forms have to be recognized here for adjectives (and participles), and two separate genitives for nouns. The parallelism between adjectives and nouns is captured terminologically by referring to the contrast in both instances as one between direct and oblique forms.

3.3. Agreement In the languages of the Andi group, in Tabasaran, and in Tsakhur, nouns in the genitive follow the example of adjectives and agree with their head nouns in class and/or number. Such genitival nouns are therefore closer to adjectives than they are in Indo-European languages. Thus, in Chamalal (a member of the Andi group), adjectives agree with their head nouns in six classes and two numbers, with both prefixes and suffixes reflecting the nominal contrasts: (14)

a. w-asak'-u hadam ISg-short-ISg man, (Sg)

Genitives and Adjectives as Attributes in Daghestanian

235

b. j-asak'-i jah IISg-short-IISg womann (Sg) c. j-asak'-uje zin-e IIIPl-short-IIIPl cowin-Pl 'short man/woman/cows'

Some Chamalal nouns have two ways of forming genitives: one with the invariable marker -L, and the other with markers differing according to the class and number of the head, just like adjectival agreement affixes. (15)

a. diw wac-i haqo'-ud zin buR my brother,-GenIISg (Sg) wife,,-Erg (Sg) cowIII (NomSg) sell=Past b. diw wac-uL haq.'-ud zin buR my brother I Gen (Sg) wif I I -Erg (Sg) cowIII (NomSg) sell=Past 'My brother's wife has sold the cow'

(16)

a. haqo'-ud wacu-L zin buR wifeII-Erg (Sg) brother,-Gen III e (Sg) cowIII (NomSg) sell=Past b. haqo'-ud wacu-b zin buR wifeII-Erg (Sg) brother,-Gen (Sg) cowIII (NomSg) sell=Past 'The wife has sold brother's cow'

(17)

a. wac-uje zin-e brotherI-GenIIIP1 (Sg) cowIII-Pl (Norn) b. wac-uL zin-e brother,-Gen (Sg) cowIII-Pl (Nom) 'brother's cows'

Nouns in the variable genitive thus share with adjectives the inflectional category of class and number agreement. Unlike adjectives but like all nouns, they are in addition inherently specified for their noun class, and they express a referential contrast of number (with the singular genitives in (15)-(17) contrasting with plural ones, referring to more than one brother). In another dialect of Chamalal, spoken in the village of Upper Gakwari, there are also two ways of forming genitives, but (according to the data in Bokarev 1949: 37—47) both distinguish agreement number, with singular -L contrasting with plural -Le: (18)

a. hadam-L c'at man,-Gensg (Sg) horseIII (NomSg) 'the man's horse' b. hadam-Le c'at-e man,-Genpl (Sg) horseIII-Pl (Nom) 'the man's horses'

236

The Caucasus

A similar pattern is reported for Tsakhur by Kibrik in the preceding chapter. In Akhvakh, Tabasaran-D'ubek, and Tabasaran-Kondig, only the genitive marked for restrictivity agrees with its head in class and number.

3.4. Restrictivity Eight Daghestanian languages of the fifteen investigated—Akhvakh, Chamalal, Bezhta-Tl'adal, Khvarsh, Lak, Dargva-Chirag, Tabasaran-D'ubek, and Tabasaran-Kondig-—morphologically distinguish restrictive attributes and attributes neutral with respect to restrictivity. Restrictive attributes delimit the set of entities denoted by their head nouns by emphasizing the contrast with other possible delimitations (e.g. 'father's sister, not mother's', 'fine houses, not ugly ones'; see further Boguslavskaja 1989). In four languages—Akhvakh (19), Khvarsh (20), Tabasaran-D'ubek (21), and Tabasaran-Kondig (22)—this inflectional category also pertains to nouns in the genitive. ("Restrictive" is abbreviated R in glosses, and in translations the corresponding forms are in boldface.) (19) a. dada jaci father,=Gen (Sg) sisterII (NomSg) 'father's sister' b. haje gude dada-je jaci it is fatherI-RGenIISg (Sg) sisterII (NomSg) 'It is father's sister' c. haje gude dada-we waci it is father I RGen ISg (Sg) brother, (NomSg) 'It is father's brother' (20) a. isti-s Rine brother-Gen wife (NomSg) 'brother's wife' b. ob-udi isti-s-so Rine goLe father-Erg brother-GenDct-R wife (Nom) call=Pres 'father is calling brother's wife' (21) a. jas ildas jasamis-sulda?as k'ak'li-n Xula-?a my friend lives wood-Gen houseII-Loc (Sg) 'My friend lives in a wooden house' b. jas ildas jasamis-suldapas k'ak'li-n-suv Xula-?a my friend lives wood-Gen-RIISs housen-Loc (Sg) 'my friend lives in a wooden house' (22) a. haljvni-n jik ip'in horse-Gen meat (Nom) eat=Imp 'Eat the horse's meat!' b. haljvni-n-u jik ip'in horse-Gen-R meat (Nom) eat=Imp 'Eat the horse's meat!'

Genitives and Adjectives as Attributes in Daghestanian

237

In Akhvakh the restrictivity markers -je and -we (19b/c), also occurring with adjectives and participles, are part of a full paradigm of class and number agreement forms (-be, -re, etc.). In Tabasaran-D'ubek the restrictivity marker, likewise exhibiting class and number agreement, is not cumulated with genitive. In Khvarsh and Tabasaran-Kondig the restrictivity markers are invariable, precluding agreement. Frequently restrictive adjectives have an additional agreement slot vis-avis restrictivity-neutral adjectives. Thus in Dargva-Chirag only restrictive adjectives agree in number with their heads: (23)

a. bat'ar-ze q'ale fine-Adj house (NomSg) 'fine house' b. bat'ar-ze q'al-be fine-Adj house-Pi (Nom) 'fine houses'

(24) a. bat'ar-kan q'ale fine-RAdj house (NomSg) 'fine house' b. bat'ar-kan-e q'al-be fine-RAdj-Pl house-Pi (Nom) 'fine houses'

As was seen above (19b/c), restrictive genitives in Akhvakh also agree with their head nouns while restrictivity-neutral ones do not. Agreeing nouns in the restrictive genitive are further evidence for an affinity to adjectives. Again, principles of case distinction like those advocated by Zaliznjak (1967) and Uspenskij (1957) suggest that restrictive and neutral genitives, in languages that have this contrast, should be considered separate cases. The agreement of such cases, as in Chamalal, multiplies case distinctions still further, since each of these forms occurs in a different syntactic context, with class and number of the head as the determining factor. This result seems undesirable, and one might alternatively consider excluding Chamalal wordforms such as wac-i, wac-ub, wac-uje (15-17) as well as restrictive genitives from noun paradigms and classifying them as adjectives. However, apart from clashing with traditions of Daghestanian grammar and being typologically inappropriate, this solution would be at odds with the referentiality criterion: such genitival "adjectives" co-occur with possessive and demonstrative pronouns, much as genuine nouns do.

3.5. Shared Exponents In two of the languages examined—Rutul and Tsakhur—the inflectional marking of the genitive is not distinguished from that of adjectives, with the two kinds of attributes thus coinciding morphologically.

238

The Caucasus

4. Summary Fifteen Daghestanian languages, mostly unwritten ones, have been investigated as to the affinity between nouns in the genitive and adjectives. Table 6.1 summarizes the results. The five criteria taken into consideration are numbered as in Section 3: 1 possibility of a non-referential interpretation 2 opposition of direct and oblique forms, agreeing with heads in this category 3 (suffixal) agreement in class and/or number 4 opposition of restrictive and restrictivity-neutral forms 5 inflectional marking of genitive identical with that of adjectives The symbols +, ---, 0 in Table 6.1 register respectively properties that genitival nouns share with adjectives, properties limited to adjectives, and properties not possessed by either kind of attribute. The arrangement of languages in Table 6.1 is according to the extent of the similarity between genitival nouns and adjectives, decreasing from top to bottom. Languages with the closest analogues to Suffixaufnahme are to be found at the very top of the table (Tsakhur, Khvarsh—with Rutul, where neither genitival nouns nor adjectives show directness agreement, coming in between) or halfway down (Bezhta, Bezhta-Tl'adal), after another small group of languages without directness agreement in both genitives and adjectives. The five properties are arranged from left to right so as to reflect the decreasing likelihood of their being shared by genitival nouns and adjectives. Disregarding properties not possessed by either kind of attribute, the implicational relations between different languages and between different properties Table 6.1. Properties of Genitival Nouns in Comparison with those of Adjectives. Properties

1

2

3

4

5

Tsakhur Rutul Khvarsh Akhvakh Tabasaran-D'ubek

+ + + + +

+

+ 0 + 0 0

0

+ 0 0 + +

0

0 0 + + +

+ + —

Chamalal Bezhta Bezhta-Tl'adal Aghul Budukh Kryz Khinalug Lak Dargva-Chirag

+ + + +

0 + + 0

+ 0 0

--

Tabasaran-Kondig

+ --

0 0

---

--

---

+ +

0 0

0 0

0 0

+ + +

0 0 0

0 0 -

--

--

-

-

0

-

Genitives and Adjectives as Attributes in Daghestanian

239

are virtually perfect (as indicated by contiguous pluses and minuses, allowing for intervening zeroes): a shared property in one line tends to imply a shared property also in the line above, and a shared property in one column tends to imply a shared property also in the column to its left. Variation among Daghestanian languages as to the relationship between genitival and adjectival attributes is thus not at all random.

References Benveniste, Emile. 1962. Pour 1'analyse des fonctions casuelles: le genitif latin. Lingua 11. Reprinted in his Problemes de linguistique generate, vol. 1, 140-148. Paris: Gallimard, 1966. Boguslavskaja, Ol'ga Ju. 1989. Struktura imennoj gruppy: Opredelitel'nye konstrukcii v dagestanskix jazykax. Dissertation, Moskovskij gosudarstvennyj universitet. Bokarev, Aleksandr A. 1949. Ocerk grammatiki camalinskogo jazyka. Moscow and Leningrad: Nauka. Finck, Franz Nikolaus. 1910. Die Haupttypen des Sprachbaus. Leipzig: Teubner. Klimov, Georgij A., and Mixail E. Alekseev. 1980. Tipologija kavkazskix jazykov. Moscow: Nauka. Kurylowicz, Jerzy. 1949. Le probleme du classement des cas. Biuletyn polskiego towarzystwa j^zykoznawczego 9,20-43. Reprinted in his Esquisses linguistiques. Wroclaw and Cracow: Ossolineum, 1960. Plank, Frans. 1990. Suffix copying as a mirror-image phenomenon. Linguistics 28, 1039-1045. Tesniere, Lucien. 1959. Elements de syntaxe structurale. Paris: Klincksieck. Uspenskij, Vladimir A. 1957. K opredeleniju padeza po A. N. Kolmogorovu. Bjulleteri ob'edinenija po problemam masinnogo perevoda 5. Moscow. Zaliznjak, Andrej A. 1967. Russkoe imennoe slovoizmenenie. Moscow: Nauka.

This page intentionally left blank

IV INDO-EUROPEAN

This page intentionally left blank

7 Indo-European o-Stems and Feminine Stems in -I Francisco ViJJar 1. Introduction Although, strictly speaking, the phenomenon of Suffixaufnahme did not exist in the common Indo-European language, there do exist two nominal paradigms whose origins offer some analogies to it. These are the o-stems and the feminines in -*i. Neither of these is exactly coincident with Suffixaufnahme, but they serve to throw light on the diachrony of Suffixaufnahme and kindred patterns of attributive marking. 2. Thematic Genitive, Genitive in -os, and Denominal Adjectives

2.1. Outline of the Question and Review of Relevant Research Stated briefly, the problem is as follows: the genitive is a case which, in the Indo-European family, is both very uniform and very stable. By very uniform I mean that the ending -*osl-*esl-*s is present in all the dialectal groups without exception, and that it was the Indo-European morpheme for all the inflectional models. By very stable I mean that there are hardly any dialectal innovations to be found: the genitive ending has remained operative through thousands of years and has only been eliminated to the extent that inflection itself has been eliminated, disregarding phonetic accidents. However, this overall picture of uniformity and stability does have one outstanding exception, extending to both aspects of the general behavior of the genitive mentioned above. In the thematic type, leaving aside the Anatolian languages for the moment, the genitive shows a wide range of dialectal forms rather than the common ending -*os/-*es/-*s. Dialectal diversity in this particular respect is so intense that there are often differences even within a single branch. Thus, within Italic, the situation is different in Latin and in Osco-Umbrian, and there are differences between Old Prussian and Leto243

244

Indo-European Table 7.1. Indo-European Nominatives and Genitives. Nominative Hittite Sanskrit Greek Gothic Latin Celtiberian

-*os -*os -*os -*os -*os -*os

(> (> (> (> (> (>

-as) -as) -05) -s) -us) -os)

Genitive -*os -*osk> -*osio -*eso -*I -*6t

(> (> (> (> (> (>

-as) -asya) -ou) -is) -I) -o)

Lithuanian within Baltic, between Gaulish and Celtiberian within Celtic, between Gothic and Nordic within Germanic, and so on. Hittite data, becoming available more recently, have proved to be of major importance for the problem at issue. Indeed, the Hittite o-stem genitive is unlike any of the dialectal forms previously known and thus provides us with a new situation: it has the form -as, which happens to be identical to that of the nominative. Before continuing, however, it will be of use to compare nominatives and genitives in various Indo-European languages, as set out in Table 7.1. If before it could already be deduced that the thematic genitive must have had a problematic morphology in Indo-European, the evidence from Hittite may be said to have furnished the key to an understanding of the specific cause of its propensity for innovation. That is, if we were not acquainted with the Hittite evidence, we would be confined to mere guesswork as to the reasons for the multiplicity of dialectal innovations; but Hittite saves us from the danger of unfounded speculations by presenting—historically attested in its thematic genitive—a problematic situation capable of provoking the actual dialectal innovations. Note that in Hittite the thematic genitive and nominative are identical: both have the common form -as < -*os. This coincidence of form was already noticed and given due importance by a number of scholars as far back as the 1930s, when Anatolian linguistics was in its early stages (Borgstr0m 1934: 121-128; Kurylowicz 1935: 260; Milewski 1936; Pedersen 1938: 26). Indeed, in the years that followed and even to the present day, it has been agreed on by the great majority of Indo-European scholars,1 who generally interpret this as an archaism preserved in the Hittite language—which is of course the same as saying that in Indo-European, as in Hittite, thematic nominative and genitive were identical.2 Going a step further, Haudry (1982: 33) has drawn attention to several phenomena that reflect the same situation in Indo-European languages other than Hittite. The first of these is Mycenaean, where the same formal coincidence seems to have existed: alongside a nominative te-o ( = 0eos) we have two genitives, te-o-jo ( -*!, -*i[s]) (> -*im/-*iom) (> *-ibs) (> *-iei) (> *-ies) (> -*i6m)

Table 7.9. Imitative Inflectional Model of Hypostasis of Genitive. Nominative Accusative Genitive Dative Nominative Plural etc.

-*I -*i + -*i + -*i + -*i +

m [-*i + am] as ai as

(> (> (> (>

-*im, -*ilm) -*i^s) -*iai) -*i3s)

258

Indo-European

Most of the historical Indo-European languages adopted the second procedure, which indeed is the only inflectional solution existing in Greek, Germanic, Baltic, and Slavonic; it is also the one we find in the Vedic type dev'i. All these languages coincide in imitating the feminines in -a/3 in the oblique cases of the singular and throughout the plural. They differ, however, in the solution adopted for the accusative singular, where some opt for -* iam (Gothic-/a, Lithuanian -ja, Old Church Slavonic -io) and others for -*m (Old Prussian-zn, Sanskrit -im, Avestan -lm). Other languages, especially Greek and Celtic, have their own patterns which need not concern us here.22 Only Old Indie presents the two inflectional solutions at the same time. To attribute them both to the common language, as Wackernagel and Lohmann did and as their followers continue to do, is one more episode in the unfortunate pan-Sanskritistic reconstruction, which is the true original sin of IndoEuropean studies. The mechanism for the formation of these paradigms has been to a great extent dialectal, and the formation of two alternative inflectional varieties is a dialectal trait of Indie alone. 4. Summary There remains only to summarize the main points of my thesis and to make some of its corollaries explicit. (1) The Indo-European genitive expressed noun-noun determination. This does not mean that the denominal adjective did not perform similar functions. The duality of domuspatris vs. domus patria was as Indo-European as it is Latin or Greek. But, contrary to what Wackernagel believed, denominal adjectivalization is more recent than the genitive—so much so that the former was constructed on the basis of the latter by a procedure of the Suffixaufnahme type. (2) In Indo-European there was a genitive ending in -*l whose main function, like that of the other genitive endings, was to express noun-noun determination. (3) The genitive ending -*l was accented, as is often the case with other genitive endings, especially -*6s. (4) The ending -*l underwent hypostasis when expressing various adnominal relations, most importantly when expressing the condition of female, particularly of animals, in elliptic form on the basis of phrases of the type *[dh&lus] ulkri 'female of wolf. The frequency of this type of expression and of its elliptic form is to be assessed within the context of a language with no masculine/feminine grammatical gender, although it would be equally plausible in a language possessing such a gender distinction but lacking a feminine variant for all animals, as is the case in some modern languages. (5) Once the -*z form had been adopted, by hypostasis, as the nominative of the corresponding female noun, it was necessary to provide it with full inflection through a process similar to the one known as Suffixaufnahme, although differing from it in various aspects.

Indo-European o-Stems and Feminine Stems in -I

259

(6) The new paradigm could have been constructed in various ways, and there is indeed a wide range of variation in the different dialects. In Vedic the two inflected forms (vrkl and devi) arise respectively from the possibilities of adding the standard endings or of analogically imitating the paradigm of the feminines in -ah. In spite of Wackernagel's contention,23 there is no comparative basis for supposing that both types existed as differentiated paradigms in Indo-European, and they are in all likelihood a short-lived Indie innovation. (7) Most languages opted for the inflected form imitating analogically the paradigm of the feminines in -ah. The original -*l (without the addition of -a) remained in the nominative and sometimes also in the accusative. (8) As a productive genitive ending, -*z was generally replaced by -*6s. It remained only in Latin, Venetic, and a few Celtic dialects (Irish, Gaulish, Lepontic, but not Celtiberian). In Sanskrit it remained fossilized as an adverb in expressions of the type mithunl karoti. In the other languages it was simply eliminated. (9) The reason why -*z historically appears linked to the thematic type in Sanskrit (as an adverb24) and in Latin and Celtic (as a genitive ending) lies in circumstances affecting the nature and development of this inflectional type, as explained above. (10) The use of -*l as a genitive marker is therefore an archaism. Consequently, its presence in Italic and Celtic as such lacks dialectal significance. (11) As a corollary from all considerations above, it may be deduced that we should not look for an explanation of the vrkl and devi paradigms in ancient apophonic differences, as is usually done at the cost of anachronism when very recent paradigms, to a great extent dialectal ones, are explained by means of accentual phenomena that actually belong to an early phase of the common language. (12) The feminine derivational suffix -*i, created from the genitive case ending -*I, formally resembles Suffixaufnahme insofar as a second case marker is added to the genitive ending. However, while hypostasis may play a role in genuine Suffixaufnahme as well, the feminines in -*l take the second case marking not by virtue of agreement but owing to the need of markers for their own syntactic functions. (13) As far as the origin of the thematic inflection is concerned, we must distinguish two very different results. On the one hand, the hypostasis of the genitive case ending -*os has resulted in thematic nouns of the type of Sanskrit rathas 'cart'; for this hypostasis the same comments apply as for the hypostasis of -*l forms. On the other hand, thematic denominal adjectives have been created from the same genitive in -*os owing to the necessity of agreement with their heads, thus resembling the Suffixaufnahme pattern. Nevertheless, in the creation of the thematic paradigm, for both nouns and denominal adjectives, the formal procedure has not been that of double case marking but that of acquiring simple inflections mainly by analogy. (14) Therefore, neither of the Indo-European paradigms analysed in this chapter are instances of Suffixaufnahme in the classical sense. The differences from genuine Suffixaufnahme are functional in the case of the feminines in -I,

260

Indo-European

formal in the case of denominal adjectives, and both formal and functional in the case of thematic nouns. Notes 1. Among others Szemerenyi (1957: 102); Berg (1958: 224); Watkins (1963: 16 n. 1, 28 n. 2) and (1966); Garcia Calvo (1964: 113); Bailey (1970: 41); Gil (1968: 37); Villar (1974: 107ff.); Rosenkranz (1979: 221); Schmidt (1977); Lehmann (1981); Haudry (1982: 33); Beekes (1985: 185f., 192-194); Adrados (1988: 25f.). 2. Of course, as will be seen subsequently, it is not the same to say, as did Wackernagel (1930: 37f.), that the thematic type did not have a genitive (which would be an absurd situation) as to say that genitive and nominative had the same form. Watkins (1966: 28-39) states that the thematic type had no genitive; but no doubt this must be understood as meaning that genitive was identical to nominative, as he explicitly says elsewhere (1963: 16 n. 1, 28 n. 2). Attempts to explain this identity have been few and unsuccessful. Probably the best known is that of Kronasser (1962-1966: 163), for whom the Hittite nominative derives from the Indo-European thematic masculine nominative -*os and the Hittite genitive from the Indo-European feminine genitive of stems in aid, -*as. Another possibility is to take the Hittite thematic genitive in -s as being analogical to the consonantal stems. 3. The existence of this Mycenaean thematic genitive was first discovered by Luria (1957: 321f.). Lejeune (1965) was the first to suggest that this might be the same phenomenon as in Hittite. But the first definite proposal was that of Haudry, which was accepted by Adrados (1990) in a detailed study of Mycenaean. 4. Indras (. . .) Varunas (. . .) kratur bhdvati ukthyds 'The power of Indra, of Varuna is worthy of being praised' (RV 1.17.5). 5. See a complete list in Brugmann (1905: 342-347); cf. also Brugmann (1909: 661ff.). 6. It is hypothetical that the Latin suffix -osus can be identified with -* uent; cf. Leumann (1963: 231). 7. The Hittite example is transliterated from syllabic cuneiform; the hyphens here do not segment the forms of the morphological analysis given in the gloss. 8. Beekes (1985: 191-195) has recently reached a similar conclusion, although he starts from ergativist assumptions to explain the coincidence of nominative and genitive. 9. See Villar (1981: 147-149, 164-166, 196f.; 1985: 31-48) for extensive further discussion. 10. Sommer based his hypothesis on the idea, put forward shortly before by Michels (apud Wackernagel 1908: 126, and Sommer 1902: 371), that the feminine suffix -fwould originally have meant 'possession' (Zugehorigkeit), but without relating it as yet to the Latin-Celtic genitive. Thus, for Michels * ulk^l would originally have meant 'zum Wolf gehorige'. 11. It was also accepted by Sommer himself (1914: 341). 12. Wackernagel (1908: 128) declared that he followed Brugmann in accepting -fas the old ending of the thematic genitive: "Das Zutreffendste hat bis jetzt Grundriss II 568f. [=1909: 568] insofern geaussert, als er die Moglichkeit offen lasst, dass -f die alteste Genitivendung der o-Stamme gewesen sei." Of course Wackernagel took this genitive in -f, like all the others, to be an adverbial case. But his views must have evolved during the following years so that eventually (1930: 37f.) he considered the thematic

Indo-European o-Stems and Feminine Stems in -\

261

type to have lacked a genitive in Indo-European and -Fto have been an adverbial suffix: "Auch die ig. o-Stamme scheinen ursprunglich beim Nomen keinen ausgebildeten Gen.Sg. gehabt zu haben ... die iibrigen Sprachen verwenden entweder die Ablativform (bait, -o, slav. -a aus idg. -*od) oder eine ursprunglich adverbartige Form auf -F." 13. A certain precedent for attributing these two inflectional modalities to IndoEuropean may be found in Brugmann (1905: 348; 1909: 208ff.). 14. Among those who have written on the subject recently, e.g. Schlerath (1988). 15. Recently Prosdocimi (1989: 158-160) has declared himself in favor of having the three items derive from the same Indo-European morpheme. 16. "Erstens ist die Bedeutung 'zum Gott gehorig' von Michels nur konstruiert" (Wackernagel 1908: 126). 17. As is normally done in Indo-European comparative linguistics, I am using the term "suffix" in its restricted meaning for "derivational suffix," and "ending" for the final suffix, having inflectional values, in particular that of case marking. 18. The theory that the feminine gender is recent in the Indo-European family has a long tradition, dating back as far as Brugmann (1897). It gained acceptance with the work of Meillet (1926) and was further consolidated as evidence from the Anatolian languages became available (Neu 1969; Laroche 1970; Carruba 1972; Brosman 1976, 1978; etc.). 19. Meillet (1964: 285): "Dans les substantifs une formation en -F sert souvent a designer des femelles;" Hirt (1912: 1): ". . . dafi -F als femininbildendes Element bei Adjektiven konsonantischer Stamme verwendet wird, und vor allem weiter bei der Bildung movierter Feminina Anwendung findet." 20. An extensive list of the different formations existing in the various historical languages based on the -F element can be found in Hirt (1912); cf. also Brugmann (1909: 208-216, 599-602). 21. Side by side with this type of words in -F there can be found others in -F whose relation to them is not immediately apparent, a complication pointed out by Lohmann himself (cf. Brugmann 1909: 209f.). 22. Concerning the paradigm of feminines in Celtic, Prosdocimi (1989) has recently concluded that in this branch of Indo-European the feminines present a mixture of -F, - ia/3 and -at3 stems, not as the result of a secondary confusion but dating back to the very formation of feminine paradigms. 23. Wackernagel (1930: 164): "Diese Doppelheit der -F-Flexion ist ererbt. Sie kann nicht innerhalb des Indischen aufgekommen sein, weil die indische Entwicklung gerade zur volligen Vermischung beider Typen gefiihrt hat." Although this is obviously not a sufficient reason for transferring the duality of types to the common language, it is the sole reason provided by both Wackernagel and Lohmann (1932: 67). 24. This link is not entirely clear, however, since this type of adverb in -Fcan also be formed from various other stems (-a, -i, -r, -an, -as). Wackernagel (1908: 127f.) seeks to demonstrate that only the thematic type of such adverbs would have been old ones.

References Adrados, Francisco R. 1988. Archaisms in Anatolian nominal inflexion. In A Linguistic Happening in Memory of Ben Schwartz, ed. by Yoel L. Arbeitman, 13-40. Louvain-la-Neuve (BCILL 42).

262

Indo-European

Adrados, Francisco R. 1990. El genitivo en -o en micenico y chipriota. In Studia Indogermanica et Paleohispanica, ed. by Francisco Villar, 175-181. Salamanca: Universidad de Salamanca. Bailey, Charles-James N. 1970. Inflectional pattern of Indo-European nouns. Working Papers in Linguistics (University of Hawaii) 2, 12-16. Beekes, Robert S. P. 1985. The Origins of the Indo-European Nominal Inflection. Innsbruck: Innsbrucker Beitrage zur Sprachwissenschaft. Beekes, Robert S. P. 1990. Le type gotique bandi. In La reconstruction des laryngales, ed. by Jean Kellens, 49-58. Liege: Les Belles Lettres. Berg, Nils. 1958. Einige Betrachtungen iiber den indogermanischen Komparationskasus. Norsk Tidsskrift for Sprogvidenskap 18, 202-230. Bloch, Alfred. 1960. Kann der lateinisch-keltische Genetiv auf -I der 6-Stamme gleichen Ursprungs sein wie die altindische Praverbialform auf -I? Zeitschrift fur vergleichende Sprachforschung 76, 182-242. Borgstr0m, Carl. 1934. The thematic genitive singular in Indo-European. Norsk Tidsskrift for Sprogvidenskap 7, 121-128. Brosman, Paul W. 1976. The Hittite gender of the cognates of PIE feminines. Journal of Indo-European Studies 4, 41-59. Brosman, Paul W. 1978. Hittite evidence and the ilya adjective. Indogermanische Forschungen 83, 124-137. Brugmann, Karl. 1897. The Nature and Origin of the Noun Genders in Indo-European Languages. New York: C. Scribner's Sons. Brugmann, Karl. 1905. Abrege de grammaire comparee des langues indo-europeennes. Paris: Klincksieck. Brugmann, Karl. 1909. Grundriss der vergleichenden Grammatik der indogermanischen Sprachen, vol. 2, part 2. Strassburg: Triibner. Carruba, Onofrio. 1972. II problema del genere in anatolio e in indoeuropeo. In Le lingue dell'Europa: Atti del V Convegno Internazionale di Linguisti, 175-192. Brescia. Garcia Calvo, A. 1964. Algunos problemas en el estudio funcional de los casos y las partes de la oration en las lenguas indoeuropeas. In Adas del II Congreso Espanol de Estudios Cldsicos, 109-120. Madrid: C.S.I.C. Gil, Juan. 1968. El genitivo en -iy los on'genes de la declination tematica. Emerita 35, 25-43. Gotze, Albrecht. 1937. Transfer of consonantal stems to the thematic declension in Hittite. In Melanges linguistiques offerts a M. Holger Pedersen, ed. by Louis Hjelmslev, 488-495. Copenhagen: Universitetsforlaget. Gren-Eklund, G. 1986. The survival of a hypothesis: Sanskrit cvi and Latin gen.sg. -i. In On the Dignity of Man: Festschrift F. Rundgren, 157-166. Stockholm. Hammerich, Louis L. 1956. Contribution 3. In Proceedings of the Seventh International Congress of Linguists, 175-176. London: Wilson. Haudry, Jean. 1982. Prehistoire de la flexion nominal indo-europeen. Lyon: Institut d'etudes indo-europeennes de 1'Universite Jean Moulin (Lyon III). Hirt, Hermann. 1904/1905. Uber den Ursprung der Verbalflexion im Indogermanischen. Indogermanische Forschungen 17, 36-84. Hirt, Hermann. 1912. Zur Bildung auf -f im Indogermanischen. Indogermanische Forschungen 31, 1-23, Knobloch, Johann. 1950. Zur Vorgeschichte des indogermanischen Genitivs der oStamme auf -sjo. Die Sprache 2, 131-149.

Indo-European o-Stems and Feminine Stems in -i

263

Krahe, Hans. 1948. Germanische Sprachwissenschaft, 2. Formenlehre. Berlin: de Gruyter. Krause, Wolfgang. 1968. Handbuch des Gotischen. 3d ed. Munich: Beck. Kronasser, Heinz. 1962-1966. Etymologic der hethitischen Sprache. Wiesbaden: Harrassowitz. Kurylowicz, Jerzy. 1935. Etudes indoeuropeennes, vol. 1. Cracow: Polska Akademia Umiejetnosci, Prace Komisji Jezykowej No. 21. Laroche, Emmanuel. 1970. Etudes de linguistique anatolienne III. Revue hittite et asianique 28, 22-71. Lehmann, Winfred P. 1981. The genitive singular ending in -syo: How an IndoEuropeanist works. In Bono Homini Donum, ed. by Yoe'l L. Arbeitman, 178— 188. Amsterdam: Benjamins. Lejeune, Michel. 1965. Essais de philologie mycenienne IX: Le genitif singulier thematique. Revue de philologie, 39 14-20. Leumann, Manu. 1963. Lateinische Laut- und Formenlehre. Munich: Beck. Lohmann, Johannes. 1932. Genus und Sexus. Gottingen: Vandenhoeck & Ruprecht. Lommel, Hermann. 1912. Studien uber die indogermanischen Femininbildungen. Gottingen: Hubert. Luria, Solomon. 1957. Uber die Nominaldeklination in den mykenischen Inschriften. La Parola del Passato 12, 321-332. Mansion, Joseph. 1937. A propos de la declinaison du hittite. In Melanges linguistiques offerts a M. Holger Pedersen, ed. by Louis Hjelmslev, 480-487. Copenhagen: Universitetsforlaget. Meillet, Antoine. 1926. Le genre feminin dans les langues indo-europeennes. Reprinted in Linguistique historique et linguistique generate, by Antoine Meillet, vol. 2, 24-28. Paris: Klincksieck, 1937. Meillet, Antoine. 1931. Caractere secondaire du type thematique indo-europeen. Bulletin de la Societe Linguistique de Paris 32, 194-203. Meillet, Antoine. 1964. Introduction a Vetude comparative des langues indo-europeennes. Reprinted, University, Alabama: University of Alabama Press. Milewski, Tadeusz. 1936. L'indo-hittite et I'indo-europeen. Cracow: Bulletin de 1'Academie Polonaise (Suppl. 2). Neu, Erich. 1969. Rezension von W. Kastner, Die griechischen Adjektive zweier Endungen au/-O2 (Heidelberg: Winter, 1967). Indogermanische Forschungen 74,235-241. Pedersen, Holger. 1938. Hittitisch und die anderen indoeuropaischen Sprachen. Copenhagen: Levin & Munksgaard. Prokosch, Eduard. 1939. A Comparative Germanic Grammar. Philadelphia: Linguistic Society of America. Prosdocimi, Aldo. 1989. L'iscrizione gallica del Larzac e la flessione dei temi in -a, -i,ja. Con un 'excursus' sulla morfologia del lusitano: ace. crougin, dat. crougeai. Indogermanische Forschungen 94, 190-206. Rosenkranz, Bernhard. 1972. Entwicklungsgeschichte der idg. Nominalflexion. Cologne: Institut fur Sprachwissenschaft der Universitat. Rosenkranz, Bernhard. 1979. Archaismen im Hethitischen. In Hethitisch undlndogermanisch, ed. by Erich Neu and Wolfgang Meid, 219-229. Innsbruck: Innsbrucker Beitrage zur Sprachwissenschaft. Rundgren, Frithiof. 1960. Der Genitivus aestimativus im Lateinischen. Einige Bemerkungen. Eranos 58, 51-65.

264

Indo-European

Schlerath, Bernfried. 1988. Nochmals zu den cvi-Bildungen. In Studio Indogermanica et Slavica: Festgabefitr Wolfgang Thomas zum 65. Geburtstag, ed. by P. Kosta, 37-47. Munich: Otto Sagner. Schmidt, Gernot. 1977. Uber indogermanische nominale Relativkonstruktionen. Indogermanische Forschungen 82, 61-74. Sommer, Ferdinand. 1902. Handbuch der lateinischen Laut- und Formenlehre. Heidelberg: Winter; 2d ed., 1914. Sturtevant, Edgar H. 1940. O-stem adjectives from declinated genitives. Transactions of the American Philological Association 71, 373-378. Sturtevant, Edgar H. 1964. A Comparative Grammar of the Hittite Language. New Haven and London: Yale University Press. Szemerenyi, Oswald. 1957. The problem of Balto-Slav unity: A critical survey. Kratylos 2, 97-123. Szemerenyi, Oswald. 1989. Einfuhrung in die vergleichende Sprachwissenschaft. 3d ed. Darmstadt: Wissenschaftliche Buchgesellschaft. Tischler, Johann. 1983. Hethitisches etymologisches Glossar, vol. 1. Innsbruck: Innsbrucker Beitrage zur Sprachwissenschaft. Thurneysen, Rudolf. 1903. Etymologien. Indogermanische Forschungen 14, 127-133. Villar, Francisco. 1974. Origen de la flexion nominal indoeuropea. Madrid: C.S.I.C. Villar, Francisco. 1981. Dativo y Locativo en el singular de la flexion nominal indoeuropea. Salamanca: Universidad de Salamanca. Villar, Francisco. 1985. El dativo tematico indoeuropeo. In Symbolae L. Mitxelena, ed. by J. L. Melena, vol. 1, 31-48. Vitoria: Veleia (Anejo 1), Universidad del Pai's Vasco. Wackernagel, Jacob. 1908. Genetiv und Adjektiv. In Melanges de linguistique offerts a M. Ferdinand de Saussure, 125-152. Paris: Societe Linguistique de Paris, Librairie Honore Champion. Wackernagel, Jacob. 1930. Altindische Grammatik, vol. 3. Gottingen: Vandenhoeck & Ruprecht. Watkins, Calvert. 1963. Preliminaries to a historical and comparative analysis of the syntax of the Old Irish verb. Celtica 6, 1-49. Watkins, Calvert. 1966. Italo-Celtic revisited. In Ancient Indo-European Dialects, ed. by Henrik Birnbaum and Jan Puhvel, 29-50. Los Angeles: University of California Press. Watkins, Calvert. 1967. Remarks on the genitive. In To Honor Roman Jakobson: Essays on the Occasion of His Seventieth Birthday, vol. 3, 2191-2198. The Hague: Mouton.

8 Slavonic's Closest Approach to Suffixaufnahme: The Possessive Adjective Greville G. Corbett 1. Introduction One of the aims of this volume on Suffixaufnahme is to see the phenomenon in terms of a broader typology of attributive constructions. It is in this area that the evidence of the Slavonic possessive adjectives is particularly important. Possessive adjectives clearly show similarities to Suffixaufnahme constructions; indeed, as Plank (1990: 1039) points out, some researchers have tried to treat even prototypical examples of Suffixaufnahme as adjectives. I shall show that Slavonic possessive adjectives, to varying degrees, do indeed show similarities to undisputed instances of Suffixaufnahme, but that there are also important differences. These can be explained by reference to the nature of the morphological systems of the Slavonic languages. In a nutshell, I claim that some Slavonic possessive adjectives come as close to Suffixaufnahme as is possible without contravening major regularities in the morphology. First the essential background data are presented (Section 2). Then we compare Slavonic possessive adjectives with a notion of the ideal case of Suffixaufnahme (Section 3). We finally examine why it is that Slavonic possessive adjectives are prevented from giving rise to prototypical Suffixaufnahme constructions (Section 4). 2. Background

2.1. The Slavonic Languages The Slavonic family is a fairly conservative branch of Indo-European. It is traditionally divided into three groups—East, South, and West Slavonic— though there are shared changes which cut across these boundaries. Word

266

Indo-European

order within the clause is generally determined by information structure, with old, shared information coming before new. To the extent that this effect can be neutralized, the most common basic word order is subject-verb-object.1 Within the NP the order is more rigid, though there is variation among languages. Slavonic preserves a rich inflectional morphology, primarily fusional in nature. Most of the family retain around six cases, including a genitive case with various functions (such as marking government by verbs and prepositions), and there are distinct declensional types. There is little difficulty in establishing a word-class of adjectives and a word-class of nouns, though certain items may be seen as transitional between the two.

2.2. The Possessive Adjective The possessive adjective is widely used in Slavonic. A detailed discussion and extensive bibliography can be found in Corbett (1987).2 We repeat the salient points here, drawing material especially from those languages where the possessive adjective has the widest range, particularly from Upper Serbian.3 For example, in Upper Serbian the possessive adjective is the normal method of expressing what is conveyed by the genitive in many other languages. It would be unusual to find a one-word adnominal genitive referring to a definite person, as in (1): (1)

?

kniha Jan-a book Jan-GenSg 'Jan's book'

The normal expression in the spoken language would be:4 (2) Jan-ow-a knih-a Jan-Poss-NomSgFem book-NomSgFem5 'Jan's book'

Here we have a possessive adjective, formed from the noun Jan, by means of the suffix -ow-. In Upper Serbian, as in other Slavonic languages, possessive adjectives are formed from nouns by suffixation. There are two suffixes, -in-/ -yn-l-n- and -ow- in Upper Serbian, and -ov-l-ev- for the latter suffix in certain other languages. The distinctions within each are partly phonological and partly orthographic and need not concern us. We shall refer to the pair as -in-/ -ov-. There are other suffixes used in Slavonic for forming denominal adjectives, but these will not concern us here. The selection of the possessive adjective suffix is interesting. In Upper Serbian, the criterion is gender: feminine nouns take -in-l-yn- (thus, zona 'woman' gives zoniny 'woman's'), and masculine nouns take -ow- (for example, nan 'father', nanowy 'father's'; starosta 'headman', starostowy 'headman's'). 6 Some other languages, like Russian, preserve an older situation (as found in Old Church Slavonic, the oldest recorded Slavonic language) in

Slavonic's Closest Approach to Suffixaufnahme:

The Possessive Adjective

267

which the determining feature was declensional class. For most Slavonic nouns gender is predictable from the declensional class, so that in these instances both criteria give the same result. Simplifying somewhat, we may say that nouns with no ending in the nominative singular and which follow a particular declension (declensional class I) are masculine, and form the possessive with -ov-; nouns with the ending -a in the nominative singular follow a different declension (II), are feminine, and form the possessive adjective with -in-. However, there is a group of nouns like Russian papa 'Daddy' which follow declension II but which are masculine. Originally the possessive adjective suffix was determined by the declensional class, giving forms like Modern Russian papin 'Daddy's'. In those Slavonic languages that have gone over to the gender criterion, such nouns take the -ov- suffix (as in Upper Sorbian starosta 'headman', starostowy 'headman's'). These are Czech, Slovak, Upper and Lower Sorbian in the West Slavonic branch (but not Polish), Slovene, Bulgarian, with Macedonian moving in this direction in South Slavonic (but not Serbo-Croat), and Belorussian in East Slavonic (but not Russian and Ukrainian). This is a surprising list of languages to undergo a similar change. (There is considerable though not complete overlap with the list of languages that modify the declension of masculine class II nouns to make them more like the majority of masculines, the class I nouns.) While following one or the other criterion, languages may show sporadic exceptions. Thus Bulgarian has gone over to the gender criterion, but preserves bastin 'father's' from basta 'father'. 7 The important point is that the selection of the suffix is determined by the "underlying" noun; these suffixes do not attach freely. Returning to example (2), we note that besides the possessive suffix the adjective carries an agreement marker for gender, number, and case. Throughout Slavonic,8 the possessive adjective marks just the same distinctions in agreement as do other adjectives. In Upper Sorbian, the endings of the possessive adjective are phonologically identical with those of other adjectives; in some other Slavonic languages, like Russian, the possessive adjective has some endings that are phonologically distinct from other adjectives, though marking the same distinctions. There is an obvious contrast between (1) and (2): the adnominal genitive follows the main noun, while the possessive adjective precedes. There is variation across Slavonic, but the pattern here is the most common one, and again the possessive adjective shows the same behavior as normal attributive adjectives. 2.3. Restrictions on the Possessive Adjective There are several interacting restrictions on the possessive adjective, which apply to varying degrees in the different Slavonic languages. In some instances it is difficult to determine exactly where the restriction should be stated. However, the restrictions can be classified into two types: restrictions on formation and restrictions on use.

268

Indo-European

2.3.1. Restrictions on Formation There are restrictions on formation of a straightforward morphological type: certain stem types preclude the addition of the possessive adjective suffix. Thus in Upper Serbian female surnames and names in -ki do not allow the formation of the possessive adjective: given a name like Cisinski, no possessive adjective can be formed (Fasske 1981: 382). The precise restrictions vary from language to language; see, for example, Huntley (1989: 25) for the restrictions in Old Church Slavonic. There are other, more interesting restrictions on the formation of the possessive adjective. In Upper Sorbian, the possessive adjective can be formed when the referent is human, and occasionally also when it is an animal. Furthermore, the referent must be singular and it must be definite. These possessive adjectives cannot be used with a plural referent (Fasske 1981: 383): (3) *nas-ich muz-ow-e praw-o our-GenPl husband-Poss-SgNeut right-SgNeut 'our husbands' right'

(We return to the case of attributive modifiers like naSich in Section 3.2 below.) In such an instance the genitive must be used and not the possessive adjective: (4) praw-o nas-ich muz-ow right-SgNeut our-GenPl husband-GenPl9 'the right of our husbands'

As just mentioned, Fasske states that a definite referent is required for the use of the possessive adjective. Thus the possessive adjective muzowy must indicate a specific husband; it cannot be used generically. Nor can it have an indefinite referent: (5)

*nekajk-eho muz-ow-e praw-o some-GenSg husband-Poss-SgNeut right-SgNeut 'some husband's right'

Again, the adnominal genitive must be used. The particular restrictions vary from language to language, but one remains constant: with the suffixes we are discussing the referent must be singular in all the Slavonic languages. More generally in Slavonic, the possessive adjective is more likely to be formed when the referent is human than when it is an animal. Some languages, like Upper Sorbian, allow the formation in both cases, but favor a human referent. In others, such as Russian, the possessive adjective is normally formed only when the referent is human. If the possessive adjective can be formed when the referent is an animal (as in Serbo-Croat and Slovene), then usually hieher nnimals are involved Occasionally inanimates are in

Slavonic's Closest Approach to Suffixaufnahme: The Possessive Adjective 269

eluded; Stevanovic (1974: 185) gives suncev 'sun's' and mesecev 'moon's' as possible in Serbo-Croat. A definite referent is a requirement in Upper Serbian, and also in Modern Russian (Trubetzkoy 1939: 82) and in Bulgarian (Andrejcin 1978: 262). In Serbo-Croat, when common nouns are involved and there is a definite human referent, the possessive adjective is much the more frequent. If the referent is indefinite, the possessive adjective may still be used, but so may the genitive (Stevanovic 1974: 183-184). It is significant that the inanimates given as having a possessive adjective are 'sun' and 'moon', which typically have definite referents. This factor also has an effect on the choice between the possessive adjective and the genitive in Czech and Slovak. The conditions are not equally well studied in the different Slavonic languages (but see especially Huntley 1984 on Old Church Slavonic and Ivic 1986 on Serbo-Croat). On the basis of the existing evidence, two hierarchies may be proposed (Corbett 1987: 324): (6) a. human > animal > inanimate b. definite > indefinite

The higher the referent on the hierarchies, the more likely the possessive adjective is to be formed, the prototypical case being reference to a definite human. Of course, these hierarchies are familiar from other studies. Different languages have different cut-off points for the formation of the possessive adjective. Upper Serbian requires a definite referent; Serbo-Croat and Slovene do not. Russian requires a human proper noun or other human referent while Serbo-Croat does not. The history of the competition between the possessive adjective and the genitive can be seen as a progressive tightening of the restrictions on the possessive adjective in terms of the two hierarchies above, though their influence was evident even in Old Church Slavonic. When a particular noun can never fall within the constraints operating at the time, it may be said normally not to form the possessive adjective, though the latter may be possible in special circumstances. 2.3.2. Restrictions on Use We have examined the conditions under which a possessive adjective may be formed. However, its use, rather than that of the genitive, depends on two further sets of conditions. First, as is well known, the genitive can be used for a wide variety of roles. The possessive adjective is not available in all of these to an equal degree. Thus in nominalization constructions, we find subjective and objective genitives. Staying with Upper Serbian (Fasske 1981: 386) we find that the possessive adjective can take on either role: (7)

Hilz-in-y wopyt Hilza-Poss-NomSgMasc visit (NomSg) 'Hilza's visit'

270

Indo-European

(8) Jur-ow-y pohrjeb Juro-Poss-NomSgMasc burial (NomSg) 'Juro's burial'

In Slovene, in contrast, informants accepted the possessive adjective in the subjective role but were very reluctant to accept it in an objective role (where the genitive was much preferred). More generally "the range and frequency of use of the possessive adjective for the subjective genitive are at least as great as for the objective genitive" (Corbett 1987: 330). The other type of restriction on use appears to be clearly syntactic; the question is whether the possessive adjective may have dependents of certain types or not. Consider the following Upper Serbian example (Michalk 1974: 510): (9) moj-eho bratr-ow-e dzec-i my-GenSgMasc brother-Poss-NomPl child-NomPl 'my brother's children'

Bratrowe is a possessive adjective formed from the noun bratr 'brother'. The ending -e shows agreement with the head noun dzeci. The problem is the form mojeho; clearly this does not agree with dzSci, since it carries the wrong features. It seems rather that its agreement controller is the possessive adjective bratrowy,10 which in some sense includes the noun bratr 'brother', which is masculine singular. This means that the possessive adjective may be used in Upper Serbian even when the corresponding expression with the adnominal genitive includes an adjectival form dependent on the genitive noun: (10)

dzec-i moj-eho bratr-a child-NomPl my-GenSgMasc brother-GenSgMasc 'the children of my brother'

Besides controlling an attributive modifier as in (9), the possessive adjective may control relative and personal pronouns. Upper Serbian is unusual in this respect, with the possibilities of most of the Slavonic languages being more restricted. The behavior of the possessive adjective in (9) is unusual both within Slavonic and beyond it. Normally in Slavonic the presence of an adjectival modifier as in (10) means that the genitive is the required form. Another way of looking at the problem is to say that the suffix (-ow- in this instance) marks the phrase (moj bratr 'my brother') rather than just the noun to which it is attached. It is this construction which is most closely related to Suffixaufnahme, and so will be discussed more fully in Section 3. Before moving to that, it should be said that there are other types of syntactic constraint: for example, Old Church Slavonic allowed a possessive adjective to be conjoined with an adnominal genitive, while Upper Serbian does not; for this and other such restrictions see Corbett (1987: 332-337).

Slavonic's Closest Approach to Suffixaufnahme: The Possessive Adjective 271 Table 8.1. Frequency of Use of the Possessive Adjective (Based on Data from Ivanova 1975, 1976).

East Slavonic Russian Belorussian Ukrainian South Slavonic Slovene Serbo-Croat West Slavonic Polish Czech Slovak

overall

in situations of choice only

10% 36% 23%

22.3% 64.6% 48.9%

66% 52%

98.2% 93.1%

3% 51% 42%

5.8% 94.3% 83.0%

2.3.3. Relative Frequency of Possessive Adjective and Adnominal Genitive Given the different interacting conditions on the formation of the possessive adjective and on its use, it is not surprising that its frequency relative to the adnominal genitive varies widely from language to language. The comparative data available are less good than might be hoped for, but they are suggestive. Ivanova (1976: 9-10) gives figures based on contemporary literature, criticism, and journalism. For each language investigated she scanned 1,000 pages (counting 2,000 characters as a page). She gives the approximate frequencies of use of the possessive adjective,11 which are expressed as a percentage of the total instances of the possessive adjective and of the genitive (without preposition) in the first column of Table 8.1. These figures illustrate the difference in usage of comparable constructions in related languages (bear in mind that in some instances the possessive adjective would be excluded by the restrictions discussed above). In an earlier publication (Ivanova 1975: 151) we find comparable figures for instances where the use of the possessive adjective is theoretically possible in these languages (for a singular referent; with no modifier in the corresponding genitive phrase;12 and not expressed by an adjectival noun, which could not form a possessive adjective). These data are given in the righthand column of Table 8.1. Naturally, the possessive adjective achieves a higher frequency under these conditions, but the differences among the languages investigated are equally clear.

3. The Possessive Adjective and Suffixaufnahme We now consider the characteristics of Suffixaufnahme in turn, and examine whether the possessive adjective can be said to exhibit Suffixaufnahme.

272

Indo-European

3.1. Double Marking of Case The defining characteristic of Suffixaufnahme is "for nouns in an attributive relation to agree with their head nouns in case" (Plank 1990: 1039). Specifically, the attributive relation is marked by case (normally genitive) on the attributive noun and this has a second, agreeing case marker; that is to say, it has a double inflectional marking. Let us look again at one of our Upper Serbian examples (2) in this light: (11)

Jan-ow-a knih-a Jan-Poss-NomSgFem book-NomSgFem 'Jan's book'

The -a ending on Janowa does indeed mark agreement in case with the head noun kniha, and this part of the requirement is met.13 The difficulty is whether we have precisely a noun showing agreement in case. This problem relates directly to the means of marking of the attributive relation, namely the suffix -ow-, which is not the normal case marker. The straightforward case marker would here, by coincidence, be -a, as in (1), repeated as (12). (12)

?

kniha Jan-a book-NomSgFem Jan-GenSg 'Jan's book'

For forms like Jan-ow-a there are, then, two related questions to be asked: the first is whether we have a noun here, and the second is whether we have two inflections or one. Correspondingly, we must decide between two possible structures: (X) Jan NOUN

(Y) Jan [NOUN ROOT

-ow

-a

INFLECTION

INFLECTION

-ow DERIVATION]^

-a INFLECTION

If structure (X) is correct, then we would seem to have an instance of Suffixaufnahme, since we would have a noun marked inflectionally for case to show its attributive function and marked a second time to show agreement with the head. If (Y) is correct, then this is not an instance of Suffixaufnahme, since there is no noun directly involved and there is only one instance of inflectional marking of case. There is a long-running debate about these structures, though not exactly in these terms. The debate goes back to Trubetzkoy (1937: 16), who examined the situation in Old Church Slavonic and claimed that since every noun denoting an animate has a possessive adjective, the latter should be considered a

Slavonic's Closest Approach to Suffixaufnahme: The Possessive Adjective 273

part of the paradigm of the noun (just as participles were considered to be part of the paradigm of the verb). The conclusion from this line of argument (essentially that of productivity) would be that the forms called possessive adjectives should be treated as a matter for inflectional rather than derivational morphology, which would favor analysis (X). The subsequent debate is reviewed in Corbett (1987: 304-307), and only the main points will be repeated here. The second main line of argument, namely the syntactic argument, goes back to Lotzsch (1965); he considered data from various Slavonic languages and claimed that certain syntactic characteristics of the possessive adjective, in particular control facts, are shared with the noun. Consider the following example, which shows that the possessive adjective is able to control a relative pronoun in Upper Serbian14 (Lotzsch 1965: 378; for more examples see Fasske 1981: 385): (13)

slysetaj Wicazowy hlos, kotryz je zastupil [they] hear Wicaz's voice, who is gone=in 'They hear Wicaz's voice, who has gone in'

The relative pronoun kotryz is masculine singular; the sense shows that its antecedent is Wicaz, the noun underlying the possessive adjective, and not hlos (which is also masculine singular). This is behavior more like that of a noun (strictly an NP) rather than that of an adjective, favoring analysis (X). And this line that emphasizes the syntactic facts is in accord with Anderson's view that "inflectional morphology is what is relevant to the syntax" (1982: 587). The control possibilities of possessive adjectives in Slavonic show that their formation is relevant to syntax. The third argument is that in Upper Serbian the suffixes -in-l-ow- may mark NPs, even though they attach just to nouns. We consider this phenomenon in Section 3.2 below. At this point it is only necessary to note that marking a phrasal feature on a single item is something which regularly occurs in inflectional morphology. Against this, and again without repeating the detail given in Corbett (1987), there are three strong arguments in favor of analysis (Y). The first concerns word-class membership. If we look first at syntax, then, as noted earlier, Upper Sorbian attributive adjectives typically precede the head noun, while nouns in an attributive relation follow (as in (4)). The possessive adjective precedes the noun, copying the behavior of undisputed adjectives. And when we turn to morphology, then we note that the possessive adjective agrees with the noun in gender, number, and case. This is what we expect of adjectives, rather than nouns. However, Suffixaufnahme is unusual in precisely this respect, so the mere fact of agreement does not make the point. But the purely morphological data do support the adjective analysis (Y); the point is that noun endings and adjective endings are not identical in Upper Sorbian, and when they differ, the possessive adjective takes endings identical with those of other adjectives:

274

Indo-European

(14)

wucerj-ow-eje dzowk-i teacher-Poss-GenSgFem daughter-GenSg 'of the teacher's daughter'

(15)

stars-eje dzowk-i elder-GenSgFem daughter-GenSg 'of the elder daughter'

>

This identity is not always found in other Slavonic languages, and there are instances where particular endings are as those of nouns; however, I believe that the patterns of syncretism within the paradigms always match those of other adjectives and not those of nouns. There are, then, good arguments that possessive adjectives are indeed adjectives, and so in favor of structure (Y); this would mean that the formation of possessive adjectives involves a change of category, which many take as a sufficient condition for a phenomenon to be recognized as derivational in nature (Anderson 1982: 586). A second argument in favor of the derivational account concerns the ordering of affixes. It is well known that derivational affixes are typically found "inside" inflectional affixes (closer to the root). Moreover, in Slavonic we generally find a single inflectional affix per stem. Since -in-l-ow- occur inside the undisputed inflectional affixes, this supports the claim that they are derivational. The third argument in favor of the derivational analysis is that derivational morphology allows idiosyncratic gaps and idiosyncratic possibilities to a much greater extent than does inflectional morphology. Thus nouns may decline or not decline, but we do not find nouns in Slavonic which have, say, only a dative singular form, nor which have all forms except a dative singular.15 But there are countless examples of items which allow or do not allow particular derivational formations. Richter (1980: 116-117) points out that Upper Serbian has a small number of indeclinable nouns, such as abbe 'priest'. These nouns cannot take inflectional endings, but they may form possessive adjectives, as in the case of abbeowy 'priest's', which again suggests that possessive adjectives are derivational formations. Given these strong but conflicting arguments, it would appear that (as suggested in Corbett 1987: 327-329) the possessive adjective shows features both of inflectional and of derivational behavior, to varying degrees in different Slavonic languages. This illustrates a more general claim that there is no clear dividing line between inflectional and derivational morphology (compare Plank 1991). Rather than being able to assign the possessive adjective to structure (X) or (Y), we need to place it somewhere in between, varying according to language. Fasske (1981: 381-388) assigns it to a separate part of speech; certainly we are dealing with an unusual object, and the lack of items with similar behavior in the relevant languages makes it hard to classify. It seems clear, however, that we cannot assign the possessive adjective in Upper Serbian uniquely to structure (X). It does not therefore represent a prototypical case of

Slavonic's Closest Approach to Suffixaufnahme: The Possessive Adjective 275

Suffixaufnahme. The rest of this section will be concerned with other aspects of its behavior measured against the standard of typical Suffixaufnahme. 3,2. Noun or Noun Phrase? A question that was alluded to in the discussion of (13) above concerns the root of the possessive adjective, specifically whether this root is a noun or an NP. Since case marking typically affects NPs, then a construction would be deemed closer to prototypical Suffixaufnahme if whole NPs were involved rather than just nouns. So we must ask whether possessive adjectives can be "formed from" NPs rather than just from nouns. Consider this Upper Serbian example (Fasske 1981: 385): (16)

To je nas-eho wucerj-ow-a zahrodk-a. Won wjele w njej dzela. that is our-GenSgMasc teacher-Poss-NomSgFem garden-NomSgFem. He=NomSgMasc a=lot in it works. 'That is our teacher's garden. He [our teacher] works in it a lot.'

The personal pronoun won takes as its antecedent the NP nas wucer 'our teacher', which underlies the phrase headed by the possessive adjective (something that is not possible for other types of adjective). In other words, in Upper Serbian the suffixes -in-l-ow-, though they attach to nouns, may mark NPs. This phenomenon, the marking of a single item for a feature which rightly belongs on a whole phrase, is often found in inflectional morphology (Carlson 1983: 73): case may be marked just on the noun which heads an NP, definiteness may be marked just on an adjective but not on its head noun, and so on. If a construction allows the attributive relation to be marked in this way, we should consider it closer to prototypical Suffixaufnahme than one in which only bare nouns were involved. There is a terminological problem here: the Slavonic languages regularly allow possessive adjectives to control anaphoric pronouns, though most do not allow examples like (16); yet it is normally assumed that NPs, not nouns, control anaphoric pronouns. It would be more precise, therefore, to say that the formation of the possessive adjective is restricted in many languages to NPs consisting of a bare noun, while some, notably Upper Serbian, allow it to be formed from more complex NPs. We have already seen an example of control of a relative pronoun ((13) above). The remarkable thing about Upper Sorbian is that it allows the possessive adjective to control an attributive modifier (equivalently, it allows the possessive adjective to be formed from an NP consisting of attributive modifier and noun). We can see this again in the next example (from Fasske 1981: 382-383): (17)

moj-eho muz-ow-a sotr-a my-GenSgMasc husband-Poss-NomSgFem sister-NomSgFem 'my husband's sister'

276

Indo-European

In (17), the suffix -ow- may be thought of as marking the phrase moj muz 'my husband'. To it is added the inflectional marker for nominative singular feminine, showing agreement with the head noun sotra 'sister'. The particularly interesting form is mojeho; this is masculine because muz 'husband', which "underlies" muzowa, is masculine. It is singular for the same reason (recall that the formation of the possessive adjective requires a singular referent). However, its case is genitive. The genitive is the typical attributive case, but this case is not shown directly on the head of mojeho. There is no possibility of a marker showing the case of the main head of the NP being added to mojeho (*moj-ehoa). Thus the possessive adjective is possible in Upper Serbian, even though the NP consists of more than a bare noun; but the case of the main head is marked by the possessive adjective itself, and not by its modifiers. Looking at (17) in a slightly different way, we may say that the phrase mojeho muzowa 'my husband's' does show a type of Suffixaufnahme, in the sense that the modifier mojeho marks the attributive (by means of the genitive), while the possessive adjective muzowa marks agreement in case with the head noun; thus two members of the phrase "cooperatively" show Suffixaufnahme. This construction in Upper Sorbian is indeed remarkable; the only other modern Slavonic language which has it, and to a more limited extent, is Slovak. Control of the relative pronoun, as in (13) above, is much more common, while control of the anaphoric pronoun, as in (16), is general (except for being limited in Polish). There is an implicational hierarchy of targets for the possessive adjective: (18) attributive < relative pronoun < personal pronoun This control hierarchy can in turn be subsumed under the Agreement Hierarchy (Corbett 1987: 318-319).

3.3. Attraction We saw in the last section that an attributive modifier controlled by a possessive adjective stands in the genitive, irrespective of the main head noun, and that it cannot take an additional suffix-copying marker to agree in case with the main head noun. In this respect, the phenomenon known as attraction is especially relevant. The following example is as expected, comparable to (17) above: (19) w nas-eho nan-ow-e chez-i in our-GenSgMasc father-Poss-LocSgFem house-LocSgFem16 'in our father's house' Besides such examples, it is possible for the attributive modifier to take on the same features as those of the possessive adjective and head noun, thus showing "attraction." This example is from Sewc-Schuster (1976: 27); see also Sewc-Schuster (1968: 103) and Fasske (1981: 384):17

Slavonic's Closest Approach to Suffixaufnahme: The Possessive Adjective 277

(20)

w nas-ej nanow-ej chez-i in our-LocSgFem father-Poss-LocSgFem house-LocSgFem 'in our father's house'

In (20) naSej agrees in case (and gender and number) with the main head noun chezi (which is locative, singular, and feminine).18 Note that in order to do this, it "drops" its normal agreement inflection and so does not agree with the noun underlying the possessive adjective. Thus (20) is closer to a prototypical Suffixaufnahme construction than is (19), but the fact that one suffix is dropped to make way for the other means that it is certainly not a prototypical Suffixaufnahme construction. It is worth comparing the possibilities shown by (19) and (20) with prototypical Suffixaufnahme. If we have an NP in an attributive relation to another NP, then in a language with prototypical Suffixaufnahme we would expect this structure: (21)

[NP, [Np2 ADJ-CASE2-CASE1 N-CASE2-CASE1] N-CASE1]

The attributive NP (NP2) "should" be marked by case, indicated as 'CASE2' (probably the genitive). The main NP (NPJ "should" also be marked by case (CASE1), according to its role in the sentence. This same case should appear on the attributive NP, outside the attributive case marking, thus showing a mirror-image effect (Plank 1990: 1042-1043). What we actually get in Upper Serbian is a marker of the attributive relation on the noun in NP2, but not one which is a normal case marker, as discussed earlier; this is followed by the marker of CASE1. The adjective usually marks the attributive relation (CASE2: genitive) as in (19) above. It may instead show CASE1, in the attraction examples such as (20). But it must show one or the other case; it may not show both.

3.4. Recursion A final yardstick against which we can measure the possessive adjective construction is that of recursion; the most developed Suffixaufnahme systems allow recursion. This is very rare with the possessive adjective; however, Lotzsch (1965: 378) quotes the following Upper Serbian example: (22)

pfez Mar-in-eje macer-n-u smjerc through Marja-Poss-GenSgFem mother-Poss-AccSgFem death (AccSgFem)19 'through Marja's mother's death'

The possessive adjective macernu controls the preceding possessive adjective Manneje, which is therefore genitive singular. Informants accept (22) but say that the use of the adnominal genitive would be much more likely, especially in the spoken language.20 This is perhaps the severest test—examples are rare

278

Indo-European

even in undisputed Suffixaufnahme languages—and so it is not surprising that such recursion is rare in Slavonic. 4. The Problem We have seen that the possessive adjective construction in Slavonic, and particularly in Upper Sorbian, shares several features with Suffixaufnahme, and yet could not be called an instance of prototypical Suffixaufnahme. It appears that this is not only the closest that Slavonic21 comes to a Suffixaufnahme construction but also the closest it can come. We now consider why. There are two types of reason, though they are related: the first concerns the morphological type of the Slavonic languages, and the second concerns general morphological and phonological patterns. The Slavonic languages are typical fusional languages (though less so in the cases of Bulgarian and Macedonian). Individual inflectional formatives are associated with particular categories (such as noun or adjective), and even within a category they are often restricted to particular sets of stems (inflectional classes). They are not free to attach promiscuously to items of different types. There is no genitive inflection as such: rather there is an inflection marking genitive and singular on nouns of inflectional class I, another for class II, two more for the plural, and so on. This is an unwelcoming environment for Suffixaufnahme. However, the suffixes -in-l-ov-, which are an unusual pairing of suffixes, partially compensate for the problem of fusional morphology. The distribution of the two according to the gender or inflectional class of the host noun varies within Slavonic, as discussed earlier. But they are not a complete pair; for example, as the possessive adjective has declined in Russian, the forms in -inhave been better preserved than those in -ov-. They do not really fulfill the role of "catalytic affix" (Plank 1990:1040), since they do not attach freely and since they form a stem which in turn accepts formatives only of the appropriate type. They cannot escape from the fusional nature of the morphological system. The only other way in which Slavonic might achieve Suffixaufnahme is by attaching a second set of case endings directly to the genitive ending. This would be totally contrary to the underlying grain of the morphology, as already described. Nevertheless, Suffixaufnahme is unusual in any case, and so this possibility should not be rejected out of hand. But besides the general type of reasons already advanced, which apply equally here, there is a second, more specific, type of reason why Suffixaufnahme of this nature could not appear in Slavonic. Nominal stems in Slavonic typically end in a consonant (compare Stankiewicz 1986: 279). Inflections typically begin with a vowel and end in a vowel or consonant, more usually a vowel (and certainly genitive endings are usually vowel-final, often consisting of a single vowel). Combinations of vowel plus vowel are avoided, to varying degrees in different languages. Thus the possibility of adding a second ending to a genitive ending is all but excluded: the genitive does not provide a stem for further inflections.

Slavonic's Closest Approach to Suffixaufnahme:

The Possessive Adjective

279

This can only be done by formatives like -in-l-ov- and, as we have already seen, there are insuperable problems with this route as far as prototypical Suffixaufnahme is concerned. 5. Conclusion Constructions with possessive adjectives vary considerably within the Slavonic languages. In languages where they have the greatest possibilities, notably Upper Serbian, there are constructions which share characteristics with Suffixaufnahme. However, these do not have all the features we associate with prototypical Suffixaufnahme. This is because the fusional morphology of these languages prevents the free combining of formatives, which is one of the requirements of Suffixaufnahme. Notes / am grateful to Dr. Gerald Stone and to participants at the Suffixaufnahme Conference/ am grateful to Dr. Gerald Stone and to participants at the Suffixaufnahme Conference (Konstanz, September 1991) for helpful comments on an earlier version of this paper. Errors are mine. 1. In Upper Serbian, which will figure prominently, in the unmarked order the verb stands in final position (in main and subordinate clauses) except that auxiliaries, which are clitics, stand in second position (see Stone 1993: 652-656). 2. Additions to the bibliography there include: Ermakova (1986), Ivic (1986), Molosnaja (1987), Brauer (1988), Golovceva, Ivanov, Molosnaja, Nikolaeva and Svesnikova (1988), Huntley (1989), and Ivanov (1989). 3. Upper Serbian belongs to the West Slavonic group and is spoken in Lusatia, in the eastern part of Germany. Its cultural center is Budysin (Bautzen). Taking Upper Serbian together with the related Lower Sorbian, the total number of speakers is around 70,000. All adult speakers are bilingual with German, though some speak Sorbian better than German. However, the construction under consideration—in particular the control of adjectival modifiers by the possessive adjective—is not found in German but is found in other Slavonic languages, so there is no reason to suggest it is of German origin. For data we rely heavily on Fasske (1981: 381-388). 4. The genitive as in (1) is possible if the logical accent falls on Jana, or in high literary style. There is a third alternative, which was judged the second most common variant, after (2); it is a colloquial and dialectal form: kniha wot Jana book of Jan 'Jan's book'

The inclusion of the preposition wot is a Germanism. 5. Kniha is a feminine noun, with which Janowa agrees. However, while the -a ending is nominative and singular, it is not necessarily feminine: most, but not all, nouns of this declensional class are feminine. For simplicity we shall continue to gloss gender as in this example, and this caveat for nouns should be borne in mind.

280

Indo-European

6. The stems are nan- and starost- respectively; in the nominative (citation) form, nan- takes -£f while starost- takes -a. 7. For simplicity we omit original i-stems; few form possessive adjectives. 8. In some southern and western dialects of Czech discussed by Vachek (1954; 1961: 29-31) the possessive adjective has become indeclinable. 9. In this example the -ow is an inflectional marker (the genitive plural for this paradigm) and not the suffix for forming the possessive adjective. 10. This is the citation form, the nominative singular masculine. 11. The number of actual instances is not given by Ivanova. 12. This restriction does not apply fully to Slovak, as mentioned in Section 3.2, but Ivanova does not comment on this. 13. In this instance the agreeing case marker on the attributive adjective matches exactly that of the head; this phonological identity is not a requirement for Suffixaufnahme (Plank 1990: 1040), and there are other instances in Upper Serbian where the agreeing case marker is not identical to the marker on the head noun. 14. Around half the Slavonic languages allow this; see Corbett (1987: 319). For recent discussion of Slovene see Oresnik (1992). 15. There are isolated instances of Russian nouns that have no genitive plural. 16. While we have translated nas nan as 'our father' it may also mean 'my father'. Even an only child can use it with this meaning, which may well be connected with the previously widespread use of the plural wy 'you' as a form of respect for father or mother (for which see Stone 1976: 190-191). Moj nan 'my father' occurs, but some Sorbs find it odd (Gerald Stone, personal communication). 17. Fasske (1981: 384) suggests that attraction requires that the possessive adjective (and therefore its head noun) be in the singular. But Sewc-Schuster (1968: 103) gives examples with the plural, like nase nanowe pola 'our father's fields'. For attraction in Old Church Slavonic see Huntley (1989: 28) and for rare dialect examples of attraction in Slovak see Stole (1966: 26). 18. The preposition w 'in' is phonetically null, though it has a variant we which is pronounced (Stone 1993: 614-615). 19. The accusative singular of smjerc 'death' has a null ending, like the nominative; it is feminine. 20. See Huntley (1989: 28) for examples of how the stacking of possessive adjectives is avoided in Old Church Slavonic. 21. According to Watkins (1967: 2195), within Indo-European the possessive adjective is best represented in Slavonic and in Aeolic Greek. See also Plank's introduction for similar patterns in Romany and Anatolian Indo-European languages.

References Anderson, Stephen R. 1982. Where's morphology? Linguistic Inquiry 13, 571-612. Andrejcin, Ljubomir D. 1978. Osnovna balgarska gramatika. Sofia: Nauka i izkustvo. [First edition 1944; minor corrections in the 1978 edition.] Brauer, Herbert. 1988. Die possessiven Adjektiva auf -ov und -in des Russischen der Gegenwart und ihre Possessivitat: Teil II. Zeitschrift fur Slavische Philologie 48, 1-61. Carlson, Greg N. 1983. Marking constituents. In Linguistic Categories: Auxiliaries and

Slavonic's Closest Approach to Suffixaufnahme:

The Possessive Adjective

281

Related Puzzles, vol. 1: Categories, ed. by Frank Heny and Barry Richards, 6998. Dordrecht: Reidel. Corbett, Greville G. 1987. The morphology/syntax interface: Evidence from possessive adjectives in Slavonic. Language 63, 299-345. Ermakova, M. I. 1986. Kategorija posessivnosti v serboluzickom jazyke: Nekotorye osobennosti ee vyrazenija. In Slavjanskoe i balkanskoe jazykoznanie. Problemy dialektologii: Kategorija posessivnosti, ed. by L. E. Kalnyn' and T. N. Molosnaja, 220-229. Moscow: Nauka. Fasske, Helmut. 1981. Grammatik der obersorbischen Schriftspmche der Gegenwart: Morphologie. Bautzen: VEB Domowina Verlag. Golovaceva, A., V. Ivanov, T. Molosnaja, T. Nikolaeva, andT. Svesnikova. 1988. Tipologija funkcionirovanja posessivnyx konstrukcij v slavjanskix jazykax. In Slavjanskoe jazykoznanie: X Mezdunarodnyj s"ezd slavistov: Sofija, sentjabr' 1988 g.: Doklady sovetskoj delegacii, ed. by N. I. Tolstoj, 103-116. Moscow: Nauka. Huntley, David. 1984. The distribution of the denominative adjective and the adnominal genitive in Old Church Slavonic. In Historical Syntax, ed. by Jacek Fisiak, 217-236. Berlin: Mouton. Huntley, David. 1989. Grammatical and lexical features in number and gender agreement in Old Bulgarian. Paleobulgarica 13, 4, 21-32. Ivanov, Vjaceslav V. (ed.) 1989. Kategorija posessivnosti v slavjanskix i balkanskix jazykax. Moscow: Nauka. [Contributions by A. V. Golovaceva, V. V. Ivanov, T. N. Molosnaja, T. M. Nikolaeva, and T. N. Svesnikova.] Ivanova, T. A. 1975. Nekotorye aspekty sopostavitel'nogo analiza posessivnyx konstrukcij. (Na materiale sovremennyx slavjanskix literaturnyx jazykax.) Slavjamkaja filologija (Leningrad) 3, 148-152. Ivanova, T. A. 1976. K voprosu o sootnosenii upotrebljaemosti posessivnyx konstrukcii v sovremennyx slavjanskix jazykax. Voprosy filologii (Izdatel'stvo Leningradskogo universiteta) 5, 3-10. Ivic, Milka. 1986. On referential strategies: Genitivization vs. adjectivization in SerboCroatian. Linguistische Arbeitsberichte (Leipzig) 54-55, 23-27. Lotzsch, Ronald. 1965. Das sog. Possessivadjektiv im Slawischen, speziell im Sorbischen, und seine Stellung im System der Redeteile. Forschungen und Fortschritte 39, pt. 12, 377-379. Michalk, F. 1974. Kratkij ocerk grammatiki sovremennogo verxneluzickogo literaturnogo jazyka. In Hornjoserbski-ruski slownik, compiled by K. K. Trofimovic, 472-511. Bautzen/Budysin: Ludowe nakladnistwo Domowina. Molosnaja, Tat'jana. 1987. Pritjazatel'nye prilagatel'nye v slavjanskix jazykax. Sapostavitelno ezikoznanie (Sofia) 12, No. 6, 5-12. [This material is expanded in Ivanov 1989, 133-148.] Oresnik, Janez. 1992. O neki skladenjski posebnosti slovenskih svojilnih pridevnikov. Miklosicev zbornik 1992, 263-267. Plank, Frans. 1990. Suffix copying as a mirror-image phenomenon. Linguistics 28, 1039-1045. Plank, Frans. 1991. Inflection and derivation. EUROTYP Working Papers VII/10. Richter, Heinz. 1980. Die Possessivadjektive im Sorbischen unter Berucksichtigung der benachbarten slawischen Sprachen. Dissertation zur Promotion A, Karl-MarxUniversitat, Leipzig. Sewc[-Schuster], Hinc. 1968. Gramatika hornjoserbskeje rece, 1. zwjazk: Fonematika a morfologija. Bautzen/Budysin: Ludowe nakladnistwo Domowina.

282

Indo-European

Sewc-Schuster, Hfinc]. 1976. Gramatika hornjoserbskeje rece, 2. zwjazk: Syntaksa. Bautzen/BudySin: Ludowe nakladnistwo Domowina. Stankiewicz, Edward. 1986. Slavic morphophonemics in its typological and diachronic aspects. In The Slavic Languages: Unity in Diversity, by Edward Stankiewicz, 266-300. Berlin: Mouton de Gruyter. Stevanovic, M. 1974. Savremeni srpskohrvatski jezik (gramaticki sistemi i knjizevnojezicka norma), 2: Sintaksa. 2d ed. Beograd: Naucna knjiga. Stone, Gerald. 1976. Pronominal address in Serbian. Letopis Instituta za serbski ludospyt (Bautzen) Rjad A, 23/2, 182-191. Stone, Gerald. 1993. Serbian (Upper and Lower). In The Slavonic Languages, ed. by Bernard Comrie and Greville G. Corbett, 593-685. London: Routledge. Stole, Jozef. 1966. Vyjadrovanie posesivneho vzt'ahu v slovencine a jeho slovanska suvislost'. Slavica Slovaca 1; 3, 259-267. Trubetzkoy, N. S. 1937. O pritjazatel'nyx prilagatel'nyx (possessiva) starocerkovnoslavjanskogo jazyka. In Zbornik lingvistickih ifilologskih rasprava A. Belicu o cetrdesetogodisnjici njegova naucnog rada posvecuju njegovi prijatelji ucenici, 15-20. Beograd: Mlada Srbija. Trubetzkoy, N. 1939. Le rapport entre le determine, le determinant et le deflni. In Melanges de linguistique offerts a Charles Bally, 75-82. Geneva. Vachek, Josef. 1954. K problematice ceskych posesivnich adjektiv. Studie a Prace Linguisticke 1, 171-189. Vachek, Josef. 1961. Some less familiar aspects of the analytical trend of English. Brno Studies in English 3, 9-78. Watkins, Calvert. 1967. Remarks on the genitive. In To Honor Roman Jakobson: Essays on the Occasion of His Seventieth Birthday, vol. 3, 2191-2198. The Hague: Mouton.

9 Inflecting Postpositions in Indie and Kashmiri John R. Payne

1. Incidence An interesting form of Suffixaufnahme involving inflecting postpositions can be found in the majority of the modern Indie languages and dialects, including varieties of Sindhi, Lahnda, Punjabi, Gujarati, Marathi, Konkani, Rajasthani (Marwari and Jaipuri), Hindi (standard Hindi, Braj, Awadhi, Bundeli), Central Pahari (Kumauni, Garhwali), Western Pahari (Jaunsari, Sirmauri, Baghati, Kiunthali, Kului, Mandeali, Chameali, Bhadrawahi), Parya, and Romany. It is also found in Kashmiri, genetically a Dardic language but bordering on the Indie area. The languages and dialects where Suffixaufnahme is not found are essentially those belonging to the Eastern Indie group (Bengali, Assamese, Bhojpuri, Maithili, and Oriya), but they also include Nepali and the Eastern Hindi dialect Chhattisgarhi, which borders on Oriya, as well as Sinhalese and Maldivian.

2. The Construction The Indie Suffixaufnahme construction can be illustrated by the standard Hindi examples in (1) and (2). (1) a. Rani ka bhal Rani of=DctSgMasc brother=DctSgMasc 'Rani's brother' (direct) b. Ram ke bhal Rani of=OblSgMasc brother=OblSgMasc 'Rani's brother' (oblique) 283

284

Indo-European c. Rani ke bhal Rani of=DctP1Masc brother=DctPlMasc 'Rani's brothers' (direct) d. Rani ke bhaiyo Rani of=OblP1Masc brother=OblPlMasc 'Rani's brothers' (oblique)

(2) a. Rani kl bahan Rani of=DctSgFem sister=DctSgFem 'Rani's sister' (direct) b. Rani ki bahan Rani of=OblSgFem sister=OblSgFem 'Rani's sister' (oblique) c. Rani kl bahne Rani of=DctP1Fem sister=DctPlFem 'Rani's sisters' (direct) d. Rani kl bahno Rani of=OblP1Fem sister=OblPlFem 'Rani's sisters' (oblique)

Nouns in Hindi belong to one of two gender classes (masculine or feminine) and can show inflectional distinctions of case (direct, also known as absolute, or oblique) and number (singular or plural). Some syncretism is involved: except for masculine nouns ending in -a or -d, which have distinct forms for the oblique singular (-e and -e respectively), the forms of the direct and oblique singular are identical. For masculine nouns, the form of the direct plural coincides with that of the oblique singular; i.e., nouns ending in -a or -a have the endings -e and -e respectively, while nouns like bhal 'brother' have the same form for direct singular, oblique singular, and absolute plural. Feminine nouns always have a distinct form for the direct plural, and all nouns, whether masculine or feminine, have a distinct form for the oblique plural. While a noun in the direct case may occur in isolation, for example as the subject of an intransitive sentence or as the indefinite or non-specific object of a transitive sentence, nouns in the oblique case are invariably governed by one of a basic set of postpositions that serve to differentiate the general oblique function. These are: ne ko se me par tak ka

agent in the ergative construction definite or specific object, indirect, dative subject instrumental location inside location on top of as far as, up to possessor and other genitive functions

Inflecting Postpositions in Indie and Kashmiri

285

These postpositions have been termed "Layer II" case markers, as opposed to the "Layer I" morphological case distinction of direct versus oblique (Zograf 1976, Masica 1991). In possessor constructions like those in (1) and (2), the modified noun (bhai 'brother' or bahan 'sister') inflects with the Layer I case and number appropriate to the whole construction. The modifier NP (Rani) is governed by the genitive postposition kaV. Unlike any of the other Layer II postpositions, however, ka itself inflects for case, number, and gender in agreement with the modified noun. The inflectional forms are ka (direct singular masculine), ke (oblique singular/plural masculine and direct plural masculine), and ki (feminine). In an example like (la), therefore, the phrase Rani ka 'Rani's' simultaneously contains a genitive marker (the postposition ka itself), and a marker of the case of the whole construction (the -a inflection of ka agreeing with the direct case of the modified masculine noun). The crucial examples demonstrating the presence of Suffixaufnahme, insofar as Suffixaufnahme is a phenomenon of multiple case marking, are the examples with a masculine modified noun in (1), since the feminine form of the genitive postposition ki shows no case distinctions between direct and oblique. Phrases like Rani ki in (2a) have just the genitive marking of the postposition and lack overt case agreement with the modified noun. If, however, the notion of Suffixaufnahme is extended to categories other than case, for example number and gender, the number and gender inflections of ka become relevant. It might be argued that a phrase like Rani ke 'Rani's' in (1c) contains two manifestations of number (the singular number of the modifier Rani and the plural number of the modified noun expressed in the inflectional ending -e of ke), and two manifestations of gender (the feminine gender of the modifier Rani and the masculine gender of the modified noun expressed in the inflectional ending -e of ke). The modifying phrase can therefore assume the number and gender, as well as the case, of the noun it modifies. The Indie Suffixaufnahme construction is an interesting combination of morphological and phrasal marking. The direct and oblique case endings in Hindi can be clearly shown to be morphological; i.e., they are internal to the word that is marked. For example, when two nouns are conjoined, both must be separately marked for direct or oblique case. A single ending cannot serve as a marker of the whole phrase: (3) a. bahno aur bhaiyo sister=OblPlFem and brother=OblPlMasc b. *[bahan aur bhai]-yo sister and brother-OblPlMasc 'sisters and brothers' (oblique)

Also, declinable NP-internal modifiers such as adjectives bear independent inflectional endings:

286

Indo-European

(4) a. kale ghore black=OblSgMasc horse=OblSgMasc b. *[kal ghor]-e black horse-OblSgMasc 'black horse' (oblique)

By contrast, the postposition ka is phrasal. It can attach at the end of a conjoined phrase, and it does not independently mark NP-internal modifiers such as adjectives: (5) a. Ram aur Rani ke bhai Ram and Rani of=OblSgMasc brother=OblSgMasc 'Ram and Rani's brother' (oblique) b. *kale ki ghore ki akh black of=DctSgFem horse of=DctSgFem eye=DctSgFem 'black horse's eye' (direct)

In coordinate structures such as (5a), whether the postposition ka attaches to the conjoined phrase or to the individual conjuncts can have a semantic effect: (6) a. [Ram aur Rani ke] bhaiyo Ram and Rani of=OblP1Masc brother=OblPlMasc 'Ram and Rani's brothers' (oblique) b. [Ram ke] aur [Rani ke] bhaiyo Ram of=OblPlMasc and Rani of=OblPlMasc brother=OblPlMasc 'Ram's and Rani's brothers' (oblique)

Whereas the only possible interpretation of (6a) is that the individuals referred to are simultaneously brothers of Ram and Rani (i.e., Ram and Rani are themselves brother and sister), (6b) permits either this interpretation or an interpretation in which the referents are either brothers of Ram or brothers of Rani (i.e., Ram and Rani are not necessarily siblings). One consequence of this mixture of inflectional and postpositional marking is that in iterated genitive constructions, the case of the whole phrase cannot be carried beyond the most immediate modifier: (7) [[Ram ke] bhai ka] bhatija Ram of=OblSgMasc brother=OblSgMasc of=DctSgMasc cousin=DctSgMasc 'Ram's brother's cousin' (direct)

In (7), the most immediate modifier of bhatija 'cousin' is the phrase Ram ke bhai ka 'Ram's brother's', in which the Layer I inflection of the final postposition ka marks agreement in case, number, and gender with bhatija. The postposition ka itself governs the oblique case of bhai 'brother'. The internal modifier of bhai, Ram ke 'Ram's', then agrees via the inflection of the postposition ke with the oblique case, singular number, and masculine gender of bhai. The

287

Inflecting Postpositions in Indie and Kashmiri

direct case of the whole phrase is not transferred to the internal modifier Ram ke, which agrees solely with the noun bhai. Essentially, this follows from the fact that the postposition ka can have one and only one Layer I ending. 3. Variations In all the Indie languages in which it occurs, the Suffixaufnahme construction follows essentially the pattern illustrated in standard Hindi: an inflected Layer II postposition linked to a modifying NP shows Layer I agreement with a modified noun. This pattern is also found in Kashmiri, though here it is not the only Suffixaufnahme pattern, some agreeing genitives being constructed by purely word-internal means. The range of variation is limited to the role and form of the postposition, the case governed by the postposition, and the categories of agreement. 3.1. Indie All the Indie languages that have agreeing postpositions have an agreeing postposition with possessor and general genitive functions akin to Hindi ka. This postposition also invariably governs the oblique case. For those languages that have two genders (masculine versus feminine), the direct and oblique case forms are shown in Table 9.1. Three-gender systems (masculine, neuter, feminine) are found in the southwest of the Indie area (Gujarati, Marathi, and Konkani), with the forms of the genitive postposition illustrated in Table 9.2. It will be noted that the distinction between direct and oblique is invariably present in the masculine (and neuter) singular. Elsewhere there is wideTable 9.1. Genitive Postpositions in Languages with Two Genders.

Punjabi Lahnda Sindhi Marwari Jaipuri Parya Kumauni Stand. Hindi Braj Awadhi Bundeli Romany (Slovak)

DctSg Masc

OblSg Masc

DctPl Masc

OblPl Masc

DctSg Fem

OblSg Fern

DctPl Fem

OblPl Fem

da da; na, na jo ro ko ko ko ka kao ker ko kero/ gero

de de; ne je ra ka ka ka ke kae ke(re) ke kere/ gere

de de; ne ja ra ka ka ka ke kae ke(re) ke kere/ gere

de (dia) dia; nea, nia ju ji ra ka ka ka ke kae ke(re) ke kere/ gere

didi di; ni, ni ji ji ri i ki ki ki ki ki ki ki keri/ geri

di di; ni, ni ji (jia) ri ki ki ki ki ki ki ki kera/ gera

dia dia; nia je (ji) ri ki ki ki ki ki ki ki kere/ gere

dia dia; nia je, jini, juni ri ki ki ki ki ki ki ki kere/ gere

288

Indo-European Table 9.2. Genitive Postpositions in Languages with Three Genders.

Gujarati

Marathi Konkani

Dct Sg Masc

Obl Sg Masc

no

na

ca lo,

gelo

cya lya

Dct Pl Masc

Obl Pl Masc

Dct Sg Neut

Obl Sg Neut

na

na

nu

na

ce le,

gele

cya lya

ce le

cya lya

Dct Pl Neut

Obl Pl Neut

Dct Sg Fem

Obl Sg Fem

na

na

ni

ni

ci li

cya lya

ci li

cya lya

Dct Pl Fem

ni

cya lyo

spread syncretism: the direct-oblique distinction is rarer in the plural and feminine paradigms, but it is nevertheless preserved to some extent in the languages in the west of the Indie area (Marathi, Konkani, Lahnda, Sindhi, and marginally in the masculine plural in Punjabi). The distinction between direct and oblique is also preserved in the feminine singular in a limited number of Romany dialects. According to Ventcel' and Cerenkov (1976: 308), these are the dialects of the Slovak and Hungarian Gypsies, the dialect of the Vlakh Gypsies of Moldavia and Rumania, the Kelderari dialect (originally spoken on the Rumanian-Hungarian language border in the Austro-Hungarian Empire) and the Lowari dialect. The latter two dialects have spread throughout Eastern and Western Europe. Romany dialects are unusual in that the genitive postposition not only agrees with the case, number, and gender of the modified noun, but also has two allomorphs depending on the singularity or plurality of the modifier. In most dialects, including the dialect of the Slovak Gypsies shown in Table 9.1, the forms have initial k- and g- respectively. An illustrative paradigm of the direct forms (Ventcel' and Cerenkov 1976: 305) is given in (8). (8) a. [romes kero] chavo Gypsy=OblSgMasc of=Sg=DctSgMasc son=DctSgMasc '(male) Gypsy's son' (direct) b. [romes kere] chive Gypsy=OblSgMasc of=Sg=DctPlMasc child=DctPlMasc '(male) Gypsy's children' (direct) c. [romes keri] chai Gypsy=OblSgMasc of=Sg=DctSgFem daughter=DctSgFem '(male) Gypsy's daughter' (direct) d. [romn'a kero] chavo Gypsy=OblSgFem of=Sg=DctSgMasc son=DctSgMasc '(female) Gypsy's son' (direct) e. [romn'a kere] chave Gypsy=OblSgFem of=Sg=DctPlMasc child=DctPlMasc '(female) Gypsy's children' (direct)

Obl Pl Fem

ni

cya lya

Inflecting Postpositions in Indie and Kashmiri

289

f. [romn'a keri] chai Gypsy=OblSgFem of=Sg=DctSgFem daughter=DctSgFem '(female) Gypsy's son' (direct) g. [romern gero] chavo Gypsy=OblPlMasc of=Pl=DctSgMasc son=DctSgMasc 'Gypsies' son' (direct) h. [romen gere] chave Gypsy=OblPlMasc of=Pl=DctPlMasc son=DctPlMasc 'Gypsies' sons' (direct) i. [rome geri] chai Gypsy=OblPlMasc of=Pl=DctSgFem daughter=DctSgFem 'Gypsies' daughter' (direct)

Variants in other Romany dialects are kir-lgir-, kar-/gar-, kor-lgor-, kr-lgr- and t'ir-/d'ir-. Contrasting forms corresponding to the singular and plural of the governed noun are in general a feature of Romany Layer II postpositions, for example ke/ge (dative), telde (locative), tar/dar (ablative), and ha/ca (instrumental) in the Slovak dialect. This can be treated as an interesting formal confirmation of the status of the genitive as a Layer II postposition in Romany. In some Indie languages, agreeing Layer I case forms of the genitive postposition other than the basic direct and oblique are cited. Punjabi used to have an ablative (do or deo), but this is clearly archaic in modern Punjabi (Smirnov 1976: 380). Gujarati, Marwari, and Jaipuri appear to have an agentive (ne, kai, and re respectively), and Sindhi has a distinct vocative in hte masculine singular (ja) and feminine plural (jya) in addition to the direct form (ju). The rarity of such additional case forms follows simply from the fact that the main functional load of case marking in the Indie languages that have Suffixaufnahme is carried by Layer II postpositions rather than by Layer I cases. Even rarer in the Indie languages is an agreeing Layer II postposition with a function other than genitive. Smirnov (1976: 383-385) cites the possible existence in Punjabi of agreeing forms of locative postpositions such as vicla 'in' and nd rla 'inside' in examples like (9). (9) a. [p njab vicli] hal t Punjab=Obl in=DctSgFem situation=DctSgFem 'situation in the Punjab' b. [des nd rle] s maj k s mb ndh country=Obl inside=DctPlMasc social relationship=DctPlMasc 'social relationships inside the country'

In addition to these agreeing forms, Punjabi possesses ordinary invariant postpositions such as vicc 'in' and nd r 'inside'. What is more, the agreeing forms can themselves act as simple adjectives with a meaning corresponding to 'central' or 'internal' in examples like vicla morca 'central front' and nd rla ghol 'internal conflict'. It is possible therefore that examples such as (9a) and

290

Indo-European

(9b) should be translated as 'situation internal to the Punjab' and 'social relationships internal to the country', with vicla and nd rla treated as normal adjectives governing the oblique case of the preceding noun. In this case, Indic would have no examples of agreeing postpositions other than the genitive. 3.2. Kashmiri Kashmiri is unusual in that the form and type of genitive marker depends not only on the case, gender, and number of the modified noun, but also on inherent distinctions of the head noun of the modifying phrase, viz. common versus proper, masculine versus feminine, animate versus inanimate, and singular versus plural. The form un occurs with animate masculine singular proper nouns, uk occurs with inanimate masculine singular common nouns, und occurs with animate masculine singular common nouns, and hund occurs with feminine and plural masculine nouns. Of these four forms, hund and und behave in all respects like agreeing Layer II postpositions. There are four Layer I cases in Kashmiri, traditionally termed direct (or absolute), agentive, dative, and ablative. Given their functional diversity, dative and ablative are perhaps more appropriately designated oblique I and oblique II (Zaxar'in and Edel'man 1971). The paradigm of hund is then as in Table 9.3 (Zaxar'in and Edel'man 1971: 113). An example of hund attached to a conjoined phrase is given in (10): (10)

[m l'is ti m ji hund] kar father=OblI and mother=OblI of=DctSgMasc case=DctSgMasc 'father and mother's case'

The postposition hund governs the oblique I case of the modifying nouns m l' 'father' and m j 'mother' and itself agrees with the absolute case, masculine gender, and singular number of the modified noun kar 'case, affair'. It should be noted that the choice of the form hund (as opposed to und) is determined by the final noun in the conjoined NP. As an animate masculine singular common noun, m l' 'father' in isolation requires the postposition und (m l is Table 9.3. Inflection of Kashmiri Genitive Postposition hund.

Masculine

Feminine

Direct Oblique I Oblique II Agentive Direct Oblique I Oblique II Agentive

Singular

Plural

hund hind'is hind'i hind' hinz hinz'i hinz'i hinz'i

hind' hind'an hind'aw hind'aw hinzi hinzan hinzaw hinzaw

Inflecting Postpositions in Indie and Kashmiri

291

+ und 'father's'), whereas the feminine noun m j 'mother' in isolation takes hund (m ji + hund 'mother's'). Diachronically, und derives from the attachment of hund to the oblique I ending of masculine nouns, e.g. m l'is + hund > m l's + und 'father's', curas + hund > c rts + und 'thief's'. Synchronically, und might be considered as a morphologically conditioned allomorph of hund in the immediate environment of animate masculine singular common nouns. However, because of the sandhi phenomena that have taken place, und itself must be analysed as governing a morphologically conditioned variant of the original oblique I ending of the adjacent modifying noun (-s instead of -is, -is instead of -as). It should be noted that Zaxar'in and Edel'man (1971: 114) treat the final -s of the oblique I ending as belonging to the postposition (this then taking the form sund rather than und). This seems perverse and unnecessary, given that the oblique I always ends in -s for singular masculine nouns. The paradigm of und is identical to that of hund, giving examples such as (11). (11)

[m'on'is m l's ind'i] gari my=OblISgMasc father=OblISgMasc of=ObllISgMasc house=OblIISgMasc 'my father's house' (oblique II)

In (11), und is selected rather than hund because m l' is an animate masculine singular common noun. It takes the form ind'i in agreement with the modified noun gari 'house'. Und governs the oblique I case of m l' 'father'. That the -s case ending of m l' is a conditioned variant of the usual -is of the oblique I is shown by the agreement of the possessive m'on'is 'my'. While hund and und seem to be phrasal markers of the genitive, the two remaining genitive markers un and uk appear to have more of the characteristics of agglutinative suffixes than of genuine postpositions. The clearest manifestation of this is the fact that they obligatorily attach to all the conjunct nouns in a coordinate NP, and cannot be postposed just to the last conjunct (Zaxar'in and Edel'man 1971: 116): (12)

a. sahr-ik' ti gam-ik' sur' town-Gen=DctPlMasc and village-Gen=DctPlMasc children=DctPlMasc 'children of the twon and the village' b. Ram-un ti Naranun gari Rama-Gen=DctSgMasc and Narayana-Gen=DctSgMasc house=DctSgMasc 'Rama and Narayana's house'

Un and uk are traditionally regarded as governing the oblique II case of the modifying nouns. However, because this analysis requires the postulation of complex sandhi phenomena, it may be preferable to treat un and uk as simple genitive suffixes, adding a new genitive case to the four etymologically ancient Layer I cases (Masica 1991: 243). The full paradigms are as in Table 9.4.

292

Indo-European Table 9.4. Paradigm of Kashmiri Genitive Suffixes -un and -uk.

Masculine

Feminine

Singular

Plural

Direct

-un -(i)n'is -(i)n'i -in'

-(i)n'

-(i)n' -(i)n'an -(i)n'aw -(i)n'aw

Oblique I Oblique II Agentive

-(i)n'i -(i)n'i -(i)n'i

-(i)n'an -(i)n'aw -(i)n'aw

-uk

-(i)k'

-(i)ci

-(i)caw

Direct Oblique I Oblique II Agentive

Masculine Direct

Feminine

Oblique I Oblique II Agentive Direct Oblique I Oblique II

Agentive

-(i)k'is -(i)k'i -ik' -ic -(i)ci -(i)ci

-(i)n'i

-(i)k'an -(i)k'aw -(i)k'aw -(i)ci -(i)can -(i)caw

Unlike the etymologically ancient Layer I endings, the genitive suffixes are restricted to nouns (of the appropriate classes) and do not form part of the paradigm of agreeing adjectives and genitive phrases. These assume case endings that might be treated as agreeing genitives, but that are syncretic with the endings of the oblique II: (13)

m l's ind'i gar-uk bror father of=GenSgMasc house=SgMasc-Gen=DctSgMasc cat=DctSgMasc 'the cat of father's house'

In (13), the genitive noun gar-uk 'of the house' agrees with the direct case, singular number, and masculine gender of bror 'cat'. The possessive phrase m l's ind'i 'father's', which contains the agreeing postposition ind'i, agrees with the genitive case, singular number, and masculine gender of gar-uk (the ending being identical to that of the oblique II singular masculine). To summarize, the form of the genitive in Kashmiri is determined by the class of the final noun in any NP. The genitive marker may be an agreeing postposition (hund with feminine nouns and masculine plural nouns, or und with animate masculine singular common nouns). These govern the oblique I case of the preceding noun and its immediate modifiers; they agree in case, number, and gender with the modified noun; and as phrasal markers they can govern an entire coordinate NP. The genitive marker may also be a genitive suffix (-un with animate masculine singular proper nouns, or -uk with inanimate masculine singular common nouns). Like the postpositions, these agree in case, number, and gender with the modified noun. However, as suffixes, they must attach to everv noun in a coordinate NP

Inflecting Postpositions in Indie and Kashmiri

293

Table 9.5. Inflection of the Kashmiri Postposition k'ut 'for'.

Masculine

Feminine

Direct Oblique I Oblique II Agentive Direct Oblique I Oblique II Agentive

Singular

Plural

k'ut k'it'ts k'it'i k'it' k'ic k'ic'i k'ic'i k'ic'i

k'it' k'it'an k'it'aw k'it'aw k'ici k'ican k'icaw k'icaw

From this it follows that the structure of a genitive coordinate NP consisting of nouns belonging to different classes depends on the class of the final noun. If the final noun requires a postposition, then any preceding conjunct nouns may simply occur in the oblique I case, even though in isolation they might require one of the genitive suffixes -uk or -un. This is illustrated in (14), where the postposition hund, selected because of the plurality of gam 'village', governs the oblique I case of sahr 'town'. (14)

[sahri ti gaman hind'] luk town=OblI and villages=OblI of=DctPlMasc people=DctPlMasc 'people of the town and villages'

On the other hand, if the final noun is one that requires a genitive suffix, as in examples (12a) and (12b), the preceding conjuncts must be separately and appropriately marked. Kashmiri also possesses a further agreeing postposition k'ut 'for', occurring in examples such as (15), whose paradigm is given in Table 9.5 (Zaxar'in and Edel'man 1971: 114-115). (15) a. [paranas k'ut] gari reading=OblI for=DctSgMasc house=DctSgMasc 'house for reading' (direct) b. [cur'an k'it'aw] gur'aw thieves=OblI for=OblIIPlFem horse=OblIIPlFem 'horses for thieves' (oblique II) This appears to be the best case in the Indian subcontinent of an agreeing postposition with a function other than that of genitive. 4. Postpositional or Adjectival? One of the perennial disputes about the Indie genitive postpositions is whether they are genuine postpositions which ( a p a r t from the fact that they

294

Indo-European

Table 9.6. Adjective and Genitive Postposition Paradigms in Hindi and Sindhi. Hindi Adjective DctSgMasc OblSgMasc DctPlMasc OblPlMasc DctSgFem OblSgFem DctPlFem OblPlFem

Sindi Postposition

Adjective

Postposition

k-a k-e k-e k-e k-i k-i k-i k-i 'of

vad-o vad-e vad-a vad-ane vad-i vad-ia vad-iu vad-iane 'big'

j-o j-e j-a j-e (j-i) j-i j-i (j-ia) j-u j-e, j-ini, j-uni 'of

acch-a acch-e acch-e acch-e ac acch-i ac acch-i ac acch-i ac acch-i 'good'

have agreement paradigms) parallel other case formants in the language concerned, or whether they have the function of deriving attributive adjective phrases (a view taken for example by Bloch 1965: 180-181). The most obvious argument in favor of the adjectival hypothesis is the fact that the agreement paradigms of the postpositions typically coincide with one of the adjectival agreement paradigms in the language concerned. The formal correspondence is, however, not always exact. For example, the Sindhi postposition jo has a paradigm which, although similar to that of adjectives, is nevertheless different (Egorova 1966: 37). In Table 9.6, compare the Hindi adjective and genitive postposition paradigms, which are formally identical, with the corresponding Sindhi paradigms, which are similar but distinct. Distributional arguments also speak against the adjectival hypothesis. First, in a full description of NP structure, the position of genitive phrases may differ from that of standard adjective phrases. In Hindi, for example, the basic order of elements in Possessor-Determiner-Numeral Adjective-Noun (Kachru 1980: 41): (16)

[Sharmila ki] ve do lal kitabe Sharmila of those two red books 'those two red books of Sharmila's'

It can be seen that genitive phrases precede the determiner, while standard adjective phrases follow the determiner. Second, the genitive phrase may function in Hindi as the subject or locative in a participial construction (Kachru 1980: 71-72): (17)

a. [Ram ki likhi hul] kitab Ram of written book 'book written by Ram' b. ye [banaras ke parhe hue] hai he Banaras of studied is 'he was educated in Banaras'

Inflecting Postpositions in Indie and Kashmiri

295

In (17a) the participial is attributive, and in (17b) it is predicative. The predicative example is particularly interesting, since there is no possibility here of arguing that the genitive phrase is modifying a noun rather than acting as a locative within the participial construction. The role of subject or locative is a typical case function, which an arbitrary adjective phrase could not fulfil. The construction can also be illustrated from Punjabi (Smirnov 1976: 379): (18)

[gri rs n dia dsssia] 854 bolia Grierson of described 854 languages '854 languages described by Grierson'

Third, in some languages (e.g., Hindi and Punjabi), the genitive can occur in the position of the subject NP in existential expressions denoting partwhole relationships, non-temporary possession, and kinship (Masica 1991: 359-360). What is more, in Hindi (and possibly other languages), when the relationship is one of kinship, an invariant form of the genitive can be used: (19)

a. [kamre ki] tin hi divare hai room of=DctPlFem three only wall=DctPlFem are The room has only three walls' b. [Matadin ke] do gae hai Matadin of=DctPlMasc two steer=DctPlMasc are 'Matadin has two steers' c. [us ke] sirf ek bahan hai he of only one sister=DctSgFem is 'He only has one sister'

In the Hindi examples (19a) and (19b), which denote a part-whole relationship and a non-temporary possession relationship respectively, the genitive postposition shows agreement with the possessed noun. In (19c), on the other hand, the genitive postposition assumes an invariant form ke (the same as the masculine oblique), rather than the expected feminine ki. Standard adjective phrases do not have these functions. Fourth, genitive phrases can be governed by a further set of postpositions which typically have more specialized meanings than the basic Layer II postpositions. Zograf (1976: 117-122) calls these Layer III postpositions. Layer III postpositions are a widespread phenomenon throughout the Indie languages, and their construction with the genitive reflects the origin of such postpositions as nouns. The derivation is often totally transparent. In Gujarati, for example, puthe 'behind' derives from the locative of puth 'back' (Cardona 1965: 147). Puthe then governs a genitive phrase with the postposition ni (reflecting the feminine gender of puth); i.e., 'behind' derives from 'at the back of. In many cases, however, there is synchronically no noun to which the Layer III postposition is related, and the form of the genitive postposition is subject to considerable variation. Written Gujarati, for example, uses diachronically

296

Indo-European

based ni (feminine), na (neuter oblique), and ne (historically locative); however, many speakers always select the invariant form na, reflecting the transition from noun to postposition. In Punjabi, Layer III postpositions that govern the genitive always select an invariant form of the genitive postposition (the masculine oblique form de): (20)

[p njab de] vice Punjab of in 'in the Punjab'

The force of such examples is that the analysis of vicc 'in' as a postposition forces the analysis of p njab de as a genitive phrase, since postpositions do not govern adjectives. Fifth, genitive phrases can themselves be governed by adjectives. Examples can be cited from Gujarati in which the adjectives samu 'in front of, jewu 'like', sdrkhu 'like', and phdrthu 'around' govern the genitive. For no obvious synchronic reason, the forms of the genitive postposition are invariably ni with samu, na with jewu and s rkhu, and ne with pharthu (Cardona 1965: 146). The adjectives themselves agree with the noun which they modify or of which they are predicated: (21) [bag ne] ph rthu kundalu che garden of around=DctSgNeut circle=DctSgNeut is 'There is a circle around the garden'

As in the previous examples in which a postposition governs a phrase in the genitive, here it can be argued that an adjective is governing a phrase in the genitive. One would not expect an adjective to govern another adjective phrase. All the arguments presented above appear to demonstrate that the genitive postpositions in Indie and Kashmiri create phrases with functions typical of case-marked NPs rather than attributive adjective phrases. In other words, the formal identification of the genitive postpositions as Layer II case markers is correct, and the Suffixaufnahme phenomenon in Indie and Kashmiri genuinely involves the superposition of two cases. 5. Historical Origins What are the historical origins of the agreeing postpositions in Indie and Kashmiri? They appear to derive principally from the participles of IndoAryan verbs such as kr- 'do', as- 'be', and 'da- 'give' (Chatterji 1970: 751-759). These participles agreed with the nouns they modified, and in many languages they preserved the Layer I case, number, and gender distinctions as their relationship with the original verb paradigms became eroded. The present participle of as 'be' in the form santa(+ka) is found in the

Inflecting Postpositions in Indie and Kashmiri

297

Nasik inscriptions of the first and second centuries BCE in expressions such as pitu-santaka 'of the father' (literally 'father-being'). As satka in inscriptional Sanskrit of the Second Modern Indo-Aryan period, this participle is found all over Northern India, attaching to the base of singular nouns and the genitive of plural nouns. It survived in early Assamese in the form -saka, as in amasdka 'our', but it does not survive into modern Assamese. Its principal modern survival is in fact the agreeing postposition hund of Kashmiri. Participial forms of the verb kr- 'do' are the main source of the present Indie agreeing postpositions. The most popular genitive postposition in Modern Indo-Aryan literature is kera (or its variant form kela), which derives from the participle karya. It survives most obviously in the Romany dialects and Awadhi, but, in a vernacular form kajja, also results in the Sindhi postposition jo. As a form which has lost all agreement, it survives in Eastern Indie, for example as the Bengali genitive suffix -er. The source of the Hindi postposition ka is the past participle krta of the verb cr-, which straight forwardly gives Modern Indo-Aryan ka. The use of the participle krta in what can easily be interpreted as reinforcing a genitive function is apparent in the Transitional Modern Indo-Aryan period from Sanskrit expressions such as tasya krte dattam 'gift to him, his gift' (literally 'gift done of him'). Sanskrit expressions of this form, in place of tasmai dattam 'gift to him' with the dative of the pronoun, are evidently based on a Prakrit or vernacular use. The past participle *dita of the verb da- 'give' (in place of the Sanskrit form ditta with reduplication) is probably the source of the Punjabi and Lahnda postposition da. The failure of genitive postpositions to agree in Indie languages such as Sinhalese and Maldivian is not due to the loss of agreement but to the origin of these postpositions in nouns rather than agreeing participles, with the Sinhalese animate genitive ge and the Maldivian general genitive ge deriving from the locative singular gehi 'in the house' (Geiger 1938: 110; de Silva 1970: 147). 6. Typological Correlates

6.1. Word Order All the Indie languages have basic word order Subject-Object-Verb, and the correlation that Suffixaufnahme implies this order is therefore satisfied. Kashmiri is more problematic: it has Suffixaufnahme, but its word order is far from simple. According to Zaxar'in and Edel'man (1971: 125), word order is relatively free, but the standard order is Subject-Finite Verb-Object. There is a prima facie case, therefore, that Kashmiri constitutes a counterexample to the correlation between Suffixaufnahme and SOV order. The position is complicated by the behavior of analytic verb forms, where the non-finite section may either directly follow the finite form or follow other post-verbal sentence constituents, including the object. A marked word order may also have the

298

Indo-European

finite verb in initial position. The problem of basic word order in Kashmiri is therefore similar to that in German.

6.2. Ergativity Some degree of ergativity is a feature of all the Indie languages that display Suffixaufnahme, and of Kashmiri. For the agent of transitive sentences either a Layer II postposition may be used, or a Layer I non-direct case. Indeed, the correlation almost holds in the reverse direction: the languages that lack ergativity (basically the East Indie languages, and Sinhalese and Maldivian) also lack Suffixaufnahme. The exceptions are Assamese and Nepali, which lack Suffixaufnahme but do preserve some ergativity. Nepali does however possess a genitive postposition that (marginally) maintains number and gender agreement. The historical reasons for these correlations are complex, but it is in the Eastern languages that we see a greater tendency to lose the original Indie Layer I cases. These Layer I distinctions were the basis for the ergativeabsolutive distinction in the case marking of subjects (with subsequent further differentiation by Layer II markers in some languages), and their preservation is simultaneously a prerequisite for the origin of the Indie and the Kashmiri Suffixaufnahme construction. References Bloch, Jules. 1965. Indo-Aryan from the Vedas to Modern Times. (Translated by Alfred Master.) Paris: Adrien-Maisonneuve. Cardona, George. 1965. A Gujarati Reference Grammar. Philadephia: University of Pennsylvania Press. Chatterji, Suniti Kumar. 1970. The Origin and Development of the Bengali Language. Volume 2. Woking and London: George Allen and Unwin (first published 1926, Calcutta University Press), de Silva, M. W. Sugathapala. 1970. Some observations on the history of Maldivian. Transactions of the Philological Society 1970, 137-162. Egorova, Raisa P. 1966. Jazyk sindxi. Moscow: Nauka. Geiger, Wilhelm. 1938. A Grammar of the Sinhalese Language. Colombo: Royal Asiatic Society. Kachru, Yamuna. 1980. Aspects of Hindi Grammar. New Delhi: Manohar. Masica, Colin. 1991. The Indo-Aryan Languages. Cambridge: Cambridge University Press. Smirnov, Jurij A. 1976. Grammatika jazyka pandzabi. Moscow: Nauka. Venicel', Tat'jana V., and L. N. Cerenkov. 1976. Dialekty cyganskogo Jazyka. In Jazyki Azii i Afriki, vol. 1, Indoevropejskie jazyki, ed. by Nikolaj I. Konrad, 283-332. Moscow: Glavnaja Redakcija Vostocnoj Literatury. Zaxar'in, Boris A., and Dzoj I. Edel'man. 1971. Jazyk kasmiri. Moscow: Nauka. Zograf, Georgij A. 1976. Morfologiceskij stroj novyx indoarijskix jazykov. Mocow: Nauka.

V CHUKCHI-KAMCHATKAN

This page intentionally left blank

10 Possessive and Relational Forms in Chukchi Maria Koptjevskaja-Tamm 1. Introduction This chapter is concerned with the morphology and syntax of two nominal forms in Chukchi,1 the Possessive (1) and the Relational (2). These forms, which are mainly used as adnominal attributes and thus correspond to European genitives, optionally agree with their head in case, as in (la) and (2a), or in number, as in (Ib) and (2b). (1) a. Rult -n-ine-k tumg- k Rultyn-Sg-Poss-Loc friend-Loc 'at Rultyn's friend' b. ninqej- -ine-t tumg-at boy-Sg-Poss-AbsPl friend-AbsPl '(the) friends of (the) boy' (Sk I: 233) (2) a. kupre-kine-k ilg- k net-Rltn-Loc strap-Loc 'on the strap of the net' (Sk I: 271) b. aroq veem-kine-t vykv- t three river-Rltn-AbsPl stone-AbsPl 'three stones from a/the river' (Sk I: 269)

In (1b) and (2b) the plural suffix on the attribute is triggered by agreement with the head noun, which refers to a plural object. In addition to agreeing in number, many nominals in the Possessive may realize an inherent category of number, distinguishing between the singular and the plural of their own referent. Thus, although this occurs very rarely, one and the same Possessive may simultaneously be specified for two different number categories, inherent number and agreement number. This is illustrated in (3), which contrasts with (1b) in inherent number.

302 (3)

Chukchi- Kamchatkan inqej- rg-ine-t tumg- t boy-Pl-Poss-AbsPl friend-AbsPl '(the) friends of (the) boys' (Sk I: 233)

In constructions like those above the attribute is thus marked both for its own relation and for agreement with the head, manifesting pattern No. 7 of Plank's survey of attributive constructions (this volume). Since the agreement categories include the head's case, Chukchi Possessives and Relationals are clearly reminiscent of doubly case-marked attributes in prototypical Suffixaufnahme languages such as Old Georgian and Hurrian, and also of possessive adjectives in Slavonic. (Besides case and number, agreement in Chukchi NPs involves the category of predication, on which see Section 3.2 below.) Possessives and Relationals exist in all the Chukchi-Kamchatkan languages, but I have not been able to determine to what degree their behavior is similar across this family. No other Paleosiberian language outside the ChukchiKamchatkan family has forms comparable to Possessives and Relationals. Dealing mainly with the morphological status of the Possessive and the Relational, this chapter will show that both have inflectional as well as derivational characteristics.2 Their morphosyntactic behavior is not unique but is shared to a considerable degree by other words with overt attributive marking. Although they share some features, the Possessive and the Relational also differ in a number of respects.

2. Nominal Inflection, Possessive, Relational, and the Bare Stem In analysing Possessive and Relational forms, particular attention has to be paid to how their morphology is related to nominal inflectional morphology in general.

2.1. Outline of Nominal Inflection Chukchi is a predominantly agglutinative language whose words are inflected by suffixes, prefixes, and circumfixes. There are many complex phonological and morphonological processes, including various vowel harmony rules, consonant alternations, and the insertion of the epenthetical vowel 3. Chukchi is especially famous for its incorporation, which involves not only the arguments and adjuncts of verbs but also the attributes of nouns (as illustrated extensively in Sections 2.4, 3.1, and 4). Word order in clauses is relatively free; in NPs, attributes normally precede their head. Chukchi is exclusively dependentmarking within NPs and double-marking on the clause level, where dependents are marked for case and the verb agrees with its intransitive subject (S) or transitive subject (A) and object (P). In case marking Chukchi is consistently ergative, while its verbal morphology combines nominative-accusative and ergative features; its syntax is fairly neutral. 3

303

Possessive and Relational Forms in Chukchi

The main nominal categories are case, number, and predication. The major morphological case distinction is between the absolutive and the various oblique cases, and the numbers distinguished are singular and plural. With the exception of the so-called associative and the comitative (the latter characteristic only of nouns referring to non-humans), which are marked by circumfixes (whose prefixal part is ga-), all cases are marked by suffixes. Plural is marked by the suffixes -t (with allomorph -r- in word-medial position) or -(n)ti. Depending on their degree of animacy, nominals differ in the extent to which they manifest number (cf. Comrie 1979, 1981): all NPs obligatorily distinguish singular and plural in the absolutive case; NPs referring to non-humans do not distinguish number in any other case; personal pronouns, other pronouns, and some kinship names (for near relatives) obligatorily distinguish number in all cases except for the destinative; and other human NPs may distinguish singular and plural in the oblique cases, but normally they do not. When a noun is marked for both plural and oblique, normally each of these two categories is expressed separately and the case suffix follows the number suffix. The morphology of the absolutive case provides a major deviation from this generally agglutinative pattern in that here case and number have cumulative exponence. Thus, while absolutive singular is indicated in a number of different ways (e.g., by the ending -n, by partial reduplication of the stem, or by the absence of any inflection), absolutive plural is expressed by the ending -t or -(n)ti. Table 10.1 illustrates the interplay between the expression of case and number for the personal noun tumg -tum 'friend' (with an optional number distinction in all oblique cases) and for the non-personal noun kej - n 'bear' (with number opposition only in the absolutive case). The class of nouns includes various demonstrative, interrogative, indefinite, and quantificational nouns (such as otqen 'this (one)', me in 'who?', qol 'another', maljo 'all'). Personal pronouns may, with some reservations, also be considered a subclass of nouns.

Table 10.1. Case Paradigms of Nouns Referring to Humans and Non-humans in Chukchi. Sg Absolutive Instrumental Locative Ablative Dative Orientational Comitative Associative Destinative

tumga-turn tumg-e tumg- k tomg-ep tomg-eta tumga-gjit ga-lomga-ma

P1 ll tumg-3t tumg- r- k tumg- r- k tomg- rg- p tomga-ra-ka tumg -r -gjit ga-tomga-ra-ma tumgt-u kej 'friend'

Sg

P1

kejjn-an kej - t kej -e kej - k kej -ep kej - ta kejg -gjit ge-kej -e ge-kej a-ma ke -u 'bear'

304

Chukchi- Kamchatkan

2.2. The Possessive: Meaning and Form Possessives are mainly used as adnominal attributes and cover most of the meanings traditionally ascribed to adnominal genitives. Thus they are used in NPs expressing possession (4a), kinship and other social relations (4b), body parts (4c), part-whole relations (4d), and material (4e), to mention only the most important relations. (4) a.

b.

c.

d.

e.

np nacg- -en / Tato-n-en milger old=man-Sg-Poss / Tato-Sg-Poss gun (Abs) 'the old man's/Tato's gun' (Sk I: 226) Rulta-n-in(e-t) mirg- t Rultyn-Sg-Poss(-AbsPl)grandfather-AbsPl 'Rultyn's grandfathers' (Sk I: 232) r rk-en va q t-ta walrus-Poss tusk-AbsPl 'a/the walrus's tusks' (Sk I: 227) jej-in kanotka-n mountain-Poss top-Abs 'the top of a mountain' vakv-en jara ne stone-Poss house (Abs) 'a house of stone' (Sk I: 249)

The Possessive can also be involved in certain types of relative clauses (Nedjalkov 1979). Thus, the (lexical) transitive agent of passive participles in attributive use, when relativizing on the (lexical) direct object, may alternatively appear in the ergative case (the more common option) or with the Possessive suffix (5a). In attributive constructions that relativize on various oblique NPs and involve verbs in the purpose form (with suffix -kin(e)/ -ken(a)), the intransitive subject must, and the transitive agent may, appear with the Possessive suffix (5b). (5) a. ?orawetl? -ta m l o na-tw - -n n -k rg-in / rga-nan walom-yo-tte wetgaw-at man-Ins/Erg all (Abs) 3PlA-say-Aor-3SgP he-Dat they-Poss / they-Erg hearPassPrtcpl-AbsPl word-AbsPl 'People told him everything (that) they had heard (lit. all words heard by them)' (Nedjalkov 1979: 248, (24)) b. iwtalet-kin nelwal?-in r?et ge-rk le-lin nemaqey muul?-e come=down-Attr herd-Poss road (Abs) Perf-follow-Perf3Sg also convoy-Ins 'Along the road by which the herd came down followed also the convoy' (ibid., (26'))

Possessives are formed absolutely regularly for any nominal, including personal pronouns. As shown in (3), many nominals in the Possessive form

Possessive and Relational Forms in Chukchi

305

may distinguish between the singular and the plural of their own referents. Nominals that obligatorily or optionally distinguish singular and plural in all cases—i.e., personal pronouns and nouns referring to humans—maintain this distinction in their Possessives as well. Thus the difference between singular and plural referents is reflected in distinct Possessive markers, -(n)in(e)/ -(n)en(a) vs. -rgin(e)/-rgen(a) respectively (6a/b),4 where -r- in -rgin(e)/ -rgen(a) is the common plural marker for all oblique cases. The -nin(e)/nen(a) variant of the singular Possessive suffix is chosen by nominals with obligatory number distinction in the oblique cases (proper names, names of near relatives, interrogatives, and personal pronouns); -in(e)/-en(a) ) attaches to nominals with optional number distinction in the oblique cases. The word ama/jo 'all', which is inherently plural and always follows the plural case paradigm, takes only the plural Possessive suffix -rgen(a), as in (6c). (6) a. ninqej- -in tumg-atum boy-Sg-Poss (AbsSg) friend-AbsSg 'a/the friend of a/the boy' (Sk I: 233) b. jinqej- rg-in tumg-atum boy-Pl-Poss (AbsSg) friend-AbsSg 'a/the friend of (the) boys' (Sk I: 233) c. igyr m l?o-rg-ena-k jeekke-k evir?- t n -tur-qine-t now all-Pl-Poss-Loc daughter-Loc clothes-Pi Stat-new-3Stat/AbsStat-Pl 'Now the clothes of daughters of all people are new' (Sk I: 253)

Personal pronouns for 1st and 2nd person provide a particularly interesting example of number opposition in the Possessive, since the distinction between singular and plural is manifested both in the root (with suppletive roots for the different numbers) and in the Possessive suffix, as in (7a/b). The 3rd person pronoun follows the normal nominal pattern (7c/d). The structure of possessive forms for personal pronouns is again in line with their inflectional paradigm (cf. g m - k I-Loc and mu-r-ak we-Pl-Loc vs. -n- k he/she-Sg-Loc and a-r- k he/she-Pl-Loc). (7) a. gam-n-ine-t ekke-t I-Sg-Poss-AbsPl son-AbsPl 'my sons' b. mu-rg-ine-t qaa-t we-Pl-Poss-AbsPl deer-AbsPl 'our deer (plural)' (Sk I: 412) c. -n-ine-t tumg- t he/she-Sg-Poss-AbsPl friend-AbsPl 'his/her friends' d. -rg-ine-t tumg- t he/she-Pl-Poss-AbsPl friend-AbsPl 'their friends'

306

Chukchi- Kamchatkan

Possessives for non-humans, like all oblique cases, do not distinguish between singular and plural referents and are formed with the same possessive marker -in(e)/-en(a) (8a). In instances of personification, however, the plurality of the referent may optionally be reflected in the use of the possessive suffix -rgin(e)/-rgen(a), as in (8b). Note that the choice between the variants ) is more or less in line with the general divi) -nin(e)/-nen(a) ) and -in(e)/-en(a) sion of Chukchi nouns into declensions, where the allomorphic variation for some cases involves the presence vs. absence of initial n—cf. -nal-ne and -al-e (-tal-te after vowels) for the instrumental, and -na and -etal-eta for the dative. (8) a. vopq-en rann- t elk-Poss antler-AbsPl 'antlers of an/the elk, of (the) elks' (Sk I: 227) b. rarka-rg-en vanq- tte n-?omr -qena-t 0-ikv- -?i tla qajomqa-gta walrus-Pl-Poss tusk-AbsPl Stat-strong-3Stat/AbsPlStat 2/3SgS-say-Aor-2/ 3SgS mother (Abs) white=bear=cub-Dat '(The walruses' tusks are strong,—a white bear mother said to her white bear cub' (ibid.)

To summarize, Possessives are in certain respects well integrated into the nominal inflectional paradigm. They are built in absolutely productive and regular ways, so that both their forms and meanings are completely predictable. Furthermore, Possessives maintain the typically nominal category of inherent number, realized differently for different types of nominals, following the general pattern of oblique cases in Chukchi. 2.3. The Relational: Meaning and Form Relational are also used mainly as adnominal attributes. As such, they describe an object as being "related" to or pertaining to a certain place (9a), time (9b), or object (9c). They are formed primarily from inanimate nouns of the corresponding semantics by means of the suffix -kin(e)/-ken(a) (on the choice among variants, see note 4). (9) a. jara-k qaca enmec aroq veem-kine-t v kv- t n -tva-qena-t house-Loc near already three river-Rltn-AbsPl stone-AbsPl Imperf-be-3S-Pl 'There are already three stones from the/a river near the house' (Sk I: 269) b. l?eler -kin evir?-an winter-Rltn clothes-AbsSg 'clothes for winter' c. m c kv -ken mumk l shirt-Rltn button (AbsSg) 'a button from the/a shirt' (Sk I: 268)

The same suffix -kin(e)l-ken(a) can also be followed by certain oblique (mainly locative) cases, marking the standard of comparison in certain types

Possessive and Relational Forms in Chukchi

307

of comparative constructions. Thus in (10a) the standard of comparison is "complex" (corresponding to the possessive NP 'the boy's cap'), which is shown by the use of the suffix -kine with the nominal. In (10b), the comparison involves actions. When used in comparative constructions, -kin(e)/-ken(a) is not restricted to any particular class of nouns. (10)

a. ekkeqej- -in k?eli inqej-kine-k ta -a va-l?-an girl-Sg-Poss (AbsSg) cap (AbsSg) boy-Rltn-Loc good-Comp be-Prtcpl-3Sg/ AbsSg 'The girl's cap is better than the boy's' b. ig r ekkeqe-gti inqej-kine-k ta -a ) ge-migciret-line-t today girl-AbsPl boy-Rltn-Loc good-Comp Perf-work-3SPerf-Pl Today the girls have worked better than the boys' (Sk I: 268)

It is worth noting that the Chukchi usage of the Relational suffix is restricted as compared to other closely related languages. In Alutor, nominals with this suffix can even be governed by verbs, as in (11), where Chukchi would resort to the ablative, a case lacking in Alutor.5 In this example, rarakina-jgdm (house-Rltn-lSg) is formally a secondary predicate to the subject ('I') and therefore bears the predicational suffix of the corresponding person. (11)

g mm t - tu-tk- n rara-kina-jgam I (Abs) 1SgS-go=out-Pres-lSgS house-Rltn-lSg 'I am going out of the house'

While there are possible diachronic links between the different uses of the suffix -kin(e)l-ken(a) in Chukchi, synchronically it seems reasonable to treat its use in comparative constructions (10) as quite distinct from "relational" attribution (9). In the latter use, the Relational is confined mainly to nominals referring to non-animate objects and is formed absolutely regularly by means of -kin(e)l-ken(a). Since these nouns manifest number distinctions only in the absolutive case, the Relational is again in line with the nominal inflectional system in lacking the category of inherent number. Somewhat surprisingly, possessive pronouns may also attach the Relational suffix in the form -ka-ken(a)l-ke-kin(e), the first of which is common to most of the oblique cases of personal pronouns. Such forms characterize objects as being within or coming from a particular personal sphere, defined by the personal pronoun—e.g., murd-ke-kin qora na we-Obl-Rltn deer (Abs Sg) 'a deer which has some relation to us (coming from the place where we live, or one we have to take care of, or being of the kind we keep for breeding, etc.)' (Sk I: 415). However, the same suffix (or its homonyms) appears elsewhere in Chukchi morphology, where it cannot be classified as a nominal inflectional suffix. First, -kin(e)/-ken(a) is not restricted to nominal stems, but is also used completely regularly with verbal stems to produce a special attributive form (called the "purpose form" in Skorik 1977), as seen in (5b). This form is

308

Chukchi- Kamchatkan

regularly used for relativizing adjuncts (nominals denoting time, location, and manner) in intransitive clauses (Polinsky 1994). This usage with both nominal and verbal stems is, however, not unique for -kin(e)l-ken(a), but is fairly common for suffixes in Chukchi, which generally shows a neat correspondence between the system of non-finite verbal predication and the nominal system of oblique cases. Second, in a number of words, used primarily as nouns or as adnominal attributes, the same suffix -kin(e)/-ken(a) is clearly derivational. Synchronically, the derivational source of these words lacks this suffix and is no longer a full-fledged noun. Thus, ajve-ken 'yesterday's' and yutke-kin 'here's, pertaining to here' correspond to the adverbs ajve 'yesterday' and utku 'here', while the derivational source for c cet-kin 'relative' is maintained only as an element of compounds such as cacet-nutenut 'fatherland' (lit. 'native-land'). From the available sources it is difficult to judge the degree of productivity of such formations from adverbs. In general, words with -kin(e)/-ken(a) appear to represent a more or less productive category of attributes pertaining to time, location, and perhaps some other circumstances whose bases are different parts of speech.

2.4. Nominal Stems as Incorporated Attributes Attaching the Possessive or Relational suffix to a nominal is not necessary for its use as an attribute. When incorporated into another noun, bare nominal stems are interpreted as providing a general characterization of the head, or, in other words, as categorial (non-referential, characterizing) attributes. Thus, a qa-g nne-k (sea-animal-AbsSg) denotes an animal classified as marine, regardless of its physical position, in contrast to a qa-ken gannl-k (seaRltn animal-AbsSg), which refers to an animal actually in the sea or coming from the sea. In the same way, I eley-micgir 'winter-work' (as in the sentence 'Winter-work with a herd is very difficult') provides a qualificational characteristic of a particular kind of work, while l?ele -kin micgir (winter-Rltn work) refers to work which has to be done in winter, independently of its nature (as in 'Only yesterday we finished the work planned for the winter').6 For obvious semantic reasons, personal pronouns cannot be incorporated as bare stems. Nor can those words for which -kin(e)l-ken(a) is clearly derivational be incorporated without this suffix. As attributes, nominal stems must be incorporated and can never be external. Possessives and Relationals, on the other hand, can appear as both external and incorporated attributes, as will be shown presently. 3. Morphosyntactic Behavior of Possessive and Relational As we have seen in the preceding section, Possessives and Relationals follow basically the pattern of oblique nominal cases with respect to inherent number. In contrast to oblique cases, however, they may combine with other

Possessive and Relational Forms in Chukchi

309

nominal inflectional markers. The morphosyntactic behavior of the Possessive and the Relational will now be examined in connection with their syntactic functions.

3.1. Patterns of Attribution of Possessive and Relational The pattern illustrated in (l)-(3), with Possessive and Relational attributes agreeing with their heads, competes with two other patterns. Used as adnominal attributes, the Possessive and the Relational may alternatively appear in three "modes:" as a separate word; agreeing (12a) or not agreeing (12b) with the head; or incorporated into the head nominal (12c). In all these modes the relevant suffixes are present independently of the exact syntactic status of the attribute. (12)

a. Rult -n-ine-k tumg-sk Rultyn-Sg-Poss-Loc friend-Loc b. Rult -n-in tumg- k Rultyn-Sg-Poss friend-Loc c. Rult -n-ine-tumg- k Rultyn-Sg-Poss-friend-Loc 'at Rultyn's friend' (Sk I: 229-231)

The choice among the three modes of attribution is governed by at least two factors: the case of the head, and the pragmatic structure of discourse, with the case factor normally superordinate to the pragmatic factor. To begin with the case factor, the absolutive, comitative, or associative cases of a head noun, unlike the other cases, considerably restrict the possible modes of its Possessive and Relational attributes. As will be shown in Section 4, the same restrictions apply to other types of adnominal attributes. Morphologically, comitative and associative have a rather peculiar place in the nominal inflectional paradigm, in that their markers consist of two parts— a prefix (ga-) and a suffix (-(t)a/-(t)e e for comitative and -ma/-me for associative). Any attributes to nouns in one of these cases, including Possessives and Relationals, must be incorporated into the head noun immediately following the prefix ga-. Examples (13a/b) illustrate this rule for Possessive attributes: (13)

a. qolet? lo muri n-ivini-muri ga-np nnacg-rg-ena-akka-ta other=day we (Abs) Imperf-hunt-1PlS with-old=man-Pl-Poss-son-Com 'The other day we were hunting (in the sea) together with the old men's sons' b. ?aacek- t - t- -g?e-t g tg-et ga-np nacg- rg-ena-kopra-m young=man-AbsP1 2/3S-go-Aor-3S-Pl lake-Dat with-old=man-Pl-Poss-netAss 'The young men went to the lake with the old men's net' (Sk I: 226)

310

Chukchi-Kamchatkan

A nominal in the absolutive case, by contrast, cannot incorporate its Possessive and Relational attributes, which may thus appear only as external attributes. When the case of the head noun is compatible with more than one attributional mode, the choice among them is governed by pragmatic considerations, such as what constitutes the speaker's primary focus of interest—the object itself, its additional characteristics (like being possessed by someone or being "related" to something), or both.7 Possessives appear normally as external non-agreeing attributes, somewhat less frequently as incorporated attributes, and fairly rarely as external agreeing attributes. Relationals, by contrast, appear normally as incorporated attributes and somewhat less frequently as external agreeing attributes. Instances of external Relational attributes not agreeing with their heads are extremely rare. The influences on the choice between external and incorporated attribution, as formulated above, demonstrate the important distinction between what Irina Murav'eva (1994) calls semantic and syntactic incorporation. When syntactic conditions permit incorporation but do not necessarily require it, the application of incorporation has special semantic or pragmatic consequences. This can be characterized as semantic incorporation, as opposed to syntactic incorporation, which is required by certain syntactic conditions and has no independent semantic or pragmatic motivation. Obligatory incorporation of Possessives and Relationals by a head noun in the comitative or associative case provides an example of syntactic incorporation, while incorporation by nouns in other oblique cases is semantic. We will return to syntactic incorporation in Section 4. Let us now specify exactly what is meant by agreement with the head in NPs with Possessive and Relational attributes. All examples up to now have involved case-number agreement, the rules for which can be formulated as follows: (a) When the head is in an oblique case, its attribute may optionally agree with it in case only by attaching the corresponding case suffix after the attributive marker (i.e., the Possessive or the Relational suffix); (b) when the head is in the absolutive plural, its attribute may optionally agree with it in number by attaching the plural suffix -t after the attributive marker. Note that number agreement is only possible when the head noun is in the absolutive case, even though the noun itself may distinguish between the singular and plural in the oblique cases as well. A restricted group of Possessives may also distinguish between the absolutive and all the oblique cases of their head: (c) Words with the Possessive suffix -nin(e)/-nen(a) may optionally truncate this suffix to -n when the head is in the absolutive. Besides case and number, Possessive and Relational attributes agree with their heads in the category of predication: (d) When the whole NP is used as a predicate, both the head and its attributes attach the corresponding predicational suffixes. The next subsection will be devoted to this last type of agreement.

Possessive and Relational Forms in Chukchi

311

3.2. Nominal Predicates and Agreement Nominal predication in Chukchi is accomplished by means of special predicational suffixes that are attached to the stem of the predicate nominal and agree with the subject of the clause. For 1st and 2nd persons these suffixes are basically identical to the corresponding pronominal stems, preceded in the singular by -i-l-e-l-j- (which, according to Skorik I: 443, is related to the verb root it/et 'be'); the 3rd person lacks any overt suffix, and the whole form is thus identical to the absolutive case. Such nouns are used as matrix predicates in identifying (14a), classifying (14b), and, in combination with the prefix ga'with', in Possessive (14c) nominal clauses. (14)

a. gam Terjanto-jgam I (Abs) Tynganto-lSg 'I am Tynganto' b. turi cavcava-tore you (Abs) nomad-2Pl 'You are nomads' (Sk I: 216) c. igar muri amarj ga-qaa-more now we (Abs) also with-deer-1Pl 'Now we also have deer' (Sk I: 219)

When Possessives and Relational are used as attributes within a predicational NP, they attach the corresponding predicational suffix and thus agree with the head; in (15c) the head of the predicational phrase takes both a Possessive and a Relational attribute: (15)

a. gam vopq-ena-jg m tung qej-igam -ikv- -?i ?ina I (Abs) elk-Poss-lSg friend-lSg 2/3S-say-Aor-2/3S wolf (AbsSg) ' "I am the elk's friend," said the wolf (Sk I: 265) b. gam umka-kine-jgam g nni j tt- l?-ig m I (Abs) forest-Rltn-lSg hunt-Prtcpl-lSg 'I am a hunter from a/the forest' c. g t g m-n-ine-jgat katur-kine-jgat remk-al?-igat thou (Abs) I-Sg-Poss-2Sg last=year-Rltn-2Sg visit-Prtcpl-2Sg 'You are my guest from last year' (Sk I: 277)

The same forms, but without the prefix ga- 'with', are also frequently used as appositions, mainly to the absolutive NP in a clause: (16)

a. muri, qut- rg-ine-muri t letumg -muri, ergatak in?e mat-re-kvet-g?e we (Abs), other-Pl-Poss-1Pl fellow=traveler-lPl, tomorrow morning 1P1SFut-Ieave-lPlS 'We, the others' fellow-travelers, we will leave tomorrow in the morning' (Sk I: 261)

312

Chukchi-Kamchatkan b. turi, epeqej- rg-ine-turi inentej ev- l?- turi, mor-g nan m t-re-vinrent k you (Abs), grandmother-Pl-Poss-2Pl feed-Prtcpl-2Pl, we-Erg 1PlA-Futhelp-2PlP 'We will help you, your grandmothers' bread-winners' (Sk I: 264)

It is arguable that such constructions do not represent straightforward apposition. The use of predicational suffixes here suggests that the relation between the two NPs involves predication instead. The restriction on the case of the first NP may be seen as additional support of this analysis: subjects of nominal predicates appear normally in the absolutive case.8 Thus it may be suggested that the use of predicational suffixes allows a nominal predication about some NP to be incorporated into a larger verbal or nominal clause, which is typical of "secondary predication." However, such constructions do not present cases of bona fide secondary predication like 'That cat eats grasshoppers alive', in that the truth of the incorporated predicate is not asserted within the time frame established by the main predication (cf. Dench and Evans 1988: 14). Judging from the available sources, predicational agreement of Possessive and Relational attributes with their head is mandatory, as opposed to optional case and number agreement.

3.3. Headless Possessive and Relational Up to now we have considered the attributive uses of Possessives and Relationals. They may also appear without an overt head nominal in typical nominal functions and with typical nominal inflectional categories, such as number (17a), case (17b/c), and predication (17d/e). In such uses, Possessives and Relationalsmust attach the corresponding inflectional markers in the same way as normal nouns do. (17) a. mik-an-(in) el?o-nte, mirg- t, qaa-t, milger-ti? Rult- n-ine-t, inqej- -ine-t who-Sg-Poss (Abs) grandfather-AbsPl, uncle-AbsPl, deer-AbsPl, gunAbsPl? Rultyn-Sg-Poss-AbsPl, boy-Sg-Poss-AbsPl 'Whose grandfathers, uncles, deer, guns? Those of Rultyn, those of the boy (Sk I: 233) b. gat-am ge-mika-n-ine-tumg-e ni-vini-jgat? ge-Rult -n-ine-te, ge-jic?emittumg- -ine-te thou-Emph (Abs) with-who-Sg-Poss-friend-Com Imperf-hunt-2SgS? with-Rultyn-Sg-Poss-Com, with-brother-Sg-Poss-Com 'With whose friend do you hunt? With Rultyn's, with (my) brother's' (Sk I: 233) c. r?e-kine- 'lg-e ge-kvut-lin ? tt?-an? ? tv -kine-te what-Rltn-belt-Ins Perf-tie-3SgS dog-Abs? boat-Rltn-Ins

Possessive and Relational Forms in Chukchi

313

'With a belt pertaining to what is the dog tied? With that from a boat' (Sk I: 274) d. njiv- -ine-jg t uncle-Sg-Poss-2Sg 'You are uncle's' (i.e. 'You belong to the uncle') (Sk I: 261) e. mnu -kine-jg m tundra-Rltn-lSg 'I am from the tundra' (Sk I: 275)

3.4. Possessive, Relational, and Case In Section 3.3 we have seen that Possessive and Relational suffixes may combine with number, case, and predication markers, showing behavior that is unique in the realm of nominal inflection in Chukchi. There are no other instances of double case or (relevant only for the Possessive) double number marking of nouns in the language. The ability of Possessives and Relational to combine with predicational suffixes is also remarkable, since predicational suffixes never appear on a nominal marked with the oblique case suffix. Thus, to express location, a nominal in the locative case cannot take predicational suffixes directly, but must instead combine with the copula verb vak 'to be' (whose stem in word-medial position is -tva-), as in (18) (see also (9a) above). (18)

igar mu-rg-in n n n- t k njora-k na-tva-qena-t now we-Pl-Poss child-AbsPl nursery-Loc Imperf-be-3S-Pl 'Our children are at the nursery now' (Sk II: 243)

Such peculiarities of Possessives and Relationals in Chukchi are comparable to the behavior of genitives, which in many languages—among them those showing Suffixaufnahme—are the only cases taking further case marking. As has been argued repeatedly (e.g. by Dench and Evans 1988 and by Kibrik in this volume), this type of agreement in nominals with a special attributive marker per se cannot be taken as evidence against the casehood of such markers. Cross-linguistically much more peculiar, however, is the occurrence of Chukchi Possessives and Relationals as both external and incorporated attributes. No other case marker shows a comparable range of patterns, and no other case marker is preserved in incorporation. In this respect, the Possessive and Relational suffixes clearly differ from case markers, and in fact from all inflectional markers, as will be discussed in the next section. 4. Possessive, Relational, and Derivation 4.1. Incorporation as a Criterion for Derivation In the preceding sections we considered the Possessive and the Relational against the background of nominal inflection Section 2 provided evidence

314

Chukchi- Kamchatkan

supporting the view that the Possessive and, to a lesser degree, the Relational are integrated into the nominal inflectional paradigm. Still, as Section 3 demonstrated, certain morphosyntactic peculiarities of these forms distinguish them from oblique cases. However, the role of these forms in incorporation provides a very strong argument for regarding them as derivational. Incorporated nominals do not normally retain inflectional categories of their own, such as case or number. They may well retain their derivational affixes, though, including augmentative or diminutive suffixes. The same is true for verb incorporation: although Chukchi has a well-developed system of verbal inflection, which among other things includes many non-finite forms, verbs are incorporated into other verbs only as bare stems, as in luur anp gev ga-vje-peqetat-len (suddenly old=woman (AbsSg) Perf-die-fall3SgS) 'Suddenly the old woman fell dying/dead' (Sk II: 234). Both the Possessive and the Relational suffixes are retained in incorporation, with the result that nominal attributes can be incorporated in three different forms: as bare stems (see 2.4) or with one or the other appropriate attributive suffix.

4.2. Types of Stems and Attribution Now, if incorporability is considered a strong argument for classifying Possessive and Relational suffixes as derivational, what part(s) of speech do they derive? As we have seen in Section 3.3, words with these suffixes may appear in typical nominal functions with typical nominal inflections. However, they differ from nouns in their ability to appear as external (nonincorporated) attributes (see 3.1). As shown in Section 2.4, nouns without Possessive and Relational suffixes cannot modify other nouns without being incorporated into them. It is fairly reasonable, and more or less in line with Skorik's analysis, to account for these facts by taking the attributive usage as the primary function of words with Possessive and Relational suffixes and by explaining their nominal behavior as following from the deletion of the head in attributive NPs. As we will see presently, Possessives and Relationals are not unique in having such properties but share them to a considerable degree with two other types of forms: participles and ordinal numerals. To ascertain the place of Possessives and Relationals in the system of Chukchi parts of speech, let us now compare the ability of different classes of words, or rather stems, to be used as attributes to nouns. The major distinction here is between those stems that can never appear as external adnominal attributes and those that can. While the former, used as attributes, are always incorporated into the head noun and thus present clear cases of syntactic incorporation, the latter show up as both external and incorporated attributes. This latter class, however, is not homogeneous as regards the details of the choice between external and incorporated attribution. With respect to their patterns of attribution Chukchi stems fall into two classes, each with two subclasses.

Possessive and Relational Forms in Chukchi

315

A. Steins that can be incorporated by a noun in the absolutive case. a. Stems that never appear as external adnominal attributes: nouns (without Relational and Possessive suffixes, see 2.4) and (non-quality-state) verbs (as in nn tta-kupr- n catch=fish-net-AbsSg 'a fishing net', Skorik 1975: 135). Both nominal and verbal stems used as incorporated attributes provide categorial, characterizing attribution of the head noun. b. Stems that can appear as both external and incorporated attributes to a noun in the absolutive case: so-called "quality state" words (Sk I: 421-429), or adjectives (Comrie 1981), which among other things include words with cross-linguistically "prototypical" adjectival meanings. According to the older norm, quality state words were used as predicates and as external attributes to absolutives. In such cases the stem is preceded by the prefix n- and is followed by the corresponding predicational suffix or by the suffix -qin(e)/-qen(a) (when used as a predicate to 3rd person (19a) or as an attribute (19b)). As attributes to nouns in oblique cases, quality state stems are obligatorily incorporated into the head noun without any markers (i.e., as bare stems, (19d)), which is also possible with absolutive head nouns (19c). (19)

a.

otq- n ? tt?- n n-erme-qin- , nqen- m n -rul-qinthis-AbsSg dog-AbsSg Stat-strong-3Stat-Sg, that=one-Emph Stat-weak3Stat/AbsStat-Sg 'This dog is strong, but that one is weak' (Sk I: 428) b. ajve n-ilg- qin- qorans -g ntekv- -?i yesterday Stat-white-3Stat/AbsStat-Sg deer (AbsSg) 2/3S-run=away-Aor-2/ 3SgS 'Yesterday a white deer ran away' c. ajve elg -qoran a0 -g ntekv- -?i yesterday white-deer (AbsSg) 2/3S-run=away-Aor-2/3SgS 'Yesterday a white deer run away' d. elg -qora-gt a white-deer-Dat 'to a white deer, to white deer' (Sk I: 429)

In the modern language, however, quality state words may also be used as agreeing external attributes to nominals in oblique cases (other than the morphologically peculiar comitative and associative): (20)

n -mej -qine-k ? tv- k -vak?o-£f-g?a-t k lganken ?oravetl?a-t Stat-big-Stat-Loc boat-Loc 2/3S-sit-Aor-3S-Pl fifteen man-AbsPl 'Fifteen men sat down in a large boat' (ibid.)

In this behavior, quality state words approach the words discussed below under (Bb).

316

Chukchi-Kamchatkan

B. Stems that are never incorporated by a noun in the absolutive case and thus appear only as external attributes to absolutives. a. Stems that never appear as external attributes to nouns in oblique cases and thus may only be incorporated attributes to such nouns: cardinal numerals. Thus, in the following example, the numeral 'one' has two different manifestations, being used as an external attribute to a noun in the absolutive in (21 a) and as an incorporated attribute to a noun in the instrumental case in (21b): (21)

a.

nnen milger one gun (AbsSg) 'one gun' b. nan-pojg-a one-spear-ins 'with one spear'

b. Stems that can be both external and incorporated attributes to nouns in oblique cases (other than comitative and associative): Possessives, Relationals,9 participles, and ordinal numerals. The category of participles, as described by Skorik, is a peculiar Chukchi phenomenon and includes words derived from verbs and quality state words (e.g. r ju-l - n graze-Prtcpl-AbsSg 'grazing, one who grazes'; gati - l - n (be) beautiful-Prtcpl-AbsSg 'beautiful, one who is beautiful'), as well as from nouns with the meaning 'having X' (e.g. ? tt?- l?- n dog-Prtcpl-AbsSg 'having dogs, a dog-owner'). Unlike nouns and pronouns, both participles (22) and ordinal numerals (23) are used primarily as adnominal attributes, either incorporated as in (22a) and (23a), or external as in (22b) and (23b). But in addition they may be used in typical nominal functions with nominal inflectional characteristics, such as case as in (22c) and (23c), and predication as in (22d) and (23d). In fact, participles referring to humans are used in nominal functions at least as often as in attributive functions. (22)

a. ganan ga-p ker- l a-cavc va-ta q -v?jat-g na-t qaa-t thou (Erg) with-come-Prtcpl-nomad-Com 2SgAImp-unharness-3P-PlP deerAbsPl 'Unharness the deer together with the nomad who has come' b. uvicvet- l?-at inqeg-ti n -lgiqulil?et-qine-t play-Prtcpl-AbsPl boy-AbsPl Imperf-loudly=shout-3S-PlS The playing boys were shouting loudly' (Sk I:353) c. g -r tku-l?-e veem- k na-tva-qena-t ninqeg-ti with-shoot-Prtcpl-Com river-Loc Imperf-be-3S-Pl boy-AbsPl 'On the river there were boys together with those who were shooting' (Sk I: 351) d. muri, to-l? -more, luur inqe-e ne-nce attev- -m k we (Abs), go=out-Prtcpl-1Pl suddenly boy-Erg 3A-frighten-Aor-lPlP 'Suddenly the boy frightened us as we were coming out' (Sk I: 359)

Possessive and Relational Forms in Chukchi (23)

317

a. naro-qav-nalg-a three-Ord-skin-Ins 'by the third skin' (Sk I: 400) b. nqor np nacg- n aro-qav-a kupre-te - nntt tt o- 0 -g?e then old=man-AbsSg three-Ord-Ins net-Ins 2/3S-catch=fish-Aor-2/3SgS 'Then the old man started catching fish with a/the third net' (Sk I: 401) c. t?ar-qav-orv-ak n -tva-qen- ? rjaro-qav-sk which-Ord-sledge-Loc Imperf-be-3S-Sg? three-Ord-Loc 'On which sledge is (s)he? On the third one' (Sk I: 400) d. turi qon?acgan-qav-tore you (Abs) nine-Ord-2Pl 'You are the ninth' (Sk I: 401)

"Headless" participles and ordinal numerals may in turn attach Possessive suffixes, as in (24). (Ordinal numerals can take Relational suffixes as well.) Note that in this usage both participles and ordinal numerals distinguish between the singular (24a/c) and plural (24b/d) of their referents, in line with other nouns referring to humans. (24)

a. r ju-l?- -in(e-t) caat-te graze-Prtcpl-Sg-Poss-Pl lasso-AbsPl 'the herd's (herds') lassoes' (Sk I: 371) b. gat e-etk-al?- rg-ine-jg t ekke-jg t thou (Abs) Neg-come-Prtcpl-Pl-Poss-2Sg son-2Sg 'You are a son of those who have not come' (Sk I: 370) c. t?er-qev- -ine-jg t ekke-jg t? ire-qev- -ine-jg m which-Ord-Sg-Poss-2Sg son-2Sg? two-Ord-Sg-Poss-lSg 'Whose son are you? Of the second one' d. t?er-qev- rg-in(e-t) nenene-t? m tl rj-qav- rg-en(a-t) which-Ord-Pl-Poss 3Pl/AbsPl? five-Ord-Pl-Poss-3Pl/AbsPl 'Whose children are they? Of the fifth ones' (Sk I: 402)

The next example shows that participles can sometimes attach Possessive suffixes in agreement with their overt head. Judging from Skorik's description (Sk I: 371), such cases are extremely rare. (25) r ju-l?- -in inqej- -in caat graze-Prtcpl-Sg-Poss boy-Sg-Poss lasso (Abs) 'the grazing boy's lasso' (Sk I: 371)

4.3. A Cross-cutting Category of Attributive Words The morphological structure of all words discussed under (Bb) above involves overt attributive markers, although different markers for different words— the Possessive and the Relational suffixes, the suffix -qav for ordinal numerals, and the suffixes - l - and -yo- for participles. In this way all words with

318

Chukchi-Kamchatkan

overt attributive marking tend to manifest similar functional and morphological properties, and it is reasonable to assume a cross-cutting category of attributive words10 in Chukchi. In this connection, the behavior of quality state words in modern Chukchi is especially interesting. Their ability to be used as agreeing attributes to nouns in oblique cases shows a gradual shift in their categorial status from mainly predicational words to words that combine both attributive and predicative functions. They are thus approaching the status of words with overt attributive markers, and in fact—at least in attributive functions—they too have an overt attributive marker, the suffix -qin(e)/-qen(a), as seen in (20) (the status of the prefix n- is less clear). It may be suggested that this suffix is gradually undergoing a reanalysis from a predicational (inflectional) suffix for 3rd person to an attributive (derivational) marker. A possible scenario is as follows. Predicational forms for 3rd person are identical to the absolutive case. As described in Section 3.2, nominals (nouns and other types of nonverbal words, like Possessives, Relationals, participles, and ordinal numerals) with predicational suffixes are used not only for straightforward predication of the kind 'You are boys', but also as appositions of the kind 'You boys go home'. In the 3rd person, their predicational form is identical to their absolutive case. Therefore, one and the same morphological form enters into two paradigms: that of case forms (used as arguments and adverbials), and that of predicational forms (used as predicates and appositions). If a word has an overt attributive marker, in the absolutive this marker is either the final element of the word or the penultimate one, followed only by -n in the singular or by -t in the plural. Quality state words originally did not have a category of case, being a subclass of verbs. However, since all quality state words referring to 3rd persons bear the suffix -qin(e)/-qen(a), the whole form is reinterpreted as being identical to the absolutive case and as involving a derivational attributive marker.11 Another curious piece of evidence for the reanalysis of the suffix qin(e)/ -qen(a) as a derivational (attributional) marker is provided by quality state words with augmentative and diminutive suffixes, which always attach to the stem after the suffix qin(e)/-qen(a). Thus in (26), the whole quality state word bears both a predicational suffix for 2nd person and the suffix -qena. (26)

n-ekv -qena-cg-egat Stat-long-Stat-Aug-2Sg 'You are very long' (Sk I:424)

The major differences between Possessives, Relationals, quality state words, participles, and ordinal numerals emerge in the ways they show up as external attributes, or more specifically in whether or not they agree with their heads in number and/or case. Words with the Possessive suffix normally do not agree with the head, although they may do so; words with the Relational suffix normally agree with the head, but may sometimes skip agree-

Possessive and Relational Forms in Chukchi

319

ment; quality state words, participles, and ordinal numerals must agree with their head. The following scale visualizes these differences: Absence of agreement frequent Agreement rare

Agreement frequent

Possessive forms

Relational forms

Agreement obligatory

Absence of agreement rare

quality state words, participles, numerals

Of all these kinds of forms, quality state words, participles, and ordinal numerals are purely qualificational and non-referential, comparable to adjectives in many languages. As such, they also share the cross-linguistic tendency of attributive adjectives to agree with their head. Possessives, on the contrary, are the most referential of these word types: they may be formed from proper names, personal pronouns, and demonstratives, and they may have their own category of inherent number. On the whole, they have the strongest link to nominal inflection (see 2.1) and conform to the general nominal tendency not to be double-marked for one and the same (or similar) category, like case or number. The status of Relationals is less clear; at least, the information available is not sufficient to determine their referential characteristics. 5. Conclusions As has become clear, non-incorporated (external) adnominal attributes in Chukchi, at least when attributed to non-absolutives, bear an overt attributional marker. Insofar as they also tend to agree with their head in case, number, and predication, they simultaneously mark the attributive relation and the function of the head, a process facilitated by the predominantly agglutinative morphology of Chukchi. Agreement of adnominal Possessors, which would constitute the core case of Suffixaufnahme (see Moravcsik's summary in this volume), is relatively infrequent. It has been demonstrated that the morphological status of Possessives is not clear-cut. They are well integrated into the nominal inflectional system, being absolutely regular and maintaining inherent number, but they also show features of derivation, mainly insofar as the Possessive suffix is retained in incorporation. Hence their behavior is governed by two competing principles: to be like other nominal inflections (case) on the one hand, and to be like other words with overt attributional markers on the other. The differences and similarities among attributes in Chukchi corroborate general cross-linguistic tendencies observed elsewhere in this volume: (a) if Possessors agree with their head with respect to a given agreement category, then adjectives agree with their head in the same category; and (b) words with

320

Chukchi-Kamchatkan

attributive markers do not show agreement with their head unless they are capable of taking ordinary noun inflections when forming an NP of their own. It has been argued above that words with attributive markers used as external attributes in Chukchi form a fairly homogeneous category of their own that cross-cuts the other major word classes. It is worth asking whether there are similar tendencies for other functions in Chukchi—i.e., whether there are other cross-cutting patterns that neutralize distinctions of parts of speech. A brief mention of two possible candidates must suffice here. One parallel to overt attributive forms may be found within the system of nonfinite verb forms, which roughly correspond to the nominal system of oblique cases. Thus a considerable number of adverbial relations are treated similarly for both nouns and verbs. Another possible parallel is the more or less uniform treatment of non-dynamic predication, where predication involves nominals, quality state words, and numerals, as well as verbs in the imperfect and perfect.

Notes In preparing this paper I have benefited greatly from stimulating discussions with Jan Anward, Brita Berman, Osten Dahl, and Irina Murav'eva, as well as with the participants of the Suffixaufnahme conference. Frans Plank deserves special thanks for his valuable comments. Responsibility for all remaining errors is, of course, mine. 1. Chukchi is a minor and endangered language spoken in the Far East of Russia. Together with Alutor, Kerek, Koryak, and Itelmen it belongs to the ChukchiKamchatkan family, which is sometimes included in an areal group referred to as Paleosiberian. The data for the present article come mainly from the extensive grammar of Skorik (1961/1977, subsequently referred to by the abbreviation 'Sk' followed by volume and page number), as well as from Comrie (1979, 1981), Nedjalkov (1979), and Polinsky (1994). 2. Similar problems are discussed for Australian languages by Dench and Evans (1988) and Austin (this volume), and for Slavonic by Corbett (this volume). 3. See Comrie (1979), Nedjalkov (1979), and Polinskaja and Nedjalkov (1987). 4. The choice between i and e in the Possessive and Relational suffixes is governed by complex vowel harmony requirements. The final ela of these suffixes disappears in word-final position. Thus, the non-truncated variant, with e/a, appears when the whole suffix is followed by either a number-case suffix or another stem (when the nominal to which it attaches is incorporated into another word). 5. I am indebted to Irina Murav'eva for drawing my attention to these cases. 6. These examples and explanations are taken more or less directly from Skorik I: 283. 7. See Skorik I: 241,271 for details. 8. Sometimes appositions can refer to an ergative NP, as in the following example: mor-ganan, qut rga-muri, msn-tejk-sne-t ? lra-t we-Erg, other-1Pl, 1PlASubj-make-3P-PlP snow=house-AbsPl 'We, the others, let us make snow houses' (Sk I: 217)

Possessive and Relational Forms in Chukchi

321

However, my sources do not provide any examples of Possessives or Relationals for cases like this one. 9. "Purpose forms," mentioned in Section 2.2, seem to show the same behavior. According to Polinsky (1994), however, they are normally incorporated by nouns in oblique cases. 10. Following Smirnickij (1959), Irina Murav'eva (1991) distinguishes different representations of a particular part of speech if there are subclasses of forms each with its own grammatical categories. She accordingly classifies Possessives and Relationals as "adjectival representations" of nouns. 11. According to Skorik I: 425-429, quality state words still have a flavor of predication even as dependents within an NP.

References Comrie, Bernard. 1979. Degrees of ergativity: Some Chukchee evidence. In Ergativity: Towards a Theory of Grammatical Relations, ed. by Frans Plank, 219-240. London: Academic Press. Comrie, Bernard. 1981. Languages of the Soviet Union. Cambridge: Cambridge University Press. Dench, Alan, and Nicholas Evans. 1988. Multiple case-marking in Australian languages. Australian Journal of Linguistics 8, 1-47. Murav'eva, Irina A. 1991. O kategorii reprezentacii. In Tipologija grammaticeskix kategorij, 31-33. Leningrad: Institut jazykoznanija AN SSSR. Murav'eva, Irina A. 1994. Inkorporacija. Moscow: Nauka. Nedjalkov, Vladimir P. 1979. Degrees of ergativity in Chukchee. In Ergativity: Towards a Theory of Grammatical Relations, ed. by Frans Plank, 241-262. London: Academic Press. Polinskaja, Maria S., and Vladimir P. Nedjalkov. 1987. Contrasting the absolutive in Chukchee. Lingua 71, 239-269. Polinsky [Polinskaja], Maria S. 1994. Relativization in Chukchi. In Papers from the 7th Conference on the Non-Slavic Languages of the USSR, ed. by Howard Aronson. Chicago: University of Chicago Press. Smirnickij, Aleksandr I. 1959. Morfologija anglijskogo jazyka. Moscow: Nauka. Skorik, Petr Ja. 1961/1977. Grammatika cukotskogo jazyka. Moscow and Leningrad: Izdatel'stvo akademii nauk SSSR, 2 vols. Skorik, Petr Ja. 1975. Grammaticeskie kategorii i strukturnyj tip jazyka (na materiale inkorporirujuscix jazykov). In Tipologija grammaticeskix kategorij, 120-139. Moscow: Nauka.

This page intentionally left blank

VI CUSHITIC

This page intentionally left blank

11 Genitival Agreement in Awngi: Variation on an Afroasiatic Theme Robert Hetzron

1. Awngi Agreement Awngi, the southernmost branch of Agaw (Central Cushitic), is a strict SOV language with consistent qualifier-head order and with a reasonably transparent agglutinative morphology. Masculine nouns end in -i/-i or a consonant, and feminine nouns have the suffix -a/-a. Plural is normally marked by -ka, although some frequent nominals form plurals idiosyncratically (e.g. by reduplication, as in (2b/c)). Masculines in -i/-i lose this ending before -ka, but mere dropping of the vowel may suffice to express plural. The plural ending -i exists only in demonstratives. The nominative case is unmarked. The accusative ending is -ol-o after consonants and -wa after vowels other than i; for nouns ending in -il-i a merger takes place yielding -el-e. Dative is marked by -s(i); locative 'in, at' by -dal-dd (also marking the possessor in what corresponds to 'have'constructions); ablative 'from' by -desl-des; comparative 'like' by -to; directive 'toward' by -so; adverbial 'as, in the manner of by - a (perhaps more a derivative suffix than a case ending); invocative 'for the sake/love of by -jas (used only in implorations). The choice of tone depends on the preceding noun.1 The agreement rules of Awngi are easy to establish. Within an NP, determiners and adjectives agree with the phrase-final head noun in gender, number, and case:2 (1) a. an-des c nkut-des n n-des that-Abl nice-Abl house-Abl 'from that nice house' b. ann-i-des c nkut-ka-des n-ka-des that-Pl-Abl nice-Pl-Abl house-Pl-Abl 'from those nice houses'

326

Cushitic

Gender, distinguishing masculine and feminine,3 and number, distinguishing singular (unmarked) and plural, constitute a triangular system of "numbercum-gender" as the plural has no gender distinction (see Palmer 1957: 133). Genitival nouns also exhibit obligatory agreement in number, gender, and case with the subsequent head noun. The same is true of the relative construction that is formally closely connected to the genitive construction (Hetzron 1969: 17-19). The same endings are used in both, with -w, -t, or -kw after possessors and relative verbs being chosen depending on whether the head noun is masculine, feminine, or plural, and with the head's case suffix being copied onto the possessor or relative verb. Here are some examples: (2) a. muri-w aqi village-GenMasc manMasc 'the man of the village' b. muri-t una village-GenFem womanFem 'the woman of the village' c. muri-kw aq(ka) / una una village-GenP1 menPl / womenPl 'the men / women of the village' (3)

a. woliji-w-des aqi-w-des n n-des old-GenMasc-Abl man-GenMasc-Abl houseMasc-Abl 'from the old man's house' b. woliji-w-sta aqi-w-sta n-ta old-GenMasc-Comp man-GenMasc-Comp houseMasc-Comp 'like the old man's house'

(4) a. desa-w-sta n-ta study(1Sg)-RelMasc-Comp houseMasc-Comp 'like the house in [unexpressed] which I study' b. desa-kw-sta n-ka-ta study(lSg)-RelPl-Comp house-Pl-Comp 'like the houses in [unexpressed] which I study'

In addition to a focus system operating with a focus suffix and the relative form of the verb, but accompanied by a change of word order, Awngi has two cleft constructions, distinguished through the agreement pattern (Hetzron 1969: 13):4 (5) a. an desa-w-ka nn-i-da n-ka-da I study(lSg)-RelMasc-RefArt this-Pl-Loc house-Pl-Loc b. an desa-kw-ka nn-i n-ka I study(lSg)-Rel Pl -RefArt this-Pl house-Pi 'It is in these houses that I study'

Genitival Agreement in Awngi: Variation on an Afroasiatic Theme

327

The relative verb is followed (probably obligatorily) by a referential article, which is normally used in the sense of 'the aforementioned'. In (5a) the relative verb is followed by a neutral masculine ending, whereas in (5b) it has a plural ending in agreement with the plural cleft complement. In (5a) the cleft complement has its locative case ending, whereas in (5b), as in relative constructions in general, case is not expressed, allowing interpretations such as 'it is in/from/to' etc., to be determined by pragmatics. The endings of three cases have special allomorphs after a genitive ending when copied therefrom head nouns: accusative -e turns into -sa, comparativefa into -sta, and genitive -wl-tl-kw into -su-l-stl-skw. The origin of the added -sis open to speculation (see end of Section 2.3 below); it is homophonous with the dative ending. For expressions like 'A of B of C' the genitive is recursive, although for pragmatic reasons more than two occurrences are avoided. In such an event the first possessor will have two genitive endings: one marks its genitival relationship with the possessee, itself in the genitive; the other, following it, is required by case agreement with the second, lower possessee, which imposes this ending on all that precedes. In the following example, intentionally overloaded for purposes of illustration,5 different genders are used to make the picture somewhat clearer: (6) gud-a-w-skw-da yuna-w-skw-da c nkut- kw-da n- kw-da wodel-ka-da abjel-kada good-Fem-GenMasc-GenPl-Loc womanFem-GenMasc-GenPl-Loc nice-GenPl-Loc houseMasc-GenPl-Loc large-Pl-Loc doorway-Pl-Loc 'in the large doorways of the nice house of the good woman'

Figure 11.1 represents the case-assignment process of (6) through a somewhat unconventional tree structure. As a shorthand notation, categories or exponents of the categories are placed here as the heads of the nodes. Category heads directly govern the lexical elements. What are represented here as exponents exhibit a left-branching structure with a gradually increasing field of dominance. The broken arrow lines ascending toward the left connect the

Figure 11.1. Recursive Genitives in Awngi.

328

Cushitic

source and the target of the agreement. The bottom instructs and the top accepts. The node labels express the result of the agreement, and the slanting straight lines coming out of them hit the node under which everything is affected by the agreement. The first step consists in carrying out agreement in number and gender on the NP3 level, imposing the three bottom nodes on the lexical elements they govern: gud becomes guda as required by u a (which is inherently feminine); the next two elements under 'Masc' remain unchanged; and the last two acquire the plural ending -ka. Now the higher nodes start imposing their agreement elements gradually from the bottom to the top. Since these nodes are constituents of NP1, the grammatical relation between the lower NPs must be the genitive. The NP that is the next sister of the NP governed by the genitival node will establish the gender of the genitive. Within NP2 the masculine genitive, -w, as required by the head 'nice house', will be added to 'good woman' after its feminine ending. One step higher, inside NP1, the genitive is in the plural because of the head 'large doorways', and the appropriate ending -kw is appended to the first two NP3s—in the case of the first, after another genitive ending already assigned. Finally, because the whole NP is marked as a locative because of its function in the clause, the ending -da will appear as the last constituent of every word in NP1, since it is the last to be attached. Another source of double endings is relative clauses whose head nouns are themselves in the genitive, as in (7). (7) kanta-kw-su n-ka-w I=see-GenPl-GenMascch house-Pl-GenMasc 'of the houses which I see' (with the possessee understood to be masculine)

The genitive stands out within the case system in the following respects: (a) it is governed by another noun, whereas other cases are assigned in accordance with the role of their NP in the clause; (b) it exhibits gender-cumnumber agreement with its head; and (c), the main topic of this chapter, it admits other case markers after it. It would nevertheless be inappropriate to regard the genitive as a means of adjectivalization rather than as genuine noun inflection. What argues against an adjective analysis, in addition to the intuition of native speakers (which is not to be belittled), is this. First, genitive endings, unlike derivational ones, may be added to masculine, feminine, or plural nouns as needed. Second, genitival nouns take qualifiers of their own, such as adjectives (as in (3)) and even other genitives (as in (6)); such recursiveness would be quite unusual for a derivational suffix. Third, and most importantly, as (6) and Figure 11.1 show, the genitive endings, just like other case suffixes, participate in the agreement rules which require that a qualifier take the case marker of its head. They are attached by concord to an adjective qualifying the genitival noun (as in (3) etc.), or even to a genitival noun qualifying another genitival noun (as in (6)), producing a sequence of two genitive endings. This would be quite uncharacteristic of derivational formatives. (I strongly doubt that structures like 'quitely importantly' or 'alwaysly

Genitival Agreement in Awngi: Variation on an Afroasiatic Theme

329

importantly' could exist even in languages with very highly developed agreement.) Fourth, while in the case of even the most productive derivational suffixes some formations may be excluded, and while derivational formations may also deviate from the original meaning by being something else than just the sum of their components, genitives in Awngi are one hundred percent productive and semantically absolutely regular. 2. Origin and Developments

2.1. Deictic Origin It is more than plausible that the genitive endings -w/-t/-kw come from original deictics. These survive in a clearer shape in the Awngi presentative pronouns nniku/ nniti/ nnikoni 'here is (Masc/Fem/Pl)' (comparable to French vvoid). The kltlk~h system for masculine/feminine/plural determiners (article and/or demonstrative) is found in the Rift Valley group (Highland East), Oromo, Somali, and other Eastern Cushitic languages as well as in Chadic (Mubi, with traces elsewhere). The original forms may quite reasonably be reconstructed as *ku/ti/kun (see Hetzron 1980: 3.2.1.1, 3.2.2, with further references).

2.2. Agreeing Genitives Elsewhere in Afroasiatic As to other Agaw languages, Appleyard (1975: 325, 338) mentions agreeing relative verbs only for Kemant, and states (forthcoming, Section 4) that he has "not recorded any examples of agreement in genitive phrases." In Bilin, genitival agreement involving "number-cum-gender and case" (Palmer 1958: 390; see Palmer 1957: 158 for the relative verb) occurs only when the possessee precedes the possessor (carrying the agreement ending); the more frequent order here is possessor-possessee, and it is marked by another set of suffixes conditioned by gender-cum-number and class membership of the possessor to which they are attached. Elsewhere in Agaw, alternations in the suffix of the possessor in agreement with the possessee do exist, but they are limited to gender-cum-number and exclude case. If the genitival complex is marked for case, the suffix expressing it occurs only once, in phrase-final position or after the qualifier, as shown for Kemant in (8a/b).6 (8) a. xaya- garwa-s good-GenMasc manMasc-Acc b. xaya- - s g rwa good-GenMasc-Acc manMasc 'the good man' (as object of a transitive verb)

In Khamtanga, "indefinite possessive noun phrases employing the regular possessive case may also optionally show concord between the constituents in respect of the head noun. . . . These same concord suffixes are obligatory

330

Cushitic

when the noun phrase is definite" (Appleyard 1987: 260; on relative verbs, 483ff.). Without giving examples, Appleyard indicates that in Khamtanga the case marker is attached to the possessor noun in a possessor-possessee construction, preceding the concord suffix (see below, Note 12). Outside of Agaw, genitive agreement in number and gender and (with limitations) in case is found in only two of the five members of the Rift Valley group of Eastern Cushitic—Burji and Darasa (Hudson 1976: 254). Exemplifying from Darasa, the genitive suffixes are -nki/-nti if the possessor is nominative (masculine/feminine) and -nkal-nta if it is accusative (masculine/feminine), with the suffixal -n- element dropped if the possessor is a proper name:7 (9) a. jisso-ka dulla Jisso-Gen stick (Acc Masc) 'Jisso's stick' (accusative being the basic form) b. jisso-ki dull-i Jisso-Gen stick-Nom (Masc) 'Jisso's stick' (nominative) (10)

a. warsi-nka dulla dog-Gen stick (Acc Masc) 'the dog's stick' b. warsi-nta isso dog-Gen teeth (Acc Fem) 'the dog's teeth'

In Sidamo the genitive endings vary with the gender of the possessor to which they are suffixed (rather than with the possessee). The two remaining Rift Valley languages, Hadiya and Kambata, use juxtaposition with no segmental genitival morpheme. It is likely, therefore, that the more arbitrary (i.e. less "natural") Burji-Darasa situation is an archaism vis-a-vis the crosslinguistically common Sidamo alternation. Turning to Lowland (East) Cushitic, in Yaaku "nominal gender is not marked overtly;" the distinction of masculine and feminine "can only be inferred from the [gender-marked] dependent categories that accompany the noun," viz. c-/t-/k-x- for masculine, feminine, and plural of genitives, and likewise of adjectives, near demonstratives, and relatives, in agreement with the preceding head noun (Heine 1974: 2.31, 2.331, 2.336). Order in Yaaku is possessee-possessor, and the genitive ending is attached to the possessee, according with its inherent gender (or number): (11)

a. keden c'i yie' treeMasc of = Masc man 'the man's tree b. ai t'i yie' houseFem of=Fem man 'the man's house'

Genitival Agreement in Awngi: Variation on an Afroasiatic Theme

331

The Dullay group also has the order possessee-possessor, but gendersensitive genitive endings are added to the nominal possessor in agreement with its own gender and number (also marked on the noun itself; see Amborn, Minker, and Sasse 1980: 90f.). Gollango is a member of this group: (12)

a.

asko mannee-te thatch housesPl-GenPl 'the thatch of the houses' b. soote kaas-ilo trumpet hornMasc-GenMasc 'a horn-trumpet, trumpet made of horn'

In pronominal possession, on the other hand, the possessive ending follows the possessee and agrees with it, as illustrated from Harso: (13)

a. harko h-aayyu handMasc GenMasc-lSgPro 'my hand' b. nahte c-uusu wifeFem GenFem-3SgMascPro 'his wife'

In either case the choice of the gender of the genitive suffix is determined by the preceding noun, whatever its status in the construction.8 Genitive suffixes that differ according to the gender of their hosts and originate from deictics are abundantly attested in Chadic, although this may be a specific Chadic development (see Skinner 1975). With a preceding head noun/possessee Mubi exhibits -g-/-d-/-h- for masculine/feminine/plural demonstratives, adjectives, and genitive particles (Lukas 1937: 160, 164, 167; cf. also Hetzron 1980: §3.2.2). 2.3. From Deixis to Possession If the genitive and relative endings were originally deictic elements, agreement with the subsequent noun is by no means surprising. Such agreement was motivated at the deictic stage and became frozen, persisting even after these agreeing elements lost their deictic meaning and became genitive markers.9 What one must ask, then, is how deictic elements could become genitive markers. Homonymy of existing demonstratives and genitive/relative particles is indeed attested in Afroasiatic. In Semitic (see Pennacchietti 1968), the masculine singular demonstrative is based on , which also forms the marker of the longer type of genitive (an alternative to the construct-state type, where juxtaposition is the main device but the phrase-initial possessee is also somewhat shortened) and of relative constructions; compare Aramaic di:/zy (Segert 1975: §85.5, 6.2.5) and Ge'ez

332

Cushitic

z- (Diakonoff 1988: 82f.).10 Biblical Aramaic even uses di: as a complementizer (Segert 1975: 6.5.3.5, 7.5.6). As for the Akkadian domain (von Soden 1952: 46, 137, 164), the relative and longer-genitive particle sa may also act as a complementizer in Old Assyrian and Late Babylonian. Old Akkadian still declines this particle, making it agree in gender-cum-number with the preceding possessee (su/sa:t/su:t Masc/Fem/Pl), and even retaining some traces of case marking (with su/sa/si for Norn/Acc/Gen MascSg, as in (14), and very rarely sa:tu Nom and sa:ti Gen for FemSg). (14)

W. ma:ri K. si ata:nim W. son K.Masc of=Masc she=donkey=Gen 'W., son of K., the "she-donkey man" '

The possessor following this inflected particle is in the genitive case. But this inflecting particle was soon replaced by invariable sa. In Akkadian, the demonstratives are not formed from s-, but the 3rd person independent pronouns are, though with different augments (su: 'he', si: 'she', sunu 'they (Masc)', sina 'they (Fem)'). In late Hebrew, both the relative and the complementizer are -, to be compared with the genitive marker l, a compound of and dative /-. Returning now to the question of how a deictic element became a marker of the genitive, the original construction is generally interpreted as a "descriptive, attributive phrase," roughly comparable to expressions such as 'possessor who/that is of the possessee' or 'possessee that-one of the possessor' (cf. Pennacchietti 1968: 55ff.; or more recently Diakonoff 1988: 82). This interpretation attributes precedence to the interclause relative/demonstrative function.11 This is certainly appropriate for Hebrew constructions with ?as r I- (Biblical) and l (post-Biblical), where dative /- serves as a marker of possession. In Old Aramaic, the particle di: seems to have no pronominal conjugation (it had one in Babylonian Aramaic), and therefore dative pronouns have to be used to mark pronominal possession: zy-ly thatto=me 'my, mine', di: leh that to=her 'her' (Segert 1975: §6.2.5.6). However, there may have been yet another path from deictic to possessive, and it may have been the one taken by Cushitic. Cross-linguistically, possessees tend to be definite. In Classical Semitic, a possessee in the construct state never has the definite article; rather, unless it is part of a compound noun, it is syntactically definite, as shown by agreeing adjectives, which carry the definite article if they belong with the possessee. In English genitive constructions (the old man's house), the possessee is definite without any formal marking whatsoever.12 Rather than having clause-initial particles, Agaw and some other Cushitic languages render clauses relative by changing the morphology of the verb into something like a genitive (at least in Agaw). Thus, relative expressions can hardly have been the starting point for genitive constructions. Deictics do qualify, however. One finds in Agaw and Rift Valley Cushitic that, while

Genitival Agreement in Awngi: Variation on an Afroasiatic Theme

333

qualifiers including demonstratives are prenominal, articles are postnominal. We might accordingly hypothesize that ku/ti/kun started out as prenominal demonstratives and determiners. Originally they would have been practically obligatory markers of the possessee's deflniteness, rather than expressing the genitive. With the possessee supposedly definite anyhow (that is why these markers are obligatory in Khamtanga!), they would have been functionally redundant, and could then have been reinterpreted as genitive markers. Moreover, they became suffixes attached to the preceding noun, the possessor. There were, or still are, two more sets of nominal suffixes: articles (also deriving from *ku/ti/kun) and case endings. Their relative order is articlecase, as in Khamtanga firz -y-l horse-Art-Loc (Appleyard 1987: 257). The original definitizer of the possessee, now suffixed to the possessor as a genitive marker, came to occupy the slot of the article, with case markers thus appearing after it.13 This may provide an explanation for the element -s- that follows the genitive and precedes an accusative, another genitive, or a comparative in Awngi, as noted above. This additional element may have been the dative ending that can also be used to mark the possessor. In this scenario, departing from an original construction such as (15), the definitizer of the possessee became a genitive marker of the possessor, while continuing to agree with the possessee in number-cum-gender and case, and it switched positions with the case marker, which later diappeared ( s-ku > -ku-s > -ku), except before three other case markers (accusative, comparative, and genitive). (15)

Possessor-sDat_Gen kuDef Possessee

Pace Aristar (this volume), the Bilin order of possessee preceding possessor (plus agreeing genitive) would then have to be viewed as an innovation, reversing the old order. Typological considerations would also seem to support the originality of the sequence qualifier-qualified, since this is the typical order of OV languages. This reconstruction, with all the hypothetical elements in it, provides a reasonable set of developments, to be considered seriously until challenged by a better one or shown to be flawed.

Notes 1. See Hetzron (1978) for further details. 2. In the transcription, c stands for the affricate [ts], c for the affricate [ts], q and are uvulars (voiceless/voiced), and acute accent marks high tone. 3. The feminine is regularly used for women, females of domesticated animals, some small animals, the sun, the moon, and the stars. In a few cases the same root in the two genders expresses two distinct though related meanings: rft/ rfa (Masc/Fem) 'month'/'moon' (see Hetzron 1978: 124). All other nouns are inherently masculine. However, feminines may be formed from any of these for a diminutive.

334

Cushitic

4. When doing fieldwork (in 1965-1966), I had no time to examine possible functional differences among the three focus constructions. The existence of the cleft constructions may be due to the influence of the Amharic adstratum. 5. It is an elicited example; its correctness, like that of other examples, was approved of by informants. I had some spontaneous double genitives, but these were less "loaded" constructions. 6. Presumably the "adjective" here is a relative verb; in principle, at any rate, the rules should be the same as for the agreeing genitive. The Bilin material of Leo Reinisch has examples of the case marker added to the end of the phrase; see Hetzron (1976: 4.2.1.2.2.2, 4.3.1.3, 4.3.2.2). 7. It is curious that the etymological t (the feminine marker) appears here as a glottalized t in Darasa, but not in other contexts nor in Burji. Likewise, Khamtanga has plural genitive -ikw where -kw is expected (cf. Appleyard 1987: 248f. on free variation t ~ t). It seems that t may be reduced to t, but t need not become t. Thus, they are two separate phonemes. 8. Note that in Romance languages possessive pronouns agree in gender with the possessee (cf. French son pere/sa mere), as against their gender being determined by the possessor in Germanic (his/her father, his/her mother). 9. I do not think that the kind of construction has changed significantly, e.g. from loose apposition to close-knit attribution, as the demonstrative was reinterpreted as genitive. This appears to be one difference between Aristar's account (later in this volume) and mine. 10. These members of the Semitic family, which was originally VSO, have the orders possessee-possessor and head-relative clause. 11. The close affinity between these two categories is only too clear in general. In English that is also used as a relative marker (besides interrogative pronouns), and Hungarian uses a combination of a demonstrative and an interrogative as a relative pronoun. 12. Naturally there exist situations where the possessee must be conceived of as indefinite, as in English a house of the old man's. Classical Semitic languages use another construction with a dative marker for the possessor in such cases. But whatever the formal means, definiteness is felt to be unmarked for possessees. 13. The Khamtanga construction possessor noun + definite marker + case marker + concord suffix—possessee noun must be a later development, being a manifestation of the tendency to place phrasal suffixes after the first word of the NP, as in Amharic y h- n saw this-Acc man 'this man' (object of a transitive verb).

References Amborn, Hermann, Gunter Minker, and Hans-Jiirgen Sasse. 1980. Das Dullay: Materialien zu einer ostkuschitlschen Sprachgruppe. (Kolner Beitrage zur Afrikanistik 6.) Berlin: Reimer. Appleyard, David L. 1975. A descriptive outline of Kemant. Bulletin of the School of Oriental and African Studies 38, 316-350. Appleyard, David L. 1987. A grammatical sketch of Khamtanga. Bulletin of the School of Oriental and African Studies 50, 241-266, 470-507. Appleyard, David L. Forthcoming. The position of Agaw within Cushitic. In Karel Petracek Memorial Volume. Prague.

Genitival Agreement in Awngi: Variation on an Afroasiatic Theme

335

Diakonoff, Igor M. 1988. African Languages. Moscow: Nauka. Heine, Bernd. 1974. Notes on the Yaaku language (Kenya). Afrika und Ubersee 58, 27-61. Hetzron, Robert. 1969. The Verbal System of Southern Agaw. (University of California Publications, Near Eastern Studies, 12). Berkeley: University of California Press. Hetzron, Robert. 1976. The Agaw languages. Afroasiatic Linguistics 3, 31-57. Hetzron, Robert. 1978. The nominal system of Awngi (Southern Agaw). Bulletin of the School of Oriental and African Studies 41, 123-141. Hetzron, Robert. 1980. The limits of Cushitic. Sprache und Geschichte in Afrika 2, 7— 126. Hudson, Grover. 1976. Highland East Cushitic. In The Non-Semitic Languages of Ethiopia, ed. by M. Lionel Bender, 232-277. (Monograph No. 5, Occasional Papers Series, Committee on Ethiopian Studies.) East Lansing: African Studies Center, Michigan State University. Lukas, Johannes. 1937. Zentralsudanische Studien. Hamburg: Friederichsen. Palmer, Frank R. 1957. The verb in Bilin. Bulletin of the School of Oriental and African Studies 19, 131-151. Palmer, Frank R. 1958. The noun in Bilin. Bulletin of the School of Oriental and African Studies 21, 376-391. Pennacchietti, Fabrizio A. 1968. Studi sui pronomi determinativi semitici. (Ricerche 4.) Naples: Istituto Orientale di Napoli. Segert, Stanislav. 1975. Altaramaische Grammatik. Leipzig: Enzyklopadie. Skinner, Neil. 1975. The Hausa genitive morpheme as an exponent of gender: A query. In Hamito-Semitica, ed. by James and Theodora Bynon, 389-398. The Hague: Mouton. Soden, Wolfram von. 1952. Grundri der akkadischen Grammatik. (Analecta Orientalia 22.) Rome: Pontificium Institutum Biblicum.

This page intentionally left blank

VII AUSTRALIAN

This page intentionally left blank

12 Suffixaufnahme and Related Case Marking Patterns in Australian Languages Fritz Schweiger 1. Introduction The aim of this paper is to survey the occurrence of Suffixaufnahme and related patterns of double case marking in Australian languages. Following the classical prototype in Old Georgian, two conditions give rise to a case marking pattern called Suffixaufnahme: first, an adnominal possessor is marked by a case suffix usually labeled genitive (although in Australian languages syncretism of genitive and dative is common; see Blake 1977); second, the entire NP is marked for its case role by attaching an appropriate suffix to the head (=possessed) and to the dependent (=possessor), with the dependent thus receiving double case marking. Alternatively, in some Australian languages the NP containing a possessor phrase is marked for its case role by attaching an appropriate case suffix to just one of its words. If this case marking suffix is attached to a possessor already in the genitive, this is another instance of double case marking, although not one involving agreement in case. The following case marking patterns within the NP may be distinguished (see Blake 1987; Dench and Evans 1988): I. Complete concord (=word-marking): All words in the phrase are marked for case. If an NP containing a possessor has word-marking, then Suffixaufnahme in the classical sense, meeting the two conditions stated above, is possible. A typical example is found in Anguthimri (Crowley 1981): (1) lu watayi- amra-ma patra-ma pae-ni he=S old=man-Gen-Abl canoe-Abl come=out-Past 'He got out of the old man's canoe' II. Phrase-marking: The case marker is attached to some word of the NP. Three subtypes can be distinguished here:

IIa. Final-marking: The last word of the NP is marked for case. (In Blake 339

340

Australian

1987 the notion of phrase-marking is restricted to final-marking, which in fact seems to be the most common type of phrase-marking in Australian languages.) This type is exemplified in Guugu Yimidhirr (Haviland 1979; on Lig(ative) see below): (2) biiba yarrga-aga-mu-n gudaa gunda-y father boy-Gen-Lig-Erg dog (Abs) hit-Past 'The boy's father hit the dog'

IIb. Head-marking: The head of the NP is marked for case. Here is an example from Ngandi (Heath 1978; MascSg is a noun-class prefix for masculine singular): (3) a-watu-?nayi-tu ni-jawulpa-gu agura-ga-baA-dog-his-Erg MascSg-old=man-Gen it=bit=me 'The old man's dog bit me'

IIc. Free-marking: Any word of the NP may be marked for case. The following example is from Nyigina (Dench and Evans 1988): (4) gudyarra-ni wamba mug yirrinymirri yila two-Erg man hit they=did=it dog 'Two men hit the dog'

If the NP containing a possessor has phrase-marking, double case marking may occur; this is reminiscent of Suffixaufnahme, except that the second case on the possessor is not triggered by agreement. It should be mentioned that some languages have both complete concord and phrase-marking. If in the above sentence (2) from Guugu Yimidhirr the dependent is separated from its head, both constituents are marked: (5) yarrga-aga-mu-n gudaa gunda-y biiba-ngun boy-Gen-Lig-Erg dog (Abs) hit-Past father-Erg 'The boy's father hit the dog'

In some languages peripheral (local) cases may also have an adnominal function. Since these occurrences are even less well-documented than multiple case marking with genitive, I have included such examples only sporadically. I will not include in my survey two other suffixes that are virtually universal in Australia and are related to the semantic notion of possessor, namely the proprietive or "having" suffix (admirably surveyed in Dixon 1976) and its antonym, the privative or "lacking" suffix (less well-studied cross-linguistically). Traditionally categorized as derivational, but sometimes with good reason treated as adnominal case markers (Dench and Evans 1988), these suffixes actually appear to show properties of both derivation and inflection (Blake 1991). Furthermore, some languages commonly have possessive suffixes that

Suffixaufnahme

and Related Case Marking Patterns in Australian Languages 341

mark nouns as being "possessed" by someone. In Pitta-Pitta, for instance, wara and -mala mark human nouns as being possessed by a 3rd and 2nd person respectively (Blake 1979: 200): (6) a. titi-wara-lu piti-ka wa i-wara-na older=brother-Poss-Erg hit-Past younger=sibling-Poss-Acc 'The older brother hit the younger' b. titi-mala-lu anya piti-ka older=brother-Poss-Erg 1Sg=Acc hit-Past 'Your brother hit me'

A number of Australian languages that allow multiple case marking require the insertion of an additional morpheme ("ligative" or "linking morpheme") between genitive and further inflection. In Djabugay these ligative forms, some of which appear in the following examples (Patz 1991: 269), are -mu- for ergative and dative, -munggu- for purposive/allative, and -mundafor aversive; no linking morpheme is used before instrumental/locative. (7) a. ngawu bibuy-ngun-mu-nda gurra:-nda wa:-ng ma: 1Sg=A child-Gen-Lig-Dat dog-Dat give-Pres food (P) 'I give the child's dog (some) food' b. ngawu yarrn.ga-ng bibuy-ngun-munda-lan gurra:-lan 1Sg=S fear-Pres child-Gen-Lig-Avers dog-Avers 'I'm afraid of the child's dog'

A separate topic would be the type of multiple case marking Libert (1988) calls an "arbitrarily required case," where a case marker is obligatorily attached to a noun stem already bearing a case marker which serves as a ligative or linking suffix. Typical examples are from Kalkatungu (Blake 1969: 33): Causative Ablative Allative I Allative II

= = = =

Operative plus - u Locative I plus - u Genitive plus -na Locative II plus -na

Kalkatungu actually invalidates Libert's first and second generalization. His first generalization states that "arbitrarily required cases" must be syntactic. In Kalkatungu operative (ergative) is clearly syntactic, and genitive (=dative) is syntactic for a restricted range of verbs, but locative I and II are not reported to be syntactic. Libert's second generalization implies that there should at most be one "arbitrarily required case" preceding another case. An alternative analysis would assume different stems or roots required by these case suffixes; this proposal is especially apt for the analysis of pronoun inflection. Thus in Wargamay (Dixon 1981) we find the following pronominal forms for 1st person singular: ayba intransitive subject, adya transitive subject, anya transitive object, aygu genitive, and oblique stem aygun-, to which

342

Australian

dative/allative -gu, locative/aversive -da, and ablative -iny are added. (One might assume that dative aygungu can be analysed as genitive plus -ngu; but such an analysis causes problems for 1st person dual ali with genitive ali u and oblique stem alin-.) For general information about case systems in Australian languages the reader is referred to Dixon (1980) and Blake (1987), both excellent introductions to Australian linguistics as a whole. 2. The Sources and Presentation of Data Section 3 presents the results of a literature search using a sample of fifty Australian languages. Although it includes languages from all parts of Australia and from a range of Pama-Nyungan and non-Pama-Nyungan groups, the sample is not a representative random sample in the technical terms of statistical inference. (In fact, I doubt that it is possible to obtain representative random samples of languages at all.) The availability of reliable grammatical descriptions being limited, I concentrated my search on those published more recently, a fair proportion of which were examined. However, in numerous instances the material, often obtained from the very last speakers of the language concerned, is scanty and does not mention whether multiple case marking has been observed. There are two notable reasons why Suffixaufnahme may have escaped attention: (a) inalienable possession is typically expressed by apposition (Djabugay being a remarkable exception); and (b) in many Australian languages nouns follow an ergative-absolutive case marking pattern—intransitive subjects and transitive objects are typically unmarked, and no overt case marker is therefore copied onto their possessors. In Section 3 the languages surveyed are arranged in alphabetical order, with a broad indication of their genetic affiliation and geographical location. Classifications basically follow Wurm (1972) but take into account revisions proposed by Blake (1988, 1990). If multiple case marking has been reported, at least one example sentence is given, with the page or example number in the source. If the grammatical description explicitly states that no multiple case marking (within the limited scope of the present survey, as defined in the introduction) is allowed, this is recorded here. The comment "no data on Suffixaufnahme" means that I was unable to find either a positive example or an explicit negative statement in my source. The responsibility for any erroneous attribution is completely mine. Occasional reference is made to morphological sequence conditions as manifesting themselves in so-called case haplology, case spacing, inner case substitution, inner case deletion, outer case deletion, case reordering, or case portmanteaux, dealt with in detail in Dench and Evans (1988). Although I have basically retained the original orthography of the sources, some symbols have been substituted for practical reasons. Interlinear glosses have sometimes been slightly altered in the interest of uniformity. The boundary symbol = is used when glosses deviate from the morphemic

Suffixaufnahme

and Related Case Marking Patterns in Australian Languages 343

segmentation of the example. Zero-marked categories are sometimes indicated in parentheses. 3. The Data ALYAWARRA (Pama-Nyungan, Northern Territory; Yallop 1977) is phrasemarking and allows Suffixaufnahme: (8) ayliyla artwa ampu-kinh-ila boomerang man old-Gen-Ins 'with the old man's boomerang' (p. 117) (9) aringka atjinhila akira inika dog 1Sg=Gen=Erg meat (Nom) get=Past 'My dog got the meat' (p. 118)

ANGUTHIMRI (Pama-Nyungan, Cape York Peninsula, Queensland; Crowley 1981): "Where the possessor NP is in a non-zero case inflection, the genitive NP can agree with it" (p. 182). (10)

lu watayi- amra-ma patra-ma pae-ni he=S old=man-Gen-Abl canoe-Abl come=out-Past 'He got out of the old man's canoe' (ex. 64)

BAAGANDJI (Pama-Nyungan, New South Wales; Hercus 1982) is primarily a phrase-marking language, although its rules seem not very strict in this respect (pp. 99, 102). Complete concord may occur when a free pronoun is part of the NP (pp. 100f.). No data on Suffixaufnahme. BIDYARA / GUNGABULA (Pama-Nyungan, Queensland; Breen 1973): "If an inflected noun is an adjective phrase it fills a position in a sentence that could be filled by a noun or adjective and it should therefore be able to be inflected just as a noun . . . i.e. to take a second suffix. However, there are no examples of this" (pp. 52f.). DHARAWAL / DHURGA (Pama-Nyungan, New South Wales; Eades 1976): "We have no evidence of a form inflected for one case being further inflected, as for example, a noun inflected for possessive case being further inflected as a transitive subject. This could be because the languages do not permit such double case inflections, or because R. H. Mathews [in his grammatical sketch of 1901] was unaware of such phenomena" (p. 48). DIYARI (Pama-Nyungan, South Australia; Austin 1981): Normally, only the last non-pronominal constituent of an NP is marked for case of the phrase (final-marking). The two exceptions to this rule are the separation of constituents and special emphasis. Suffixaufnahme is regular for pronouns: "They typically follow the possessed head noun in an NP and hence take the case marking of the NP as a whole" (p. 62).

344 (11)

Australian nawu-ya ama-yi ura akana-ni 3Sg-Near sit-Pres camp 2Sg=Dat-Loc 'He is sitting in my camp' (ex. 37)

Dative case pronouns can therefore be viewed as possessive pronouns. There seems to be no Suffixaufnahme for other NPs: "Other genitive NPs must precede the possessed element" (p. 138). (12)

nulu kudu paku-yi wila-ya wana-li 3Sg hole (Abs) dig-Pres woman-Dat digging=stick-Erg 'He is digging a hole with a woman's digging stick' (ex. 283)

DJABUGAY (Pama-Nyungan, Queensland; Patz 1991): "Genitive formations in Djabugay are inflected for case according to their syntactic function. . . . If an inflectional suffix is attached to a genitive stem a linking morpheme has to be inserted between stem and suffix for all cases except instrumental/locative" (p. 269). (13)

ngawu bibuy-ngun-mu-nda gurra:-nda wa:-ng ma: 1Sg=A child-Gen-Lig-Dat dog-Dat give-Pres food (P) 'I give the child's dog (some) food' (ex. 20)

I could not find examples of inflected pronominal genitives in Patz's data. However, pronominal genitives obviously inflect like genitives of nouns: on p. 276 a table of pronominal genitive stems is given. DJAPU (Pama-Nyungan, Arnhem Land, Northern Territory; Morphy 1983) is a word-marking language. "Every constituent of an NP generally bears a case-marker, with some regular exceptions" (p. 82). Suffixaufnahme is therefore quite normal. In fact, complex rules apply, which may be summarized as follows: (a) If the head NP is in the accusative or absolutive case (i.e. in P and S function), the attribute is marked by a case form labeled "(simple) Gen(itive)." (b) If the head NP is marked ergative/instrumental/causative or is a common noun (with no human referent) marked by a local case, the attribute is marked by a case form labeled "Obl(ique)." (c) In all other instances a case form labeled "Obl(ique)St(em)," which can be interpreted as oblique plus ligative, plus a case marker reflecting the marking of the head, is used. (14)

arra nhina dhiyal wa a- ur Milyin-gal yapa-'mirri u-wal arra-kala u-wal 1Sg=Nom sit (Unmarked) here (Loc) camp-Loc Milyin-Obl elder=sisterPropkin-Obl 1Sg-OblSt-Obl 'I am sitting at my sister Milyin's camp' (ex. 42)

Suffixaufnahme (15)

and Related Case Marking Patterns in Australian Languages 345

arra-ny utha-n-mara- al yumurrku-n yapa-'mirri u-w arra-kalar u-w 1Sg=Nom-Prom grow-Unmarked-Caus-Perf children-Acc elder=sister-PropkinDat 1Sg-OblSt-Dat 'I have raised my elder sister's children' (ex. 41)

(Propkin is the kinship proprietive, Prom is a prominence clitic.) For a more detailed discussion of these rules see Sections 4-8 in Morphy (1983). DJINANG / DJINBA (Pama-Nyungan, Arnhem Land, Northern Territory; Waters 1989): ". . . there is sometimes marking of each nominal constituent of an NP with the same case, and at other times there is not—usually just one nominal being marked for case (often not the head nominal)" (p. 18). "We are forced to predict that double case marking sometimes occurs at the abstract level of the case system, while, at the level of surface realisation, overt double case marking is quite rare (and, in fact, is avoided as much as possible)" (p. 19). But "note that Gen marking, when used adnominally in a possessive context, may take further case suffixation" (p. 80), for example: (16) gurrbi-ngir ngirr-ang-ngir camp-Abl 1Sg-Gen-Abl 'from my camp' (ex. 150) DJINGILI (Barkly, Northern Territory; Chadwick 1975): "The fact that adjectives take suffixed concord with nouns for number, class and case is the main structural characteristic which separates Djingili . . . from other languages in North Australia" (p. 19). Chadwick distinguishes three classes of adjectives: possessive adjectives, demonstratives, and other adjectives. Possessive NPs are formed with the help of the dative form of 3rd person pronouns: (17) a. gurnju anu baina skin 3Sg=Neut=Dat man=Dat 'the man's skin' (p. 21) b. gu nju ai i gabalamana skin 3Sg=mi=Dat yam=Dat 'the skin of the yam' (p. 21) (Djingili has four genders: masculine, feminine, neuter, and a mi-class, usually accommodating vegetable food.) No Suffixaufnahme found. DYIRBAL (Pama-Nyungan, Queensland; Dixon 1972) has two genitives, with the "simple genitive" - u indicating a relation of present possession, and the "general genitive" (Gengnrl) -mi, not available for pronouns, indicating a relation of past ownership (p. 108). Both genitives may take further case inflections. The simple genitive - u is used with a ligative affix -(ny)dyin. Note that Dyirbal has full concord. The items glossed as 'there' are noun-classifiers.

346

Australian y

y

(18)

balan d ugumbil aygud indu ba gul galbindu balgan There=Nom=II woman (Nom) 1Sg=Gen=Erg There=Erg=I son=Erg hit=Pres/Past 'My son hit the woman' (ex. 309)

(19)

ayguna ba gul wa aru yaramigu minban 1Sg=P There=Ins=I boomerang=Ins man=Gen gnrl =Ins hit=Pres/Past 'I was hit by the boomerang that belonged to the man' (ex. 329)

"Sometimes . . . a possessive phrase in genitive case, with or without affix dyin, but without any further case inflection, may modify a non-topic NP" (p. 106). "In the case of dual or plural pronouns the genitive - u ~ -nu affix can optionally be omitted in -dyin forms" (p. 244). A "possessor of a possessor" is reduced to the form of a "possessor:" (20)

aygu bulgu u bala gadyin 1Sg=Gen wife=Gen There=Nom=IV yamstick 'my wife's yamstick' (ex. 317)

The ablative case suffix - unu is formally identical with a stem-forming nominal affix - unu with a related meaning: (21)

adya nyinany yara ga yi ari unuga 1Sg=S/A sit=Fut man=Loc hole=Abl=Loc 'I'll sit by the man [who had been pulled out] from the hole' (ex. 598)

GIDABAL (Pama-Nyungan, New South Wales; Geytenbeek and Geytenbeek 1971): "Actor and Object (but not Indirect Object) suffixes are optional" (p. 13). Complete concord as well as final-marking seem possible. No evidence for Suffixaufnahme has been found, as the following examples show (gender marking with -gay is optional): (22) njule- a: gamay-gay djagabar agam-bu yi a-ni a-nji 3Sg-Poss big-Neut black dog-Actor bite-Aor 1Sg=Obj 'His big black dog bit me' (p. 34) (23) njule- a: bana:m-a yaraman-gu munu wuda 3Sg-Poss younger=brother-Poss horse-Ben that grass 'That grass is for his younger brother's horse' (p. 34)

GOONIYANDI (Bunuban, Western Australia; McGregor 1990): Double case marking (with case markers referred to as postpositions, see p. 175) seems quite unusual: "On one occasion only, -binyi Per[lative] was followed by the Erg postposition—on repetition of the example, the Erg was omitted" (p. 187). The relevant example is: (24)

balyoowa-binyi-ngga gilbawidi thinga behind-Per-Erg they=found=him foot 'They found his footprints from behind' (ex. 3-97)

Suffixaufnahme

and Related Case Marking Patterns in Australian Languages 347

However, on p. 176 it is stated that the case markers -nhingi Abl1, -binyi Per[lative], and -ngaddi Com[itative] (which is similar to a "having" suffix) may be followed by -ngga Erg. To avoid Suffixaufnahme, a "topicalization" pattern is employed: (25)

ngooddoo-yoo yoowooloo thadda nganyi nhoowoo-ngga thadda wirdnginbini that-Dat man dog 1Sg his-Erg dog it=bit=me 'That man's dog bit me' (ex. 4-82)

GUMBAYNGGIR (Pama-Nyungan, New South Wales; Eades 1979) is wordmarking (=complete concord) and has Suffixaufnahme. (26)

an y undi ma:la dyamay barway nyami-gundi-ya 1Sg=Gen hand (S) Particle big (S) woman-Gen-Loc 'My hand is bigger than the woman's' (ex. 16)

"It appears that a genitive form can optionally take a further case inflection" (p. 317): (27)

yara ni:gar gudyu: niya gidam gayi anyundi-ya gagu: umbala Demonstrative man (S) Loc=Deictic 3P1=S talk (Pres) 1Sg=Gen-Loc brother=Loc (ex. 157) 'Those men over there are talking to my brother'

(28) ma:na a:lu ni:gar-gundi nyami:gu take=Imp water (P) man-Gen woman=All 'Take water to the man's wife' (ex. 158) (29)

ni:gar-gundi-yu w a n y i : d y u yi:nydya gi:bar man-Gen-A dog=A bite=Past boy (P) 'The man's dog bit the boy' (ex. 162)

GUUGU YIMIDHIRR (Pama-Nyungan, Queensland; Haviland 1979): "Possessive expressions in Guugu Yimidhirr, as in many Australian languages, accept further case specification. That is, genitive suffixes form, from a noun N, a further nominal stem (meaning 'belonging to N') which modifies another noun (the 'thing possessed') and which must agree with it in case" (p. 56). The following morphological rules apply: (a) genitive followed by absolutive is expressed by dative -bi/-wi; (b) genitive followed by any other case is expressed by goal -:ga followed by ligative -mu and the case concerned. Furthermore, for case = Erg/Ins, Dat/Loc/All, or Abl/Caus, there are special case forms (p. 56). Similar complications arise for pronouns. Here are some examples: (30)

ngayu dhada-y biiba-aga-m-i 1Sg=Nom go-Past father-Gen-Lig-All 'I went to my father's [place]' (ex. 36)

348

Australian

(31)

gudaa ngaanhdhu-wi biini dog (Abs) woman-Gen (Abs) die=Past 'The woman's dog died' (ex. 33)

(32)

nhayun bubu ngadhu-um-i biiba-wi that land (Abs) 1Sg=Gen-Lig-Gen (Abs) father-Gen (Abs) 'That is my father's homeland' (ex. 391)

If the NP is not discontinuous, final-marking is the rule; if the head noun follows the genitive expression or is separated from it, both head and modifier carry inflection: (33)

a. biiba yarrga-aga-mu-n gudaa gunda-y father boy-Gen-Lig-Erg dog (Abs) hit-Past b. yarrga-aga-mu-n gudaa gunda-y biiba-ngun boy-Gen-Lig-Erg dog (Abs) hit-Past father-Erg 'The boy's father hit the dog' (ex. 37, 38)

IWAIDJA (Iwaidjan, Coburg Peninsula, Northern Territory; Pym 1979): No Suffixaufnahme is found, as is to be expected since Iwaidja has a complex system of prefixes and suffixes within the verb system and virtually no nominal inflection. KALKATUNGU (Pama-Nyungan, Queensland; Blake 1979): "Since case marking is applied to all words in the noun phrase and since the dative has adnominal usage, it is possible that case marking will need to be added to the constituents of a noun phrase where one is already marked by the dative" (p. 49). A ligative is employed (case spacing): (34)

ai a-ci-wa-tu kunkuyu-lu anya kuu 1Sg=Nom 1Sg-Dat-Lig-Erg daughter-Erg give=Past water 'My daughter gave me water' (ex. 3.73)

However, Suffixaufnahme is not obligatory (if the outer case is ergative): (35)

a-ci matu-yu tuku layi-na 1Sg=Dat mother-Erg dog hit-Past 'My mother hit the dog' (ex. 3.77)

An antipassive construction exhibits an example of a ligative being inserted between two datives: (36)

a-tu caa nanya marapai matu-inyci-i-ya-ku tu-yi-nyin maa-ci 1Sg=Erg here see=Past woman mother-her-Dat-Lig-Dat cook-Antipass-Prtcpl food-Dat 'I saw the woman cooking food for her mother' (ex. 3.75)

Suffixaufnahme

and Related Case Marking Patterns in Australian Languages 349

The view of the genitive as derivational is suggested by examples like (37), where the head of the adnominal phrase is omitted or in some sense incorporated: (37)

kupa uru-u-ya-tu ai icayi old=man-Dat-Lig-Erg 1Sg=Nom bite=Past 'The old man's [dog] bit me' (ex. 3.74)

KUKU-THAYPAN (Pama-Nyungan, Queensland; Rigsby 1976): "The schema for a common type of alienable possessive phrase is (possessor noun)possessed noun-genitive pronoun, in which the genitive pronoun agrees in person and number with its antecedent 3rd person possessor noun, but in case with the possessed noun. Simple deictic or anaphoric possessor pronouns appear only in genitive case-form, which may take further case-inflection to agree with the possessed noun" (p. 262). This means that Suffixaufnahme is restricted to pronouns, possibly indicating a development of possessive pronouns. (38) lpu-lkal to iw-anda te-n nay old-man dog his-Erg bite-NonFut me 'The old man's dog bit me' (p. 262) (39)

to- ga pigipigi may-n taw dog-Erg pig find-NonFut my 'My dog found a pig' (p. 263)

(It is not clear why sentence (39) could not be translated 'The dog found my Pig'.) MALAK MALAK (Daly, Northern Territory; Birk 1976): In the expression of alienable possession the possessor noun follows its head. (40)

muyin yinya-no dog man-Poss 'the man's dog' (ex. 342)

Since adnominal affixes are suffixed to the final word of the constituent, multiple case marking could occur, but no Suffixaufnahme has been found. MANGARAYI (Mangarayan, Northern Territory; Merlan 1982): "A genitive noun is cross-referenced by a possessive pronominal suffix on a possessed noun: (41)

na-bugbu -gu -banam-nawu a-wa-b MascGen-old=man-Gen NeutAbs-camp-his 1Sg/3Sg-visit-Past 'I visited the old man's camp'

350

Australian

Genitive nouns are not declined to agree with the case of the possessed noun: (42)

na-bugbu -gu -banam-nawu -galama ga- a-yag MascGen-old=man-Gen NeutAH-camp-his-All 3-1Sg-go 'I am going to the old man's camp' " (p. 66).

Case prefixes in Mangarayi, and similarly in Mara (where there is an additional class prefix), also express gender. MARA (Maran, Arnhem Land, Northern Territory; Heath 1981) has some evidence for multiple case marking; compare these two examples: (43)

a. n- a-radbur n-jawuru Class-NeutNom-camp Class-3Sg=Gen 'his camp' b. na-radbur-yu(r) na-yawuru-yu(r) NeutObl-camp-All/Loc NeutObl-3Sg=Gen-AIl/Loc 'to his camp' (p. 64)

Unfortunately, the data available lack an allative analogue to the NP in the following example: (44)

na-'riyi-mar n- a-radbur (n-jawuru) man [Oblique form] Class-NeutNom-camp (Class-3Sg=Gen) 'the man's camp' (p. 64)

MARGANY and GUNYA (Pama-Nyungan, Queensland; Breen 1981a): "This category [genitive] applies only to personal pronouns and denotes ownership. . . . A genitive pronoun is a derived noun (as it can be inflected as a noun)" (p. 308). (45)

aya wabangu atyu- ga bama- ga 1Sg go=Purp 1Sg=Gen-Loc brother-Loc 'I'm going with my brother' (ex. 25M)

"This [dative] marks ownership (except with personal pronouns) or indirect object. . . . There is no evidence on whether a dative noun can, like a genitive pronoun, function as a derived noun stem" (p. 309). However, complete concord seems to be the norm: (46)

inu- gu wanba- gu uda(- gu) ana bada:ni 2Sg=Gen-Erg big-Erg dog(-Erg) 1Sg=Acc bite=Recent 'Your big dog bit me' (ex. 210)

MAUNG (Iwaidjan, Goulbourn Island and Arnhem Land; Capell and Hinch 1970): No data on Suffixaufnahme.

Suffixaufnahme

and Related Case Marking Patterns in Australian Languages 351

MURUWARI (Pama-Nyungan, New South Wales/Queensland; Oates 1988): "Dixon (1980: 300, 321) has suggested that possession (usually labeled genitive case) should be classed as a derivational suffix rather than an inflection. However, the genitive in Muruwari is clearly [a] case" (p. 51). Muruwari has word-marking or free-marking (with a preference for head-marking in ergative NPs). "[W]hen a genitive noun phrase occurs in function A as the subject of a transitive verb, the possessor noun is marked with the genitive, the possessed noun with the ergative" (p. 68). NGADJUMAJA (Pama-Nyungan, Western Australia/South Australia; Brandenstein 1980): The material is scanty. There seems to be complete concord: (47)

midyal-da alba-gga water-Loc big-Loc 'on the big water (sea)' (p. 21)

No Suffixaufnahme was found. NGALAKAN (Gunwingguan, Arnhem Land, Northern Territory; Merlan 1983): "There is no 'double' case marking; genitive nouns are not further suffixed to express case functions of the entire NP" (p. 44). NGANDI (Gunwingguan, Arnhem Land, Northern Territory; Heath 1978): "The Genitive is not 'declined' to agree with the case of the modified noun" (P. 43). (48)

a-watu- nayi-tu ni-jawulpa-gu agura-ga-baA-dog-his-Erg MascSg-old=man-Gen it=bit=me 'The old man's dog bit me' (p. 43)

If two nouns form an NP by apposition, it is possible to mark both (MascSg and Gu being noun-class prefixes): (49)

a-rudu- i gu-rer- nayi-gic ni-jawulpa-gic I=went Gu-house-his-All MascSg-old=man-All 'I went to the old man's house' (p. 44)

NGANKIKURUNGKURR (Daly, Northern Territory; Hoddinott and Kofod 1988): "It is the NP rather than the nominal which is marked for case in Ngankikurungkurr. Case is indicated by the addition of a clitic to some constituent of the NP, usually the final one. Case marking is not however obligatory" (p. 70). Only combinations of locative and directional markers have been found, but no Suffixaufnahme. NGAWUN (Pama-Nyungan, Queensland; Breen 1981b): No Suffixaufnahme is reported for noun stems: "The genitive suffix is - u which occurs with and following the allative suffix -pir. It marks the owner of a person (in kinship) . . . or of an animal or object. . . . There are no examples of further

352

Australian

suffixation of a genitive form" (pp. 44f.). A similar remark applies to pronouns. Ngawun seems to be a free-marking language, but at least for the ergative complete concord has also been recorded. NGIYAMBAA (Pama-Nyungan, New South Wales; Donaldson 1980): "Normally every constituent of an NP is marked for the case function of NP as a whole" (p. 232). Suffixaufnahme is attested for nouns, where dative -gu is used for possessive function, and for pronouns, where 1st and 2nd person "free form" pronouns use an oblique case form for possessive function. (50)

adhu giyanhdha-nha idji-la winar-gu-dhi miri-dji 1Sg=Nom fear-Pres this=Circumstative-Est woman-Dat-Circumstative dogCircumstative 'I am frightened of this woman's dog' (ex. 4-56)

(51)

adhi:-gu-gila miri-gu girbadja gadha-ra 1Sg=Obl-Erg-Hypothetical dog-Erg kangaroo (Abs) bite-Pres 'I reckon my dog (can) catch kangaroos' (ex. 9-70)

('Est' is short for 'established reference'.) NUNGGUBUYU (Nunggubuyan, Arnhem Land, Northern Territory; Heath 1980a, 1984): Genitive is marked by a suffix /-yinyung/, which has a wide range of uses. It is worth mentioning that /-yinmns/ may be added to a noun which already has a non-zero case marker, such as pergressive (Per): (52) mana:-'mi:-'la-yung manggarabi ama-madhala-ba-jinyung nirima=n g u-ni Class=coming root=nut Class-beach-Per-Rel we=ate=it 'We ate root nuts along here, on the coast' (166.24)

When the head noun is in a non-zero case category "we have a case-spreading rule by which the case suffix of the head noun may also show up on the 'Relative' noun/pronoun with the Relative suffix /-yinyung/ deleted to make room for it. This spreading (or agreement) rule is optional but rather common" (1984: 546). This spreading can be seen as an inner case deletion, since the suffix /-nyinyung/ is normally attached to the so-called presuffixal oblique, which may have a long final vowel. There are three possibilities, illustrated respectively by the following three examples: complete concord, headmarking, and marking on the pronoun. (53)

wu: = jarari-ny nga niga-wi-ruj a:-'nga-ruj they=moved and=then Masc3Sg-Obl-Loc Class-camp-Loc 'They went on until they reached his camp' (40.7)

(54)

nuru=buri- nuri-'-n y in y un g ana-lha:l-waj 1PlExcl=sat 1PlExcl=Rel Class-country-Per 'We [plural exclusive] were staying in our country' (113.1)

Suffixaufnahme

and Related Case Marking Patterns in Australian Languages 353

(55) wu: = ya-nggi aba ngaya-wi-wuy a-lha: l they=went then 1Sg-Obl-All Class-country 'They went then to my country' (41.5)

(The use of the "punctual" prefix a- on nouns of the ana-class such as 'camp' and 'country' may suggest the covert presence of a non-zero case suffix.) The choice of the noun-class prefix, with several forms available for each class, is extremely complicated (1984: 163). Some strong correlations between forms and case categories are apparent. The demonstrative allative adverb yu:-guni occurs in this example: (56))

nga ngaya-wi yu:-gu-ni Cape Barrow-wuy

and=then 1Sg-Obl to=there Cape Barrow-All 'and then to my [country] Cape Barrow' (47.21)

NYAWAYGI (Pama-Nyungan, Queensland; Dixon 1983): "The data for Nyawaygi do not contain any inflective genitives; but the corpus is too scanty for any significance to be attached to this omission" (p. 456). PANYJIMA (Pama-Nyungan, Western Australia; Dench 1991): "Unlike many Australian languages, Panyjima shows an especially high degree of multiple case-marking. Nominal suffixes are distributed to all constituents of a marked phrase or clause" (p. 125). (57)

nyinkutharntu-ku wangka-yu ngatha mirlimirli-la wantha-larta 2Sg=Gen-Acc language-Acc 1Sg=Nom paper-Loc put-Fut 'I'll put your language down on paper' (ex. 19)

(58) ngatha purlpi-yayi-ku kurrjarta-ku mantu-tharntu-ku 1Sg=Nom want-Inch-Pres spear-Acc meat-Gen-Acc 'I want a spear [to use] for meat' (ex. 46)

Local cases may also carry further case markers: (59)

ngaliya yana-nha wiya-larta ngunha-kutha-ku marlpa-kutha-ku wangka-ka-ku 1DuExcl go-Past see-Fut that-Du-Acc man-Du-Acc word-Loc-Acc 'We [dual exclusive] went to see those two men talking' (ex. 26)

PITJANTJATJARA (Pama-Nyungan, Central Australia; Glass and Hackett 1970) has final-marking. No data on Suffixaufnahme. PITTA-PITTA (Pama-Nyungan, Queensland; Blake 1979) is word-marking, and Suffixaufnahme is quite common (note that genitive is labeled Purp): (60)

atu kat y u-na wat y ama-ka kiniyari- a-na 1Sg= Erg clothes-Ace wash-Past girl-Purp-Acc 'I washed the girl's clothes; I washed the clothes for the girl' (ex. 13)

354 (61)

Australian tita- a-lu yaramana-lu tawi-kala anay white=man-Purp-Erg horse-Erg throw-Lest 1Sg=Acc 'The white man's horse might throw me' (ex. 102)

(62) Thana-nga-arri-inu karnta nganyu they-Purp-Far-All go=Fut 1Sg=FutSbj 'I am going to go to theirs [their camp]' (Blake 1991)

(FutSbj is a special case for subjects and instruments in the future.) Local cases may likewise undergo further inflection: (63)

wara-inya-na intu mari-ka kati-na who-Abl-Acc 2Sg=Erg get-Past meat-Acc 'Where did you get the meat from?' (Blake 1979, ex. 63)

(64) kalpurru-inya- u anyu kanta-li a Boulia-Abl-FutSbj 1Sg=FutSbj go-Purp 'I'm going to leave Boulia' (ex. 103)

Note that Pitta-Pitta has two special suffixes marking human nouns as being "possessed" by a 2nd or 3rd person, viz. -mala and -wara. RITHARNGU (Pama-Nyungan, Arnhem Land, Northern Territory; Heath 1980b): Multiple case marking is avoided by inner or outer case deletion. "A special kind of Case-Agreement transformation is applicable to possessive constructions when the possessed NP is in the Locative, Allative, Ablative, or Pergressive cases. In such constructions the possessor NP loses its usual Genitive suffix and assimilates to the case category of the possessed NP. Thus compare the usual possessive type (65)

nu- u dawal [2Sg-Gen country (Nom)] 'your country'

with the following sentence: (66) wa: n-i+nu+ra nu: -kala-li? dawal-li? go-Fut+now+1Sg 2Sg-Lig-All country-All ['I will now go to your country']

[Human and, less rigorously, "higher" animate nouns require a ligative -galabefore some local cases.] This rule applies only when the possessed NP is in one of the four cases listed above" (pp. 103f.). If the possessed NP is in another case (nominative, accusative, genitive/dative, instrumental), no multiple case marking is observed, which may formally be seen as outer case deletion:

Suffixaufnahme (67)

and Related Case Marking Patterns in Australian Languages 355

yaku-n -gu+na+ra la-na madalu gu-y this=Aug=Gen=3Sg=Acc=1Sg spear-Past hook=spear-Ins 'I speared him with this [man's] hook spear!' (p. 104)

UNGARINJIN (Wororan, Western Australia; Rumsey 1982): "Grammatical case relations are signalled only by cross-reference on the verb . . . and nongrammatical cases are signalled mainly by the nominal postpositions treated below" (p. 61). Rumsey sees postpositions as phrase- or clause-bound forms which characteristically occur on the last word of the phrase. In this sense Suffixaufnahme may occur: (68)

dambun ni-na ga-ra camp 1Sg-Gen-Loc 'at my camp' (p. 141)

URADHI (Pama-Nyungan, Cape York Peninsula, Queensland; Crowley 1983): "Case marking in the noun phrase is applied obligatorily to the head constituent and it optionally applies concordially to the bracketed constituents" (p. 371). Suffixaufnahme is quite normal (all examples from Atampaya dialect): (69)

ulu awutyi- u wutpu-:namu- u intya 3Sg=Nom house-Obl old=man-Gen-Obl live=Pres 'He lives in the old man's house' (ex. 128)

(70) ura atumu:-namu pana-:namu mina this (Abs) 1Sg=Gen-Gen (Abs) friend-Gen (Abs) meat (Abs) This is my friend's meat' (ex. 130) (71)

ula ani wutpu-:namu-ntu utyanpi-mpu aru-ny 3=NonSg (Nom) 1Sg=Acc old=man-Gen-Ins club-Ins hit-Past 'They hit me with the old man's club' (ex. 131)

WAALUBAL (Pama-Nyungan, New South Wales; Crowley 1978) is a Bandjalang dialect and seems to be word-marking. Suffixaufnahme is quite regular. (72))

(73)

mali-yu daba:y-dyu bunybi-ni dyina yalany-dyu nyula- a:-yu

that-A dog-A lick-PastDef foot (P) tongue-Ins 3SgMasc-Gen-Ins 'The dog licked his foot with his tongue' (ex. 106) anya:-yu ma:ma -gu buma-ni guruman 1Sg=Gen-A father-A kill-PastDef kangaroo (P) 'My father killed the kangaroo' (ex. 38)

"The locative case is unusual in Waalubal in that it can take the full range of case affixes after itself" (p. 67). The vowel -a is lengthened before further inflection:

356

Australian

(74) muru anya:b a n y a d y a - r igulu bay-dyu bawur-a:-yu nose (P) 1Sg=Gen (P) cover-PastDef cold-Ins head-Loc-Ins 'My nose is blocked up with a cold in the head' (ex. 97)

WALMATJARI (Pama-Nyungan, Western Australia; Hudson 1978) basically has complete concord and there are clear examples of Suffixaufnahme. ('Modal' is a modal root which has pa, ma, and ngu as its three allomorphs.) (75) kanya pa- -ja ngamaji-rlu ngaju-kura-rlu carried Modal-3Sg=S-1Sg=P mother-Erg 1Sg-Poss-Erg 'My mother carried me' (ex. 24) (76)

kunyarr-warnti-rlu pa-lu-nya karnanganyja-warnti- pajani malji-warntipurlka-warnti-rlu ngaju-kura-warnti-rlu dog-Pl-Erg Modal-3Pl=S-3Pl=P emu-Pl-Nom bit male-Pl-Nom big-Pl-Erg 1SgPoss-Pl-Erg 'My big dog bit [caught] the male emus' (ex. 26)

Double possession is expressed as follows: (77)

minyarti pa- -rla juku-wu Mayawany-kura-wu this Modal-3Sg=S-3Sg=Dat son-Dat Mayawany-Poss-Dat 'This belongs to Mayawany's son' (ex. 82)

"Five of the non-syntactic case markers can be further inflected by other case markers" (p. 35); for example: (78) jilji-karti-rlu ma- -nya wanyjani yapa-warnti- nganpayi-rlu sandhill-All-Erg Modal-3Sg=S-3Pl=P left child-Pl-Nom man-Erg 'He left the children there on his way to the sandhill' (ex. 140) (79)

mayaru-ngurni-rlu pa- -ja nyanya house-Abl-Erg Modal-3Sg=S-1Sg=P saw 'He saw me from the house' (ex. 146)

WANGKUMARA (GALALI) (Pama-Nyungan, Queensland; McDonald and Wurm 1979): "There are a number of sentences with possessor of possessor constructions, where both NPs are marked by the dative inflection. . . . One is always a pronoun. The dative of possession is never attested with further case-marking" (p. 51). WARGAMAY (Pama-Nyungan, Queensland; Dixon 1981): "All nominals in a noun phrase must agree in case inflection" (p. 28). (80)

ayba a: bungi inunda bu: guraydya 1Sg (S) Neg sleep 2Sg=Loc snore=Loc 'I couldn't sleep for your snoring' (ex. 5)

Suffixaufnahme (81)

and Related Case Marking Patterns in Australian Languages 357

inu gu bada gu anya gunydyay waga 2Sg=Gen=Erg dog=Erg 1Sg=P bite shin (Abs) 'Your dog bit my shin' (ex. 61)

"Note that pronominal genitives do decline, taking case inflections. . . . Although, despite several attempts, no examples have been obtained of nominal genitives declining, it seems very likely that they will do so" (p. 31). WARLPIRI (Pama-Nyungan, Northern Territory; Hale 1982): "The predicational use of spatial cases is common in Warlpiri. . . . Predication is signaled by case agreement, giving rise to a full range of 'double case marking' in which a spatial case is extended by one of the grammatical cases" (pp. 267f.). (82)

a. ngarrka-ngku ka yankirri luwa-rni ngapa-ngka-rlu man-Erg Pres emu shoot-NonPast water-Loc-Erg 'The man is shooting the emu at the waterhole' (ex. 87a) b. ngarrka-ngku ka-rlajinta yankirri-ki luwa-rni ngapa-ngka-ku man-Erg Pres-3Sg=Dat emu-Dat shoot-NonPast water-Loc-Dat 'The man is shooting at the emu at the waterhole' (ex. 87b)

But other semantic cases may also be further marked: (83)

Jakamarra-rlu ka-ngalpa jakuru-pi-nyi ya-ninja-kurra-rlu Jakamarra-Erg Pres-1PlIncl Preverb-leave-NonPast go-Inf-Purp-Erg 'Jakamarra is taking leave of us [plural inclusive] with the intention of going' (ex. 131a)

Hale (1982) is especially valuable for examples of the adnominal use of locational cases. According to Dench and Evans (1988: 13) Warlpiri should allow Suffixaufnahme, but I have not been able to detect examples in Hale (1982) or other sources consulted. WARNDARANG (Maran, Arnhem Land, Northern Territory; Heath 1980c): "The basic genitive construction consists of the possessor NP in Nominative case, juxtaposed to the possessed NP which is followed by a pronominal possessive postposition" (p. 27). This is illustrated by the following example (where Wu and Ra symbolize class prefixes): (84)

wu-yilba- a-gi ra-wadabir Wu-hole-Nom its Ra-goanna 'the goanna's hole'

There is no evidence of Suffixaufnahme. WATJARRI (Pama-Nyungan, Western Australia; Douglas 1981) seems to be final-marking. Since the possessor, taking the genitive (used for alienable possession), precedes its head, no Suffixaufnahme occurs.

358

Australian

YANKUNYTJATJARA (Pama-Nyungan, South Australia; Goddard 1983) is a final-marking language. "A genitive noun precedes its head. . . . A genitive pronoun sometimes precedes . . . but usually follows the head, so that it may carry an additional case-marker on behalf of the NP as a whole" (p. 75). (85)

ngana-nya puta yana-nyi iwara ngayu-ku-ngka who-Nom what=do=you=say go-Pres road 1Sg-Gen-Loc 'Who's that walking on my road?' (ex. 3-83)

Suffixaufnahme is therefore possible only for pronouns. YIDINY (Pama-Nyungan, Queensland; Dixon 1977): " 'Genitive' is essentially a derivational affix in Yidiny (as in many Australian languages), forming a stem that functions as an adjective and takes the full set of case inflections; a genitive noun or pronoun must agree in case with the 'head noun' which it qualifies" (p. 134): (86)

gudaga gu wagalni gu anyany badya:l dog=Erg wife=Gen=Erg 1Sg=P bite=Past '[My] wife's dog bit me' (ex. 49)

However, Suffixaufnahme is not obligatory if the outer case is ergative: (87)

anyany wagudyani: gudaga gu badya:l 1Sg=P man=Gen dog=Erg bite=Past 'The man's dog bit me' (ex. 51)

YINDJIBARNDI (Pama-Nyungan, Western Australia; Wordick 1982): No Suffixaufnahme is found, but use is made of double case marking in Tcomplementizer function (in the sense of Dench and Evans 1988). YIR YORONT (Pama-Nyungan, Cape York Peninsula, Queensland; Alpher 1973) typically has final-marking, but complete concord is also observed, especially when the NP is split. No data on Suffixaufnahme. YUKULTA (Tangkic, Northern Territory/Queensland; Keen 1983): "Adjectives take case inflections in agreement with that of the noun head. The category of 'adjective' includes possessive, demonstrative, interrogative and locative, though all except the latter can function as head of a noun phrase as well as adjectivally" (p. 199). (Note that in Yukulta ergative formally coincides with locative.) (88)

tatin-ta miyal-ta kila-ka umpan-pakaran-tyi markatu-karan-tyi ampir-i that (Abs) spear (Abs) get-Imp 2Sg-Gen-Erg father's=sister-Gen-Erg humpyErg 'Get that spear from (at) your aunty's humpy!' (ex. 55)

(89)

nan-ma la:lu-mpan-ta 1Sg-Stative south-Gen-Abs 'I'm a southerner' (ex. 49)

Suffixaufnahme

and Related Case Marking Patterns in Australian Languages 359

4. Results and Conclusions Fifty languages have been included in this survey. In roughly 50 percent of the languages in the sample Suffixaufnahme is attested, though sometimes only with pronouns. Often the information available does not suffice to determine whether Suffixaufnahme is obligatory or optional. Examples of peripheral local cases which allow further case marking were found in only a few languages. Here is a breakdown of our results. (A) Suffixaufnahme, in the traditional sense of the term, is attested for nouns and pronouns in Djabugay, Djapu, Dyirbal, Gumbaynggir, Guugu Yimidhirr, Kalkatungu, Ngiyambaa, Nunggubuyu (non-Pama-Nyungan), Panyjima, Pitta Pitta, Uradhi, Waalubal, Walmatjari, Warlpiri (according to Dench and Evans 1988), Yidiny, and Yukulta (non-Pama-Nyungan). (B) Double case marking due to final-marking is attested for nouns and pronouns in Alyawarra. (C) Suffixaufnahme is attested for nouns in Anguthimri (although there is no evidence that it is excluded for pronouns). (D) Suffixaufnahme is attested only for pronouns in Djinang/Djinba, Mara (non-Pama-Nyungan), and Wargamay. (E) Double case marking due to final-marking is attested only for pronouns in Ungarinjin (non-Pama-Nyungan). (F) Suffixaufnahme is attested only for pronouns and seems to be avoided for nouns in Margany/Gunya. (G) Double case marking due to final-marking is attested only for pronouns and seems to be avoided for nouns in Diyari, Kuku-Thaypan, and Yankunytjatjara. (H) No Suffixaufnahme is found in Baagandji, Bidyara/Gungabula, Dharawal/Dhurga, Djingili (non-Pama-Nyungan), Gidabal, Gooniyandi, Iwaidja, Malak Malak, Mangarayi, Maung (the last five non-Pama-Nyungan), Muruwari, Ngadjumaja, Ngalakan, Ngandi, Ngankikurungkurr (the last three non-Pama-Nyungan), Ngawun, Nyawaygi (but remember Dixon's remarks), Pitjantjatjara, Ritharngu, Wa kumara, Warndarang (non-Pama-Nyungan), Watjarri, Yindjibarndi, and Yir Yoront. (I) Case marking after an adnominal local case is attested for Pitta Pitta, Panyjima, Waalubal, Walmatjari, and Warlpiri. (J) The data for Gooniyandi are not entirely clear. The collective Australian evidence thus suggests two potentially universal generalizations: (a) If a language has Suffixaufnahme for nouns, it also has it for pronouns. (b) If a language has peripheral local cases that allow further case marking by virtue of agreement, it also has Suffixaufnahme with possessors. References Alpher, Barry J. 1973. Son of Ergative: The Yir Yoront Language of Northeast Australia. Ph.D. dissertation, Cornell University.

360

Australian

Austin, Peter. 1981. A Grammar of Diyari, South Australia. Cambridge: Cambridge University Press. Birk, D. B. W. 1976. The Malak Malak Language, Daly River (Western Arnhem Land). Canberra: Pacific Linguistics B-45. Blake, Barry J. 1969. The Kalkatungu Language: A Brief Description. Canberra: Australian Institute of Aboriginal Studies. Blake, Barry J. 1977. Case Marking in Australian Languages. Canberra: Australian Institute of Aboriginal Studies. Blake, Barry J. 1979. Pitta-Pitta. In Dixon and Blake 1979-1991, vol. 1, 183-242. Blake, Barry J. 1979. A Kalkatungu Grammar. Canberra: Pacific Linguistics B-57. Blake, Barry J. 1987. Australian Aboriginal Grammar. London: Croom Helm. Blake, Barry J. 1988. Redefining Pama-Nyungan. Aboriginal Linguistics 1, 1-90. Blake, Barry J. 1990. Languages of the Queensland/Northern Territory border: Updating the classification. In Language and History: Essays in Honour of Luise A. Hercus, ed. by Peter Austin, R. M. W. Dixon, and Isobel White, 49-66. Canberra: Pacific Linguistics C-116. Blake, Barry J. 1991. Case after case. Manuscript, La Trobe University. Brandenstein, C. G. von. 1980. Ngadjumaja. Innsbruck: Institut fur Sprachwissenschaft, Universitat Innsbruck. Breen, J. G. 1973. Bidyara and Gungabula: Grammar and Vocabulary. Melbourne: Monash University (Linguistic Communications 8). Breen, J. G. 1981a. Margany and Gunya. In Dixon and Blake 1979-1991, vol. 2, 275-393. Breen, Gavan. 1981b. The Mayi Languages of the Queensland Gulf Country. Canberra: Australian Institute of Aboriginal Studies. Capell, Arthur, and H. E. Hinch. 1970. Maung Grammar. The Hague and Paris: Mouton. Chadwick, Neil. 1975. A Descriptive Study of the Djingili Language. Canberra: Australian Institute of Aboriginal Studies. Crowley, Terry. 1978. The Middle Clarence Dialects of Bandjalang. Canberra: Australian Institute of Aboriginal Studies. Crowley, Terry. 1981. The Mpakwithi dialect of Anguthimri. In Dixon and Blake 1979-1991, vol. 2, 147-194. Crowley, Terry. 1983. Uradhi. In Dixon and Blake 1979-1991, vol. 3, 307-428. Dench, Alan. 1991. Panyjima. In Dixon and Blake 1979-1991, vol. 4, 125-243. Dench, Alan, and Nicholas Evans. 1988. Multiple case-marking in Australian languages. Australian Journal of Linguistics 8, 1-47. Dixon, R. M. W. 1972. The Dyirbal Language of North Queensland. Cambridge: Cambridge University Press. Dixon, R. M. W., ed. 1976. Grammatical Categories in Australian Languages. Canberra: Australian Institute of Aboriginal Studies; New Jersey: Humanities Press. Dixon, R. M. W. 1977. A Grammar of Yidi . Cambridge: Cambridge University Press. Dixon, R. M. W. 1980. The Languages of Australia. Cambridge: Cambridge University Press. Dixon, R. M. W. 1981. Wargamay. In Dixon and Blake 1979-1991, vol. 2, 1-144. Dixon, R. M. W. 1983. Nyawaygi. In Dixon and Blake 1979-1991, vol. 3, 431-525. Dixon, R. M. W., and Barry J. Blake, eds. 1979-1991. Handbook of Australian Languages. 4 vols. Canberra: Australian National University Press; Oxford University Press (vol. 4).

Suffixaufnahme

and Related Case Marking Patterns in Australian Languages 361

Donaldson, Tamsin. 1980. Ngiyambaa, the Language of the Wangaaybuwan. Cambridge: Cambridge University Press. Douglas, W. H. 1981. Watjarri. In Dixon and Blake 1979-1991, vol. 2, 197-272. Eades, Diana K. 1976. The Dharawal and Dhurga Languages of the New South Wales South Coast. Canberra: Australian Institute of Aboriginal Studies. Eades, Diana. 1979. Gumbaynggir. In Dixon and Blake 1979-1991, vol. 1, 245-361. Geytenbeek, B., and H. Geytenbeek. 1971. Gidabal Grammar and Dictionary. Canberra: Australian Institute of Aboriginal Studies. Glass, A., and D. Hackett. 1970. Pitjantjatjara Grammar. Canberra: Australian Institute of Aboriginal Studies. Goddard, Cliff. 1983. A Grammar of Yankunytjatjara. Alice Springs: IAD. Hale, Kenneth. 1982. Some essential features of Warlpiri verbal clauses. Work Papers of the Summer Institute of Linguistics, AAB Series A, vol. 6, 217-315. Haviland, John. 1979. Guugu Yimidhirr. In Dixon and Blake 1979-1991, vol. 1, 27-180. Heath, Jeffrey. 1978. Ngandi Grammar, Texts, and Dictionary. Canberra: Australian Institute of Aboriginal Studies. Heath, Jeffrey. 1980a. Nunggubuyu Myths and Ethnographic Texts. Canberra: Australian Institute of Aboriginal Studies. Heath, Jeffrey. 1980b. Basic Materials in Ritharngu: Grammar, Texts, and Dictionary. Canberra: Pacific Linguistics B-62. Heath, Jeffrey. 1980c. Basic Materials in Warndarang: Grammar, Texts, and Dictionary. Canberra: Pacific Linguistics B-72. Heath, Jeffrey. 1981. Basic Materials in Mara: Grammar, Texts, and Dictionary. Canberra: Pacific Linguistics C-60. Heath, Jeffrey. 1984. Functional Grammar of Nunggubuyu. Canberra: Australian Institute of Aboriginal Studies. Hercus, Luise. 1982. The Bagandji Language. Canberra: Pacific Linguistics B-67. Hoddinott, W. G., and Frances M. Kofod. 1988. The Ngankikurungkurr Language (Daly River Area, Northern Territory). Canberra: Pacific Linguistics D-77. Hudson, Joyce. 1978. The Core of Walmatjari Grammar. Canberra: Australian Institute of Aboriginal Studies; New Jersey: Humanities Press. Keen, S. 1983. Yukulta. In Dixon and Blake 1979-1991, vol. 3, 191-304. Libert, A. R. 1988. Going from the allative toward a theory of multiple case marking. McGill Working Papers in Linguistics 5; 1, 93-129. McDonald, M., and Stephen A. Wurm. 1979. Basic Materials in Wa kumara (Galali): Grammar, Sentences and Vocabulary. Canberra: Pacific Linguistics B-65. McGregor, William B. 1990. A Functional Grammar of Gooniyandi. Amsterdam and Philadelphia: Benjamins. Merlan, Francesca. 1982. Mangarayi. (Lingua Descriptive Series 4.) Amsterdam: North-Holland. Merlan, Francesca. 1983. Ngalakan Grammar, Texts and Vocabulary. Canberra: Pacific Linguistics B-89. Morphy, Frances. 1983. Djapu, a Yolngu dialect. In Dixon and Blake 1979-1991, vol. 3, 1-188. Oates, Lynette F. 1988. The Muruwari Language. Canberra: Pacific Linguistics C-108. Patz, Elizabeth. 1991. Djabugay. In Dixon and Blake 1979-1991, vol. 4, 245-347. Pym, N. (with B. Larrimore). 1979. Papers on Iwaidja Phonology and Grammar. Darwin: Summer Institute of Linguistics. Rigsby, Bruce. 1979. Possession in Kuku-Thaypan. In Languages of Cape York, ed. by Peter Sutton, 260-268. Canberra: Australian Institute of Aboriginal Studies

362

Australian

Rumsey, Alan. 1982. An Intra-sentence Grammar of Ungarinjin, North-Western Australia. Canberra: Pacific Linguistics B-86. Waters, Bruce E. 1989. Djinang and Djinba: A Grammatical and Historical Perspective. Canberra: Pacific Linguistics C-114. Wordick, F. J. F. 1982. The Yindjibarndi Language. Canberra: Pacific Linguistics C-71. Wurm, Stephen A. 1972. Languages of Australia and Tasmania. The Hague: Mouton. Yallop, C. 1977. Alyawarra: An Aboriginal Language of Central Australia. Canberra: Australian Institute of Aboriginal Studies.

13 Double Case Marking in Kanyara and Mantharta Languages, Western Australia Peter Austin 1. Introduction It has been known for some time that Australian Aboriginal languages have what has been termed "Suffixaufnahme" (Finck 1910, Plank 1990)—i.e., agreement of case-marked attributive nominals with their head by a second layer of case affixation1—but it was not until the important work of Dench and Evans (1988) that a solid descriptive foundation was established for discussing the phenomenon. Dench and Evans catalogue the full extent of Suffixaufnahme in Australia, and the ways it is formally expressed. A recent survey covering the whole of the Australian continent is Schweiger's chapter in this volume. In this chapter I present a description of double case marking in two groups of Western Australian languages, taking as a starting point the parameters established by Dench and Evans. I will show that a further level of double case needs to be recognized: a "derivational" level where case marking takes place when there are affixal dependencies, with certain case forms serving as "founding forms" for other cases. In some instances there is diachronic evidence for this level, suggesting that double case marking has an established history in the language groups. 2. Background The region between the Gascoyne and Ashburton Rivers in the northwestern part of Western Australia, north and inland from the coastal town of Carnarvon, was traditionally occupied by speakers of two groups of Aboriginal languages: the Kanyara and Mantharta groups. The Kanyara languages, spoken along the coast, comprise Payungu, Thalanyji, and Purduna; they show a high degree of structural similarity and have 60 to 70 percent common vocabulary. Their relationships and reconstructed ancestor have been described in Austin (1981a, 1988a). The Mantharta languages were spoken in363

364

Australian

land of the Kanyara group and comprise Jiwarli, Thiin, Warriyangka, and Tharrkari. They are even closer to one another than the Kanyara languages and share approximately 80 percent common vocabulary.2 In this paper, most examples will be drawn from Thalanyji and Jiwarli as representatives of each group, but in all relevant respects, claims I make about them can be supported by data from the other five languages.3 The more distant relatives of the Kanyara and Mantharta languages include the well-known Warlpiri and Pitjantjatjara (O'Grady et al. 1966, Wurm 1972). Structurally, Thalanyji and Jiwarli show a number of morphosyntactic characteristics that are typical of languages in the immediate area, including the five following. First, there is highly productive suffixal morphology. It is of the agglutinative kind, and sequences of easily segmentable morphemes up to four or five per word are common. Second, there is a rich case marking system with eight overtly distinguished cases. It is of the split-ergative type, with syncretism of case forms that is partly ergative-absolutive, partly nominative-accusative, and partly a mixture of both. Case is marked by suffix or by suppletion (for some pronouns). All elements of what might be thought of as "noun-phrase constituents" receive case; that is, these are word-marking (Blake 1987) or completeconcord languages (Dench and Evans 1988: 4). Compare Warlpiri, where case is coded on the last of a sequence of adjacent elements.4 Third, there are no demonstrable phrasal categories (there are no secondposition auxiliaries as found in Warlpiri, for example), and syntactic structure is "flat." To the extent that the term is well-defined, Jiwarli seems to be a prototypical "non-configurational" language.5 Word order is unusually free, especially in texts. It is even freer than in Warlpiri (as described by Hale, Nash, and Simpson) because for some clause types, elements from different clauses may be freely mixed within a single sentence (Austin 1987, 1988c). Fourth, there are no voice mechanisms of the familiar sort. There is no passive or antipassive, nor is there a morphologically marked reflexive. Fifth, agreement other than case agreement (such as agreement for number or gender) is absent, and there are no cross-referencing bound pronominals either (unlike in Warlpiri, for example). There are two major word classes and three minor ones in Kanyara and Mantharta languages (Austin 1989). The major word classes are nominals and verbs. Nominal words can be marked for number and case but not for tense/mood. This category includes substantives (nouns and adjectives, not morphologically or syntactically distinguishable), names, pronouns, demonstratives, and locationals (cardinal directionals). Verbs can be marked for tense/mood but not for number; some dependent verb inflections may be followed by case suffixes, but verb stems may not take cases directly. We can distinguish between main verbs that code tense/mood, and dependent verbs that cannot carry tense/mood and take a series of inflections coding dependency and coreference type (through a switch-reference system).

365

Double Case Marking in Kanyara and Mantharta Languages

The minor word classes include adverbs, which modify verb predicates; particles, which are generally uninflected and have semantic scope over the whole sentence (examples from Jiwarli are warri 'not', pampa 'cannot', and kaji 'try'); and interjections, which form separate utterances and show no morphology. 3. Morphological Case There are eight morphologically coded case forms distinguished for most nominals. The various inflection paradigms of Jiwarli nominals are set out in Table 13.1. The paradigms for Thalanyji and the other languages are similar, except that only Jiwarli makes the distinction between two ablative cases (see below). Cases may be divided into two groups: grammatical, those borne by nominals subcategorized by the predicate (and their modifiers); and nongrammatical, also called "local." The grammatical cases are: (a)

Ergative—codes transitive subject function (abbreviated, following Dixon 1979, as A) and instrument. (b) Absolutive—codes intransitive subject (S) function. This is the unmarked root.6 Note that the 1st person pronoun alone has no separate ergative form distinguished from the absolutive.7 (c) Accusative—codes transitive object (P) function. This is formally marked by suffix for all nominals with animate reference; inanimates collapse absolutive and accusative forms. In Thalanyji all nominals are overtly marked for accusative, regardless of animacy. Table 13.1. Jiwarli Morphological Cases. 'boy' Ergative Absolutive Accusative Dative Locative Allative Ablative 1 Ablative2

wirtangku wirta wirtanha wirtawu wirtangka wirtarla wirtangkanguru wirtaparnti 'man'

Ergative Absolutive Accusative Dative Locative Allative Ablative 1 Ablative2

manthartalu mantharta manthartanha manthartawu manthartala manthartarla manthartalanguru manthartaparnti

'tree'

'hill kangaroo'

wurungku wuru wuru wuruwu wurungka wururla wurungkanguru wuruparnti

mathantu mathanma mathannha mathanku mathanta mathankurla mathantanguru mathanparnti

'you'

'I'

nhurralu nhurra nhurranha nhurrampa nhurrala nhurrarla nhurralanguru nhurraparnti

ngatha ngatha ngathanha nganaju ngathala ngatharla ngathalanguru ngathaparnti

366

Australian

(d) Dative—codes predicate complement, purposive, benefactive, and (alienable) possessor. (Inalienable possession is coded syntactically by apposition of the possessor and possessed. Depending on animacy, possessor and possessed may have the same case form; see below.) The 1st person singular has a suppletive dative case form in all Mantharta languages and in Thalanyji.

The non-grammatical cases are: (a) (b) (c) (d)

Locative—codes location in place or time. Allative—marks direction towards a place. Ablative 1—marks direction from a location in place or time. Ablative2—marks direction from a location (synonymous with Ablative1), temporal sequence ('after'), and cause. In all the languages apart from Jiwarli there is just one ablative covering all these functions.

There are two predicate types in Kanyara and Mantharta languages, corresponding to the two major categories. Nominal predicates occur in construction with a (nominal) subject inflected in the same case as the subject of an intransitive verb. There are two types of nominal predicates: simple nominals, which occur with just a subject; and extended nominals, which require both a subject and a dative casemarked nominal complement. These (and the corresponding extended intransitive verb predicates described below) have been referred to as "middle constructions" in the Australianist literature; but this term is easily confused with the middle construction of Indo-European languages (which is a particular voice form), and hence I prefer the term "extended." Verbal predicates fall into four classes according to their subcategorization frames: (a)

Intransitive verb (VIntrans)—takes a single subject argument inflected for absolutive case. (b) Extended intransitive verb (VExt)—takes a subject argument inflected like a subject of an intransitive verb and also a second argument inflected for dative case.8 (c) Transitive verb (VTrans)—takes two arguments, a transitive subject inflected for ergative case, and a transitive object inflected for accusative case. (d) Ditransitive verb (VDitrans)—takes three arguments, one inflected like a transitive subject and two inflected like transitive objects. The two transitive object-like nominals can, however, be differentiated syntactically (see Austin 1989).

The following Jiwarli examples illustrate the various predicate types and the case coding, presented synoptically in Table 13.2. The first example shows a simple nominal used as a predicate with a demonstrative as subject; (2) is an example of an extended nominal predicate; intransitive verbs are illustrated by (3); and example (4) illustrates an extended intransitive with a dative casemarked complement.9

Double Case Marking in Kanyara and Mantharta Languages

367

Table 13.2. Jiwarli Predicate Types. i. Nominals ii. Verbs

(a) (b) (a) (b) (c) (d)

plain extended VIntrans VExt VTrans VDitrans

S S S S A A

Dat Dat P P P

e.g. wirta 'boy' e.g. jirril 'afraid' e.g. puni- 'to go' e.g. yarrukarri- 'to want' e.g. nhanya- 'to see' e.g. wantha- 'to give'

(1) mantharta ngunha-pa man that(Abs)-Spec 'He is a man' [T53sl6] (2) yakara paju ngunha-pa pirrpilyangkura-wu brave truly that(Abs)-Spec rock-python-Dat 'He was truly brave about rock pythons' [T50s1] (3) ngatha parlirri-nyja-rni warlpara-nguru I (Abs) return-Past-hence south-Abl1 'I came back from the south' [T31s2] (4) ngurru-nyjarri mangkapurturri-nyja pirru-wu old=man-Pl (Abs) be=glad-Past meat-Dat 'The old men were glad for the meat' [T14s7]

Example (5) shows a transitive clause; note the three instances of the accusative case on the P nominals: (5) ngatha mana-nyja yanyja-nha mantharta-nha ngurtirti-nha I=Erg get-Past another-Acc man-Acc [name]-Acc 'I got another man Ngurtirti' [T31s7]

Transitive clauses are also illustrated by (6); notable features here are the word order (the possessor and possessed nominals are not adjacent) and inalienable possession with 1st person singular accusative. (6) juru-ngku ngatha-nha kulypa-jipa-rninyja parna sun-Erg I-Acc be=sore-Caus-Past head (Acc) 'The sun made my head sore' [T19s3]

Notice that morphological case is coded "locally" (Silverstein 1976) on nominals; the body part P parna is treated as inanimate and is in absolutive case form, but the possessor ngatha is treated as animate and bears an accusative case suffix. 10 Compare this example with (8) below, which shows alienable possession. Finally, example (7) illustrates a ditransitive verb.

368

Australian

(7) ngatha wantha-rninyja-rru pirru kamu I=Erg give-Past-now meat (Acc) hunger (Acc) 'I gave the meat to the hungry one' [ADNlpl8s7]

A Thalanyji example showing both non-subject arguments in the accusative case is: (8) ngulu murla-nha wantha-rrkin wartirra-nha that=Erg meat-Acc give-Pres woman-Acc 'He gives meat to the woman' [HHPAN4p44sl]

In both Kanyara and Mantharta languages there are certain non-finite dependent clauses that also take case markers following the dependent verb inflection. It is not possible to describe the full system here (see also Austin 1992), but I will provide some examples of imperfective dependent clauses, which function like NP-relative clauses or adverbial temporal/logical clauses. In all languages the subject of such a clause is missing, and the verb bears an affix marking switch-reference, i.e. whether the (missing) subject is to be understood as having the same reference (SS, same subject) or different reference (DS, different subject) from the subject of the clause on which it is dependent. If it is a different-subject clause (ImperfDS) then the verb bears a case marker appropriate to the grammatical function of the matrix clause non-subject. The following are two examples from Jiwarli, showing accusative and locative cases respectively on the dependent clause verb, and two from Thalanyji, showing locative and dative cases respectively: (9) payalpa-nthu-rru ngatha nhanya-nyja wirntu ngurnta-iniya-nha at=last-again-now I=Erg see-Past dead (Abs) lie-ImperfDS-Acc 'At last I saw (him) lying dead' [T14s4] (10) wuru ngunha tharrpa-rninyja ngarti-ngka kajalpu-la ngarri-ngka ngurnta-iniyala stick (Acc) that=Acc insert-Past inside-Loc emu-Loc ashes-Loc lie-ImperfDSLoc 'He inserted the stick inside the emu lying in the ashes' [T40s9] (11)

ngatha marrkarri-n nyinta-ma parnakarri-yitha-ku I (Abs) wait=for-Pres you-Dat come-ImperfDS-Dat 'I wait for you to come' [HHPAN4p46s8]

(12)

ngatha wangka-yin ngunhi-la kanyara-la nhikirntarri-yitha-la I (Abs) speak-Pres there-Loc man-Loc dance-ImperfDS-Loc 'I am talking to the man there who is dancing' [HHPAN4p47s8]

Double Case Marking in Kanyara and Mantharta Languages

369

4. Double Case Marking 4.1. Subtypes In the examples above, nominals serving as arguments of predicates are inflected for case according to their grammatical function. There are also instances in these languages of nominals carrying two case affixes. We can identify three situations where this occurs: (i)

Derivational—this is the addition of a case marker before a nominal derivation suffix or a further case marker is attached. The locative and dative cases participate in this.

(ii)

Adnominal—dative case-marked nominals serving as adnominal (genitive) modifiers take a second case affix in agreement with the modified head nominal (with some exceptions to be considered below). This is the prototypical instance of double case as Suffixaufnahme.

(iii)

Referential—nominals serving as adverbial modifiers or secondary predicates can be assigned (subcategorized) ergative, accusative or dative case following their allative or ablative suffixes.

We will exemplify each of these patterns in turn.

4.2. Derivational Double Case In Kanyara and Mantharta languages case markers are sometimes used not to code predicate-argument or nominal-nominal syntactic relations, but in a derivational way. In all the languages, certain affixes (both case and non-case) require that the nominal to which they are attached is already inflected for case before affixation takes place; additional grammatical and non-grammatical cases may then be added after this affix. Locative and dative cases participate in this phenomenon, serving as base forms. There is some evidence that this process has affected pronouns historically. 4.2.1. Case-marked Bases The locative case form in Jiwarli serves as the basis (the "founding form") for the ablative1 affix -nguru; i.e., it is appended to inflected locative nominals, as in (13). (13)

a. jurla-ngka-nguru tree-Loc-Abl1 'from the tree'

370

Australian b. purrarti-la-nguru woman-Loc-Abl1 'from the woman' c. ngatha-la-nguru I-Loc-Abl1 'from me' d. ngula-nguru that=Loc-Abl1 'from there'

The dative serves as a founding form for demonstratives in Kanyara and Mantharta languages: all demonstratives must be inflected for dative case before any non-case affixes (such as number marking, or any of a group of derivational affixes) are added. Thus, consider the following examples of the distal demonstrative ngunha from Thalanyji and Jiwarli: (14)

that (Abs) that=Erg that=Dat that=Loc that-Pi that-Ass that-side

ngunha ngulu ngurnu ngula ngurnu-nyjarri (in Jiwarli ngurnu-malu) ngurnu-nyungu ngurnu-kuji

When the suffixed demonstrative is used in a sentence, a further case affix may follow, as in these sentences from Jiwarli and Thalanyji respectively: (15)

ngatha nhanya-nyja kumpa-iniya ngurnu-malu-la mantharta-la I=Erg see-Past sit-ImperfDS that=Dat-Pl-Loc man-Loc 'I saw (him) sitting with those men' [N12p37]

(16)

ngurnu-kuji-la tharrpa-yin that=Dat-side-Loc enter-Pres '(He) goes in that side' [BHOGM47s45]

This requirement also applies to ablative case forms of demonstratives; the suffix parnti must be added to the dative-inflected demonstrative, as in Jiwarli: (17)

kajalpu ngunha-pa yurnu-parnti ngurra-parnti kanya-nyja emu (Abs) that-Spec this=Dat-Abl2 place-Abl2 carry-Past 'The emu was carried from this place' [T45s39]

In the Kanyara languages all pronoun ablative case forms are built upon the dative. That is, any ablative pronoun will carry two case affixes, the derivational dative and the ablative, as in the Thalanyji example (18).

Double Case Marking in Kanyara and Mantharta Languages

371

(18) ngunha yarrkarni-nha pirungka ngali-ya-ma-parnti that (Abs) run-Past fear we=Du-Excl-Dat-Abl 'He ran away from us two in fear' [HHPAN4p111s6] 4.2.2. Derivational Double Case: Diachronic Evidence The 3rd person singular pronoun paradigm in Jiwarli shows evidence of several historical layers of case marking. Consider the relevant forms: (19) Nominative Ergative Accusative Locative Dative

panhalu panhaluru panhalunha panhalura parnumpa

It is clear that the non-dative root panhalu is an old ergative form containing the ergative suffix -lu (widespread in Pama-Nyungan languages). A 3rd person pronoun panha is found in other Western Australian languages, including Thalanyji. Similarly, the dative parnumpa consists of parnu, an old dative form, plus the pronominal dative case affix -mpa. Interestingly, when this pronoun functions adnominally (see below) as a possessor, it can be further suffixed with an argument dative case, this time of the form -wu, as in: (20) ngatha jirrilarri-a thuthu-wu parnumpa-wu I (Abs) be=afraid-Pres dog-Dat he=Dat-Dat 'I am afraid of his dog' Here there are, diachronically, three layers of dative case forms: parnu, mpa, and wu!

4.3. Adnominal Double Case In addition to serving as arguments of predicates, nominals may also function adnominally, modifying the meaning of another nominal. The dative case may be followed by an additional case marker when it codes a genitive relationship of inalienable possession. The possessor takes dative case plus the case appropriate to the syntactic role filled by the possessed nominal. It is important to realize that the case form of the possessed nominal is not copied to the possessor nominal to follow the dative; rather, case is formally spelled out on the possessor independently. For instance, if the possessor is an animate nominal and the possessed is inanimate, then in Jiwarli the possessor will take an accusative case suffix to mark P function, even though the possessed will take no suffix and hence will be formally identical to the absolutive case. Examples (21) from Jiwarli and (22) from Thalanyji show dative plus accusative.

372

Australian

(21) warn nganaju-nha ngurra panyi-ma not I=Dat-Acc camp (Acc) disturb-Imp 'Don't disturb my camp!' [N5p74s3] (22)

kupuju-lu kaparla-nha yanga-lkin wartirra-ku-nha child-Erg dog-Acc chase-Pres woman-Dat-Acc 'The child chases the woman's dog' [N12p3]

The case marking patterns described here are those which apply in simple main clauses in Kanyara and Mantharta languages. These languages also have sets of dependent clauses marked by non-finite verb suffixes. Depending on the clause type (see Austin 1992), regular marking of transitive object as accusative or absolutive is suspended and dative or allative case is assigned. Essentially, imperfective and perfective modifying clauses (the functional equivalent of English relative clauses and adverbial clauses) assign dative to their P, while purpose same-subject clauses assign allative to their P; purpose different-subject clauses take main-clause case marking. Dench and Evans (1988) have called this the "associating function" of case. Consider the following Thalanyji example where murla 'meat', the transitive object of kuthuwa'to cook', is marked with an associating dative case: (23)

kupuju-lu kaparla-nha yanga-lkin wartirra-ku-nha nyina-yitha-ku-nha kuthuwa-lkarra murla-ku child-Erg dog-Acc chase-Pres woman-Dat-Acc sit-ImperfDS-Dat-Acc cookImperfSS meat-Dat 'The child is chasing the dog of the woman who is sitting down cooking meat' [HHPANllp3s8]

Now, in Kanyara and Mantharta languages any adnominal modifying nominal that is semantically connected to the P will take dative or allative case after its own (adnominal) dative case. This is not different in principle from main clauses but is included here for sake of completeness. Examples from Thalanyji are: (24)

mara-lkarra nyina-yin kupuju kaparla-ku wartirra-ku-ku hold-ImperfSS sit-Pres child (Abs) dog-Dat woman-Dat-Dat 'The child is sitting holding the woman's dog' [HHPANllp3s4]

(25) ngatha puni-ra-ya thaka-ru kaparla-rla jurti-rla I (Abs) go-Hort-Emph get-PurpSS dog-All I=Dat-All 'I'm going to get my dog' [MROGN7p84bs3]

Note that adnominal dative case may be followed by any other case. That is, adnominal datives may modify any nominal in (subcategorized) argument or non-argument function. There is no ban on identical case suffixation, as the following Thalanyji example shows (see also (24) above):

Double Case Marking in Kanyara and Mantharta Languages (26)

373

kupuju pirungkarri-n kaparla-ku wartirra-ku-ku nyuja-ngka wangka-yitha-kuku child (Abs) be=afraid-Pres dog-Dat woman-Dat-Dat white=man-Loc talkImperfDS-Dat-Dat 'The child is afraid of the dog of the woman talking to the white man'

However, in Jiwarli double datives occur only when the case morphemes have different phonological shapes, thus apparently being subject to haplology. The dative case added to nouns has the form -ku after consonants, -yi after i, and -wu after a and u. Phonetically, iyi is realized as long [i:] and uwu as long [u:]. Addition of a further -yi or -wu after these would produce an extra-long vowel and hence is excluded. Thus, constrast the following: (27)

juma ngunha jirrilarri-a thuthu-wu yirnu nganaju-wu purrarti-yi (*purrarti-yi-

yi)

child (Abs) that (Abs) be=afraid-Pres dog-Dat this=Dat I=Dat-Dat womanDat (*woman-Dat-Dat) 'The child is afraid of my woman's dog' [N13p55s2] (28)

juma jirrilarri-a thuthu-wu nganaju-wu yakan-ku-wu child (Abs) be=afraid-Pres dog-Dat I=Dat-Dat spouse-Dat-Dat 'The child is afraid of my wife's dog' [N13p55s3]

Interestingly, in Tharrkari, it seems that the dative form of pronouns (e.g., nganayi I=Dat) must be followed by a ligature suffix -dhi- before a further case affix is added (see Dench and Evans 1988 for other examples of this type), as in: (29)

ngadha nyirta kupa-inha nganayi-dhi-da maya-ka that (Abs) here sit-Pres I=Dat-Lig-Loc house-Loc 'I am sitting here in my house' [CYTKBp23s4]

Adnominal genitive case agreement is limited to a maximum of two, dative plus head case. The genitive of a genitive does not bear two dative cases followed by the head nominal's case. The following Thalanyji example illustrates this: (30) nhani-karta nhurra parnakarri-n jurti-karta papu-ku-karta ngarrari-karta what-All you (Abs) go-Pres I=Dat-All father-Dat-All camp-All 'Why did you come to my father's camp?' [DAOGM105s74] Note that jurti, the (suppletive) 1st person possessive, is a dependent of papu 'father', but it does not bear a further dative case in agreement (i.e. jurti-ku), merely carrying the additional allative of the head noun ngarrari 'camp'.

374

Australian

4.4. Referential Double Case Nominals marked for allative or ablative case (both ablative1 and ablative2 in Jiwarli) can function as adverbial modifiers in Kanyara and Mantharta languages, giving spatial or temporal information about the event. In their locational usage these cases must be followed by an ergative case suffix when the clause is transitive. The temporal use of the ablative ('time from . . .') can be followed by any subcategorized argument case (ergative, accusative, or dative). Allative case marks the place or thing toward which an action or motion is directed. A Jiwarli example is: (31)

ngatha parlirri-a ngurra-rla-rru I (Abs) return-Pres camp-All-now 'I am going back to the camp' [N5p74s4]

Nominals in allative case have an adverbial function; they provide information about the locational orientation of the action described by the predicate. When an allative occurs in a transitive clause providing information about the directional orientation of the (transitive) subject, then the allative case is followed by the ergative, as in (32). (32) thuthu-ngku juma-rti-nha yanga-rninyja warlpari-lu dog-Erg child-Pl-Acc chase-Past south=All-Erg 'The dog chased the children south' [N13pl8sl]

The same phenomenon is found in many central and western Australian languages, including Warlpiri (see Dench and Evans 1988 for examples). There are two ablatives in Jiwarli, one (Abl1) with primarily local functions and the other (Abl2) with local, temporal, and causal uses (see above). Ablative1 takes ergative case when modifying the directional orientation of a transitive clause, as in:11 (33)

juma-ngku ngatha-nha nhanya-nyja maya-ngka-nguru-lu child-Erg I-Acc see-Past house-Loc-Abl1-Erg The child watched me from the house' [N10p10s3]

No additional case is required in intransitive clauses. The ablative2 inflection similarly requires ergative when serving as a synonym of ablative1 in a transitive clause. The phenomenon of ergative case assignment to non-subjects in transitive clauses is not restricted to allative and ablative nominals in Kanyara and Mantharta languages, but applies to members of three other categories. First, manner adverbs semantically qualifying a transitive predicate take ergative case, as in these examples from Jiwarli and Thalanyji respectively:12

Double Case Marking in Kanyara and Mantharta Languages

375

(34) wurnta-nma nhapa pirru tharti-ngku cut-Imp this (Acc) meat (Acc) quickly-Erg 'Cut this meat quickly!' [N5p71sl] (35) jankarra-n yaparru-lu ngali puni-kurrara tie-Imp quick-Erg we=Du (Abs) go-Intent 'Tie it up quickly so we two can go!' [HHPAN4p59sll]

In intransitive clauses the adverb takes no case suffix. Second, in Jiwarli the root modal particle pampa 'cannot' takes ergative case when it has scope over a transitive clause, as in: (36)

pampa-ngku-nthi ngatha mama-lkurni-a cannot-Erg-just I=Erg get-Cont-Pres 'I just can't get them' [N13p22s2]

When the particle is used in intransitive clauses it takes no suffix. Notice that the ergative in (36) is not an instance of "case copying" with the transitive subject, because the 1st person singular subject does not bear a separate ergative case. Also, no other particles (such as the negative), adverbs, or locatives take an additional case. Third, dependent clauses that are adjoined to transitive main clauses take an ergative case suffix following their non-finite verb inflection, even if that inflection can never take any other case affix. This applies in all languages to imperfective same-subject clauses, and to a type of intentive clause (note, however, that it does not apply to purpose clauses). Examples from Jiwarli and Thalanyji respectively are the following: (37)

kuwarti kurriya purra-rninyja patha-rrkarringu-ru jiriparri-yi now boomerang toss-Past hit-Intent-Erg echidna-Dat 'Next (he) threw a boomerang to hit echidna' [T35slO]

(38)

ngatha paja-rna mantu-nha ngulha-rla-lkarra-lu-ya I=Erg eat-Past meat-Ace nothing-Fact-ImperfSS-Erg-Emph 'I ate the meat finishing it' [WJOGM46s31]

Note that in (38) the main clause transitive subject does not bear an overt ergative case affix since it is a 1st person pronoun; the dependent clause, however, must take ergative case. A further function of the ablative case (ablative2 in Jiwarli) is to serve as a temporal adnominal modifier, making a kind of secondary predication meaning something like 'from the time when . . .'. It is often followed by the temporal post-inflectional suffixes (Thalanyji -nyji, Jiwarli -purra). In this usage the ablative case-marked nominal must be followed by a second case suffix in accordance with the syntactic function of the nominal of which it is predicated. The following example illustrates this for an extended nominal

376

Australian

predicate; here the ablative is adnominally modifying the subcategorized dative nominal: (39)

ngatha nhukura juma-parnti-yi I (Abs) knowing child-Ab12-Dat 'I have known (him) from a child (i.e. since he was a child)' [Nllp24sll]

The corresponding Thalanyji sentence is (40), to be contrasted to (41). (40)

ngatha nhukura kupuju-parnti-ku I (Abs) knowing child-Abl-Dat 'I have known him from a child (i.e. since he was a child)' [N12p5sl]

(41)

ngatha nhukura kupuju-parnti-nyji I (Abs) knowing child-Abl-time 'I have known (him) from a child (i.e. since I was a child)' [N15p4]

An example with an accusative is: (42)

ngatha ngarnka-nyja-rna kupuju-parnti-nha I=Erg big-Caus-Past child-Abl-Acc 'I raised (him) from a child' [N15p3]

Such ablatives may be followed by any subcategorized case, that is, ergative, dative, or accusative.13

5. Conclusion The Kanyara and Mantharta languages of Western Australia show widespread double case marking in their case morphology. We can discern three types: derivational, adnominal (Suffixaufnahme), and referential. Synchronically, a maximum of two levels of case marking is permitted in all these languages, although evidence from Jiwarli 3rd person pronouns shows that diachronic reanalysis has taken place so that reflexes of three separate dative case markers can be found. Notes This paper is one of a series on case marking in Mantharta and Kanyara languages (see Austin 1981c, 1989, 1992). An earlier version was presented at the Franz Nikolaus Finck Memorial Symposium at the Universitat Konstanz, September 1991. I am grateful to Frans Plank, the Deutsche Forschungsgemeinschaft, and the Universitat Konstanz for giving me the opportunity to present this material at the Symposium. Attendance at the Symposium was also supported by a grant from the ViceChancellor, La Trobe University.

Double Case Marking in Kanyara and Mantharta Languages

377

For helpful comments on the earlier version of this paper I thank Barry Blake, Greville Corbett, Edith Moravcsik, and Frans Plank, none of whom is responsible for remaining errors. Fieldwork on Western Australian languages has been supported by grants from the University of Western Australia Department of Anthropology, La Trobe University School of Humanities, Australian Research Grants Scheme, and the Australian Institute of Aboriginal and Torres Strait Islander Studies. Thanks are due to G. N. O'Grady and T. J. Klokeid for sharing their data on Kanyara and Mantharta languages with me. I owe a great debt to the speakers of these languages who have attempted to enlighten me about them over the years, especially Jack Butler, Dolly Butler, and Helen Hayes. 1. See, for example, discussion in Blake (1977: 38). 2. Data on Warriyangka and Thiin were collected by G. N. O'Grady, and on Tharrkari by T. J. Klokeid; I am grateful to them for kindly making their unpublished fieldnotes and tape-recordings available. Errors of transcription or analysis of these materials are solely my responsibility. 3. Data on Jiwarli were collected between 1978 and 1985 from the late Jack Butler of Onslow, Western Australia. Jack spoke several Australian languages, but Jiwarli was his mother tongue learned while he was living a semitraditional life early this century (see Butler and Austin 1986a, b). The Jiwarli data consist of some seventy texts (of various genres and on various topics), several hundred pages of fieldnotes, and about fifteen hours of tape recordings. Thalanyji materials were collected by myself in 1978-1987 from Helen Hayes of Carnarvon, Western Australia, and by G. N. O'Grady in 1967 from a number of Thalanyji speakers. The data comprise elicited (translated) sentences and a few brief texts. All this material is held at the Australian Institute of Aboriginal and Torres Strait Islander Studies, Canberra, and copies are available from the Institute or from the author. 4. See Hale (1979), Nash (1980: 169ff.), Simpson (1983: 97; 1991: 128ff.). 5. See Hale (1979, 1983, 1985), Laughren (1989), and Nathan (1986) for discussion of this notion. 6. In Mantharta languages there is a phonotactic constraint that all words must end in a vowel. To meet this constraint the empty syllable -ma is added to roots or stems ending in a nasal, and -pa is added to roots or stems ending in a lateral or rr. This accounts for the final ma in the absolutive form of 'hill kangaroo' in Table 13.1. 7. In Tharrkari and Warriyangka the 2nd person singular pronoun nhurra also does not distinguish ergative and absolutive formally. 8. Blake (1987: 36) somewhat confusingly refers to the grammatical relation borne by the complement of an extended intransitive verb as "indirect object." 9. The transcription adopted for Kanyara and Mantharta languages is a practical orthography that follows usual Australianist conventions: th/nh/lh represent laminodental stop, nasal, and lateral respectively; j/ny/ly are lamino-palatal; rt/rn/rl are apicodomal (retroflex). The velar nasal is ng. In homorganic nasal-stop clusters the digraph for place of articulation is written once only: thus nth (not nhth) and rnt (not rnrt). Note, however, that nyj is phonetically distinct from nj. The alveolar tap is written rr, and the post-alveolar continuant as r. The three vowels are a, i, u; length is indicated by doubling the vowel symbol. Equal signs separate morpheme glosses where the language form is not segmentable. Abbreviations in square brackets following the free English translations are references to the source data. 10. See also example (21) below for another instance. 11. Recall that ablativel is attached to the locative inflected form of a nominal (see 4.2.1 above).

378

Australian

12. Adverbs differ from nominals in that they cannot be marked for number and cannot take certain derivational suffixes that attach only to nominals (see Austin 1989; for more general discussion of adverbs in Australian languages see Bowe 1991). The use of ergative case with manner adverb modifiers is found in other central and western Australian languages; see Simpson (1991: 123) on Warlpiri, and Bowe (1990: 56ff.; 1991) on Pitjantjatjara. 13. Dench (1987) shows that in the nearby Martuthunira language the temporal ablative may also modify local case-marked nominals, as in, for example, the equivalent of 'I went to town-Allative small-Ablative-Allative', meaning 'I have been going to town since it was small'. There are no examples of this type in my Kanyara and Mantharta data.

References Austin, Peter. 1981a. Proto-Kanyara and proto-Mantharta historical phonology. Lingua 54, 41-77. Austin, Peter. 1981b. Switch-reference in Australia. Language 57, 309-334. Austin, Peter. 1981c. Case-marking in Southern Pilbara languages. Australian Journal of Linguistics 1, 211-226. Austin, Peter. 1987. Word order and clause combining in Gascoyne-Ashburton languages. In Proceedings of the Third Annual Pacific Linguistics Conference, ed. by Scott DeLancey and Russell Tomlin, 1-11. Eugene: University of Oregon. Austin, Peter. 1988a. Classification of Southern Pilbara languages. Papers in Australian Linguistics No. 17, 1-17. Canberra: Pacific Linguistics. Austin, Peter. 1988b. Aboriginal languages of the Gascoyne-Ashburton region. La Trobe University Working Papers in Linguistics 1, 43-63. Austin, Peter. 1988c. Word order in a free word order language—the case of Jiwarli. Manuscript, La Trobe University. Austin, Peter. 1989. A reference grammar of the Mantharta languages, Western Australia. Manuscript, La Trobe University. Austin, Peter. 1992. Cases and clauses in Jiwarli, Western Australia. La Trobe University Working Papers in Linguistics 5. Blake, Barry J. 1977. Case Marking in Australian Languages. Canberra: Australian Institute of Aboriginal Studies. Blake, Barry J. 1987. Australian Aboriginal Grammar. London: Croom Helm. Bowe, Heather. 1990. Categories, Constituents and Constituent Order in Pitjantjatjara. London: Routledge. Bowe, Heather. 1991. Evidence of a syntactic category adverbial in Pitjantjatjara. Paper presented at the workshop on Aboriginal Languages and Theoretical Linguistics, Brisbane. Butler, Jack, and Peter Austin. 1986a. Jiwarli texts. Manuscript, La Trobe University. Butler, Jack, and Peter Austin. 1986b. Earthquake and Halley's comet. Aboriginal History 9, 8-21. Dench, Alan C. 1987. Martuthunira: A Language of the Pilbara Region of Western Australia. Ph.D. dissertation, University of Western Australia. Dench, Alan, and Nicholas Evans. 1988. Multiple case-marking in Australian languages. Australian Journal of Linguistics 8, 1-47. Dixon, R. M. W. 1979. Ergativity. Language 55, 59-138.

Double Case Marking in Kanyara and Mantharta Languages

379

Dixon, R. M. W., ed. 1987. Studies in Ergativity. (Lingua 71.) Amsterdam: NorthHolland. Finck, Franz N. 1910. Die Haupttypen des Sprachbaus. Leipzig: Teubner. Hale, Kenneth. 1979. The position of Walbiri in a typology of the base. Manuscript, Massachusetts Institute of Technology. (Circulated by Indiana University Linguistics Club, 1980.) Hale, Kenneth. 1983. Warlpiri and the grammar of non-configurational languages. Natural Language and Linguistic Theory 1, 5-77. Hale, Kenneth. 1985. On non-configurational structures. Manuscript, Massachusetts Institute of Technology. Laughren, Mary. 1989. The configurationality parameter and Warlpiri. In Configurationality: The Typology of Asymmetries, ed. by Laszlo Maracz and Pieter Muysken, 319-353. Dordrecht: Foris. Nash, David G. 1980. Topics in Warlpiri Grammar. Ph.D. dissertation, Massachusetts Institute of Technology. (Published New York:Garland, 1985.) Nathan, David. 1986. Topics in configurationality. BA honours thesis, La Trobe University. O'Grady, Geoffrey N., Carl Voegelin, and Florence M. Voegelin. 1966. Indo-Pacific fascicle six. Anthropological Linguistics 7(2). Plank, Frans. 1990. Suffix copying as a mirror-image phenomenon. Linguistics 28, 1039-1045. Silverstein, Michael. 1976. Hierarchy of features and ergativity. In Grammatical Categories in Australian Languages, ed. by R. M. W. Dixon, 112-171. Canberra: Australian Institute of Aboriginal Studies. Silverstein, Michael. 1980. Of nominatives and datives: Universal grammar from the bottom up. Manuscript, University of Chicago. Silverstein, Michael. 1981. Case marking and the nature of language. Australian Journal of Linguistics 1, 227-246. Simpson, Jane. 1983. Aspects of Warlpiri Morphology and Syntax. Ph.D. dissertation, Massachusetts Institute of Technology. Simpson, Jane. 1991. Warlpiri Morphosyntax: A Lexicalist Approach. Dordrecht: Kluwer. Wurm, Stephen A. 1972. Languages of Australia and Tasmania. The Hague: Mouton.

14 Suffixaufnahme and Apparent Ellipsis in Martuthunira Alan Dench 1. Introduction For many languages, the status of the genitive construction is the apparent key to the recognition, or otherwise, of a pattern of Suffixaufnahme, defined as that phenomenon wherein nouns in an attributive relation agree with their heads in case. Plank (1990: 1039) notes that: This . . . kind of agreement is apparently so out of the ordinary that grammarians faced with it feel tempted to annul it terminologically, by labelling case-agreeing attributive nouns adjectives solely on the strength of their agreeing in case.

This has certainly been something of a preceived tradition in the description of Australian languages, as Plank notes. But the defining away of Suffixaufnahme by recourse to the "genitive is an adjective" ploy crucially depends on there being a clear distinction between noun and adjective, and then on the principled assignment of genitive constructions to the adjective class. Of course, if there is no clear distinction between adjective and noun, then the ploy fails, in a trivial fashion. The suggestion that suffixes inside Suffixaufnahme case agreement be considered derivational—if not categorically (as has often been done in the Australian tradition) then at least functionally—is a related ploy. It has sometimes been assumed that suffixes are typically polyvalent in many languages of the world and that in regard to the relationship between derivation and inflection, there are two polar possibilities. A suffix could accordingly be considered to function as a case marker if it relates a noun phrase to a head, and as derivation if there is no (overt) head present. In this chapter I hope to shed some light on these issues by considering the structure of noun phrases in Martuthunira. First, I seek to demonstrate that there is no contrast between adjective and noun in Martuthunira. Second, I argue that it makes no sense to invoke a distinction between deriva-

Suffixaufnahme

and Apparent Ellipsis in Martuthunira

381

tional and inflectional suffix functions as a way of treating Suffixaufnahme in Martuthunira. Following this discussion I show that the flexibility inherent in the lack of a distinct adjective class and the presence of productive Suffixaufnahme allow a great deal of creativity in the establishment of reference and the structuring of NPs in Martuthunira text. 2. Noun and Adjective? In many Australian languages it is difficult to make a strict division between the classes of noun and adjective. First, there are often no clear formal criteria for a distinction among subclasses; and second, it is often difficult to make discrete classifications based on semantic/functional criteria. Dixon (1980) argues that noun and adjective classes in Australian languages are justified on semantic grounds even if it is not possible to distinguish between them on formal grounds. He notes that the division often has, in particular languages, certain syntactic (functional) correlates, but that these are "of a more/less statistical nature, rather than being either/or properties that could be used as defining criteria" (1980: 275). This leads Dixon to make statements such as the following about Warrgamay: "Nouns and Adjectives have almost the same morphological properties; there is of course a clear semantic difference" (1981: 27). Not surprisingly, not all writers share Dixon's confidence. Thus Donaldson, in her description of Ngiyambaa, briefly considers reduplication as a test for adjectivehood since only "prototypical adjectives" reduplicate. But in Ngiyambaa there are no known further differences, morphological or syntactic. . . . Syntactically, for instance, any nominal which can be a constituent part of an NP can also be the sole representative of an NP. . . . To introduce the terms 'noun' and 'adjective' as synonyms for 'non-reduplicating' and 'reduplicating' would serve no descriptive purpose elsewhere in the grammar. (Donaldson 1980: 70-71)

McGregor makes similar observations in his grammar of Gooniyandi. Here, "Entity" is the semantic head of a phrase, and "Quantifier" is an NP modifier: There seem to be no compelling reasons to identify distinct subclasses of 'Adjectives' and 'Nouns' . . . [T]he classes of words which may realise the NP roles of Classifier, Entity, or Qualifier . . . have many members in common. Nearly all the words which can realise the Entity role . . . can also realise the role Qualifier: in other words, words which can be referential can also be qualifying, and may indicate qualities or properties of things. (McGregor 1990: 141-142)

As McGregor's comments suggest, the a priori recognition of a semantically based division between noun and adjective is not without problems. The fact that a particular nominal may denote an entity in one NP and yet function as a modifier of a head, denoting properties of an entity in another NP, means that

382

Australian

we may not be able confidently to decide whether the particular lexical item is a noun or an adjective even on semantic grounds, unless we assume certain universals of human conceptualization. Given these problems, a number of grammatical descriptions of Australian languages choose not to distinguish nouns and adjectives, instead recognizing that "nominals" may have different functions in different NPs. For example, in Morphy's description of Djapu (1983) and McGregor's description of Gooniyandi (1990), the contrast between entity-denoting nominal and modifying nominal is handled by positing a system of functional slots within the NP which may be filled by different nominals. Evans's description of Kayardild (1985) and Dench's description of Martuthunira (1987) essentially follow McGregor's proposals (first made in his 1984 thesis). The Martuthunira analysis will be discussed below. By contrast, Goddard (1983) provides good arguments for the description of separate noun and adjective classes in Yankunytjatjara, a Western Desert language. Nominals are subcategorized according to semantic criteria reinforced by a consideration of the co-occurrence restrictions of particular nominals. Goddard describes a set of five syntactic frames against which nominals are compared: (a) head of an NP; (b) in apposition to another nominal (head) in an NP; (c) in apposition to another nominal in a simple ascriptive clause; (d) with a copula in a simple ascriptive clause; and (e) as a second predication on an NP in a verbal clause. Nouns are defined as those nominals which can occur as heads (restricted to slot (a)). There are two kinds of adjectives: "stative" adjectives modify NP heads (b) and require no copula in simple ascriptive clauses (c); "active" adjectives require a copula in simple ascriptive clauses (d) and function as second predications (e). The same syntactic tests can be applied to nominals in Martuthunira, but, unlike in Yankunytjatjara, the tests do not reveal mutually exclusive sets. Martuthunira nominals are distributed among the following groups: A. B.

C. D.

Nominals that almost always appear as heads (a). These are closest, on notional grounds, to nouns. Nominals that typically function as modifiers of NP heads (b), may appear as heads of phrases (a), and may occur in apposition to another nominal in a simple ascriptive clause (c). Any nominal that can appear in this last frame can also appear with a copula (d) (where the speaker chooses to provide tense/mood information about some state, for example). The class includes a large number of what might be considered prototypical adjectives. Nominals that may appear in all five slots. Most items in this group can also be recognized as prototypical adjectives. Nominals that cannot be apposed to another nominal in a simple ascriptive clause but require a copula (d), and that function almost exclusively as second predications of manner (e). These function very much like manner adverbs and are equivalent to "active adjectives" in Yankunytjatjara.

Table 14.1 shows the range of functions for a selection of nominals.

Suffixaufnahme

383

and Apparent Ellipsis in Martuthunira

Table 14.1. Range of Nominal Functions in Martuthunira.

wirra 'boomerang' tharnta 'kangaroo' wartirra 'woman' ngapala 'mud' karlarra 'hot' winya 'full' pinkarranyu 'dry' panyu 'good' jalya 'useless' kur.ta 'clever' kartarr 'firmly' nguyirri 'asleep' puwany 'hunting'

c

d

e

a head

b modifier

simple ascriptive

copula ascriptive

second predication

* * * * * * * * * *

* * * * * * *

* * * * * * *

* * * * * * * * * *

* * * * * * * *

The range of functions for two of the more flexible nominals, pinkarranyu 'dry' (a) and winya 'full' (b), is illustrated in the following examples. Head of NP: (1) a. nhuwana-yi pawulu-ngara parrani- yilangu nyina-lu pinkarranyu-la 2Pl-Voc child-Pi return-Imp here stay-PurpSS dry-Loc 'You children come back here and stay in the dry' b. winya-npa-rra-rru, pintirrijila-rru puni-marri-layi ngurra-arta-rru, pintirrijila wanti-marri-layi-rru ngurra-ngka-rru winya-ngara full-Inch-Rel-Now, scattered-Now go-Coll-Fut camp-All-Now, scattered lieColl-Fut-Now camp-Loc-Now full-Pi '(We) get full now, and then scatter to (our) camps, each to lie in (our) camps then, (we) full ones' [no mention of implicit 'we' in previous twelve clauses] Modifier in NP: (2) a. nhulaa manku- pinkarranyu kalyaran that grab-Imp dry stick 'Grab that, a dry stick!' b. nhula manyarrka wantha-rryu nganaju-wu-la parrka-ngka, kayarra mirntir winya that sugar put-Imp 1Sg=Obl-Gen-Loc tea-Loc, two spoon full 'Put that sugar in my tea, two spoons full'

384

Australian

Simple ascriptive: (3) a. nhiyu thanuwa pinkarranyu paju nganaju mungka-lwaa this bread dry Too 1Sg=Acc eat-PurpSbj=Obj1 'This bread is too dry for me to eat' b. nhiyu kayulumarnu winya-rru this waterbag full-Now 'This waterbag is full'

Ascriptive with copula: (4) ngulangu yirrala-la, jalyuru-la-rru, ngunhu-ngara pawulu-ngara nyina-layi pinkarranyu waruul-wa-rru there sheet-Loc, hole-Loc-Now, that(Nom)-Pl child-Pl be-Fut dry still- -Now 'Under the sheet, those children will stay dry then'

Second predication: (5) a. nhulaa wartirra wiru thanuwa-a thurnta-rninyji kayulu-wirriwa-a, pinkarranyu-u that [near you] woman want flour-Acc knead-Fut water-Priv-Acc, dry-Ace 'That woman wants to rub flour without water, dry' b. punga-a-rru winya-ma-rninyji puni-waa winya-rru, thuur.ta-a mungka-lalhanguru guts-Acc-Now full-Caus-Fut go-PurpSbj=Obj full-Now, fruit-Ace eat-PastAbl '(They) then fill their guts so (they) can go along full now, having eaten some fruit'

With such wide-ranging patterns of use it is clear that a classification into nominal subclasses is not nearly as neat as that suggested for Yankunytjatjara. There is no set of necessary and/or sufficient conditions which allow the definition of noun versus adjective. They cannot be distinguished by the ability of just one class to function as the head of an NP. This is not to deny the fact that individual nominals have greater or lesser ability to occur in particular frames. However, these possibilities of occurrence cannot be predicted by general syntactic or semantic rules. Such information might be stated in the lexical entry for each nominal or taken as the result of general rules of inference given particular situations. 3. Noun Phrase Structure The lack of a formal distinction between adjective and noun has some important implications for the description of NPs in Martuthunira. Given that a range of nominals—and, as it turns out, complex nominal expressions—may function as either entity-referring or modifying, how is one to identify the head of any given NP?

Suffixaufnahmeme and Apparent Ellipsis in Martuthunira

385

One possible way of handling this is to assume that NPs do not have heads. This would mean, from a semantic perspective, that there is no subconstituent of an NP which is more important than any other in contributing to the NP's reference. However, from a purely syntactic perspective, the assumption of headlessness has further implications. If a phrase has no head, then arguably it is not a phrase. Such an approach would be akin to the kinds of "flat" analyses that have been suggested for clause structure in non-configurational languages. This is certainly not a novel idea. Descriptions of a number of Australian languages, most notably Kalkatungu (Blake 1983) and Warlpiri (Hale 1983, Simpson 1983), have argued that it is not necessary to set up an NP constituent in these languages. Where the language appears to have a relatively high incidence of discontinuous (apparent) NPs, rules of case concord are required to link similarly-marked nominals together at some level of semantic/ functional structure. These same rules can be used to handle adjacent nominals, thus obviating the need for any mediating syntactic constituent such as NP. Blake suggests that "where an argument is represented by more than one word we have nominals in parallel or in apposition" (Blake 1983: 145). If we were to treat sequences of nominals in Martuthunira as appositional structures of some kind, the problem of deciding on a head for this structure could be avoided. If each element in a sequence of nominals is considered to be an NP, then there are essentially three possible analyses of the appositional construction. We either impose some internal structure (for example, assign head status to the first NP and modifier status to the following NPs), assume that the structure is headless, or assume that the structure is multi-headed (as we might analyse a sequence of conjuncts, for example). In this last case, each nominal is assumed to be the head of its own NP and provides information that can be unified, in accordance with case agreement (Suffixaufnahme), at some higher level. However, to my mind the simple form of this argument is inappropriate for Martuthunira. Martuthunira is not obviously non-configurational. First, it has a clear preference for particular word orders; at the clause level this is SVO, with a marked SOV alternative for object-focus. Second, the nominals referring to a particular entity filling the role of an argument of some predicate typically occur adjacent to one another in an order which is both predictable and meaningful.2 That is, there are structures that look like NPs. And third, since it is possible for two distinct arguments of some verbs to bear the same case (in particular, the two objects of ditransitives) there is independent motivation for a level of NP structure intervening between inflected nominal and the level of the clause. This is not to say that there cannot be a number of non-adjacent, yet co-referring, nominal expressions in a clause—this is exactly how second predications, for example, are coded in Martuthunira. So Martuthunira has NPs which consist of a sequence of categorially indistinct expressions, one of which is to be identified as the semantic head of the phrase. The identification of the NP is quite straightforward:

386

Australian

The NP is defined as a sequence of adjacent nominals over which some nominal suffix is distributed. Some fine-tuning of this definition is required, since I want to allow NPs in which some clitic or particle can intervene between otherwise contiguous nominals (as in (22) and (24) below).3 Within the NP, nominals occur in a fixed order of functional slots: (Determiner) (Quantifier)

(Classifier)

Entity

(Qualifier)

This sequence corresponds to that described by McGregor for Gooniyandi (1990) and is essentially the same as that described for Kalkatungu by Blake (1983). In Kalkatungu, the order holds for elements in discontinuous "NPs."4 The filler of the Entity slot is the semantic head of the NP. It is the nominal in this slot which makes the primary reference to some object or person. Nominals in the pre-head slots restrict the reference of the head nominal by narrowing the set of entities from which the referent is chosen. Nominals filling the Determiner slot, most often demonstratives or possessive pronouns, narrow the reference of the phrase by contextual identification of the referent. The Quantifier slot may be filled by a numeral or by a nominal functioning as a mass quantifier. Nominals in Classifier function pick out a subset of the set of items to which the head may refer. For example, the Classifier may describe a property manifested by a subset of the class of objects denoted by the head, or it may name a specific type of entity which forms a subset of the generic class denoted by the head. Some examples of simple NPs are given below. (6) nganaju yaan yungku-lha murla-a yartapalyu-u kanyara-ngara-a 1Sg=Gen wife give-Past meat-Ace others-Ace man-Pl-Acc 'My wife gave meat to the other men' Det Head DetDet Head (7) purra-lyarra parla-marta yarta-ngka pilyi-ngka parla-ngka hit-Rel stone-Prop other-Loc flat-Loc stone-Loc '. . . hit with a stone on another flat stone' Det Class Head (8) ngunhu kartatha-lalha marruwa-a wirra-a that (Nom) chop-Past snakewood-Acc boomerang-Acc 'He chopped out a snakewood boomerang' Class Head (9) thathu-rnu ngaliwa ngurnu tharnta-a murla-a send-Pass 1Pl=Incl that=Acc kangaroo-Ace meat-Ace 'We were sent that kangaroo meat' Det Class Head

Suffixaufnahme

and Apparent Ellipsis in Martuthunira

387

Nominals in the post-head Qualifier position have a non-restrictive modifying function: they provide some additional information about the entity picked out by the NP. All nominals that may precede the head of an NP may also function as fillers of the Qualifier slot, and apparently the converse also holds. It appears that a number of qualifying expressions may follow the head. However, an extended sequence of qualifiers does not usually occur under the same intonation contour; instead, successive nominals are separated by a noticeable pause. These come more and more to resemble "afterthought" constructions. The analysis of the NP as including an optional post-head Qualifier, given that heads are not categorically distinct from other elements in the NP, implies a degree of indeterminacy in the analysis. How is one to decide which of the nominals is the semantic head? Thus (9) above, for which it is suggested that 'meat' is the head (in the Entity slot) and 'kangaroo' a Classifier, could have an alternative reading in which 'kangaroo' is the head and 'meat' a post-head modifier. The two readings would be: a. We were sent that meat of the kangaroo kind. b. We were sent that kangaroo, which is meat. There are other alternatives. The initial demonstrative might be the head, followed by a post-head modifying NP which contains in turn the two nominals 'kangaroo' and 'meat' in various alternative interpretations. c. We were sent that thing, which is kangaroo meat. d. ... I see no particular problem with this degree of indeterminacy and am happy to accept that the grammar has the flexibility to generate these alternative and essentially, though usually very subtly, ambiguous structures. To summarize the argument so far, I have suggested that Martuthunira has no distinction between nouns and adjectives and that the constituents of an NP are not categorially distinct. Yet despite this, NPs have heads—even despite the lack of a structurally defined head position. The next section describes the use of adnominal inflections, such as the genitive. These produce complex nominal expressions which, like simplex nominals, may function as either heads or modifiers in NPs. 4. Nominal Attributive Constructions and Suffixaufnahme Nominal suffixes in Martuthunira have a wide range of the functions described by Dench and Evans (1988). While for the purposes of the discussion at hand it is mainly the adnominal and derivational functions which are of interest,

388

Australian Table 14.2. Functions of Nominal Suffixes in Martuthunira. Derivational

Effector Accusative Ablative Locative

Associative Proprietive Genitive(s) Privative

* * (*)b

Adnominal

Relational

* *

* *

* * * *

*

* (*)a

Referential *

* *

Complementizing C-comp T-comp *

» *

» *

Notes: a. The proprietive has an important function as the marker of instrumental NPs and in this function it typically agrees in case with the actor in the clause. Since this agreement is most often with the unmarked nominative subject of an active clause, the proprietive appears to have a relational function as the "instrumental" case. This (developing) pattern is accentuated by the occasional omission of effector case agreement when the instrument agrees with a passive agent (usually corrected on quizzing). The development of the relational function of the associative no doubt followed a similar route. b. The derivational function is restricted to the minor genitive suffixes -wura Belong and -waya Owner.

Table 14.2 demonstrates the full range for a selection of the more common nominal suffixes. The suffixes fall into a number of groups. The true syntactic cases, accusative and effector (which marks passive agents), cannot be used adnominally.5 The local suffixes have the widest range of functions but are not transparently involved in lexical derivation. A set of suffixes—including the productive genitive (and two minor genitive suffixes), the common Australian proprietive or "Having" suffix, and the privative or "Lacking" suffix—are restricted to derivational and adnominal functions. By derivational function, I mean here that the suffix appears in lexical items for which the meaning is not completely predictable from the meaning of the root and that of the suffix. Examples include: (10)

a. karla-marnu fire-Ass 'place on thigh for punishment spear' b. mirntiri-marta claw-Prop 'Gould's yellow sand goanna' c. warnany-kura rain-Belong 'rivergum tree'

The fixed lexical status of some derived forms can be established by their behavior under Suffixaufnahme. There is a general ban on sequences of identical nominal suffixes in Martuthunira. Thus a sequence of two adnominal

Suffixaufnahme

and Apparent Ellipsis in Martuthunira

389

proprietives is not allowed, although a privative followed by a proprietive is acceptable: (11)

ngunhu wartirra puni-lha ngurnu-marta kanyara-marta tharnta-wirriwa-marta / *tharnta-marta-marta that (Nom) woman go-Past that=Obl-Prop man-Prop kangaroo-Priv-Prop / * kangaroo-Prop-Prop 'That woman went with the man who is without a kangaroo / *who has the kangaroo'

The constraint does not apply where one of two adjacent suffixes is part of a derived lexeme: (12)

ngunhu wartirra puni-lha ngurnu-marta mirntiriniarta-marta that (Nom) woman go-Past that=Obl-Prop goanna-Prop 'That woman went along with a goanna (carrying it)'

However, not all cases are as straightforward as this. The word for 'soap' is usually given as ngawurrmarta, but in example (13) the form does not take an expected proprietive; thus the (presumed) derivational form is transparent to the rule blocking agreement in this instance. (13)

wilyiwilyi-ma-lalha-nguru ngurnu-marta ngawurrmarta(*-marta), wara-ngara-a, wantha-rninyji wilyiwilyi-ma-rnu-ngara-a wantha-rninyji yakarrangu-lu kampa-rra clean-Caus-Past-Abl that=Obl-Prop soap(*-Prop), clothes-Pl-Acc, put-Fut clean-Caus-Pass-Pl-Acc put-Fut sun-Eff heat-Rel 'Having cleaned the clothes with that soap, put the cleaned ones out to be heated by the sun'

The following examples further illustrate the adnominal/attributive functions of the proprietive. In each of these examples the nominal bearing the suffix clearly functions as a one-place predicate. The patterns are generalizable to the other "adnominal" suffixes: (14)

ngunhu kanyara jawurta-marta that (Nom) man beard-Prop 'That man has a beard'

(15)

nhartu nhulaa marra-marta? what that wing-Prop 'What's that thing with wings?' ('What kind of bird is that?')

(16)

nhawu-layi ngurnaa kurryarta-marta-a-rru see-Fut that=Acc spear-Prop-Acc-Now '(I'll) see that that one now has a spear (in it)'

390

Australian

Two genitive suffixes are illustrated in the following sentences. The minor genitive suffix, glossed as Belong (19, 20), is attached to nominals denoting entities which exert some controlling possessive relationship over another entity: (17)

ngunhu ngurra tharratal-yu thungkara-la wantha-rnu that (Nom) camp bird-Gen ground-Loc put-Pass 'The nest of the tharratal bird is built on the ground'

(18)

yirna-tharra-wu kanyara-tharra-wu, ngunhu ngurra parlu-ngka parla-ngka this=Obl-Du-Gen man-Du-Gen, that (Nom) camp top-Loc hill-Loc 'These two men's, that camp is on top of a hill'

(19)

nhiyu kanparr-wura parla-marta this (Nom) spider-Belong stone-Prop 'This thing belonging to the spider [nest], has a stone [for a doorway]'

(20)

warruwa-ngara-wura Walter. kanyara-wura Karlinpangu European-Pl-Belong Walter. Aboriginal-Belong Karlinpangu 'His European [name] is Walter. His Aboriginal [name] is Karlinpangu'

Where such predications occur within NPs, the expected patterns of Suffixaufnahme occur. The following examples are chosen to illustrate not only regular Suffixaufnahme but also the fact that adnominal ascriptives can appear in isolated NPs (shown in boldface). (21)

a. ngayu kangku-lha mayiili-marnu-ngu kulhampa-arta I take-Past Son's=Son(my)-Group-Ace fish-All 'I took a group of my grandchildren for fish' b. nganangu-ngara pawulu-ngara? who=Gen-Pl child-Pi 'Whose children are they?' c. ngurnungara-a yaan-wirriwa-wura-a that=Pl-Ace spouse-Priv-Belong-Ace '(I took) the ones who belong to the one who is without a spouse' d. ngaa, purrkuru, pala. ngarraya-ngu-ngara-a Yes, Okay, It. niece-Gen-Pl-Acc 'Yes, Okay, That's it. (You took) niece's ones'

(22)

a. ngunhu muyi yanga-lalha pawulu-tharra-a that (Nom) dog chase-Past child-Du-Acc 'That dog chased two children' b. ngana-ngura-tharra-a yanga-lalha? who-Belong-Du-Acc chase-Past 'Whose two did it chase?' c. yirna-tharra-wura-a this=Obl-Du-Belong-Ace '(It chased) the ones belonging to these two'

Suffixaufnahme

and Apparent Ellipsis in Martuthunira

391

Examples of the adnominal use of the locative, with Suffixaufnahme, are given in (23) and (24). ngali panyu-ngka-a warra kalyaran-ta-a thuur.ta-a manku-layi lDu=Incl good-Loc-Acc Cont tree-Loc-Acc fruit-Ace get-Fut 'We'll get fruit that's in a better tree' (24) jinkayu-rru, warnanykura-la-ngara-a kunkuwarra-a nyina-layi kartatha-l.yarrarru up=river-Now, river-gum-Loc-Pl-Acc hive-Acc sit-Fut chop-Rel-Now 'Up river now, (she'll) sit and chop many hives in a river gum tree' (i.e. a number of hives in one tree)

(23)

It is the adnominal suffixes that have been described as pre-case suffixes by Blake (1987: 88), and as derivations by Dixon (1980: 300) and others. The knock-down argument here is that if the criteria were to be applied strictly, the Martuthunira locative—since it can be used adnominally inside relational accusative marking, as in (23) and (24)—should also be in this class. This is clearly at odds with its relational, referential, and complementizing functions (see Table 14.2). Thus adnominal suffixes are not derivational suffixes (see Dench and Evans 1988: 10 for further discussion). An alternative is to treat the adnominal suffixes as functionally bivalent; they may be either derivational or inflectional, depending on context. Where the marked nominal functions as a head, the suffix is assumed to have a derivational function deriving a new lexeme of the category of normal heads. Where the marked nominal occurs as a modifier of a head, it is taken to be an inflectional adnominal suffix. Such a treatment would stretch the notion of "derivation" to a point at which the concept becomes quite meaningless in a language like Martuthunira. As we have seen, there is no categorial difference between heads and modifiers. There is no reason to believe that all marked nominals appearing as heads have undergone some process of lexical derivation; the suffixation is entirely consistent with processes of regular syntactic (and semantic) inflection. 5. Apparent Ellipsis Examples (21) and (22) above illustrate a common pattern in which elliptical responses in a conversation preserve the case suffixes appropriate to the full clause. Thus in (21c) and (21d) the response maintains the accusative case marking of (21 a), and similarly in (22b) and (22c). However, not all examples of adnominally inflected heads are quite so easily explained by a regular pattern of ellipsis. In (25), the two principal protagonists are referred to by transparent adnominal expressions (the Owner suffix is a minor genitive), neither of which can be thought of as lexical derivations. The expressions thaluwaya and murlamarnu are chosen to emphasize the particular relationship existing between a man and his dog. The man is described solely as the

392

Australian

dog's owner, the dog is described by a generic expression often used for grouping hunting implements. (25)

thalu-waya panyu-npa-wala-rru wiru ngurnulangu-lu-rru murla-marnu-lu kanarri-yangu pet-Owner good-Inch-PurpDS-Now feelings that=Gen-Eff-Now meat-Ass-Eff come-Pass The pet owner will be getting good feelings now that his meat-getter [dog] has

come to him' A speaker may choose to use an adnominally expressed attributive as a head where there may not be a specific referent known to either speaker or addressee. Thus, in (26) it turns out that the first speaker has a 'camel bush' thorn in his foot, but this is not known until later in the conversation. Until then, the thorn is referred to by the ascriptive property that it is 'in the foot'. (26)

a. jina karta-rnu wanthala? foot poke-Pass somewhere '[Your] foot got poked somewhere?' b. ngawu, thuulwa-rninyji-nu jina-ngka-a ngathala-a? yes, pull-Fut-Quot foot-Loc-Acc 1Sg=Loc-Acc 'Yes, how about pulling out [the thing] in my foot?' c. nyina- kartu ngathala yilangu, ngayu thuulwa-rninyji-la jina-ngka-a sit-Imp you me=Loc here, I pull-Fut-Loc foot-Loc-Acc 'You sit here by me while I then pull out [the thing] in [your] foot'

There are two examples of apparently elliptical expressions in (27). Here the narrator asks a group of people to shop for him. First, in (27b) the NP thawun-ta town-Loc, referring to the things bought 'in town', provides a generic reference to a collection of groceries which are itemized later in the text (when it is discovered that some things have been forgotten). The second expression, thanuwa-marta-ngara food-Prop-Pl in (27c), referring to the people 'with the food', picks out a particularly relevant property of the shoppers who were sent back to town for the forgotten items. (27)

a. nhuwana puni-rra wii thawun-mulyarra, nganaju wuruma-rninyji yurntura-a manyarrka-a-thurti wii parrka-a wii? . . . 2P1 go-Rel if town-All, 1Sg=Acc do for-Fut flour-Ace sugar-Acc-Conj maybe tea-Ace maybe 'If you go to town will you get for me some flour and sugar maybe, and maybe tea? b. nganarna manku-lha-nguru-rru thawun-ta-a, wuruma-l.yarra kartungu, parrani-lha-ma-rninyji-rru kartungu-mulyarra . . . 1Pl= Excl get-Past-Abl-Now town-Loc-Acc, do-for-Rel 2Sg = Acc, returnPast-Caus-Fut-Now 2Sg=Obl-All 'Now having got [the things] in town for you, we then brought them back to vou . . '

Suffixaufnahme

and Apparent Ellipsis in Martuthunira

393

c. ngayu marlara-a karri-nguru nhawu-rra. purrkuru waruul, nhula-ngara murna-ngka-rru ngaliwa-a, thanuwa-marta-ngara 1Sg (Nom) road-Acc stand-Pres watch-Rel. true still, that[near you]-Pl close-Loc-Now 1Pl=Incl-Acc, food-Prop-Pl 'I stand watching the road. True enough, they are close to us now, [those people] with the food'

Once again, these adnominal expressions cannot be considered instances of lexical derivation. Instead, the ascriptive use of adnominal suffixes is employed as a productive device allowing a great deal of creativity in the construction of narrative. An entity can be referred to by any of the properties it is known to have, whether these are inherent—for instance, determine its being considered a token of a particular nominal type (e.g. 'dog' or 'man')— or whether they are dependent on a particular context (that dog's owners use them in hunting meat). Similarly, once a participant has been introduced into a text, it can be named by any of the characteristic properties that served to identify it in the first place, by the properties attributed to it by way of qualifying expressions, or by characteristics or properties acquired through its role in the events recounted in the text. As the following example shows, this pattern of using ascriptive predicates as the apparent heads of nominal referring expressions is not restricted to adnominally marked nominals. In (28) the expression in boldface is a verb bearing plural marking and the accusative suffix. A similar example appears in (13) above. (28) thanuwa-ngara-marta nyina-layi wangkarnu-marra-rru nhartu-ngara-a mankuIha-nguru wuruma-l.yarra nyina-marri-lha-ngara-a food-Pl-Prop sit-Fut talk-Coll-Rel-Now thing-Pl-Acc get-Past-Abl do-for-Rel sit-Coll-Past-Pl-Acc 'With all the food they then sit down and talk together, having got the things for [the ones who] stayed (behind) together'

Instances of apparent ellipsis, in which an ascriptive predication functions as the head of an NP, can be considered to be equivalent to the use of apparent adjectives as heads. The lack of a categorial distinction between noun and adjective is extended to adnominal expressions, and further to verbal expressions.

6. Conclusion Martuthunira illustrates a particularly pervasive form of Suffixaufnahme. Suffixaufnahme, or multiple case marking, is completely productive within NPs. Moreover, there are no grounds whatever for an argument that the patterns be "explained away" by assuming that the inside suffixes are derivational. As I have shown, there is no basis for a distinction between noun and adjective in

394

Australian

Martuthunira. Hence, it is not possible to define away Suffixaufnahme as adjective derivation. The lack of a distinction between noun and adjective has interesting implications for the analysis of NPs in the language and suggests that there is a degree of indeterminacy in the assignment of heads in NP structures. I have argued that this flexibility, coupled with productive Suffixaufnahme, allows a high degree of creativity in the establishment and maintenance of reference in Martuthunira text.

Notes 1. Switch reference in Martuthunira includes forms for subject-object identity (glossed as Sbj=Obj) in addition to same-subject (SS) and different-subject (DS) forms. 2. Nathan (1986) presents arguments for and against the "non-configurationality" of Martuthunira. His arguments are based on a statistical evaluation of major constituent orders and NP orders. He concludes that, assuming a continuum between clearly configurational and clearly non-configurational languages, Martuthunira lies somewhere in the middle. However, I disagree with some of Nathan's interpretations of the figures for major constituent order and with his assumption of discontinuous NPs in Martuthunira. By my interpretation, Martuthunira has a more rigid word order and no discontinuous NPs. Thus I see the language as more configurational. 3. By contrast, Nathan (1986) groups such "intruded" NPs together with apparently discontinuous sequences involving major word-class items, usually verbs. By the definition given here, I would treat these latter examples as separate NPs. 4. The same order is maintained within Nathan's (1986) calculations for Martuthunira NPs, including those which are considered to be discontinuous, at a frequency of greater than 90 percent. 5. Historically, the genitive and the accusative descend from a single set of forms. In the modern language their range of functions is complementary.

References Blake, Barry J. 1983. Structure and word order in Kalkatungu: The anatomy of a flat language. Australian Journal of Linguistics 3, 143-176. Blake, Barry J. 1987. Australian Aboriginal Grammar. Sydney: Croom Helm. Dench, Alan C. 1987. Martuthunira: A Language of the Pilbara Region of Western Australia. Ph.D. dissertation, Australian National University. Dench, Alan C., and Nicholas Evans. 1988. Multiple case-marking in Australian languages. Australian Journal of Linguistics 8, 1-47. Dixon, R. M. W. 1980. The Languages of Australia. Cambridge: Cambridge University Press. Dixon, R. M. W. 1981. Warrgamay. In Handbook of Australian Languages, ed. by R. M. W. Dixon and Barry Blake, vol. 2, 1-143. Canberra: Australian National University Press. Donaldson, Tamsin. 1980. Ngiyambaa: The Language of the Wangaybuwan. Cambridge: Cambridge University Press.

Suffixaufnahme

and Apparent Ellipsis in Martuthunira

395

Evans, N. R. D. 1985. Kayardild: The Language of the Bentinck Islanders of North-west Queensland. Ph.D. dissertation, Australian National University. Goddard, Cliff. 1983. A Semantically Oriented Grammar of Yankunytjatjara. Ph.D. dissertation, Australian National University. Hale, Ken. 1983. Warlpiri and the grammar of non-configurational languages. Natural Language and Linguistic Theory 1, 5-47. McGregor, William. 1984. A Grammar of Kuniyanti. Ph.D. dissertation, Sydney University. McGregor, William. 1990. A Functional Grammar of Gooniyandi. Amsterdam: Benjamins. Morphy, Frances. 1983. Djapu, a Yolngu dialect. In Handbook of Australian Languages, ed. by R. M. W. Dixon and Barry J. Blake, vol. 3, 1-188. Canberra: Australian National University Press. Nathan, David. 1986. Topics in configurationality. BA honours thesis, La Trobe University. Plank, Frans. 1990. Suffix copying as a mirror-image phenomenon. Linguistics 28, 1039-1045. Simpson, Jane. 1983. Aspects of Warlpiri Morphology and Syntax. Ph.D. dissertation, Massachusetts Institute of Technology.

15 Multiple Case in Kayardild: Anti-iconic Suffix Ordering and the Diachronic Filter Nick Evans 1. Introduction Kayardild is an Australian language belonging to the Tangkic group of nonPama-Nyungan languages, spoken on Bentinck Island in the southern Gulf of Carpentaria; only about twenty speakers remain. Other members of the group are Lardil, Yukulta, and Yangkaal, the latter two now extinct and the former with just a few very old speakers. As Figure 15.1 shows, the typological division into ergative and non-ergative languages cross-cuts the genetic classification. Yukulta is the most conservative Tangkic language, preserving ergative morphology found in proto-Tangkic; Lardil is typologically similar to Kayardild. Though genetically related to the head-marking non-Pama-Nyungan languages of Arnhem Land, which have only scanty case morphology, the Tangkic languages are typologically similar to such well-known Pama-Nyungan languages as Warlpiri, Dyirbal, and Diyari in being dependent-marking, caseusing languages, and there is evidence of ample lexical and typological influence from Pama-Nyungan languages of North Central Australia. A full grammar of Kayardild, as well a partial reconstruction of protoTangkic morphosyntax, is in Evans (1994), and a Kayardild dictionary with many textual examples is already published (Evans 1992); for information on Yukulta the reader is referred to Keen (1983), and on Lardil to Hale et al. (1981), Hale (1973), and Klokeid (1976). Klokeid (1978), McConvell (1981), and Evans (1994) discuss the evolution of accusative systems in Kayardild, Lardil, and Yangkaal from an original ergative system. Thanks to the good level of grammatical documentation of Lardil, Yukulta, and Kayardild, and the unusual (for Australia) situation in which two of the languages (Kayardild and Lardil) have evolved independently on their own islands, we have been able to reconstruct many of the changes in morphosyntax since proto-Tangkic. But the genetic distance and radical typological disparities between the Tangkic languages and their nearest relatives in Arnhem Land make it unlikely that we will

Multiple Case in Kayardild

397

Figure 15.1. Genetic and Typological Relationships of the Tangkic Languages.

be able to reconstruct much of the morphosyntactic development leading to proto-Tangkic. All the Tangkic languages have extensive multiple case marking, of which Suffixaufnahme—in the canonical sense of a genitive followed by another case that reflects the case of the head—is merely the simplest manifestation. Up to four levels of case marking may appear on nominals, as will be shown for Kayardild in the first part of this chapter. In general, these suffixes occur in the expected order, but there is one situation in which the ordering is antiiconic insofar as an outer case suffix originates in a lower constituent than the suffix immediately inside it. Owing to the complicated and unusual nature of Kayardild morphosyntax, I will go into some detail to show how this problem arises. In the second part of the chapter I discuss the general issue of the sorts of explanation linguistic theory can give for the ordering of case suffixes. I shall argue that pertinent restrictions such as the "mirror principle" (Baker 1985) or the "linearity constraint" (Sadock 1991) are too strong, positing exceptionless laws instead of mere tendencies, and that affix ordering is best explained in diachronic terms as reflecting the order of grammaticization. It is particularly important to understand what diachronic paths can lead to the rare exceptional cases of anti-iconic affix orders: if these paths can occur only in highly marked situations, we can place a "diachronic filter" on the possible exceptions to the mirror principle.

2. The Extent of Multiple Case Marking 2.1. Stacking of Adnominal and Relational Case Canonical examples of Suffixaufnahme in Kayardild, in which nouns inflect not only for their own possessive function but also for the case category

398

Australian

inherited from the possessed NP, are (1) and (2); both the genitive and ablative can indicate possession in this construction. These are comparable to adnominal genitive plus relational case constructions in Hurrian, Old Georgian, Awngi, and so on. Note that I am using the term "relational case" in the sense of Dench and Evans (1988) for cases whose function is to relate either arguments to the verb, or peripheral NPs like location or instrument to the clause as a whole. (1) thabuju-karra-nguni mijil-nguni brother-Gen-Ins net-Ins 'with brother's net' (2) dan-kinaba-nguni dangka-naba-nguni mirra-nguni wangal-nguni this-Abl-Ins man-Abl-Ins good-Ins boomerang-Ins '. . . with this man's good boomerang' Unlike Old Georgian and some other languages discussed in this volume, the further inflection of the possessive NP with its inherited case is obligatory in Kayardild and does not depend on whether the possessor NP is separated from its head; also, all words in an NP inflect for their own and any inherited case. In fact, all the case marking rules discussed in this paper are absolute and always apply regardless of word or phrase order. Although adnominal genitives are cross-linguistically the most common type of participant in Suffixaufnahme constructions, other adnominal constructions are possible in Kayardild. Thus, in (3) an adnominal use of the origin case is followed by a relational proprietive, which is an alternate means of expressing instrument role; in (4) an adnominal proprietive case (indicating 'having') is followed by a "modal locative," in agreement with the head noun 'woman' (as we shall see in Section 2.2, objects of clauses in "actual modality" take the "modal locative"); in (5) an adnominal associative case (here indicating temporary co-location) is followed by a "modal proprietive," in agreement with the head noun 'man', which is assigned the modal proprietive as object of a verb in the "potential" modality; and in (6) an adnominal privative case is followed by a modal proprietive in agreement with the head noun 'place', which as a locational complement of a verb in the potential is assigned the modal proprietive.1 (3) (darirra mardala-a-ja) mutha-wu ngunymurr-u, mutha-wu ngunymurr-u; wuranku, mak-u-n-maan-ju wuran-ku, ngimi-waan-ju wuran-ku (newborn rub-M.Act) much-Prop grease-Prop, much-Prop grease-Prop; stuffProp, bark=torch-Don-Nomin-Orig-Prop stuff-Prop, night-Orig-Prop foodProp 'The newborn child was rubbed with lots of grease, with lots of greasy stuff, with stuff obtained by the light of a bark torch, with food obtained at night'

Multiple Case in Kayardild

399

(4) niya karrngi-ja dun-kuru-ya maku-y 3Sg=Nom keep-Act husband-Prop-M.Loc woman-M.Loc 'He is living with a married (i.e. husband-having) woman' (5) ngada kurri-ju dangka-wu yubuyubu-nurru-wuru 1Sg=Nom see-Pot man-M.Prop road-Ass-M.Prop 'I will see the man on the road' (6) nga-ku-rr-a warra-ju dangka-warri-wu dulk-u 1-Incl-Du-Nom go-Pot person-Priv-M.Prop place-M.Prop 'We two will go to uninhabited places (i.e. to person-without places)'

It is also possible to use the locative case adnominally, though this is subject to morphological sequence restrictions: the locative suffix may combine with a following oblique2 case suffix into the portmanteau {-kurrka} as in (7); the modal oblique is assigned to the objects of verbs in the apprehensive modality. And it may be followed by the nominative, which has a zero exponent here, as in (8). (Objects of imperatives take the nominative.) But it cannot be followed by any other case, and in any adnominal construction where such a banned sequence would arise, such as the unacceptable (9), it is replaced by the associative case, as in (5). (7) kunawuna bilarri-nyarra nguku-ntha wuruman-kurrk child (Nom) spill-Appr water-M.Obl billy-Loc=M.Obl 'The child might spill the water in the billy' (8) bilarri-na dathin-a nguku-wa wurruman-ki- , warra-a ngukuspill-Neg=Imp that-Nom water-Nom billy-Loc-Nom, far-Norn water-Norn3 'Don't spill that water in the billy; the well's a long way off!' (9) *ngada kurri-ju dangka-wu yubuyubu-ya-wuru 1Sg=Nom see-Pot man-M.Prop road-Loc-M.Prop 'I will see the man on the road'

There has been some debate among Australianists about whether such adnominal cases permitting further case marking should be treated as inflectional or derivational; the main reason given for treating them as derivational is that they are non-final (Dixon 1980: 300). I assume they are inflectional on the basis of arguments given in full in Dench and Evans (1988). The crucial points to our argument are: (a) they are fully productive and semantically predictable; (b) they can occur on all words in an NP, as in (2), and their semantic scope is over the NP, not the lexeme; and (c) the existence of other constructions (some of which will be exemplified below) in which clearly inflectional suffixes are non-final removes the need to treat them as a special case to be explained. The relative ordering of adnominal with respect to relational or other case

400

Australian

functions (such as "modal") is always iconic, with the adnominal inside the relational.

2.2. Modal Case: The Use of Case to Signal Mood, Tense, and Aspect Kayardild has three additional, higher levels of case marking that can lead to multiple case inflection. The first, which I term the "modal case level," involves a special use of case suffixes to signal modality, tense, and aspect, working in concert (though to an extent independently) with verbal inflection. Modal case suffixes appear on most non-subject NPs4 outside other adnominal or relational cases. In (10), for example, the modal use of the ablative, which here marks past tense, follows the adnominal genitive and the relational instrumental. There is no accusative case in Kayardild, so the object takes the modal case alone. (10)

ngada yalawu-jarra yakuri-na thabuju-karra-nguni-na mijil-nguni-na I catch-Past fish-M.Abl brother-Gen-Ins-M.Abl net-Ins-M.Abl 'I caught (the) fish with brother's net'

Contrast this with (11), in which the modal use of the proprietive marks future tense; (12), in which the modal use of the oblique marks "emotive" modality; and (13), in which the modal use of the allative marks "directed" mood/ aspect, which may signal either spatial entry into a scene or inception of an event. (11)

ngada yalawu-ju yakuri-wu thabuju-karra-ngun-u mijil-ngun-u I catch-Pot fish-M.Prop brother-Gen-Ins-M.Prop net-Ins-M.Prop 'I will catch (the) fish with brother's net'

(12)

ngada yalawu-nyarra yakuri-nja thabuju-karra-nguni-nja mijil-nguni-nj I catch-Appr fish-M.Obl brother-Gen-Ins-M.Obl net-Ins-M.Obl 'I might catch (the) fish with brother's net (which would be a bad thing)'

(13)

dathin-a dangka-a jirrkaan-da warra-jir, diya-jir wuran-kir that-Nom man-Norn from=north-Nom go-Directed eat-Directed food-M.All That man is coming hither from the north eating food'; 'That man is beginning to eat food, going/coming from the north'

The two other modal cases in Kayardild are the modal locative, used for "instantiated" modality—i.e., events that are occurring or have occurred, as in (14)—and the "zero" modality, confined to the imperative (15) and to ongoing nominalizations (see below). NPs assigned no other case will receive the nominative if occurring in a clause with zero modal case; in general, the nominative is an elsewhere case assigned to NPs that receive no non-zero case at relational level or above.

Multiple Case in Kayardild

401

(14) birangkarra bi-l-da mardala-tha dangka-walath-i, ngimi-marra-y long=time 3-Pl-Nom paint-Act man-many-M.Loc, night-Util-M.Loc 'They have been painting the men for a long time, getting ready for (the dance) tonight' (15) dathin-a ngunguk-a balmbi-marra karrngi-j! that-Nom story-Norn morrow-Util keep-Imp 'Save that story for tomorrow!' "Normal," non-modal uses of the ablative and proprietive cases have already been shown. (16) illustrates a normal use of the oblique,5 (17) a normal use of the allative, and (18) a normal use of the locative. (16) dathin-a dangka-a mulurr-a niwan-inja maku-nth that-Nom man-Nom jealous-Norn his-Obl wife-Obi 'That man is jealous of his wife' (17) ngada warra-ju ngarn-kiring-ku 1Sg=Nom go-Pot beach-All-M.Prop 'I will go to the beach' (18) dathin-ki mijil-i mutha-wuru malji-wuru that-Loc net-Loc many-Prop hole-Prop 'In that net there are many holes' The examples given above represent common combinations of verbal tense/mood categories with modal case, and they might give the impression that the modal case shows some sort of "agreement" with the categories marked on the verb. While this analysis would be satisfactory for the majority of Kayardild sentences, a more exhaustive analysis shows that modal case can be used independently of verbal inflection. One reason for saying this is that verbless clauses can still show tense/mood by the choice of modal case, as in (19). (19) ngada dathin-kiring-ku kamarr-iring-ku lSg=Nom that-All-M.Prop stone-All-M.Prop 'I will (go) to that stone' Such examples could always be attributed to ellipsis of the verb, however; a stronger reason for seeing modal case as independent of verb inflection is that the same verbal category may combine with more than one modal case, according to the exact sense of the verb inflection. For example, the verbal "potential" inflection is polysemous between ability and futurity. Combining it with the modal proprietive (20) preserves this ambiguity, while combining it with the modal locative (21) eliminates the future reading, forcing an interpretation of "ability in the past." Similarly, the apprehensive inflection on verbs,

402

Australian

which expresses the undesirability of the clausal event, combines with the modal oblique when giving warnings, with the modal proprietive when issuing threats, and with the modal locative when commenting on undesirable events that are already occurring (not illustrated here). (20) ngada kurri-nangku mala-wu (balmbi-wu) 1Sg=Nom see-Neg=Pot sea-M.Prop morrow-M.Prop 'I won't / won't be able to see the sea tomorrow' (21)

ngada kurri-nangku mala-y (barruntha-y) 1Sg=Nom see-Neg=Pot see-M.Loc yesterday-M.Loc 'I couldn't see the sea yesterday'

For these reasons I treat the modal case system as a system distinct and independent from the system of verbal tense/mood inflections, even though the needs of semantic compatibility make some combinations particularly common. Table 15.1 summarizes the semantic categories encoded by modal case, and the verbal categories with which particular modal cases usually co-occur. Those given in brackets are grammatical, but less common. Here and elsewhere braces enclose the canonical forms of morphemes; most Kayardild case morphemes have a large number of allomorphs which I do not wish to discuss here. To complete this section, I would like to discuss briefly the historical origins of modal case, as this will help make sense of this typologically unusual system and will also serve as a backdrop to our eventual goal of understanding the relative ordering of modal case and other suffixes. Table 15.1. Modal Case Semantics and Verbal Tense/Mood in Kayardild. modal case Locative {-kiya}

Proprietive {-kuru} Ablative {-kinaba} Oblique {-kiya} Allative {-kiring} Zero

semantic category Instantiated

Future Prior Emotive Inceptive —

corresponding verbal categories Actual (Affirmative & Negative) Immediate [Potential (Affirmative & Negative), giving "actual ability" meaning] [Apprehensive, giving "actual undesirable" meaning] Potential (Affirmative & Negative) [Apprehensive, giving "future undesirable" meaning] Past, Almost, Precondition Apprehensive Desiderative Hortative (Affirmative & Negative) Directed Imperative (Affirmative & Negative) Continuative Nominalization

Multiple Case in Kayardild

403

Modal case appears to have arisen through a blending of two construction types in proto-Tangkic, both preserved in Yukulta, the most conservative member of the Tangkic group. Although the unmarked transitive clause had ergative-absolutive case markings on nominals (22), there were a number of detransitivized constructions that assigned a nominative-oblique case frame, in which the object took the proprietive (23), dative (24), or ergative/locative (25) case. These detransitivized constructions were triggered by several conditions: irrealis (23), negative (24), and certain "inverse" person combinations of subject and object (24, 25). (Note that Yukulta has an auxiliary cliticized to the first constituent, consisting of subject and object or indirect object clitics as well as tense; some of these are fused into portmanteaux, and morpheme boundaries within the auxiliary will not be shown here.6) As these examples illustrate, more than one such condition often occurred at once; the reader is referred to Keen (1983) for further examples. (22)

dangka-ya+karri ngawu pala-tha man-Erg/Loc+3/3Pres dog (Abs) hit-Ind 'The man is hitting the dog'

(23)

diya-da+ngarri kuruntha-wurlu eat-Des+l/3Pres barramundi-Prop 'I would like a feed of barramundi'

(24) ngarr-ma+ngarra kuli-nada kuluthu-ntha 1Du-Stat+lPres wash-Neg=Des(Irr) clothes-Dat 'We don't want to wash the clothes' (25) walirra+ngalawa-ningki bala ngalawan-ji burldamurr-i Neg+lPlObl-3SgFut hit 1Pl-Erg/Loc three-Erg/Loc 'He didn't hit us three'

The other construction type involved subordinate clauses that were complementized by cases indicating temporal or modal relations between the two clauses—what Dench and Evans (1988: 18) call "T-complementizing case." Again I shall use Yukulta examples, which can be assumed to preserve the proto-Tangkic constructions. In such subordinate clauses the subjects were typically omitted under coreference with the main clause, and both verbs and NPs in the subordinate clause took T-complementizing case. In the case of verbs, these directly followed the conjugation marker (here glossed 'Theme'); with nouns they attached directly to the stems of objects (26, 27) and locative (31) NPs, but followed other non-object cases such as the allative in (27). (26)

kurri-ja+ngarri, [murruku-ya mirrala-th-i]T Loc see-Ind+lSgPres, [woomera-Loc make-Theme-Loc] 'I'm watching him making a woomera'

404

Australian

(27) taami-ja+nganti [natha-rlu-ngkurlu warra-j-urlu]T Prop ask-Ind+lSgFut [camp-All-Prop go-Theme-Prop] 'I'll ask him to come to the camp' (28) nganh-ma+kati warra [kurlkang-kirlu jani-jani-j-irlu kuluu-kuluu-j-irlu]T All 1Sg-Stat+lSgPres go [bulrush-All search-Redup-Theme-AH dig-RedupTheme-All] 'I'm going to look for some bulrushes to dig up'

In the above examples the T-complementizing case has the same form (apart from predictable morphophonemic alterations) whether it attaches to a noun or a verb. But there are other subordinate clause types where the verb and noun take different suffixes. In Yukulta verbal forms like those in (29) and (30) can no longer be analysed synchronically, but comparison with Kayardild and Lardil suggests that in proto-Tangkic they followed the regular pattern of conjugation marker plus case suffix (added in parentheses in the glosses), except that noun and verb carried different though semantically related cases. (29) dangka-ya+kanta kurri-ja maku [kunawuna-naba jambila-th-arrba] man-Erg+3/3Past see-Ind woman (Abs) [child-Abl kick-Prior-Theme (-Conseq)] 'The man saw the woman (who had already started) kicking the child' (30) yarrarama-ja+lati [jirrma-ny.marra kunawuna-ntha] whisper-Ind+3PlPres [wake-Appr=Theme(-Util) child-Dat] 'They are whispering, lest they wake the child'

In addition to appearing on verbs and objects, T-complementizing case appears in Yukulta, and probably appeared in proto-Tangkic, on allative NPs (27), locative NPs (31), and time NPs (32); in fact, from the limited data it seems to appear on all NPs except subjects. (31) kurri-ja+nganda [kaban-jinaba jawi-jarrba] see-Ind+1Sg/2SgPast [sand-T.Abl run-Prior] 'I saw you running on the sand' (32) birlkali-ja-t-wuluwaka [yulmburr-inaba-ntha thaa-tharrba-ntha] feel=sorry-Ind+ 1Sg/3PlPres [long-T.Abl-C.Dat return-Prior-C.Dat] 'I feel sorry for them having had to walk so far'

Between proto-Tangkic and Kayardild, Lardil, and Yangkaal, a process of what I have called "insubordination" occurred: subordinate construction types were generalized to main clauses (though they continued the subordinate use as well). In the process, the semantics of the tense/mood expressed by T-complementizing case shifted: instead of being relative, between two clauses (e.g., subordinate clause at the same time as main clause), it could now be absolute, between the clause and the speech act (e.g., main clause at

Multiple Case in Kayardild

405

the same time as the speech act). It is this use to signal absolute tense/mood that I refer to as "modal case." As regards the domain of modal case marking, the insular Tangkic languages continued the basic pattern found in the subordinate clause construction, in which case was marked on all NPs except the subjects. And the unhooking of the case system on NPs from the verbal inflections themselves continued further to the point where modal case on NPs in Kayardild is to an extent independent of verb inflection. I will not discuss the reasons here in detail, but the process of insubordination seems to have been functionally motivated by the highly marked asymmetry between main clauses, which had a limited set of tense/mood marking possibilities, and subordinate clauses, for which the T-complementizing system had furnished a rich system of relative tense/mood distinctions. The existence of a number of case choices for main clause objects in protoTangkic, partly dependent on tense and mood choices, probably provided a template for the marking of tense and mood categories on main clause NPs.7 The purpose of this historical digression into the origins of modal case has been twofold. First, it is clear that, historically, modal case had clausal scope (in its older guise as T-complementizing case). Second, the position of modal case in the morphological structure of nominal words has not changed since proto-Tangkic: it followed oblique relational cases and was attached directly to object NPs.

2.3. Associating Case on Non-subject Arguments ofNominalized Verbs A suffix -n in all the Tangkic languages derives nominalizations from verbal stems; these then inflect for case, and in their citation form take the nominative case (-da in this environment in Kayardild). In all the Tangkic languages, as I show in Section 3.4, such nominalizations are a source of new lexical nouns denoting agents, names of actions or states, instruments, or locales; often the nominalization is compounded with a noun stem. Some Kayardild examples are rajurri-n-da 'toddler' from rajurri-ja 'walk around'; mijil-burldi-n-da 'spider' (lit. 'net-caster') from mijil-da 'nest' and burldl-ja 'cast, throw'; banga-ra-yii-n-da 'turtle spear' from banga-a 'turtle' and rayii-ja, the middle form of raaja 'spear'; kunawuna-wirrka-a-n-da 'initiation ground' from kunawuna 'youth' and wlrrka-a-ja, the middle form of wirrkaja 'to initiate'; and wirrka-n-da 'dance (noun)' from wirrka-ja 'dance' (another sense of this verb, this time with a monovalent argument structure). In Kayardild and Yangkaal, alone among the Tangkic languages, the nominalized form has extended its function and can be used as a participle. Unlike the original function as a lexical nominalizer, the new participial function is completely productive and can govern external arguments and adjuncts. In Kayardild, participial nominalizations may modify head nouns—for example, in complements of perception verbs (33)—and they may be used as the sole verb of a main clause to mark ongoing aspect (34). In either case, it is possible

406

Australian

for the participle to govern non-subject NPs, which are marked with an outer oblique suffix: essentially, this suffix associates these NPs with the nominalized verb which heads their clause. It seems likely that this "associating function" arose through a reinterpretation of sentences like (33) from 'Look at those women, diggers for yams!' to 'Look at those women digging up yams!'. (33)

kurri-ja dathin-a maku-walad-a [dalwani-n-da thawal-inj]! look-Imp that-Nom woman-many-Nom [dig-Nomin-Nom yam-A.Obl] 'Look at those women digging up yams!'

(34)

ngada yalawu-n-da yakuri-nja thabuju-karra-nguni-nja mijil-nguni-nj I catch-Nomin-Nom fish-A.Obl brother-Gen-Ins-A.Obl net-Ins-A.Obl 'I am catching fish with brother's net'

In both these examples no modal case is marked in the participial clause: in (33) this is because the modifier noun 'women', as object of an imperative, takes no modal case and is assigned the "elsewhere" nominative, while in (34) it is because nominalized main clauses take no modal case. The situation becomes more complicated when participial clauses inherit modal case from a higher clause; the resultant morphological complexities will be discussed in Section 2.6.

2.4. Complementizing Case: The Use of Case to Mark Interclausal Relations on Complementized Clauses What I call "complementizing" case marks clauses as (a) complements of a higher clause, as in (35a), or (b) as having the shared NP departing from the unmarked sequence where it would be the subject in both main and subordinate clause, as in (35b),8 by inflecting every word of the lower clause with an oblique or locative if this is morphologically possible (depending on the person of the subject—basically, locative if the subject includes the addressee, oblique otherwise); the complementizing case may follow modal, relational, and adnominal case markers. (35)

a. ngada mungurru, [maku-ntha yalawu-jarra-ntha yakuri-naa-ntha thabujukarra-nguni-naa-ntha mijil-nguni-naa-nth] I know, [woman-C.Obl catch-Past-C.Obl fish-M.Abl-C.Obl brother-GenIns-M.Abl-C.Oblnet-Ins-M.Abl-C.Obl] 'I know that the woman caught the fish with brother's net' b. jina-a bijarrb, [nga-ku-1-da bakiin-ki kurulu-tharra-y?] c Loc where-Nom dugong (Nom), [1-Incl-Pl-Nom all-C.Loc kill-Past-C.Loc] 'Where is the dugong, which we (i.e. we and you) all killed?'

In (35a) the subordinate clause is an object argument of mungurru; its subject is 3rd person, conditioning the choice of the oblique as complementizing case. In (35b) the common NP, 'dugong', is subject of the main clause but object of

Multiple Case in Kayardild

407

the subordinate clause, triggering complementizing case here; the subordinate subject 'we inclusive' includes the addressee, requiring the locative to be chosen as complementizing case. Note that the pronoun ngakulda here escapes the complementizing locative; this is because historically ngakulda, the ergative, from which the complementizing locative derives (see below), did not have a distinct form for pronouns, even though there is a distinct locative form in both Kayardild and proto-Tangkic. It is common in Kayardild for "complementized clauses" to be insubordinated, which results historically from ellipsis of the main clause. Such insubordinated clauses have a variety of uses which need not concern us here (see Evans 1988, 1994); but they preserve the four layers of case marking seen in (35a). An example is (36); insubordination here has the semantic effect of marking a statement based on inference about a past event. (36) [maku-ntha yalawu-jarra-ntha yakuri-naa-ntha thabuju-karra-nguni-naa-ntha mijil-nguni-naa-nth] [woman-C.Obl catch-Past-C.Obl fish-M.Abl-C.Obl brother-Gen-Ins-M.AblC.Obl net-Ins-M.Abl-C.Obl] 'The woman caught the fish with brother's net' As with modal case, we have evidence that the morphological pattern of complementizing case has, in its essentials, continued unchanged from protoTangkic, despite drastic changes in function in Kayardild and Lardil which seem to have stemmed from their abandonment of ergative constructions and a consequent unhooking of the new system of main clause case from the system inherited for subordinate clauses. Again Yukulta appears to have preserved the proto-Tangkic pattern: subordinate clauses could agree in case with antecedents in the ergative/locative (37) or the dative (38); in (38) the dative agrees not with an overt antecedent NP, but with the 2nd person indirect object clitic. (37) dangka-ya+karri ngid-a karna-ja [[makurrarra-wurlu-ya karna-j-urluya]T.Prop]c.Erg

man-Erg+3/3Pres wood-Abs light-Ind [[wallaby-Prop-Erg cook-Theme-PropErg]] 'The man lit a fire, in order to cook the wallaby' (38) birlkali-ja+baka [[miyarl-inaba-ntha darla-tharrba-ntha]T Abl]c Dat sorry=for-Ind+lSg/2IObj [[spear-Abl-Dat break-Prior-Dat]] 'I'm sorry for you breaking your spear' The ergative/locative is the etymological source of the Kayardild locative, and the dative of the Kayardild oblique. Where the antecedent was absolutive there was no marking on the subordinate clause. C-complementizing case followed T-complementizing case and, being inherited from the higher clause, naturally occurred outside it—an iconic ordering. The limitation of anteced-

408

Australian

ent agreement in proto-Tangkic to non-zero core cases, namely ergative and dative, explains the later limitation in Kayardild of complementizing cases to their reflexes, the locative and oblique.

2.5. Summary of Case Functions and Positions Figure 15.2 relates the morphological structure of the Kayardild nominal word to the syntactic source of the various types of case. I assume that associating and modal case are assigned at S level, and complementizing case at S' (sentence plus complementizer) level.9 The columns show the case options at each SYNTACTIC

(Locative) Ablative Proprietive

Genitive Associative Origin Privative Consequential

Oblique Allative Instrumental Utilitive (Genitive) (Associative) (Origin) Privative

STRUCTURE

Locative Locative Ablative Proprietive Oblique Oblique Allative

Locative Ablative Proprietive Oblique

Verbal Dative Verbal Allative Verbal Translative Verbal Evitative Verbal Donative Verbal Purposive Nominative MORPHOLOGICAL

STRUCTURE

Figure 15.2. Functional Levels of Kayardild Cases.

Multiple Case in Kayardild

409

level—adnominal, relational, modal, associating, and complementizing—and each case level is associated with its corresponding syntactic level. Note that the nominative case is shown as opposed to all levels except the adnominal: this is because the nominative appears only when no case has been assigned at any of these levels. 2.6. Recursive Ordering Within a Single Functional Level In addition to examples like the above, where each case suffix has a different functional level, one occasionally finds instances of recursion within a particular level. This is attested for adnominal, relational, and modal functions. An example of recursion of adnominal case function, here represented by the associative (here encoding 'having') and ablative (here encoding 'belonging to'), is (39). (39)

. . . maku-yarr-nurru-naba-walad . . . woman-two-Ass-Abl-many 'the many belonging to (those) having two wives'

Although speakers consider such words acceptable (albeit clumsy), they are not a normal feature of spontaneous Kayardild, and my only examples like (39) are elicited. The only construction with adnominal recursion that occurs with significant frequency in spontaneous texts involves the use of the genitive or ablative of possession on possessive pronouns, which is of course found in many languages lacking canonical Suffixaufnahme (Moravcsik, this volume). An example is (40). (40)

burdumbanyi waydbala raba-nangku ngijin-jinaba-wu ngarriju-naba-wu jardinaba-wu dulk-u ignorant (Nom) white=man (Nom) trespass=on-Neg=Pot my-Abl-M.Prop mother's=mother-Abl-M.Prop mob-Abl-M.Prop country-M.Prop 'The ignorant white man must not trespass on the country of all my mother's mothers (that came down to me from them)'

Recursion of case suffixes in relational or modal function, however, does occur in spontaneous texts. An example of recursion of case suffixes in relational function is (41). (41)

(darirra mardala-a-ja) mutha-wu ngunymurr-u, mutha-wu ngunymurr-u wuranku, mak-un-maan-ju wuran-ku, ngimi-waan-ju wuran-ku, kurdala-thirrin-ju ngimi-wan-jinaba-wu kanthathu-naba-wu (newborn rub-Mid-Act) much-Prop grease-Prop, much-Prop grease-Prop foodProp, torch-Don-Orig-Prop food-Prop, dark-Orig-Prop food-Prop, spearRes-Prop night-Orig-Abl-Prop father-Abl-Prop

410

Australian '(The newborn was rubbed) with lots of grease, lots of greasy food, with food (speared) by (the light of) a bark torch, with food (speared) at night-time, speared by (the baby's) father at night-time'

Here the relational ablative case,10 assigned to 'father' as demoted subject of the resultative nominalization kurdalathirrin-, is followed by the relational proprietive case, assigned to the head noun wuran- in the main clause. Both relational cases are then percolated down to the adnominal NP ngimiwan'(hunting) at night'. Such stacking of relational cases is possible only when the nominalized nature of the subordinate clause allows the matrix relational case to be inherited. Recursion of modal cases does not normally occur, since clauses assigning their own modal case are finite, and finite clauses do not normally inherit modal case from higher clauses. However, directed clauses, which assign the modal allative, are an exception: when used as subordinate clauses giving 'movement with a purpose', they inherit the matrix clause's modal case, resulting in two modal case suffixes on subordinate NPs, as in (42). Again, case ordering is iconic here. (42) balmb-u ngada warra-ju [bijarrba-ring-ku raa-jiring-ku]M Prop morrow-M.Prop lSg=Nom go-Fut [dugong-M.All-M.Prop spear-DirectedM.Prop] 'Tomorrow I will go to spear dugong'

In general, then, Kayardild grammar displays "unbounded concord" whereby all cases percolate downwards, not just to immediate constituents but on to their subconstituents, until the unit "word" is reached. However, there are morphologically imposed limits to the syntactic possibilities of recursion: to be grammatical, a construction must allow a correspondence with the morphological structure outlined so far. There can normally be only one exponent of each of the relational, modal, and complementizing levels. Should the syntax require more—as with (41) and (42) above—one of the exponents must be "stolen" from another level where it is not being used. For example, (41) is acceptable because the relational proprietive can be "stolen" from the modal level, since the sentence is in the unmarked modal case and the modal slot is therefore unused. And (42) is acceptable because the inner modal allative is "stolen" from the relational level; no relational case is being used in the subordinate clause. There are two additional morphological restrictions as well. The first constraint, which we saw in Section 2.1, states that the locative can be followed only by the oblique, with which it forms the portmanteau {-kurrka}.11 The second constraint, which will play a crucial part in understanding the anti-iconic ordering to be discussed below, states that the oblique can never be followed by another case. These morphological structure limitations, together with rules of association between possible morphological and syntactic representations, serve as a filter on possible syntactic structures.

Multiple Case in Kayardild

411

3. Explaining Suffix Ordering

3.1. Morphology Mirrors Syntax As the above examples show, and as was observed on a wider Australian basis by Dench and Evans (1988) and for Suffixaufnahme languages in general by Plank (1990), the ordering of multiple case suffixes is normally iconic insofar as outer suffixes register categories of higher constituents. A number of accounts attempt to build such behavior into universal grammar: in Baker's (1985) approach based on Government and Binding theory, inflections (Baker's own focus being on verbal ones) are co-generated with syntactic rules at the appropriate level; in Andrews's (1991) proposal of "inside-out unification" of inflectional features within a Lexical Functional Grammar framework, unification of morphosyntactic features proceeds morphologically by working outward from the stem, and syntactically by working outward from the terminal node of constituent structure to successively more remote layers of constituency. The most recent and best-worked-out statement of this principle is within the theory of Autolexical Syntax in Sadock (1991: 103), which sees independently generated syntactic and morphological modules, linked by words that are both syntactic and morphological entities and are subject to certain association constraints between the two modules. Relevant here is the "linearity constraint," stated in a strong (43a) and a weak form (43b): (43)

a. The associated elements of morphological and syntactic representations must occur in the same linear order, b. The associated elements of morphological and syntactic representations must occur in as close to the same linear order as the morphological requirements allow.

In fact, this formulation is not quite adequate for the Kayardild facts, since dominance rather than order is involved; a better formulation would be something like "given two affixes A and B, such that A is more peripheral than B morphologically, the syntactic element associated with A must dominate that associated with B." However, our main concern here is not the exact formulation of such rules, but their status: are they part of universal grammar or merely general tendencies?

3.2. The Nominalized Construction and Anti-iconic Case Ordering Now there is one construction in Kayardild that is incompatible with the strong version of the linearity constraint and favors its weak version, and it is this construction which forms the crux of the rest of this chapter. When "modal" and "associating" affixes co-occur in a participial construction modifying a non-nominative head, they are ordered anti-iconically: as (44) and (45) illustrate, the associating oblique case is outer even though it originates in the

412

Australian

nearer constituent. We know this because the associating oblique arises from the nominalization of the verb in its own clause, while the presence and value of the modal case is determined by the tense/mood of the main clause. (44)

ngada kurri-jarra niwan-jina [kurdama-n-kina nguku-naa-ntha] wuruman-urrunaa-nth lSg=Nom see-Past him-M.Abl [drink-Nomin-M. Abl water-M.Abl-A.Obl] billycan- Ass-M. Abl-A. Obl 'I saw him drinking the water in the billycan'

(45)

ngada balmbi-wu kurri-ju kilwan-ju [barrki-n-ku kurda-wuu-nth] lSg=Nom morrow-M.Prop watch-Fut them-M.Prop [chop-Nomin-M.Prop coolamon-M. Prop-A .Obl] 'Tomorrow I will watch them chopping (making) a coolamon'

The way in which the anti-iconic suffix ordering in (45) violates the strong version of the linearity constraint, when the lines associating the two outer morphological elements of kurdawuuntha with their syntactic sources cross, is shown in Figure 15.3. Only those syntactic positions relevant to the assignment of case to kurdawuuntha are shown. This contrasts with the behavior of resultative nominalizations (41, 46). Here the demoted subject is assigned the ablative by the nominalization; this appears inside any relational (41) or modal (46) case inherited from a higher clause:12 (46)

Wulkatharri bathin-ki durrwaa-ja ngardarrji ngakinda murrukurnangk, bathinki bal-umban-ji, niwan-jiyarrng-kinaba-ya jibarna-yarrng-kinaba-ya birdiruthirrin-ji [place] from=west-M.Loc chase-Act pregnant=dugong (M.Loc) our=Nom [triangular kin term], from=west-M.Loc west-Orig-M.Loc, his-two-AblM.Loc uncle=in=law-two-Abl-M.Loc miss-Res-M.Loc 'At Wulkatharri the one who was your father and my second cousin chased a pregnant dugong from the west, one from the west that had been missed by his two uncles-in-law'

Figure 15.3. Anti-iconic Suffix Ordering in Participial Construction.

Multiple Case in Kayardild

413

Figure 15.4. Iconic Suffix Ordering in Resultative Nominalization.

This permits the morphological and syntactic representations to occur in the same linear order, as shown in Figure 15.4. Given the existence of such instances—admittedly rare cross-linguistically—of anti-iconic orderings of case, we must seek an account of affix ordering that can include both (a) the predominance of iconic ordering and also (b) the possibility of non-iconic ordering.

3.3. Constraining Anti-iconic Ordering The existence of the anti-iconic orderings discussed above focuses our attention on the weak form of Sadock's linearity constraint (43b), requiring the associated elements of morphological and syntactic representations to occur in as close to the same linear order as the morphological requirements allow. This is a statement about the possible relations between syntactic and morphological components of grammar. But for it to have any constraining power at all, we need to set some bound on what are possible "morphological requirements." The condition "as close to the same linear order as the morphological requirements allow" will be so open as to be vacuous unless we can show that there are strict limits on what deviations can occur. I submit that the task of limiting these deviations should be left to a "diachronic filter," setting a limit on possible paths of grammaticization that would engender deviations from iconic ordering, while favoring the development of iconic ordering; it should also explain why case ordering is the way it is through reference to possible paths of development. Since this type of explanation constrains the set of possible synchronic systems by attributing the lack of certain possible types to the absence of any diachronic path that could engender them, I refer to it as a "diachronic filter" argument.13 In the case at hand, it focuses attention on the processes by which antiiconic versions of multiple case marking arise. However, it will be helpful to see the problem in a wider perspective, and for this purpose I will extend my use of "anti-iconic" to include—alongside cases in which the order of affixes does not mirror the hierarchy of syntactic constituents with which they are

414

Australian

associated—cases where the morphological ordering of affixes does not reflect their semantic scope. To be empirically substantive, diachronic filter models must be based on broad studies of the development of affix orders. To my knowledge there has as yet been no cross-linguistic survey of the diachronic paths engendering antiiconic orderings. For Uralic, however, there is an extensive literature on the evolution of the different orderings Noun-Case-Possessive (as in Finnish ystdva-lle-ni friend-to-my 'to my friend') and Noun-Possessive-Case (as in its Hungarian equivalent barat-om-nak friend-my-to).14 In considering the question of which order is original in Uralic, Comrie (1976) formulates four principles, essentially diachronic in nature, that determine suffix ordering. They can be paraphrased as follows: (a) Other things being equal, the order of suffixes will reflect the order in which the given affixes developed from separate words, i.e., the order of grammaticization. (b) Where there is an ordering discrepancy between old and new affixes within the same semantic domain, their positions relative to other affixes will tend to influence one another (i.e., old affixes may over time move into the new affix position, or new affixes may move into the old affix position). (c) Affix categories that have been partly neutralized may reestablish a fuller set of distinctions, even if this means using an order of affixes different from that which would result from (a) and (b). (d) Metanalysis (i.e. analogical remodeling) may result in a reinterpretation of an existing order of segments. Note that each of these principles is compatible with both iconic and antiiconic orderings. Without some further principle that more "relevant" or "syntactically proximate" categories must be grammaticized first,15 principle (a) is neutral as to what order particular grammaticization events occur in, and therefore what the affix order will be. Principle (b), allowing the mutual influence of semantically related inner and outer suffixes to move material in either direction, is also compatible with both orderings. Principle (c) places the actual order entirely in the hands of morphophonemic chance: what new, outer material is developed to restore a full set of categorial contrasts depends on which inner material has been eliminated by morphophonemic rules, although it would of course be possible to add a stipulation along the lines of Nichols (1973: 235) that "an overall switch in order without other conditioning factors would presumably go in the direction dictated by semantic considerations rather than in the opposite direction." Principle (d), since it does not stipulate what sorts of reinterpretations are possible, is compatible with both iconic and anti-iconic orderings. In the remainder of this chapter I would like to extend the study of possible diachronic sources of anti-iconic suffix orderings by examining, from a diachronic perspective, the three relevant Australian cases known to me; each of these illustrates a different diachronic process. An example of anti-iconic ordering of case and possessor suffixes that resembles the Uralic case comes from the Kulin group of languages, spoken in

Multiple Case in Kayardild

415

Western Victoria and documented by Luise Hercus through her work with the last speakers in the 1960s and 1970s (Hercus 1986). In Wemba-Wemba (47) and Madimadi (48), the two Kulin languages for which we have adequate morphological information, nouns could take a possessive pronominal suffix after the case suffix; there is some indication that another Kulin language, Wergaia, had the same pattern (49). (47)

a. lan-ad-ug camp-Obl-3SgPoss 'in his camp' b. lan-a- in camp-Obl-2SgPoss 'into your camp'

(48)

a. dali-ngu-rin tongue-Oper-2SgPoss 'in your language' b. dina- gu-ru foot-Oper-3SgPoss 'by his foot'

(49)

mum-r-in bottom-on-2SgPoss 'on your bottom'

Hercus (1986: 125) notes two exceptions to the above ordering in Madimadi: 'our father, God' has become a fixed locution and "was felt to be monomorphemic and case suffixes were therefore simply added," giving the order Root-Possessor-Case as in mam-ura-l father-our-locative for 'to God, with God' rather than the expected *mam-al-ura; a similar reinterpretation is recorded with the noun for 'camp'. These reanalyses, apparently based on analogical remodeling of the stem, give an iconic ordering. As yet we have no comparative reconstruction of Kulin morphology, and the limited information we have on the languages will prevent us from assembling a full account. However, there is reasonable evidence that the anti-iconic ordering here resulted from possessive suffixation developing well after case suffixation. First, at least two Kulin case suffixes appear to be inherited from their Pama-Nyungan ancestor: the "operative," with its two allomorphs -u (after consonants) and -gu (after vowels) in Wembawemba and Wergaia and its three allomorphs -u, -ru, and -rjgu in Madimadi, almost certainly continues the proto-Pama-Nyungan ergative - gu ~ -lu; the "oblique," with its allomorphs -a (after consonants) and -ga (after vowels) in all three languages, is a likely reflex of proto-Pama-Nyungan locative - ga —la16; the locative -al may also derive from the proto-Pama-Nyungan locative via a morphemic split. Second, the system of possessive pronominal suffixes, by contrast, appears to be an innovation at the level of proto-Kulin: the forms are cognate within Kulin but to my knowledge do not have any non-Kulin reflexes.

416

Australian

The order Root-Case-Possessive, then appears to reflect the diachronic order of suffixation, with a very old case system, at least in part, inherited from proto-Pama-Nyungan, inside a more recently developed system of pronominal possessives. It is likely that the pronominal possessor suffixes developed via cliticization of possessive pronominals, either to the first constituent or to the NP itself; possessive pronominal clitics to the first constituent are found in Ngiyambaa (Donaldson 1980), spoken just to the north of the Kulin languages, and possessor suffixes that are clearly reduced forms of possessive pronouns are found in Bagandji, another nearby language (Hercus 1982). While the Kulin case conforms to one of Comrie's diachronic principles, the second case we shall examine—the ordering of aspect and subject affixes in Jingili (Chadwick 1978)—necessitates a further possible explanatory principle: anti-iconic orderings can arise through the morphological merger of two words with different orders of affix and root. In Jingili, verbs have the structure Root-Subject-Aspect, as seen in (50). (50) wambaya-nga-(w)adi talk-I-Habitual 'I always talk'

As noted by Blake (1990: 55),17 this order is typologically marked, violating the order predicted by Bybee's notion of relevance (1985), namely RootAspect-Subject. Work by Chadwick (1983) and Blake (1990) on the historical development of these languages suggests that Jingili has, under areal influence, switched relatively recently from being a prefixing to an almost exclusively suffixing language, and that the form wambayangaadi arose in the following way. First nga-adi, which was originally a full verb with pronominal prefixes, was reduced to the status of auxiliary in Wackernagel's position, probably under areal influence from neighbors such as Mudburra which employ second-position auxiliaries; the closely related language Wambaya still has second-position auxiliaries (Nordlinger 1993). Then it further reduced from auxiliary to verbal suffix, again under the areal influence of neighbors like Mudburra, which are entirely suffixing. The Jingili case thus shows a further possible diachronic source of antiiconic ordering: suffixation of an erstwhile prefix following a shift in preferred direction of affixation.18 We now return to the Kayardild case, which exemplifies a third and more complicated diachronic path: the development of new syntactic functions for existing morphology, resulting in an originally iconic suffix ordering becoming anti-iconic in its new function.

3.4. The Evolution of the Tangkic Nominalized Construction and Kayardild Associating Case We have seen that anti-iconic case sequences in Kayardild are confined to the associating use of the oblique in nominalized clauses. I shall argue in this

Multiple Case in Kayardild

417

section that the relevant construction is a recent development, at the level of Kayardild or perhaps proto-South Wellesley, and that it resulted from the syntactic extension of existing, fixed morphological possibilities. There is evidence that the use of a nominalizer -n attached to the verb root to derive new nominal lexemes goes back to proto-Tangkic. Such uses are attested in Kayardild, Yangkaal, Yukulta, and Lardil. In both Yukulta and Lardil, however, the nominalizer is confined to this noun-deriving function: verbs suffixed with -n cannot take external NP arguments and have limited and lexically determined semantics. This suggests that the development of its syntactic uses in Kayardild, as described in Sections 2.3 and 3.2, is a recent one.19 Particularly relevant to the question of why the associating oblique is ordered anti-iconically in Kayardild is the fact that the syntactic uses of the nominalized form, both as a participle and to signal continuous aspect, developed after the morphological constraints on the use of the case marker {-inja} had been established. In Yukulta, -n is derivational and non-productive: it derives agentive nominalizations, stative adjectives, instrument nominalizations, and names of actions from verb stems. As Keen (1983: 233) observes, "a noun may be derived from a verb by the addition of the stem-forming suffix -n to the verb; the derived nominal stem then takes normal case inflection"—e.g. kapanta 'hunter' (Abs) from kapa 'find' (transitive verb) and -n. An exhaustive search of Keen's Yukulta vocabulary reveals the following types: (51)

a. agentive nominalizations rlajurrinda 'toddler' < rlajurrija 'learn to walk, toddle' b. instrument nominalizations mirlanda 'lasso, headrope' < mirlaja 'return' c. derived stative adjectives kalkanda 'sick' < kalkatha 'feel sick'; ngamanda 'hungry' < ngamaja 'be hungry'; burlbanda 'full (with tucker)' < burlbaja 'feel full (with food or drink), swell'; rdulanda 'fat (person)' < rdulaja 'get fat' d. names of actions bilmanda 'movement' < bilmaja 'move (of thing)' e. result nominalizations darlanda 'broken' < darlatha 'break'

In Keen's collection of three Yukulta texts, which is the only textual material we have in Yukulta, the nominalizing suffix occurs only three times, all in the same word birlji-n-da be=alive-Nomin-Nom 'alive', and always in assertions of the type 'X (is) alive'. A typical use is in (52). (52) ngijin-da kunathu birlji-n-da! my-Abs younger=brother (Abs) alive-Nomin-Abs 'My little brother, alive!' (line 58, text 2 in Keen 1983)

418

Australian

The functions of -n in Lardil are similar: it derives new nominal lexemes denoting actions, agents, or instruments from verb stems. The main difference from Yukulta is that, in addition to active verb stems, it can attach to passive verb stems when deriving instrument nominalizations. Some examples, taken from Hale et al. (1981), are in (53). (53)

a. simple nominalization (agent; action) were 'throw' — weren 'thrower, throwing' keli 'jump' — kelin 'jumper, jumping' wungi 'steal' — wungin 'stealer, stealing' b. nominalization, compounded with object noun (habitual agentive) be-tha 'bite' — dangka-been 'person-biter' la-tha 'spear' — werne-laan 'food-spearer' kubari 'make' — wangal-kubarin 'boomerang-maker' netha 'hit' — dangka-neen 'person-hitter' c. passive, compounded with object noun (instrument nominalization) dangka-neyin 'fighting boomerang' (by means of which person is hit) werne-layin 'spear' (by means of which food is speared)

All these types of lexical nominalization are attested in Kayardild (see 2.3) and Yangkaal as well, and we can assume that, with the possible exception of the "instrument nominalization," these constructions go back to protoTangkic. What is significantly absent from Yukulta and Lardil is the participial use of the nominalization, whether in subordinate clauses or as an indicator of ongoing action. Given that Yukulta and Lardil come from separate branches of Tangkic and show no evidence of mutual influence or convergent drift, it is highly probable that they are representative of proto-Tangkic in this regard. The most plausible diachronic scenario, then, is to see the Kayardild use of nominalized verbs in subordinate clauses as an innovation, resulting from the extension of the nominalizing derivational suffix to a productive participial use. In extending the functional range of the -n verbal suffix, some means of marking NP arguments of the participles had to be found, and the functions of the oblique were extended to include the marking of non-subject NPs in participial clauses, giving rise to the new construction type:

(54) A. second Kayardild innovation was the development of resultatives in THirrin-. Unlike standard nominalizations, whose form is attested in all Tangkic languages, the resultative is attested only in Kayardild, so its origins remain obscure. Rather than propose three independent processes of loss in Lardil, Yukulta, and Yangkaal, I shall assume that it is a Kayardild innovation, albeit of unknown origin.20 Like standard nominalizations it probably began as a derivational suffix deriving nouns of result from verb stems; this lexical use is exemplified by such nominal lexemes as darathirrinda 'circum-

Multiple Case in Kayardild

419

cised man', literally 'one who has been broken', from daratha 'break', and danathlrrinda 'divorced person, one that has been left' from danatha 'leave'. Later it would have been extended to the resultative participial use exemplified in (41) and (46), and the ablative extended to mark demoted subject arguments. Note that this syntactic innovation did not necessitate the creation of new morphological structures for nominals, since words of the form NP-(Case)Ablative already existed, these structures being used in constructions with the modal ablative case. As the new plain participle construction developed, by contrast, the inherited morphological structures for nouns seem to have limited the way in which the "new" cases could be ordered, in the sense that each had to appear in a slot compatible with existing word structures, and this ruled out the possibility of the associating oblique being ordered iconically. I will return to this point below, but first let me justify my claim that the inherited templates for Kayardild noun morphology were so rigid that they ruled out the development of a new and iconic ordering in the dependent participial constructions. It is a striking fact about comparative Tangkic morphosyntax that the morphological structure of the nominal word had already been fixed at the stage of proto-Tangkic and underwent no further expansions in Kayardild or Lardil. The existence of identical morphological structures in Lardil, Kayardild, and Yukulta makes it clear that nominal words in proto-Tangkic could take up to four case suffixes, and that the possible choices within each slot were limited according to the diagram in Figure 15.2. The morphological limitations relevant to later developments appear to have already been in place in proto-Tangkic: the outermost slot could only contain the ergative/locative (> Kayardild locative) or the dative (> Kayardild oblique); the dative could never be followed by another case; the ergative/locative could only be followed by the dative, in which case it yielded the portmanteau {-kurrka}; and only the five cases that functioned as Tcomplementizers (locative, ablative, dative, allative, and proprietive) could follow relational cases. Evidence for each of these proposed limitations in proto-Tangkic can be found in Kayardild, Yukulta, and Lardil. The Kayardild manifestations of these constraints are discussed in Evans (1994: §4.2.3). The Yukulta limitations are implicit in the presentation of nominal morphology in Keen (1983), and the entire Yukulta corpus contains no counterexamples, such as examples of the dative followed by another case. The Lardil evidence is less obvious, since sweeping phonological changes resulting in the loss of final syllables have obscured the nature of its inherited nominal morphology. However, the following two facts about Lardil morphosyntax are relevant. First, Lardil grammar avoids constructions in which subordinate clauses in the unmarked tense (which would have objects marked with the unmarked objective case, which reflects the proto-Tangkic dative) depend from main clause objects; the "marked non-future" construction must be used instead. More generally, the unmarked objective case in Lardil is alone

420

Australian

among the set of three objective forms in not having a "protected allomorph" when the subordinate clause in which it occurs has an antecedent which is object in the main clause. The other two object cases, the markednon-future objective (ObjMNonFut) and the future objective (ObjFut), each have protected allomorphs that descend from earlier structures in which they were followed by a complementizing case (material found only in "protected form" is in parentheses): (55) a. final syllable retained:

proto-Tangkic, Yukulta21 Kayardild

Conseq(-C.Dat)

Prop(-C.Dat)

-ngarrba(-ntha) -ngarrba(-ntha)

-kurlu(-ntha) -kuru(-ntha)

b. loss of final syllable: Lardil

ObjMNonFut(C.Obj) -ngarr(ba)

ObjFut(C.Obj) -kur(u)

The avoidance of constructions in which clauses in the unmarked tense are subordinated, and the lack of a protected allomorph of the unmarked objective, are synchronically an unmotivated exception in Lardil grammar. But they make sense if the ancestral dative case could never have been followed by another suffix, and thereby be protected. The constructions that would have contained the banned sequence would have been disallowed on morphological grounds, and protected allomorphs of the dative could thus not have arisen. Second, the Lardil locative case is never followed by another case; should the grammar assign a modal case to locative NPs, the modal case alone appears and the locative is not marked.22 This is the only semantic case to behave this way. There is thus considerable evidence that proto-Tangkic did not allow ergative/locatives to be followed by any case except the dative, and that the dative itself could never be followed. It appears that all the Tangkic languages continued this constraint on morphological structure. Why should the morphological structure have become fixed at the stage of proto-Tangkic? Apart from the fact that it may have been reaching the limits of processability (no other language in the world allows so many layers of nominal cases), I believe there was a more important reason: case assignment rules in subordinate clauses did not change significantly, and they became detached from the new main clause case rules. The reorganizations of the Kayardild and Lardil main clause systems of core case marking away from an ergative system meant that the principles for assigning complementizing case no longer followed from straightforward rules of antecedent agreement. Thus the complementizing ergative no longer corresponded to a relational ergative in main clauses in Kayardild or Lardil, which no longer had ergative case frames. And the complementizing dative lost its role as marker of indirect objects in Kayardild, where its relational uses were sharply curtailed and it was used mainly as a modal or complementizing case. In

Multiple Case in Kayardild

421

Lardil the complementizing dative was used just in those modalities—future and marked non-future—where the object antecedent did not take the dative, again obscuring the original reasons for complementizing case in simple antecedent agreement. The Lardil changes were further complicated by the operation of phonological rules that meant outer datives were themselves deleted under apocope, being detectable only by the phonological protection they afforded to inner suffixes. In both Kayardild and Lardil, then, the original rationale for the complementizing case system became completely obscured. At the level of modal case, too, growing formal differences between nouns and verbs made it harder for speakers to analyse modal case as an operator with clausal scope, and hence to expand the paradigm; this tendency would have been corroborated in Kayardild by the growing independence of modal case from verbal tense/mood. In fact, all modal case forms can be traced back to proto-Tangkic T-complementizers: neither Kayardild nor Lardil added any cases to the proto-system. In sum, it was only possible to keep expanding the nominal morphology while the rationale for the system rested on straightforward rules of case assignment over different levels (N', N", V, V"). This was possible in protoTangkic and allowed the development of the four-level system inherited by all modern Tangkic languages. But once changes in the grammar of Kayardild and Lardil obscured the agreement basis for assigning higher level cases, no further morphological developments were possible. Despite this, the morphological structure was sufficiently transparent to allow the several layers of case access to syntactic rules. With this background of fixed morphological possibilities, we can see what would have happened when the Kayardild participial construction developed, with its assignment of the oblique (< proto-Tangkic dative) to NPs in the participial clause. An associating oblique would have been compatible with existing morphological possibilities as long as no case was inherited from the higher clause; we saw examples of this in (32) and (33). But where the participial clause inherited a modal case from the higher clause, as in (44) and (45), an iconic ordering of associating oblique inside the modal case would have been incompatible with the existing morphological possibilities, which made it impossible for another case to follow the oblique. The conflict was resolved by co-opting the nearest existing morphological sequence, namely that in which a modal case suffix was followed by an outer oblique; this led to the anti-iconic ordering we have discussed. In fact, since the morphological ordering already existed and it was the syntactic construction which was new, it would be more appropriate to speak of an anti-iconic association of the suffix ordering with a new construction. It is interesting to compare this problem-ridden development with that of the ablative in the equally new resultative constructions. In the case of the ablative, it was perfectly possible for it to be followed by modal case, since templates for the sequence Ablative-Modal Case already existed; they had already been evolved to deal with adnominal possessives modifying objects, for example, as in (56) and (57).

422

Australian

(56)

jungarrba-naba-ya dangka-naba-ya wungi-ja wuran-ki big-Abl-M.Loc person-Abl-M.Loc steal-Act food-M.Loc '(The boys) stole (stingray) food from the adults', or '(The boys) stole the adults' food'

(57) nyingka karna-jarra ngamathu-naba-na wunkurr-ina? you=Nom light-Past mother-Abl-M.Abl grass-M.Abl 'Did you set fire to mother's grass windbreak?'

This meant that, when resultative nominalizations were embedded in higher clauses assigning modal case, as in (46) above, the ordering of case suffixes would iconically reflect the syntactic structure.

3.5. Summary Our motivation for considering the behavior of participial clauses in comparative Tangkic was to understand how the anti-iconic ordering of associating obliques with respect to modal case may have arisen in Kayardild. The scenario I have proposed can be summarized as follows. First, the morphological structure of the nominal word had already been fixed at the stage of proto-Tangkic and underwent no further expansions in Kayardild or Lardil. Words could take up to four case suffixes, and the possible choices within each slot were limited according to the diagram in Figure 15.2. The morphological limitations relevant to later developments were that the outermost slot could contain only the locative or the dative; that the dative could never be followed by another case; that the locative could only be followed by the dative, yielding the portmanteau {-kurrka}; and that only the five modal cases (locative, ablative, dative, allative, and proprietive) could follow relational cases. Second, the development of two new participial constructions in Kayardild, involving the plain and resultative nominalizations, each brought new rules for case assignment. Plain nominalizations, when used as participles, assigned an associating oblique to their non-subject NPs. Where such participles modified NPs bearing a non-zero modal case, the new construction would, if case ordering had been iconic, have generated case sequences such as A.Obl-M. Abl (44) and A.Obl-M.Prop (45) that would have been incompatible with inherited constraints on morphological structure. The morphological constraints won out, so that a morphologically permitted but anti-iconic suffix ordering was used in the new construction. Resultative nominalizations used as participles assigned the ablative to their demoted agents. Where these participles modified NPs that bore a modal or relational case in a higher clause, an iconic ordering of suffixes would generate sequences like Abl-Prop (41) or Abl-Loc (46). Since these were compatible with existing morphological constraints, no conflict arose, and case ordering could be iconic in the new construction. Each of the three examples of anti-iconic ordering that have been examined here illustrates a different diachronic path by which anti-iconic orderings

Multiple Case in Kayardild

423

can arise. In the Kulin languages the ordering of possessive suffixes outside the case suffixes simply reflects the order in which the categories were grammaticized. In Djingilu verbs, the anti-iconic ordering of pronominal subject suffixes inside aspectual suffixes results from a shift in morphological type from prefixing to suffixing, which saw an erstwhile verb, prefixed for subject, being reduced to auxiliary status and then suffixed to a new lexical verb. In Kayardild, the anti-iconic ordering results from morphological stasis rather than morphological development: a new function for the associating oblique was developed as part of a new participial construction, but the morphological possibilities had already frozen, so that an iconic ordering of multiple case would have been incompatible with inherited constraints on morphological structure. The conflict was resolved by using an existing, anti-iconic suffix sequence in the new construction. For generalizations about the relation between morphology and syntax to have any force, we need to characterize and constrain the set of empirically attested exceptions to the ideal of iconic association between the two components. Until we have studied the diachrony of anti-iconicity in more detail we cannot know how many other diachronic paths to anti-iconicity are attested in the history of human languages. This chapter has done no more than sketch a few such irritants to linguistic Panglossianism. But hopefully I have made it apparent that such inconveniences can be minimized if we can sufficiently constrain the set of possible diachronic paths so that history, rather than a baroque set of formal constraints, does the filtering for us.

Notes / would like to thank Frans Plank for inviting me to the conference and for his helpful editorial comments, and Guido Manzelli, Maria Koptjevskaja-Tamm, and Evgenij Xelimskij for examples and references on Uralic languages. I would also like to thank the many Kayardild speakers whose patience and insight have, over the years, made this study possible, in particular the late Darwin Moodoonuthi, Roland Moodoonuthi, and Alison Dundaman. 1. Another unusual feature of Kayardild and Lardil is the presence of a set of "verbal cases," which have a similar function to normal cases, show concord over the NP, are subcategorized for by verbs, and so forth, but have the peculiarity that they convert the words that host them from morphological nouns to morphological verbs. To avoid further burdening the reader with grammatical oddities I have restricted the number of examples with verbal case in this paper, but makunmaanju in (3) is to be understood in the following way: mak- 'bark torch' is inflected for the verbal donative case, one of whose functions is to mark instruments. Normally this would give the morphologically verbal word makuja 'use a bark torch', but since the NP is embedded and the verbal case could not otherwise take further case inflection, it is renominalized here to makun-, which then takes the adnominal origin and relational proprietive cases; the whole word is to be understood as 'with (food) obtained by the light of a bark torch'. We may note in passing that the development of verbal cases in Kayardild and Lardil (it is rather restricted in Yukulta) was probably a result of the functional

424

Australian

overextension of the other cases as they accrued modal, associating, and complementizing functions (on which see below; in glosses these functions are indicated by M., A., C., and T., prefixed to case labels). 2. Note that "oblique" here refers to a specific case (etymologically an old dative), not to a set of non-nominative cases. 3. In the interlinear glosses for this example I have explicitly shown certain zero allomorphs of the locative. Elsewhere in the chapter, zero morphemes are omitted for brevity's sake, so that the last word of the example would normally be glossed nguku water (Nom). 4. The main non-subject NPs that are exceptional in not taking modal case are "anticipated objects" taking the proprietive, secondary predicates on the subject, and NPs giving body parts of the subject. In Evans (1993) I argue that the absence of modal case marks a syntactic link with the subject NP in the same way that ergative agreement with transitive subjects marks such links in morphologically ergative Australian languages like Warlpiri and Pitjantjatjara. 5. The Kayardild oblique case is the reflex of the dative in proto-Tangkic; in proto-Tangkic it was the normal means of marking indirect objects, and this pattern is preserved in Yukulta. But in Kayardild the relational uses of the oblique are highly restricted, and indirect objects are normally marked with either the proprietive or the verbal dative. 6. In glosses combinations like 1/3 mean '1st person subject acting on 3rd person object'; + marks a clitic boundary. 7. For fuller discussion the reader is referred to chapter 10 of Evans (1994). 8. In fact the conditions governing the use of complementizing case in Kayardild are very unusual, but they are irrelevant to the discussion here. They can be broken down into (a) marking complement clauses, which are syntactically the objects of main clause predicators like 'know', 'see', 'rejoice', and so forth; (b) marking subordinate clauses obeying "odd pivot" conditions (essentially, when the coreferential NP is not subject of both clauses; this may include main object-subordinate subject, main subject-subordinate object, main object-subordinate instrument, and various other combinations). In addition, complementizing clauses may themselves be "insubordinated," in which case they are interpreted either as having ellipsed main clause predicators, or being in a marked discourse relation to what precedes. For details see Evans (1988, 1993, 1994). 9. One could make an argument that both modal and associating case are assigned at verb-phrase level, given that they do not appear on subjects, and are to some extent dependent on values of verbal categories. However, (a) there is no other evidence motivating a verb phrase constituent in Kayardild; (b) we have seen that the choice of modal case is not completely determined by the verb, since one verbal category may combine with up to three different modal cases; and (c) the verb phrase analysis would still not explain the failure of modal case to appear on "anticipated objects" and some other NPs that one would normally expect to be inside a verb phrase. So I prefer to analyse them as assigned at S level, but blocked from appearing on certain constituents such as subjects, secondary predicates, and so on. 10. In fact, it is not straightforward to decide that the ablative here is being used at the relational level. One might consider it adnominal, linking one NP to another (nominalized) NP, or associating, marking the NP as part of a (resultative) nominalized clause. I treat it as relational here for three reasons. First, the resultative is not a prototypical nominalization in Kayardild, and in some ways is structurally parallel to regular, finite passives, which would also assign the ablative to their demoted subjects.

Multiple Case in Kayardild

425

Second, although one can argue on formal grounds that resultative THirrin- can be segmented into THirri plus nominalizing -n (which can then be inflected, just like a nominalization), THirri is a cranberry morph, not combining elsewhere in the grammar with any other morpheme; moreover, it does not decline like the regular nominalizer. Third, the ablative in resultative constructions does not appear on the same range of NPs as the associating oblique. In sum, the behavior of -THirrin- is rather squishy, as is typical with nominalizations, and our classification of the ablative case it assigns as adnominal, relational, or associating will likewise be squishy. 11. This portmanteau is formed whatever the level of the locative and the oblique, as long as the locative is at a lower level. For example, the sequence could be adnominal locative plus complementizing oblique, modal locative plus complementizing oblique, modal locative plus complementizing or associating oblique, and so forth. See Evans (1994: §4.2.3). 12. The exact status of the ablative with respect to my typology of functional levels here is unclear, as I noted in Note 10. But the point stands whether it is adnominal, relational, or associating: the ablative, assigned inside the resultative clause, is ordered iconically with respect to the modal case inherited from a higher clause. 13. For similar types of argument applied to the rarity of metathesis in morphology see Janda (1984) and Anderson (1992). 14. The reader is referred to Serebrennikov (1962), Nichols (1973), and Xajdu (1985 [1966]) for further discussion. 15. Comrie considers the need for a further principle "that there is some preferred order, established as a general linguistic principle, for the relative order of case and possessive suffixes" (1976: 8). However, he goes on to reject it on the grounds that both orders are widely and independently attested. Support for this skeptical position comes from the data in Bybee (1985), whose major cross-linguistic study of affix ordering on the verb mentions a number of cases of anti-iconic orderings in verbal morphology (though it does not survey the diachronic paths engendering them). Of course, the issue of iconicity is a somewhat broader one here, since it is not clear that the morphological exponents of such verbal categories as aspect or mood should be associated with syntactic elements. But similar questions arise regarding the diachronic source of anti-iconic orders. 16. Further evidence that these suffixes originally commenced with a nasal comes from an otherwise unmotivated morphophonemic rule in Wemba-Wemba by which the underlying sequence -ga, where the r is part of the noun root, becomes -na, as exemplified by the locative lana- in 'into your camp', from lar 'camp' (Hercus 1986: 30); this can be interpreted as the result of a feature blend containing the retroflexion of the stem final and the nasality of an old suffix initial. There are insufficient data to tell if this process also applied to the operative suffix. 17. "Typologically the sequence [Verb-Subject Suffix-Aspect] is quite marked since normally person marking is more peripheral than aspect with respect to the stem" (Blake 1990: 55). 18. Of course prefixation of a suffix would theoretically be equally possible. 19. Our knowledge of Yangkaal is insufficient to determine the exact grammatical status of nominalized verbs. Hale's field notes contain a couple of examples of a participial-like use, e.g.: ngada bijilkurri-ju niwan-ju kali-n-ku lSg=Nom watch-Fut 3Sg-M.Prop jump-Nomin-M.Prop 'I will watch him jump'

426

Australian

which suggests that Yangkaal, like Kayardild, had extended its function beyond the derivation of agentive and other nominals, but there are no examples with dependent NPs that would show whether it assigns an associating case in these constructions. 20. Two further reasons for assuming it to be a Kayardild innovation are (a) the absence of even any frozen lexical forms reflecting -THirrin from all other Tangkic languages, and (b) the presence in Kayardild of a couple of plain nominalizations with resultative meaning, such as bulba-n-da 'full' from bulbaja 'to become full'; these do not take the resultative suffix one would expect. Presumably these are relics from an earlier phase when the plain nominalization, through lack of an opposing resultative construction, had a wider range of meaning than it does now, and as in Yukulta included resultative in its semantic range. 21. Yukulta has lost the consequential case, so does not attest the -ngarrba(ntha) sequences directly. But the "prior" verbal form -tharrba, etymologically composed of verb theme -th plus an allomorph arrba of the old consequential, participates in the sequence -tharrba-ntha when its clause has a main clause dative antecedent. 22. There is a Lardil trace of the only proto-Tangkic construction in which the locative preceded another case, namely when the sequence Loc-Dat was realized by the portmanteau -(k)urrka. I refer here to the Lardil "contemporaneous construction," in which the verb bears the suffix -jirr and its object takes the future objective form -ku(r). The suffix -jirr derives from proto-Tangkic -THurrka, historically analysable as verb theme TH plus -(k)urrka. The original Loc-Dat sequence comprised Tcomplementizing locative plus C-complementizing dative with "obviative" function, since contemporaneous clauses normally had different subjects to the matrix clause. The assignment of the future objective -ku(r) is strange synchronically, but it makes sense diachronically if it continues the old locative plus dative portmanteau -kurrka, which marked objects of contemporaneous clauses in proto-Tangkic and would descend in Lardil as -ku(rr); it would not be difficult for this suffix, which would be synchronically restricted, to be confused with the common future objective and hypercorrected to -ku(r).

References Anderson, Steven. 1992. Amorphous Morphology. Cambridge: Cambridge University Press. Andrews, Avery D. 1991. Inside-out unification and multiple case marking in Kayardild. Manuscript, Australian National University. Baker, Mark. 1985. The mirror principle and morphosyntactic explanation. Linguistic Inquiry 16, 373-415. Blake, Barry. 1990. Languages of the Queensland/Northern Territory border. In Language and History: Essays in Honour of Luise A. Hercus, ed. by Peter Austin, R. M. W. Dixon, Tom Dutton, and Isobel White, 49-66. Canberra: Pacific Linguistics. Bybee, Joan L. 1985. Morphology: A Study of the Relation between Meaning and Form. Amsterdam: Benjamins. Chadwick, Neil. 1978. The West Barkly Languages: Complex Morphology. Ph.D. dissertation, Monash University. Chadwick, Neil. 1983. The Relationship of Jingulu and Jaminjungan. Manuscript, Monash University.

Multiple Case in Kayardild

427

Comrie, Bernard. 1976. The ordering of case and possessive suffixes in the Uralic languages. Manuscript, King's College, Cambridge. Dench, Alan, and Nicholas Evans. 1988. Multiple case-marking in Australian languages. Australian Journal of Linguistics 8, 1-47. Dixon, R. M. W. 1980. The Languages of Australia. Cambridge: Cambridge University Press. Donaldson, Tamsin. 1980. Ngiyambaa, the Language of the Wangaaybuwan. Cambridge: Cambridge University Press. Evans, Nicholas. 1988. Odd topic marking in Kayardild. In Complex Sentence Constructions in Australian Languages, ed. by Peter Austin, 219-266. Amsterdam: Benjamins. Evans, Nicholas. 1992. Kayardild Dictionary and Thesaurus. University of Melbourne, Department of Linguistics and Language Studies. Evans, Nicholas. 1993. Code, inference, placedness and ellipsis. In The Role of Theory in Linguistic Description, ed. by William A. Foley, 243-280. Cambridge: Cambridge University Press. Evans, Nicholas. 1994. Kayardild, the Language of the Bentinck Islanders: With Elements of a Historical-comparative Tangkic Grammar. Berlin: Mouton de Gruyter. Hale, Ken. 1973. Deep-surface canonical disparities in relation to analysis and change: An Australian example. In Current Trends in Linguistics, vol. 11, Diachronic, Areal and Typological Linguistics, ed. by Thomas A. Sebeok, 401-458. The Hague: Mouton. Hale, Ken, Ann Farmer, David Nash, and Jane Simpson. 1981. A Preliminary Dictionary of Lardil. Manuscript, Massachusetts Institute of Technology. Hercus, Luise A. 1982. The Bagandji Language. Canberra: Pacific Linguistics. Hercus, Luise A. 1986. Victorian Languages: A Late Survey. Canberra: Pacific Linguistics. Janda, Richard D. 1984. Why morphological metathesis rules are rare: On the possibility of historical explanation in linguistics. In Proceedings of the 10th Annual Meeting, Berkeley Linguistics Society, 87-103. Keen, Sandra N. 1983. Yukulta. In Handbook of Australian Languages, ed. by R. M. W. Dixon and Barry J. Blake, vol. 3, 190-304. Canberra: Australian National University Press. Klokeid, Terry. 1976. Topics in Lardil Grammar. Ph.D. dissertation, Massachusetts Institute of Technology. Klokeid, Terry. 1978. Nominal inflection in Pama-Nyungan: A case study in relational grammar. In Valence, Semantic Case and Grammatical Relations, ed. by Werner Abraham, 577-615. Amsterdam: Benjamins. McConvell, Patrick. 1981. How Lardil became accusative. Lingua 55, 141-179. Nichols, Johanna. 1973. Suffix ordering in Proto-Uralic. Lingua 32, 277-238. Nordlinger, Rachel. 1993. A Grammar of Wambaya. M.A. thesis, University of Melbourne. Plank, Frans. 1990. Suffix copying as a mirror-image phenomenon. Linguistics 28, 1039-1045. Sadock, Jerrold M. 1991. Autolexical Syntax: A Theory of Parallel Grammatical Representations. Chicago: University of Chicago Press. Serebrennikov, B. A. 1962. O pricinax neodinakogo porjadka raspolozenija pritjazatel'nyx suffiksov v ural'skix i altajskix jazykax. Acta Orientalia Academiae Scientiarum Hungariae 15, 311-313.

428

Australian

Tauli, Valter. 1953. The sequence of the possessive suffix and the case suffix in the Uralic languages. Orbis 2, 397-404. Xajdu, P. 1985. Ural'skije jazyki i narody. Moscow: Progress. (Russian translation of original P. Hajdu, Bevezetes az urali nyelvtudomanyba. Budapest: Tankonyvkiado, 1966; the English translation lacks the relevant discussion.)

VIII DIACHRONY

This page intentionally left blank

16 Binder-Anaphors and the Diachrony of Case Displacement Anthony Rodrigues Aristar 1. Introduction In certain languages case marking that properly belongs to the head noun of an NP and defines the case relation which that head noun has in a higher phrase or clause appears affixed to the modifying noun or NP. In the majority of relevant languages (Moravcsik, this volume), case markers appear not merely on the dependent noun, but on the head noun as well. In this situation we have what is essentially an agreement phenomenon, and thus for these languages agreement is not limited to adjectivals and deictics; consider Uradhi (Crowley 1983: 377): (1) wutpu-:namu-mun antu-mun anta a-n old=man-Gen-Abl canoe-Abl take-Past '[they] took [it] out of the old man's canoe'

This type of displaced case marking has been called Suffixaufnahme.1 In still other languages, agreement does not occur. Instead, the dependent noun preempts all case marking of the highest NP, and the case marking appropriate to the head noun appears only once, on the modifying noun. Thus all case marking for the entire NP appears on the modifying noun. This occurs in languages such as Kanuri and Sumerian (exemplified in (2), from Thomsen 1984: 91).2 (2) e Ses lugal-ak-ak-a house brother king-Gen-Gen-Loc 'in the house of the brother of the king'

Even though there is apparently considerable syntactic difference between these two patterns, there is also an obvious commonality. In both patterns, since the modifying nominal is internal to a higher NP, what we see throughout is displaced—and often multiple—case marking. The modifying nominal 431

432

Diachrony

is marked once for the NP-internal case which defines its own dependent relation to the rest of the NP, and once for the NP-external case which defines the case relationship of the head NP to its higher clause. This chapter focuses on this commonality and offers a single diachronic explanation for the existence of both Suffixaufnahme (as in (1)) and dependent-noun case preemption (as in (2)). I will argue that both patterns arise from essentially the same source, and that the fundamental defining characteristic of both is the displacement of case marking from its logical argument and its appearance on a modifying noun or NP. I will therefore use the term "case displacement" as a general term for both these patterns. 2. The Distribution and Taxonomy of Case Displacement The number of languages in which case displacement occurs is relatively small. The distribution of the pattern is even areally constrained, for there seem to be areas where case displacement occurs, and others where it does not. As is shown in this volume, it has existed in the Caucasus region for a considerable time, and was found, for example, in Old Georgian, as it is in modern Tsez. It was even found here in ancient times in Hurrian and its relative Urartian, both spoken in the vicinity of Lake Van. It appears in Eastern Siberia—witness Chukchi—and in two isolated areas of Africa, in the neighborhood of the Ethiopian Highlands (e.g., in the Cushitic Agaw subgroup) and in the vicinity of Lake Chad (e.g., the Nilo-Saharan Kanuri). And it is common in Australian languages. As far as can be ascertained, it is found nowhere else.3 Taxonomically, languages with case displacement are not merely divided according to whether they have agreement between head noun and modifying noun. There is a further division, depending on the kinds of modifying nouns which can take case displacement. For most languages which show this phenomenon not every type of modifying noun acquires the case marking appropriate to the head. In the vast majority of languages and in all areal groupings where this type of agreement occurs, only a modifying genitive shows this agreement (Moravcsik, this volume), and case displacement cannot occur with other cases. We see this in languages like Uradhi, Gumbaynggir, and Old Georgian. In a few languages, however—these seem to occur only in Australia, where they are apparently a minority—though genitives always take case displacement, nouns in other cases do so as well. Indeed, in some of these languages a noun in any case that modifies another noun can take caseagreement suffixes. Martuthunira (Dench, this volume; Dench and Evans 1988) is an example of this: (3) warnanykura-la-ngara-a kunkuwarra-a river=gum-Loc-Pl-Ace hive-Acc 'hives in a river gum'

Binder-Anaphors and the Diachrony of Case Displacement

433

Finally, there exist two kinds of affixation patterns within case displacement. In one pattern, case affixes occur directly on the modifier's case marker, which is usually the genitive. We see this is (1) and (2) above. In other languages, the presence of the genitive is not necessarily sufficient for case displacement to occur: both the genitive case marker and a special ligature morpheme are usually or obligatorily present, as in the Agaw language Awngi and in Hurrian ((4), from Bush 1968: 151): (4) sen(a)-iffu-ue-ne-z asti-z brother-my-Gen-Lig-Erg wife-Erg 'my brother's wife'

The ligature morpheme is obligatory in most languages that have it,4 for example in Wulamba, Gugu Yalanji (Hershberger 1964), Dyirbal, Kalkatungu, Aranda (Aristar 1984: 147), and case attachment cannot occur if this morpheme is absent. Within each of these divisions two kinds of languages potentially seem to exist: (a) Languages where heads as well as the dependent NPs take the case marking appropriate to the head, that is, those which show case displacement as part of a process of syntactic agreement. These are the majority of relevant languages (e.g., Awngi, Martuthunira, or Uradhi). (b) Languages where the dependent noun alone takes case marking. Few languages do this (e.g., Kanuri, Sumerian, Agaw other than Awngi). No language where non-genitives can take case marking falls into this category.5 These latter two subtypes of case displacement are not mutually exclusive. For example, although in the Agaw language Bilin only the dependent noun usually takes the case marking for the head noun, the head noun may optionally take the case marking as well (Hetzron 1976: 37). Thus even some languages which must be classified as being in group (b) above sometimes display patterns more appropriate to (a). From this taxonomy we can draw the following generalizations: all languages which have case displacement show this phenomenon in the genitive; and case displacement which co-occurs with head-marking is heavily favored over case displacement alone. 3. Why Case Displacement?

3.1. The Generation of Genitives From Pronouns From an Indo-European point of view—or even from the point of view of orthodox theories of linguistics—such external marking of the case relations of a noun seems an oddity, and it is easily regarded as an idiosyncrasy of isolated languages. The appropriate place for the case markers whose argu-

434

Diachrony

ment is the head of an NP would seem to be the head itself, not the dependent NP of that head.6 There may be surface counterexamples to this generalization, but there are usually obvious explanations for the presence of the phenomenon, such as a deleted head.7 Even languages which exhibit case displacement seem to agree with this observation, since the vast majority of them, though they may mark dependent nouns with exterior case, insist that the head should be appropriately marked as well. It could be argued, then, that the comparative rarity of case displacement represents its atypical nature, and an explanation for the phenomenon should be sought in atypical diachronic circumstances. I will suggest here, however, that an atypical synchronic pattern does not necessarily imply atypical diachronic development. And I will suggest that an examination of linguistic situations where the historical development of the phenomenon can be traced shows that genitivally-linked case displacement has arisen as an incidental side-effect of the generation of new genitival marking by appositional pronominals. Case displacement in contexts where the genitive is not the only case involved is more problematic, and its origin is unclear. It could have arisen as an analogical extension of genitive-linked multiple case marking, or its origin may be independent. But the most pervasive variety of multiple case marking, that where the genitive is involved, is a direct result of a totally orthodox linguistic process: the binding of an NP into a construction by an anaphor. It is in fact this origin which accounts for the pervasive presence of the genitive in case displacement, and the rarity of case displacement is a natural result not of the idiosyncrasy of the process by which it arises, but rather of the fact that several conditions must be met in the course of its diachronic generation. It can arise only in contexts where the pronominal is head-referent, where the head-referent pronominal is itself case-marked, and where the order of nominals in the NP so formed leaves a head-marked pronominal adjacent to the dependent NP, and so allows attachment of reinterpreted morphemes to that NP. This is such a complex set of criteria that few languages meet them all. In order to demonstrate this, I will first examine the evidence from Agaw, a subgroup of Cushitic, where the data show clearly the process by which case displacement arose. As we will see, the process is closely related to the introduction of a new genitival formation, one based on the use of a case-marked head-referent anaphor as a means of binding a dependent to its head. Once this anaphor was reinterpreted as a genitival affix, genitives became capable of taking the case marking appropriate to their head. Then I will examine the evidence from some case-displacing Australian languages, in order to show that the Agaw process is found in unrelated languages and distant regions. The Australian evidence will also demonstrate that both languages which affix case markers only to a ligature and languages which affix them directly to the genitive have inherited case displacement from the same construction: one using head-referent anaphors. In one set of languages the anaphor has been reinterpreted as a genitive marker, and in another set as a ligature.

Binder- Anaphors and the Diachrony of Case Displacement

435

3.2. Agaw The Agaw subgroup offers important evidence with regard to case displacement, in that here we have a closely related group of languages which display both of the major types of case-agreeing genitive constructions we have been discussing. In one of these languages, Awngi, both heads and genitival modifiers are marked for the case function of the heads. In this language, then, case displacement appears as part of a generalized agreement phenomenon. All modifiers, including the genitive (through its allomorphs), agree in gender and number with all lower heads in the construction (Hetzron 1976: 37, and in this volume): (5) yuna-w-s-kw-da ce kut- kw-da n- kw-da abjel-ka-da woman-GenMasc-Dat-Genpl-Loc nice-Genpl-Loc house-Genpl-Loc doorway-Pl-Loc 'in the doorways of the woman's nice house'

In other Agaw languages, however, we find that case displacement cooccurs with equivalent head-marking only exceptionally. For example, in Bilin—unlike Awngi—there are two genitival constructions, one of which shows case displacement and one of which does not. When the case-displaced construction is used, however, the head of the genitive does not normally take case marking, which is usually affixed to the possessor (Hetzron 1976: 37): (6) ti'idad adari- w- d order lord-GenMasc-Dat 'by the order of the lord'

In all Agaw languages it is the head of this genitival construction which controls the form of its marker. That is, the marker is head-referent, as is exemplified for Awngi in (7). (7) a. aqi-w du:ri man-GenMascSg rooster (MascSg) b. aqi-t du:ra man-GenFemSg hen (FemSg) c. aqi-kw du:r-ka man-Genpl chicken-Pi 'the man's rooster / hen / chickens'

Furthermore, here we see more than mere copying of case marking onto a modifying noun. Gender is copied too, and thus genitive marking behaves in a way that is more typical of an anaphor than of a bound case marker. The explanation for this behavior shows very clearly if we examine the diachronic origin of these forms. If the differing reflexes of the agreeing

436

Diachrony Table 16.1. The Agreeing Genitive in Agaw.

Masculine Feminine Plural

Bilin

Khamir

-xw -ri -w

-w -y -k"

Quara -w -y [+round]

Kemant

Awngi

- /- " -y -w

-w/- w -t(i) -ku/-kw

genitive in the various dialects are looked at, the pattern of Table 16.1 is found. We can reconstruct these forms in a relatively straightforward manner (Aristar 1984: 46-54), and we arrive at the following pattern for Proto-Agaw: (8) Masculine *-ku

Feminine *-ti

Plural *-kun

These forms are essentially identical to a particular set of Cushitic pronouns, a set which still occurs pronominally in the Awngi presentative pronouns masculine -ku, feminine -ti, and plural -koni (Aristar 1984: 52f.). Now, if this reconstruction is correct, then the genitive markers are diachronically simply pronominals which have become reinterpreted as suffixes to nouns. That is, we can suggest that the phrases which are now interpreted simply as genitival were at an earlier stage appositional. Thus the construction given in (6) above originated in a construction equivalent to 'the order, by the lord's one'. Proof of this interpretation exists in an interesting fact. Two dialects, Kemant and Bilin, have a non-dependent-marking genitive as well as the dependent-marking genitive we have been discussing here. This secondary genitive acts in every way like an orthodox Indo-European genitive, and it displays none of the gender agreement with the head of a genitive or of the displaced case marking which the other genitive does. We can see this genitive in Kemant (Appleyard 1975: 326): (9)

ndan katam-i g mb this city-Gen walls 'the walls of this city'

Now the form the noun has when it takes the dependent-marking genitive suffixes is always identical to the form of the other, non-agreeing genitive, as we can see in Table 16.2. It seems fairly certain, then, that the agreeing genitive—the only genitive for the majority of Agaw dialects—had its origin in a construction which was a concatenation of a noun in the old, non-agreeing genitive, and a pronoun— e.g., either (10) or (11) below. (10) shows the reconstructed string for a nonagreeing dependent-marking genitive, and (11) that for an agreeing dependentmarking genitive.

Binder-Anaphors and the Diachrony of Case Displacement

437

Table 16.2. Formation of the Agreeing Genitive in Bilin.

'dog' 'man' 'men' 'woman'

Absolute Genitive

Non-agreeing Genitive

MascSg Agreeing Genitive

g da ) g rwa gr w gana

g d g rwi g r wa ganar

g d xw" g rwixw g r waxw" ganarxw

(10)

*g da g rw-i ku- d dog man-Gen ProMasc-Dat 'the dog, to the man's one'

(11)

*g da - d g rw-i ku- d dog-Dat man-Gen ProMasc-Dat 'to the dog, to the man's one'

Since of all the six Agaw languages only Awngi shows obligatory case marking on both head and dependent, there is a prima facie case that (10) above was the original Agaw construction. And since Awngi is the only Agaw language which has lost the older, non-agreeing genitive, and since these constitute the vast majority of counterexamples to the generalization in which all modifiers are capable of taking agreement (even if not required to do so), the evidence seems strong that Awngi has innovated in this regard. Whether or not this interpretation of the data is correct, however, it is clear that in Agaw both case displacement constructions—whether showing agreement or not—are a direct result of the use of a pronoun as a genitival binding strategy. This strategy clearly gave rise to both the typologically ubiquitous head-and-dependent-marking genitives and to the typologically rarer strategies which mark only the dependent nouns of the genitive. It is clear that such an origin for the Agaw genitive offers an easy and effective explanation for the behavior of the case displacement construction. The idea of a case-marked noun taking a further case marking whose argument is some other noun is theoretically highly problematic. But there is nothing peculiar about pronouns which are in apposition to a noun taking the same case marking as that noun. At least diachronically, we have two nouns, one of which happens to be a pronoun, each taking a single case marker of which it is the logical argument. We can say, then, that at least in terms of Agaw the diachronic origin sketched here is a complete explanation for the presence of case displacement in those languages.

3.3. Australian Now, it is clear that the diachronic origin of genitivals in anaphoric constructions is very common indeed (Aristar 1991), and the process we have described here for the Agaw languages is not in any sense idiosyncratic. I will

438

Diachrony

discuss here two examples from the areally and genetically unrelated Australian family: Gumbaynggir, which has case displacement on the bare genitive stem, and Dyirbal, where case displacement is allowed only if a ligature morpheme is added first. I will show that these two types of the phenomenon are essentially identical, and that they arose by a process completely analogous to that found in Agaw. In Gumbaynggir, the genitive has a varying form, depending on the class of noun to which it is suffixed. The suffix -gundi or - undi is added to most common nouns and adjectives, while -mbandi or -undi is added to pronouns and other nouns, as in (12) (Dixon's analysis, from Eades 1977: 186). (12)

ni:gar-gundi gamay man-Gen (Abs) spear (Abs) 'man's spear'

The origin of these forms becomes clear if we compare them to the relative marker and to the dative. The relative has a simple set of allomorphs: only -(a)ndi appears (Dixon 1969: 40): (13)

ai-dja guga:mgam a:wa- bua:- -andi ni:gar I-Erg emu (Abs) see-Past kill-Past-Rel (Abs) man (Abs) 'I saw the emu that had killed the man'

The forms of the dative are, like the genitive, more complex. To most common nouns and adjectives -gu or - u is added, while -mba or -u is added to pronouns and other nouns. Since these are identical to the forms found in the genitive (Eades 1977), the case which is synchronically a genitive actually arose from the following diachronic source: (14)

ni:gar-gu-ndi gamay man-Dat-Rel (Abs) spear 'the spear which (belongs) to the man'

Thus we have in Gumbaynggir a precisely analogous situation to that found in Agaw. The Gumbaynggir genitives are diachronically concatenations of an earlier case marker—in this case the dative, not the genitive—and a morpheme which is acting anaphorically. Unlike the situation in Agaw, the diachronic origin of the Gumbaynggir morpheme is unclear, and whether it descends from a pronoun is unknown. But synchronically, that same morpheme clearly has anaphoric value, and the diachronic situation is thus essentially identical to that of Agaw. Let us now compare the Gumbaynggir genitives with those of the closely related language Dyirbal. The usual marker of the genitive in Dyirbal is - u, a form very different from that found in Gumbaynggir (Dixon 1969: 36):

Binder-Anaphors and the Diachrony of Case Displacement (15)

439

badibadi- u yabi Badibadi-Gen (Abs) mother (Abs) 'Badibadi's mother'

But the syntactic difference between the two languages is more important. In Gumbaynggir the genitive can take case marking directly, without any modification. In Dyirbal, however, the genitive can take case marking only if a ligature suffix -ndji(n) intervenes. Thus in Dyirbal we have a contrast between the form of a case-marked genitive and the form of a non-case-marked one. Compare, for instance, (15) and (16) (Dixon 1969: 39). (16)} al ga guda- gu yara- u-ndjin-du badja-n child (Abs) dog-Erg man-Gen-Lig-Erg bite-Pres/Past 'The man's dog bit the child'

A comparison of the Dyirbal case-marked genitives with those of Gumbaynggir is intriguing, for the resemblance between them is much more suggestive than their non-case-marked forms would imply. For most nouns the Dyirbal case-marked genitive would be - undji(n), as compared to the general Gumbaynggir genitive, which is, in its most common form, -gundi or - undi. These forms are very similar. In fact, as I have shown (Aristar 1984: 106ff.), the forms are diachronically related. The Dyirbal genitive is simply a specialization of a form, - u, which originally marked datives in pronominals and kinship terms. This is most apparent in the fact that the idiosyncratic forms which pattern as distinctive genitives of the pronouns in Dyirbal often pattern as the dative/purposive in closely related languages (Aristar 1984: 109f.). The relationship between the Dyirbal and Gumbaynggir inflected genitives thus seems complete. The evidence suggests that they are diachronically the same. The single difference between Dyirbal and Gumbaynggir is that in Gumbaynggir the original anaphor is required in all genitivals, while in Dyirbal it appears only when case agreement occurs. 3.4. Synthesis It is clear from the evidence above that in Dyirbal and Gumbaynggir we see the same pattern as we saw in the Agaw dialect group. In Agaw we have two genitives. In some of the Agaw languages—Awngi, for example—only the agreeing genitive construction exists. Bilin and Kemant, as we saw earlier, have another, non-agreeing genitive as well. And it is only with the agreeing genitive that case displacement can occur. Thus we have the following contrast: (17)

garw-i g d n- d man-Gen dog-Dat 'to the man's dog'

440 (18)

Diachrony g d - /- d g rwi-xw- d dog-Abs/-Dat man-GenMasc-Dat 'to the dog, the man's one'

The reason for this contrast is obvious in Agaw. Only in the agreeing genitive was there, diachronically, a pronoun available to carry the extra case marker. The non-agreeing genitive lacked any such anaphor, and thus case displacement was impossible. In some Agaw dialects (e.g., Awngi) the non-agreeing genitive has been lost, leaving only the agreeing genitive construction, while in others (e.g., Kemant) both are used. But the agreeing genitive exists in all dialects, and only with the agreeing genitive is case displacement possible. In Gumbaynggir the situation is analogous to that of Awngi, in that the synchronic genitive was formed from a concatenation of an earlier genitive marker (here probably originally a dative) and an anaphoric element -(a)ndi, which still functions synchronically as an anaphor in that it marks relative clauses. Both the Agaw dialects and Gumbaynggir, then, suffix cases directly to the genitive marker, without a ligature. This lack of a ligature, however, is only synchronic. The genitive marker has incorporated into itself the morpheme which once acted as a ligature, and this morpheme had its origin in an anaphor. In Dyirbal, on the other hand, case displacement occurs only when a ligature is present. But since the ligature morpheme seems to be diachronically related to the Gumbaynggir -(a)ndi morpheme, which is clearly anaphoric in origin, we can argue that the Dyirbal ligature is itself anaphoric in origin, and the difference between languages which have case displacement in the genitive with or without a ligature is more apparent than real. Clearly case displacement in some languages with a ligature has the same origin as case displacement in languages which have no ligature: it arises from a concatenation of genitive case marking with a pronominal. It seems, then, that the distinction between languages which affix case markers directly to the genitive and languages which first append a ligature before the genitive may be case-marked is—at least in some instances—of no significance. In the languages examined here, case marking of a modifier can occur only when a morpheme derived from an anaphor is present. And since in all the languages examined, only the genitive could acquire such an anaphoric morpheme, only a genitival modifier has acquired the ability to take the case marker appropriate to the head. Finally, all these languages used anaphors which referred to the head of the genitive construction. It was this head-referent quality of the anaphors in question which allowed them to copy the cases appropriate to the head of their NP. 4. Conclusion There seems to be a clear, orthodox diachronic explanation for genitival case displacement. In the languages discussed, the generation of case displacement requires that there be an anaphor8 present. A noun cannot synchronically

Binder-Anaphors and the Diachrony of Case Displacement

441

take the case marking of another noun. And if we look at case displacement in this fashion—as a relatively natural result of the use of a pronoun to bind one noun to another—then the phenomenon seems much less exotic, and much more just one possible result of a fairly orthodox linguistic strategy. Binderanaphors used in this fashion are hardly rare or unusual. Binder-anaphors should come in two varieties, since there are two logical constituents in a genitival construction: one the "head" of the construction, and the other the modifying, dependent NP. Thus, if a pronoun marks such a construction, there are two constituents to which the pronoun can refer. If a binder-pronoun occurs in such a genitival construction, then, it should be capable of being either "head-referent" or "dependent-referent." In a genitive with 'John' as dependent and 'hat' as head, head-referent anaphors will refer to 'hat': 'the hat, John's one'. Dependent-referent pronominals, on the other hand, will refer to 'John': 'John his hat'. Now, there is a sameness about all the languages discussed above and about the case displacement they display. All of them acquired case displacement by means of anaphors which were originally head-referent. The exclusive presence of head-referent anaphors in the generation of case displacement is not a result of the rarity of dependent-referent anaphors. Indeed, they seem to be by far the most common kind of genitival binderanaphor. They are common in Europe (e.g., German; Hungarian; Abkhaz, Hewitt 1979), Asia (e.g., Turkish, Kornfilt 1987), Oceania (e.g., Palauan, Josephs 1975), Africa (e.g., Mangbetu, Larochette 1958: 81f.), and especially in America (e.g., Chickasaw, Payne 1982; Quechua, Nichols 1986), where they are the most common means of forming genitives. They even existed once in English, as numerous forms like be bishop his broper in Middle English texts like Layamon's Brut show. Indeed, these forms were so common in Early Modern English as to rival the genitive -s inherited from Anglo-Saxon (Kellner 1892: 194). As an example of such a construction in a modern language we add here the German dialectal genitive, where the genitival relationship is marked twice—once with a dative case marker on the dependent and once with a dependent-referent anaphor: (19)

dem Vater sein Buch the father-Dat his book 'father's book'

We find the same kind of dependent-referent anaphor in Hungarian, but in this language the dependent noun only optionally takes a case marker:9 (20)

az ember haz-a the man house-his

(21)

az ember-nek haz-a the man-Gen/Dat house-his 'the man's house'

442

Diachrony

Now, if we recall that case displacement is denned as a construction in which a modifying NP either copies or preempts the (NP-external) case of the head of its construction, dependent-referent anaphors obviously can never give rise to it, since they acquire their case from the nominal to which they are in apposition, and they are in apposition to the dependent nominal.10 They will still agree in case with their referent, like any appositional element; but since a dependent-referent anaphor is in apposition to the possessor nominal, it can take only the case which in that language marks possession. And this case will always be internal to the genitival phrase, since the case of the dependent marks that genitival relationship. A head-referent anaphor, on the other hand, is in apposition to the head of the NP, and the case such a constituent takes is always external to the genitive, since it defines the relationship of the NP to the clause in which it resides. Thus only head-referent anaphors will show the variation in external case which is prototypical case displacement. In the languages discussed here—the Agaw group (Awngi, Bilin, Kemant), Gumbaynggir, and Dyirbal—the pronoun is clearly headreferent, as we can see in (10), (11), and (14). The head/dependent-referent distinction is crucial, then, because it defines what kind of morphology a noun can acquire from the other noun in a construction where one nominal modifies another. A dependent-referent pronoun, since it can only mark a single, fixed relationship—the possessive—will never vary in case, and will become invariable with regard to this category. The only varying markers of the dependent that a dependent-referent anaphor can acquire are gender markers. Thus one would expect heads whose marking derives diachronically from dependent-referent anaphors to vary in gender (if gender is marked in the language and construction concerned) but not in case. Thus this kind of genitive marking will never show classic case displacement. We see precisely this phenomenon in the following Hebrew example (Hetzron 1987), where the dependent-referent anaphor agrees in gender with the dependent, and (vacuously) with its possessive case: (22)

bet-o sel ha-is house-3MascSgPoss of Def-man 'the man's house'

By contrast, if the anaphor used is head-referent, then such anaphors will potentially vary in both the case and the gender marking of the head noun, since the head varies in case and gender. Thus we would expect morphology which derives from head-referent anaphors to show both case and gender variation. Most importantly, it will show variation in external case in typical case-displacement fashion. Therefore we can suggest that, although the binding-anaphor construction seems to be a common diachronic source for genitival marking in every language group, only head-referent marking could ever produce case displacement, for only with head-referent anaphors do we find conditions in which varying, NP-external case can be copied onto an appositional pronoun.

Binder-Anaphors and the Diachrony of Case Displacement

443

The two diachronic sources for anaphor-based genitival marking are given schematically in (23).n (Note that the elements in them should be seen as unordered.) (23)

a. Head-referent marking [NP[NPDEPENDENT-Poss, ProNP] HEADNP] (e.g., Agaw, Gumbaynggir, Aramaic) b. Dependent-referent marking [NPDEPENDENT(-Poss) [NPPro-Poss, HEADNP]NP] (e.g., German, Hungarian, Aramaic)

Both choices have been made, at different diachronic stages, in the same language. At the earliest known stage of Aramaic, the Imperial Aramaic used in the time of the Persian empire, two genitives were common (Johns 1972). One, the more normal at this stage, was the original, inherited Semitic construct, exemplified in (24), where the head noun is placed in a distinctive morphological form, undergoes stress reduction, and appears immediately before the dependent noun. (24) madi:nat malk-a: province=Const king-Def 'the province of the king'

At the same time, a second genitive construction appears, which uses a headreferent pronoun in the construct: (25)

madi:na: di: malk-a: province Pro=Const king-Def 'the province which-of the king'

By the time the Aramaic portions of the Bible were written, a third genitival strategy has begun to appear as well, based on dependent-referent anaphors. One phrase of Ezra actually shows all three genitives used together (Ezra 5:11): (26)

Tabdo^hi di: ?ela:h ssmayya: slaves-3MascSgPoss Pro=Const god=Const heaven 'the slaves of the god of heaven'

This example is important in another way. It shows that not even headreferent anaphors are sufficient in themselves to lead to case displacement. Only certain orders of morphemes in a head-referent construction will give rise to case displacement. To see this, compare the Proto-Agaw construction (27) with the Proto-Aramaic (28):

444

Diachrony

(27) [NPdog [NPman-Gen ProNP]NP]-Dat 'to the man's dog' (28) Dat-NP[dog NP[Pro-Const man ProNP]NP] 'to the man's dog'

The fundamental distinction between the two languages is that Agaw casemarks nominals with NP-final affixes, and Aramaic case-marks nominals with NP-initial proclitics. Since Pro to-Agaw used Genitive-Noun order, and the head-referent anaphor is the head of the embedded dependent NP, the anaphor at issue appeared in Agaw at the rightmost boundary of the NP, where a case suffix could attach to it. This generated classic case displacement. In Aramaic, on the other hand, a proclitic could attach only to the head of the higher NP, which was the head of the whole genitival NP. The Aramaic proclitic could thus attach only to its logical argument, not to the dependent NP; this does not produce case displacement. Thus we see three constraints on case displacement. For this phenomenon to arise, these conditions must be met: (a) the anaphor strategy must be chosen to mark a genitive; (b) the anaphor which is chosen for this construction must be head-referent; and (c) the order of case markers with regard to their arguments must be such that the case marker is adjacent to the anaphor, and the anaphor is adjacent to the dependent nominal. I submit that in this set of conditions lies the reason why case displacement is so rare.12 The anaphor strategy is not inevitable: many languages will not choose it. If they do not, no case displacement can occur. Judging by comparative evidence, dependentreferent anaphors are chosen much more often than head-referent ones; and with this choice, no case displacement can occur. And, finally, only if the ordering conforms to the relevant conditions—i.e., the head-referent anaphor is adjacent to the dependent nominal, and the head's case marker is adjacent to the anaphor—will attachment of originally independent morphemes to the dependent nominal occur.13 The anaphor-binding strategy even explains the occurrence of ligature markers in many of the languages which have genitival agreement, such as Hurrian, where case agreement can occur only when an anaphor-like element is suffixed first. These elements are simply the reinterpreted remnants of binder-anaphors. If indeed such constructions are the origin of case displacement, then they are merely one reflex of a much more general process, for anaphors are commonly used to bind one element to another in a variety of constructions. As I showed earlier (Aristar 1991), such an anaphor-binding strategy figures not merely in the formation of genitives, but also in the formation of relativeclause and adjectival marking. It is the existence of this strategy, I would argue, which accounts for the relationship typologists have noted between the prevalent ordering of genitival, relative, and adjectival constructions. It is also, I believe, the cause of the idiosyncrasy of case displacement. I would suggest here that this strategy is a common linguistic means of incorporating

Binder-Anaphors and the Diachrony of Case Displacement

445

nouns into constructions, and a driving diachronic force behind much of the syntactic behavior of constructions which bind nouns together. Notes 1. The term "Suffixaufnahme" is, strictly speaking, unsuitable. There seems no principled reason why external, prefixed case displacement should not occur. The fact that examples of this are unattested or at any rate rare (see Plank's introduction to this volume) is probably a function more of the rarity of prefixed case markers than of the impossibility of prefixal Aufnahme. Indeed, I suggest later in this paper that the phenomenon by which Suffixaufnahme is generated also generates headmarking gender markers in modifying constructions, some of which—e.g., in Bantu—are prefixes. 2. Though I lack the space to enter into this here, there are at least two contrasting classes of marking in these languages. One class has to do with whether marking is globally or locally NP-final. Marking in Kanuri is locally NP-final: it is permitted only insofar as the NP does not contain another NP. Other languages (e.g., Sumerian and Moru) have global NP-final marking: case marking always goes on the final nominal of an NP, no matter how much recursion it contains. This means that nominals which occur inside other NPs are never marked, and center-embedding of a complex type is obligatory. The second class has to do with when NP-final marking occurs. In some of these languages—Kanuri is one—though no nominals other than the genitive can take the head's case markers, case marking also appears on NP-final adjectives. In languages such as Kemant and Bilin, however, NP-final marking occurs only with adjectives which are diachronically related to the genitive. The parallelism between genitives and adjectives is not surprising: adjectivals often derive diachronically from genitival modifiers (Aristar 1991). 3. Less prototypical forms of case displacement, as surveyed in the introduction and some other chapters in this volume, distribute a little more widely, however. 4. Wegner (this volume) suggests that the ligature morpheme may be optional in the singular of Hurrian nouns, and obligatory only in the plural. 5. My guess is that this absence is simply a result of an inadequate sample, and does not constitute a genuine typological generalization. 6. Note that it is the separation of case markings from their heads and their appearance on other nouns which is the problem here, not double case marking. Large numbers of languages form compound adpositions or case markers by treating casemarked nouns as arguments, e.g., 'from in the house'. The problem with case displacement is that the nominal upon which the case marker occurs is not its argument. This is fundamentally different from the situation adjectivals display in Indo-European languages, where these are in themselves not potential arguments of cases in the NPs in which they appear, even if they may be treated as such in different contexts. 7. For example, the English clitic genitive is the argument to a preposition when the head is deleted. But, while the head is absent from the surface-string, it is still present at an underlying level (e.g. of the old man's [canoe]). 8. I use the term "anaphor" rather than "pronoun" or "pronominal" since Gumbaynggir shows clearly that the crucial factor in this strategy is the use of an element with anaphoric value, whatever this may be. An element does not have to be pronominal for it to have such a value.

446

Diachrony

9. I am indebted to Edith Moravcsik (personal communication) for this observation. 10. For the sake of completeness, I note here one, vacuous, exception to this generalization: a dependent-referent anaphor which arises in a language which already has case displacement will naturally agree in case with the head, since all genitives would agree in case with that constituent in such a language. 11. I use Toss' as cover-term for the case which in the language concerned marks possession. Clearly this varies. In Gumbaynggir it was the dative, in Agaw the genitive. 12. In an attempt to make my point clear, I have ignored the many other orders of nominals and case morphemes possible in a genitival construction. I will discuss these at much more length in a paper now in progress. 13. For examples of a language where case displacement is at present being generated, but attachment of independent morphemes has yet to occur, I refer readers to Moru (see Hagege 1976).

References Appleyard, David L. 1975. A descriptive outline of Kemant. Bulletin of the School of Oriental and African Studies 38, 316-350. Aristar, Anthony Rodrigues. 1984. On the Syntactic Incorporation of Linguistic Units. Ph.D. dissertation, University of Texas at Austin. Aristar, Anthony Rodrigues. 1991. On diachronic sources and synchronic pattern: An investigation into the origin of linguistic universals. Language 67, 1-33. Bush, Frederic W. 1968. A Grammar of the Human Language. Ph.D. dissertation, Brandeis University. Crowley, Terry. 1983. Uradhi. In Handbook of Australian Languages, ed. by R. M. W. Dixon and Barry J. Blake, vol. 3, 307-428. Canberra: Australian National University. Dench, Alan, and Nicholas Evans. 1988. Multiple case-marking in Australian languages. Australian Journal of Linguistics 8, 1-47. Dixon, R. M. W. 1969. Relative clauses and possessive phrases in two Australian languages. Language 45, 35-44. Dixon, R. M. W. 1980. The Languages of Australia. Cambridge: Cambridge University Press. Eades, Diana. 1977. Gumbaynggir relative clauses and possessive phrases reexamined. Oceanic Linguistics 16, 179-192. Hagege, Claude. 1976. Relative clause, center-embedding, and comprehensibility. Linguistic Inquiry 7, 198-201. Hershberger, Henry. 1964. Case-marking affixes in Gugu Yalanji. In Papers on the Languages of the Australian Aborigines, ed. by Richard Pittman and Harland Kerr, 73-82. (Occasional Papers in Aboriginal Studies, 3.) Canberra: Australian Institute of Aboriginal Studies. Hetzron, Robert. 1976. The Agaw languages. Afroasiatic Linguistics 3, 31-75. Hetzron, Robert. 1987. Hebrew. In The World's Major Languages, ed. by Bernard Comrie, 686-704. London: Croom Helm. Hewitt, B. G. 1979. Abkhaz. (Lingua Descriptive Series, 2.) Amsterdam: NorthHolland. Johns, Alger F. 1972. A Short Grammar of Biblical Aramaic. Berrien Springs, Michigan: Andrews University Press.

Binder-Anaphors and the Diachrony of Case Displacement

447

Josephs, Lewis S. 1975. Palauan Reference Grammar. Honolulu: University Press of Hawaii. Kellner, Leon. 1892. Historical Outlines of English Syntax. London: Macmillan. Kornfilt, Jaklin. 1987. Turkish and the Turkic languages. In The World's Major Languages, ed. by Bernard Comrie, 619-644. London: Croom Helm. Larochette, Joe. 1958. Grammaire des dialectes Mangbetu et Medje. Tervuren. Nichols, Johanna. 1986. Head-marking and dependent-marking grammar. Language 62, 56-119. Payne, Doris L. 1982. Chickasaw agreement morphology: A functional explanation. In Studies in Transitivity, ed. by Paul J. Hopper and Sandra A. Thompson, 351378. (Syntax and Semantics, 15.) New York: Academic Press. Thomsen, Marie-Louise. 1984. The Sumerian Language. Copenhagen: Akademisk Forlag.

This page intentionally left blank

IX EPILOGUE

This page intentionally left blank

17 Summing up Suffixaumahme Edith A. Moravcsik 1. Introduction The purpose of this concluding chapter is to provide an overview of Suffixaufnahme as this phenomenon has revealed itself in the preceding languagespecific discussions.1 Two main issues will be explored, both of the sort that a comprehensive discussion of any linguistic pattern would need to address. They have to do with structure and distribution: What is Suffixaufnahme like? And what occasions its occurrence? The first issue involves the study of the proper place of Suffixaufnahme in the taxonomy of linguistic constructions (Section 2). We will explore the range of structure types that Suffixaufnahme is a variety of (2.1) and the subtypes it has in turn (2.2). In the former endeavor, we will be interested in sketching the structural parameters along which one notch identifies Suffixaufnahme. In the latter endeavor, we will discuss the parameters along which patterns of Suffixaufnahme itself may vary. The second main issue of the chapter has to do with the distribution of the pattern in question: the context in which Suffixaufnahme phenomena are embedded (Section 3). First, we will take a narrow view and, staying within the confines of the construction type, explore whether the occurrence of one feature of the construction may predict the occurrence of another feature of the same construction, and whether the availability of one subtype of the pattern in a language might predict the availability of another subtype in the same language (3.1). Second, we will broaden the scope of investigation to see if the presence of (some subtype of) Suffixaufnahme is linked to the presence of some other grammatical pattern in the same language (3.2). Having considered how Suffixaufnahme is anchored in sentences and languages, we will next consider how languages that have this construction are anchored in space and time—i.e., the distribution of languages with Suffixaufnahme relative to other, spatially and temporally adjacent languages and language stages (3.3). Both of these fundamental issues are discussed extensively in Plank's introductory chapter. While the present overview covers in part the same ground,

452

Epilogue

the two chapters also complement each other by adopting somewhat different points of view. The core phenomenon of Suffixaufnahme will be taken to be a pattern where an attributive nominal carries two distinct case markers: one appropriate to its own function as an attributive, and the other appropriate to the function of the NP that includes both the attributive and the head. The definitional schema of the construction is therefore as follows: (1) [A, [B, ([HeadNom, (ExtCase), C]), [AttrNom, IntCase, ExtCase, D]]NP]S

Head nominal (HeadNom) and attributive nominal (AttrNom) are both stems of nominals (i.e., nouns or pronouns), which may or may not be accompanied by associated modifiers and other satellites (variables C and D). Internal case (IntCase) and external case (ExtCase)2 may range over free or bound, segmental and suprasegmental case markers (affixes, adpositions, and tonal or accentual marking), with the internal case encoding the function of the attributive nominal and the external case encoding the function of the entire NP of which the head nominal and attributive nominal are parts. B and A are variables for additional, possibly zero material within the NP and the sentence (S), respectively. Commas indicate co-occurrence without commitment to linear order; square brackets delimit constituents; parentheses indicate optional presence. Thus, in (2), a Gumbaynggir example from Dench and Evans's seminal paper (1988: 8), ba:ba 'father' is the head nominal, junuy 'child' is the attributive nominal, genitive -gundi is the internal case, ergative -gu and -yu are manifestations of the external case, and A, B, C, and D are zero. (2) ba:ba-gu junuy-gundi-yu father-Erg child-Gen-Erg 'the child's father'

The use of the term "Suffixaufnahme" for the class of phenomena defined by schema (1) is not quite appropriate since the schema delimits a broader class of constructions than the one so designated by Franz Nikolaus Finck, who first introduced the term (see Plank, this volume, Section 2.2). One way in which the domain of the schema is more extensive than Finck's is that, while Finck used "Suffixaufnahme" in its literal sense with respect to constructions that involve suffixes, our present schema is not partial to suffixed case markers. Although multiple case marking of attributive nominals mostly involves suffixes, there are sporadic instances of prefixes, adpositions, and tonal marking (Plank, 4.6, 4.7), and the schema is designed to accommodate this fact. The second, more important way in which the present use of the term differs from Finck's, as well as from the prototype of Suffixaufnahme defined by Plank (1990 and this volume, 4.1), is that the more restrictive FinckPlankian concept stipulates an obligatory, rather than optional, head nominal and an obligatory, rather than optional, specification of this nominal for external case. Thus the Gumbaynggir example in (2), where the head is case-

Summing up Suffixaufnahme

453

marked, does fall within both the broader and the narrower concepts, but the following example from Aranda (3), cited by Plank (2.6.1), where the head is not case-marked, falls outside the narrower concept while it is still accommodated by the present schema. (3) worra ingkata-kana-la son chief-Gen-Erg 'the chief's son'

Similarly, hypostasis—the double case marking of a headless possessor, as it occurs in Basque ((4), cf. Plank, 2.7.1)—strays far from the Finck-Plankian prototype, although our schema still admits it to the category of Suffixaufnahme. (4) gizon-aren-ari man-Gen-Dat 'to that of the man'

Here are the skeleta of the three construction types whose classification is at issue (with blanks left in place to highlight differences): (5)

a. Suffixaufnahme proper

(head present and case-marked; e.g. (2)) [A, [B, [HeadNom, ExtCase, C], [AttrNom, IntCase, ExtCase, D]]NP]S b. Suffixhaufung

(head present but not case-marked; e.g. (3)) [A, [B, [HeadNom, C], [AttrNom, IntCase, ExtCase, D]]NP]S c. Hypostasis

(head absent; e.g. (4)) [A, [B,

[AttrNom, IntCase, ExtCase, D]]NP]S

As shown in Plank's historical narrative (Section 2), different linguists have come to different taxonomic decisions regarding the three constructions. Thus, in contrast with Finck, whose term "Suffixaufnahme" pertained only to the type involving a case-marked head (5a), Ferdinand Bork used the term "Suffixhaufung" to include both the Finckian head-marked pattern and also non-head-marked instances of the double-case-marked attributive pattern (5b); and, more recently, Heinrich Wagner has lumped together all three patterns, including (5c). For Finck's and Plank's concept of Suffixaufnahme, both the head and its case marker are obligatory; for Bork, the head was obligatory but its case marker optional; for Wagner, as for us, both the case marker of the head and the head nominal itself are optional. The last concept differs from the other two in that these crucial demands on the head both for its presence and for its form, while our schema is defined solely on the structure of the attributive nominal and makes no assumptions regarding the head.

454

Epilogue

It seems that as a working hypothesis, each of the three concepts— Suffixaufnahme proper, Suffixhaufung, and hypostasis—has prima facie validity. Since both the differences among the three patterns and their similarities are undeniable, the three will have to be represented in any account both as distinct types and as subtypes of the same type. What is at issue is whether they are all immediate functional subtypes of attributive structure (or, even more comprehensively, of constructions involving interrelated nominals; cf. Plank, Section 3), or whether any pair are more closely related to each other, forming an intermediate-level type within the taxonomy. The issue, as noted by Plank, is empirical, to be decided on grounds of what classes of constructions best serve the purposes of typological generalizations. Thus, the term "Suffixaufnahme"—used in this chapter for in lack of a better term, with a sense that is both noncompositional and historically inauthentic—-is meant to present a tentative hypothesis about the significance of a class of constructions for language typology. As far as I can tell, presently available evidence and theory do not adjudicate the issue. 2. The Taxonomy of Suffixaufnahme The definitional schema of Suffixaufnahme in (1) can be easily "telescoped" in and out so that it extends to more comprehensive classes of constructions, or zooms in on more specific ones. For example, if we relax the requirement that the two nominals involved be constituents of the same NP, we arrive at a broader class of constructions which includes also case agreement between a head and some non-attributive nominal satellite, such as what occurs in Walmatjari. In one of the examples cited from this language by Schweiger, meaning 'The man left the children there on his way to the sandhill', the word 'man' and the word 'sandhill' are both marked ergative, with the latter also carrying allative marking; but the two are not part of the same NP: (6) julji-karti-rlu . . . nganpayi-rlu . . . sandhill-All-Erg . . . man-Erg . . .

On the other hand, if we exploit the negative value of the option of having or not having an external case, we will have narrowed the scope by defining a particular subtype of Suffixaufnahme: Suffixhaufung. Examples of this pattern come from Aranda ((3) above) and Alyawarra (7). (7) aringka atjinh-ila dog I=Gen-Erg 'my dog'

This section will survey the range of constructions that can thus be derived as subtypes of Suffixaufnahme, and the range of supertypes, or hyperclasses, that in turn include it. We will begin with the latter. What are the genera of which Suffixaufnahme is a species?

Summing up Suffixaufnahme

2.1. Suffixaufnahme 2.1.1.

455

as Species

Morphology

First, we might view Suffixaufnahme as a morphological pattern: a construction where a nominal is doubly marked for case quite regardless of the syntactic conditions calling for such marking (8a), or, more generally, where a constituent, nominal or not, is doubly marked for the same category, whatever that category may be (8b). (8) a. [A, [Nominal, Case, Case, B], C] b. [A, [Constituent, Category1 Category1 B], C]

Double, or multiple, morphological case marking is not unusual. Assuming four morphological types of case markers—prefix, suffix, preposition, and postposition—the logical possibilities for two-member combinations are ten in number, of which only some will be exemplified here. (9) Preposition Preposition Preposition Prefix Preposition Suffix Preposition Postposition Prefix Prefix Prefix Suffix Prefix Postposition Suffix Suffix Suffix Postposition Postposition Postposition

Preposition Preposition: (10) English from near Boston; from under the table; for after nine; until after the holidays; This is a quote from "On the river" (11) Polish do po film-ie until after film-Loc 'until after the film'

Preposition Suffix: (12) German aus dem Haus(e) from the = Dof house=Dat 'from the house'

456

Epilogue

(13)

Latin ex dom-o from house-Dat 'from (the) house' Suffix Suffix:

(14)

(15)

Hungarian a. ez-t-et this-Acc-Acc 'this' (substandard; cf. standard ez-t this-Ace) b. ez a "Duna-nal"-bol van this the "Danube-at"-from is 'This is from [the poem/book entitled] "At the Danube" ' Jiwarli Juila-ngka-nguru tree-Loc-Abl 'from the tree' (cf. jurla-ngka tree-Loc 'on the tree'; *jurla-nguru tree-Abl)

Suffix Postposition: (16)

Hungarian a. a hid-on at the bridge-on across 'across the bridge' b. a Duna mellett-rol van the Danube next=to-from is 'He/she/it is from next to the Danube' c. a nyar elott-ig the summer before-until 'until before the summer'

Postposition Postposition: (17)

Hungarian a konyv "A Duna felett" mogott van the book the "Danube above" behind is 'The book is behind [the picture/book entitled] "Above the Danube" '

Adjacent case markers may hide diverse morphological and syntactic patterns. Sequences such as English from above, Polish do po, and Jiwarli -ngkanguru are simply compound case markers (for discussion see Austin and Schweiger, this volume). Expressions like from "On the river" and their Hungarian equivalents appear to involve the temporary lexicalization of a phrase

Summing up Suffixaufnahme

457

such as "On the river" with its internal structuring suspended. In spite of their syntactic diversity, there is some evidence that all of these instances, along with Suffixaufnahme, do belong to a single class. For example, resistance to a sequence of two identical case markers manifests itself across the various types. This constraint holds for Suffixaufnahme in Old Georgian (Boeder 1987: 47ff.), Jiwarli (Austin), Dyirbal (Schweiger), Martuthunira (Dench), and several other languages (Dench and Evans 1988: 36ff., Plank, 4.10), just as it holds in case marker sequences which are in all other ways unlike Suffixaufnahme. An example is (18a) from Hungarian, visa-vis (18b). (18)

a. ?a konyv a "Duna-na/"-na/ van the book the "Danube-af"-atf is 'The book is at [the picture entitled] "At the Danube" ' b. a konyv a "Duna mellett"-nel van the book the "Danube next=to"-at is 'The book is next to [the picture entitled] "By the Danube" '

Similarly, resistance to sequences of more than two case markers, as shown by some languages in Suffixaufnahme constructions even if the case markers are phonologically distinct, also holds outside Suffixaufnahme. The following examples are again from Hungarian. (19)

a. ?ez-t-et-et this-Acc-Acc-Acc 'this' (cf. ez-t and ez-t-et in (12a), both meaning 'this' in the accusative) b. ?a haz mogott-on tul the house behind-on beyond 'beyond behind the house' (cf. a haz mogott the house behind 'behind the house', a haz-on the houseon 'on the house', a haz-on tul the house-on beyond 'beyond the house')

The class within which all these examples fall, however, is likely to include not only multiple case markers but also sequences of any otherwise nonstackable constituents such as articles, conjunctions, and numerals, since their iteration, too, tends to resist phonological identity and decreases in acceptability with the increasing number of constituents involved. Possessed nouns that are marked both for their own number and for the number of the possessor, as in Hebrew or Turkish, have phonologically distinct markers for the two purposes, and so do verbs that agree with both their subject and their direct object, as they do in, say, Lebanese Arabic. From the point of view of pure form, therefore, the broadest genus that Suffixaufnahme may be subsumed under is that of stacked constituents, free or bound, phonologically identical or not.

458

Epilogue

2.1.2. Syntax Second, we may abstract away from the morphological side of Suffixaufnahme and consider its syntactic genera. Plank's comprehensive taxonomy (Section 3) does just this when it places Suffixaufnahme among subtypes of interrelated nominals. In his scheme, the internal case and the external case are functionally different: the former is a relation marker, indicating the kind of relation that obtains between that nominal and another; the latter is a relatedness marker, identifying the terms of the relation through agreement. Thus, in the Gumbaynggir example (2), repeated here, -yu, the ergative suffix on the attributive nominal, is a relatedness marker since it points at 'father' as being the nominal that 'child' is in some relation with; and -gundi, the genitive suffix, is the relation marker in that it specifies the nature of the relation between 'child' and 'father'. (2) ba:ba-gu junuy-gundi-yu father-Erg child-Gen-Erg 'the child's father (ergative)'

In a somewhat different scheme, to be explored here, both the external and the internal case are relation markers in that they both speak of the nature of the relationship between the two nominals. The relationship is twofold: on the one hand, there is an asymmetric relationship of dependency holding between the nominals; on the other, there is also a symmetric relationship of co-constituency obtaining between them. As a dependent, the attributive is a (potential) case assignee, or governee, of the head; as a co-constituent, it is a potential agreer controlled by the head, or more precisely by the NP. The corresponding general syntactic schema projected from that of Suffixaufnahme that captures this relationship is as follows: (20)

[A, [B, ([HeadNom, C]), [AttrNom, D]]NP]S

Double case is then the result of the two distinct relations that the attributive nominal is engaged in vis-a-vis the head: the internal case is the result of the dependency relation and is assigned by the head; the external case is the correlate of the co-constituency relation percolating from the NP that includes both nominals. The syntactic pattern of Suffixaufnahme is thus at the intersection of three construction types: attribution, government, and agreement. As an attributive pattern, its congeners include adjective-noun and adverb-verb constructions. As an instance of government, it is in a class with verb-complement structures. And as an instance of agreement, it is most closely related to instances of agreement-by-percolation, such as adjective-noun agreement in case and number. Suffixaufnahme is not the only kind of construction where the functions of

Summing up Suffixaufnahme

459

governee and agreer converge on the same constituent. The case of Russian and Polish numerated nominal heads is syntactically akin to Suffixaufnahme: the nominal is subject to case percolation from the NP, while its case is also governed by the numeral (Babby 1980, 1984, 1986, 1987, 1988, 1991; McCreight 1988: 56ff.). In the following example, the numeral requires plural genitive (21a) while the preposition 5 dictates that the NP be in the instrumental (21b). When both requirements are present, the resulting structure is different from Suffixaufnahme, however, in that only one case—the instrumental—actually appears (21c/d/e). (21) a. pjat' butylok five bottle=Gen 'five bottles' b. s butylkami with bottle=Ins 'with bottles' c. s pjat'ju butylkami with five=Ins bottle=Ins 'with five bottles' d. *s pjat'ju butylok with five=Ins bottle=Gen e. *s pjat'ju butylokami with five=Ins bottle=Gen=Ins

Conflicting marking demands imposed on the same constituent are not necessarily the joint result of government and agreement, and they do not necessarily pertain to case. Other patterns may involve multiple government, with a constituent being governed by more than one governor, or multiple agreement, with a constituent falling within the domain of two or more agreement controllers. An example of multiple government are direct objects of negated verbs in Russian, where the verb itself governs some case and the negator governs the genitive. Once again, only one case ends up being marked (Babby 1987, also Corbett 1978): (22)

a. Ivan vsegda est mjaso Ivan always eats meat=Ace 'Ivan always eats meat' b. Ivan nikogda ne est mjasa Ivan never not eats meat=Gen 'Ivan never eats meat' c. Ivan zaviduet bogatym Ivan envies rich = DatPl 'Ivan envies the rich'

460

Epilogue d. Ivan ne zaviduet bogatym Ivan not envies rich=DatPl 'Ivan does not envy the rich' e. *Ivan ne zaviduet bogatyx Ivan not envies rich=GenPl f. *Ivan ne zaviduet bogatyxym Ivan not envies rich=GenPl=DatPl

Other examples of multiple government are direct objects in Russian that are governed by two or more conjoined verbs, each subcategorized for a different case; or relative pronouns introducing free relatives in Polish, which are governed by both the subordinate and the matrix verb (McCreight 1988: 31ff., 25ff.). Examples of multiple agreement are adjectives modifying conjoined nouns of different numbers and/or genders, or verbs agreeing with subjects of different persons and/or numbers (Corbett 1983). Pullum and Zwicky (1986) provide a number of additional examples for feature conflict both in government and in agreement. In other cases, the multiple demands imposed upon a single constituent may not be the result of multiple relations: one of the demands may be "inherent."3 For example, possessed nouns in Hebrew, Turkish, and many other languages are doubly marked for number, with one marker indicating the number of the possessum itself and the other the number (and person) of the possessor. The pattern of sequence of tenses well-known from Latin and English might also be seen as an example of both inherent and acquired specification for the same category, viz. tense.4 In sum: syntactically, Suffixaufnahme belongs to a construction type where a single constituent is "under obligation" to signal grammatical information (case, number, person, etc.) for two or more distinct constituents. By source, the two kinds of conflicting grammatical information—schematically, A and B—may show six patterns: (23) a. b. c. d. e. f.

A is: by government by agreement by government inherent inherent inherent

B is: by government by agreement by agreement by government by agreement inherent

Suffixaufnahme belongs to type (c). This is so whether the external case is or is not marked on the head nominal. In both instances, the attributive nominal is not the "primary" receiver of the external case, and the fact that it nonetheless carries that case can be accounted for only by assuming agreement—whether sister-agreement with the head nominal, or agreement-by-percolation with the controller being the superordinate NP node.

Summing up Suffixaufnahme

461

2.1.3. Syntax and Morphology The preceding two sections have shown that, syntactically, Suffixaufnahme involves interrelated nominals of which the attributive (or secondary) nominal is both a sister and a dependent of the head, and that, morphologically, the construction involves multiple case marking. Do the two patterns necessarily go together, or can one occur without the other? That the morphological pattern of multiple marking for the same category (such as case) does not necessarily originate from multiple relations borne by the case-marked constituent was already shown in Section 2.1.1. What will be shown next is that the correlation is not necessary in the other direction, either: the pattern of a constituent bearing multiple relations does not by any means result in double morphological marking for the same category. First, it is possible that the constituent in question will carry some marking in reference to both relations, but that they will not both pertain to the same category. For example, the transitive verb in Hungarian agrees both with the subject and with the direct object but, while it shows number and person for the subject, it is oblivious (for the most part) to these features of the object and is sensitive only to the object's definiteness, a feature irrelevant in turn to subject agreement. Second, even if a constituent shows markings for two relations with respect to the same property, a sequence of two morphemes of the same type is not necessarily the only option, since the mode of the two markings may differ. For example, the English question or relative pronoun object whom is both the object and the focus of the verb; but of these two relations only one—objecthood—is encoded morphologically, with the other indicated by immediately preverbal position. Third, suppose there is a constituent which bears multiple relations, and as a result the same grammatical category is to be marked on it with both markings being of the same mode, viz. morphology. Even so, double marking of the sort Suffixaufnahme exemplifies is not the only option: alternatively, double marking may either be avoided or resolved. Given two markers A and B, both slated to appear on a constituent, the pertinent logical possibilities regarding their actual fate are as follows: (24)

a. b. c. d. e.

A appears, B does not both A and B appear a third marker C appears no marker appears the construction fails

Applying this to the marking of an attributive nominal, the options are as follows: (25)

a. a. internal case appears, external case does not P. external case appears, internal case does not

462

Epilogue b. c. d. e.

both external case and internal case appear a third case marker appears no case marker appears the construction fails

Suffixaufnahme is (b); all of the others are related to it as alternative solutions to the same problem. I have no example for (d). While the remaining four options all actually occur, some are more frequent than others. Babby (1984: 8) cites one of the few examples of an irresolvable case conflict—(e). Examples where the two cases are represented as a single portmanteau morpheme are of type (c) (cf. Dench and Evans 1988: 42), and so are the patterns in Old Georgian and some Daghestanian languages where the genitive case has different forms depending on the external case (cf. Boeder, Boguslavskaja, and Kibrik, this volume).5 The most common resolution of case conflicts is (a), the strategy also most often used in resolving other feature conflicts: one of the competing features gains precedence over the other. Dench and Evans have examples for either of the two options, the external case winning over the internal case or vice versa. Case attraction phenomena (e.g. in Latin or Polish relative clauses) show the prevailing of the external case (a ), but generally it is the internal case that appears, and the external case remains unmarked (a ): adnominal genitive attributes are normally immune to percolation (cf. Babby 1987: 94, McCreight 1988). Russian exemplifies this: (26) a. bel-aja knig-a devusk-i white-Norn book-Nom girl-Gen 'white book of the girl' (subject) b. bel-uju knig-u devusk-i white-Ace book-Acc girl-Gen 'white book of the girl' (object) c. *bel-uju knig-u devusk-u white-Ace book-Ace girl-Ace 'white book of the girl' (object) In the case of Russian numeral phrases, case competition is resolved in terms of a more complex hierarchy (Babby 1987): (27) lexical case > genitive > configurational case This was exemplified in (21) above, where the phrasepjat' butylok, when used as subject, shows the genitive form butylok rather than the nominative because the numeral governs the genitive, and genitive takes precedence over the nominative, which is a configurational case. But if the phrase includes the preposition i 'with', which governs the instrumental, this lexical case requirement will take precedence over the genitive, with the resulting construction showing the noun in the instrumental: s pjat'ju butylkami.

Summing up Suffixaufnahme

463

Another hierarchy of prioritizing case markers is based on the distinction between pragmatic and grammatical cases. In most languages that have morphological topic markers as well as grammatical case markers, such as Samoan or Tagalog, the two cannot co-ocur: topics end up getting marked for topichood (with the preposition ang in Tagalog and o in Samoan) but not for grammatical case. This is generally so for subjects and objects in Japanese as well, but oblique cases, such as locative, apparently "stick;" they may co-occur with the topic marker wa, thus giving rise to double marking (Babby 1980: 25): (28)

a. niwa-ga garden-Norn b. niwa-wa garden-Top c. * niwa-ga-wa d. *niwa-wa-ga e. niwa-de garden-Loc f. niwa-de-wa garden-Loc-Top

Suffixaufnahme now reveals itself as an unusual resolution of a familiar problem. The familiar problem—a universal one, according to Babby (1984: 13f., 1987: 138, 1991: 7)—is how to encode two distinct cases converging on the same constituent. The unusual resolution is actually the most straightforward one: by spelling out both. Thus, as observed by Dench and Evans (1988: 44f.), from the point of view of the relationship between syntax and morphology, Suffixaufnahme stands out as being maximally compositional, showing a one-to-one correspondence between syntactic relations and morphological cases. While from the point of view of languages like Russian that choose to mark only one of the two competing cases, Suffixaufnahme is a problem in need of an explanation, looking at things from the angle of Old Georgian or Gumbaynggir, one begins to wonder why all languages do not just "bite the bullet" and encode both cases rather than forcing a choice between the two. In other words, there is equal justification for asking either of these two questions: Why is 'the child's father' (ergative) in Gumbaynggir (29a) and not (29b)? Why is 'five bottles' in Russian (30a) and not (30b)? (29)

a. ba:ba-gu junuy-gundi-yu father-Erg child-Gen-Erg b. *ba:ba-gu junuy-gundi father-Erg child-Gen

(30)

a. s pjat'ju butylk-ami with five=Ins bottle-Ins b. *s pjat'ju butylok-ami with five=Ins bottle=Gen-Ins

464

Epilogue

In sum, morphologically Suffixaufnahme is an instance of multiple marking for the same category; syntactically, it is an instance of a constituent being a candidate for marking each of the multiple relations it is involved in—and, more specifically, of a construction where the multiple case relations are straightforwardly indicated by multiple morphological markers.

2.2. Suffixaufnahme

as Genus

Having surveyed some construction types under which Suffixaufnahme is subsumable, let us now narrow rather than broaden the definitional schema (1), repeated here: (1) [A, [B, ([HeadNom, (ExtCase), C]), [AttrNom, IntCase, ExtCase, D]]NP]S

How widely can patterns of Suffixaufnahme differ from one another without ceasing to be such? Possible parameters of variation can be read off the definitional schema: — terms — kinds of head nominal — kinds of attributive nominal — kinds of internal case — kinds of external case — kinds of A, B, C, D — relations — linear relations — between head nominal and attributive nominal — proximity — precedence — between attributive nominal, internal case, and external case — proximity — precedence — prosodic relations — between head nominal and attributive nominal — between attributive nominal, internal case, and external case

Let us now survey the actual variation found. The variables A and B will not be discussed since they do not appear to make a difference. The discussion parallels the contents of Section 4 in Plank's introduction; some parameters of variation are also addressed by Aristar. 2.2.1. Head Nominal and its Satellites

The head nominal may be a sole noun, or it may have modifiers, quantifiers, and determiners. These options do not seem either to encourage or to obstruct the central pattern of the double case marking of the attributive nominal.

Summing up Suffixaufnahme

465

The head itself may or may not be case-marked; as noted in Section 1, the standard notion of Suffixaufnahme assumes that it is. The head nominal of a particular attributive nominal may itself be an attributive to another head. In other words, Suffixaufnahme may be a recursive construction, as in Old Georgian (Boeder) and Awngi (Hetzron). Lehmann (1983: 359ff.) predicts that Suffixaufnahme should actually be restricted to possessors of alienable heads. He argues that while possessors of inalienable possessions are complements of the head, possessors of alienable ones are modifiers, and thus case agreement with the head, being a characteristic of modifiers, makes sense in the latter case while it does not in the former. In actuality, however, Suffixaufnahme occurs with both alienable and inalienable heads. Finally, Suffixaufnahme, interpreted broadly, occurs both with heads present and with heads absent. 2.2.2. Attributive Nominal and its Satellites Attributive nominals come in different sizes and shapes: they may be pronominal or nominal, animate/human or inanimate/non-human, personal names or other. They may or may not have modifiers and other satellites; if so, these may or may not carry the same case markings as the noun itself. An interesting case marking pattern is described by Corbett for Upper Serbian. As he points out, Upper Serbian does not have Suffixaufnahme in that there is no single attributive word showing both the external and the internal case. But the language has a construction very similar to Suffixaufnahme, where both the external and the internal case are marked in the possessor phrase, albeit not on the same constituent: the attributive noun has the external case and its modifier has the internal one. In (31), the attributive noun carrying the external case is 'brother', and its modifier showing the genitive is 'my': (31)

moj-eho bratr-ow-e dzec-i my-Gen brother-Deriv-Nom child-Nom 'my brother's children'

If the head is case-marked, the corresponding external case marking on the attributive nominal may or may not be a phonologically exact copy of the head's case marker. For example, Austin points out that in Jiwarli and Thalanyji, if in an object NP the possessor is animate and the possessum is inanimate, then the possessum will take the absolutive as inanimate objects generally do, but the possessor will take the accusative as appropriate to animate objects. Similarly, Testelec (1991) points out that in Old Georgian, the attributive nominal will show the external case even if the head belongs to a morphological class that happens to lack the case in question. These facts suggest that the case marker on the attributive nominal is not necessarily a direct copy of that of the head.6

466

Epilogue

The attributive nominal may show agreement in categories other than case. These are number (as in Chukchi, Hurrian, Old Georgian, Kashmiri, and some Indie languages), gender (as in Agaw, Kashmiri, and Indie), and predicativity (Chukchi).7 2.2.3. Internal Case Most commonly the adnominal case involved is the genitive, but often other cases may also participate in the construction. Dench and Evans (1988: 7ff.) list genitive, locative, ablative, allative, and perlative as pertinent adnominal cases in Australian languages. Examples with the locative as internal case are also provided for Martuthunira by Dench. Evenki assigns external case to some comitative complements.8 In Punjabi, locative postpositions also participate in Suffixaufnahme, and so does the postposition 'for' in Kashmiri (Payne). Indie languages and Kashmiri are also unusual in that the internal case is marked by adpositions rather than affixes. 2.2.4. External Case The range of external cases is generally not restricted. For examples of several external cases figuring in Hurrian Suffixaufnahme, see Wilhelm's chapter. In a construction type related to Suffixaufnahme where the genitive has different forms depending on the external case, apparently the distinction between direct and oblique case9 is significant, since the genitive only has two forms: one for instances where the external case is direct, and another where it is oblique.10 As noted above, the external case may or may not be the same as the internal case; if it is phonologically the same, languages may ban the two-case sequence.11 2.2.5. Ligative In some languages, such as Dyirbal and Hurrian (see Wegner, Wilhelm, and Plan, 4.11), the attributive nominal includes, in addition to the two case markers, a third element—"a special carrier suffix" (Plank 1990) or ligative— wedged in between the two. 2.2.6. Linear Relations The attributive nominal may precede or follow the head, and it may or may not be adjacent to it. The construction thus does not crucially depend on either proximity or precedence. There are, however, more preferred and less preferred orders, as will be discussed in Section 3. The case markers involved in Suffixaufnahme are generally suffixes, presumably reflecting the overwhelming cross-linguistic dominance of postposed case affixes outside Suffixaufnahme as well. Nonetheless, some instances in-

Summing up Suffixaufnahme

467

volving prefixes, adpositions, and tonally marked case have been unearthed by Plank (4.6, 4.7). The order of the case markers is generally iconic in that syntactic closeness of the constituent whose case marking is at issue is mirrored by closeness of the marker to the attributive nominal's stem (Plank 1990). The only exception seems to be Kayardild, as discussed (and historically explained) by Evans, where a higher case precedes a lower one. The ligative, if one occurs, is always between the affixes. 2.2.7.

Prosodic Relations

Even if the attributive nominal and the head are adjacent, they may be separated by a pause. There is evidence for this from one language, Old Georgian (Boeder). 3. The Distribution of Suffixaufnahme So far we have discussed what Suffixaufnahme constructions may be like. The second question to be addressed here concerns the conditions under which the construction itself and its several varieties occur. In other words, we will now attempt to anchor the very occurrence and the particular properties of Suffixaufnahme that were surveyed in Section 2 in their context. In the simplest of all worlds, the following would be true: (i) (ii)

All languages have Suffixaufnahme. All instances of Suffixaufnahme have properti(es) Y (W, Z, etc.).

Statement (i) is not true in the real world: Suffixaufnahme is far from being universal. Regarding (ii), there is only one (near-)universal property of Suffixaufnahme constructions; it pertains to the order of affixes:12 (Gl)

In almost all instances of Suffixaufnahme, the segmental marker of internal case is closer to the stem of the attributive nominal than is the segmental marker of external case.

This generalization was suggested by Dench and Evans (1988: 6f.) and Plank (1990); it holds for the linear order of both suffixed and prefixed markers (cf. Plank, 4.9). It is a special case of a more general linearization pattern proposed by Bybee (1985), Baker (1985), and Sadock (1991). The single exception is Kayardild (Evans). In view of the non-universality of the occurrence of Suffixaufnahme and of the fact that most of its particular properties—beyond the order of affixes and those present by definition—are variant, we are faced with the following tasks: since Suffixaufnahme is not universal for the domain of all languages, we want to find a subuniverse of languages where it is universal; and since

468

Epilogue

most of its properties are cross-linguistically variant, we want to find those subuniverses within which certain properties are uniformly present. In other words, we will search for statements of these two kinds—both compromised versions of (i) and (ii) above: (i') All languages that have X have Suffixaufnahme. (ii") All occurrences of Suffixaufnahme that have properti(es) X (Q, U, etc.) also have properti(es) Y (W, Z, etc.).

Depending on the type of implicans, there are several possible types of generalizations that we might find to hold: (i)

(ii)

(iii)

(iv)

(v)

properties of the construction implied by other properties of the same construction Question: Can we predict from one characteristic of a Suffixaufnahme construction some other characteristic of it? Schema of generalizations sought: If a Suffixaufnahme construction has property X, it also has property Y. subtypes of the construction implied by other subtypes of the construction in the same language Question: Can we predict from the occurrence of one subtype of Suffixaufnahme the occurrence of another subtype within the same language? Schema of generalizations sought: If a language has Suffixaufnahme of Type A, it also has Suffixaufnahme of Type B. (subtypes of) the construction implied by properties of the including sentence Question: Can we predict from some property of a sentence the occurrence or subtype of Suffixaufnahme in that sentence, or vice versa? Schema of generalizations sought: If a sentence has property X, it also has Suffixaufnahme (of Type A), or vice versa. (subtypes of) the construction implied by some property of the language Question: Can we predict from some property of the language the occurrence or subtype of Suffixaufnahme, or vice versa? Schema of generalizations sought: If a language has property X, it also has Suffixaufnahme (of Type A), or vice versa, (subtypes of) the construction implied by some property of an areally adjacent language Question: Can we predict from the fact that one language borrowed X from another that it will also have borrowed Suffixaufnahme (of Type A), or vice versa?

Summing up Suffixaufnahme

469

Schema of generalizations sought: If a language has borrowed X from another, it will also have borrowed Suffixaufnahme (of Type A), or vice versa, (vi) (subtypes of) the construction implied by some property of a temporally preceding or following language stage Question: Can we predict where Suffixaufnahme comes from and what it turns into? Schemata of generalizations sought: If a language stage has construction C, then if C changes, it must, or may, turn into Suffixaufnahme (of Type A). If a language has construction C, then if C was something different before, that different construction was, or may have been, Suffixaufnahme. If a language stage has Suffixaufnahme and it is followed by a language stage which does not have it, then that language stage will have X instead. If a language stage has Suffixaufnahme and it is preceded by a language stage which does not have it, then that language stage will have X instead.

3.1. Internal Typology Let us first consider questions (i) and (ii) above. They pertain to the internal typology of the construction, since they have to do with correlations among properties of Suffixaufnahme that may hold syntagmatically (i) or paradigmatically (ii). The discussion will be structured in accordance with the overview of the property types of Suffixaufnahme given in Section 2.2. 3.1.1. Head Nominal and its Satellites It was noted in Section 2.2 that the head nominal involved in Suffixaufnahme, interpreted broadly, may be present or absent. As pointed out by Boeder, these two subtypes have a skewed distribution, with one of the two crosslinguistically favored. In particular, while in the majority of Suffixaufnahme languages the construction occurs both with present and with elided heads, and while there are also languages—such as Archi (Kibrik) or perhaps Huallaga Quechua (McCreight 1988: 164)—where Suffixaufnahme occurs only as hypostasis, i.e. in elliptical constructions, there are no languages where Suffixaufnahme is restricted to headed constructions. Thus, following Boeder, (G2) may be hypothesized:13 (G2) In all languages, if Suffixaufnahme occurs with head present, it also occurs with head missing. It was further noted above that the head may itself be a complex phrase including a head and an attributive. Again, there appears to be a hierarchy of preferences In some Innonacrpi; both simnlf -jnH mnlti-l