Information Structure and Language Change: New Approaches to Word Order Variation in Germanic

Information Structure and Language Change

≥

Trends in Linguistics Studies and Monographs 203

Editors

Walter Bisang (main editor for this volume)

Hans Henrich Hock Werner Winter

Mouton de Gruyter Berlin · New York

Information Structure and Language Change New Approaches to Word Order Variation in Germanic

Edited by

Roland Hinterhölzl Svetlana Petrova

Mouton de Gruyter Berlin · New York

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.

앝 Printed on acid-free paper which falls within the guidelines 앪 of the ANSI to ensure permanence and durability.

Library of Congress Cataloging-in-Publication Data Information structure and language change : new approaches to word order variation in Germanic / edited by Roland Hinterhölzl, Svetlana Petrova. p. cm. ⫺ (Trends in linguistics. Studies and monographs ; 203) Includes bibliographical references and index. ISBN 978-3-11-020591-6 (hardcover : alk. paper) 1. Germanic languages ⫺ Word order. 2. Germanic languages ⫺ Syntax. I. Hinterhölzl, Roland. II. Petrova, Svetlana. PD380.I54 2009 4301.45⫺dc22 2009011896

ISBN 978-3-11-020591-6 ISSN 1861-4302 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. ” Copyright 2009 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Cover design: Christopher Schneider, Laufen. Printed in Germany.

Table of Contents

Introduction Roland Hinterhölzl and Svetlana Petrova ................................................. 1

The role of information structure in the grammar The verb-second property in Old High German: Different ways of filling the prefield Katrin Axel ............................................................................................... 17 The role of information structure in word order variation and word order change Roland Hinterhölzl ................................................................................... 45 OV languages: Expressions of Cues Þorbjörg Hróarsdóttir .............................................................................. 67 Discourse relations and word order change Ans van Kemenade ................................................................................... 91

Methodological problems of the information-structural analysis of data from historical text corpora On the methods of information-structural analysis in historical texts: A case study on Old High German Svetlana Petrova and Michael Solf ........................................................ 121 Paleographic clues to prosody? – Accents, word separation, and related phenomena in Old High German manuscripts Jürg Fleischer ........................................................................................ 161 On the “syntax of silence” in Proto-Indo-European Thomas Krisch ....................................................................................... 191

vi

Table of Contents

Information-structural categories in the main texts of the early German inheritance Word order variation and information structure in Old High German: An analysis of subordinate dhazs-clauses in Isidor Eva Schlachter ........................................................................................ 223 Information structure and word order variation in the Old High German Tatian Svetlana Petrova .................................................................................... 251 Verb placement and information structure in the OHG Gospel Harmony by Otfrid von Weissenburg Andreas Lötscher .................................................................................... 281 Translating information structure: A study of Notker’s translation of Boethius’s Latin De Consolatione Philosophiae into Old High German Rosemarie Lühr ...................................................................................... 323 Aspects of word order and information structure in Old Saxon Sonja Linde ............................................................................................. 367 Subject index .......................................................................................... 391

Introduction Roland Hinterhölzl and Svetlana Petrova

1. Motivation The present volume contains papers dedicated to the explanation of word order variation and change in the history of the Germanic languages. It relates to one of the most prominent and widely discussed topics in the syntax of the Germanic languages. In this volume, these classical topics are approached from the perspective of a novel account to grammar change (cf. Lightfoot 1999, Hróarsdóttir 2003, Hinterhölzl 2004) in which slow gradual language change paves the way for abrupt changes in grammar. In particular, Hinterhölzl (2004) proposes that one important link between language use and grammar is constituted by the grammatical means of information packaging or shortly information structure. Since information packaging represents a core means of the speaker to create novel patterns and expressions for rhetoric expressivity and particular communicative needs, it serves as one of the major sources of variation. This approach is further developed in the Research Project B4 which is anchored in the collaborative research center (SFB 632) on Information Structure at Potsdam University and the Humboldt University in Berlin. In a large-scale investigation on the interaction between information-structural and grammatical conditions in the development of word order regularities in the Germanic languages, new explanations for the diverse developments of these languages are sought. The novelty of this approach consists in three main points. First of all, it aims at describing variation in word order not as a result of language contact, as in previous research (cf. Kroch 1989, Pintzuk 1999), but rather as a language-internal phenomenon which is due to stylistic effects related to information packaging and discourse structure. Secondly, bearing in mind that information structure is a completely new field of study in historical linguistics, it develops methods for the analysis of information-structural phenomena in texts from historical corpora that are available only in written form and bear no or little cues to prosody and intonation. Thirdly, the project carries out research on a much larger scale than previous research

2 Roland Hinterhölzl and Svetlana Petrova on the development of Germanic word order, including empirical studies on large corpora of Old High German (OHG), Old Saxon (OS) and Old Icelandic (OI). The present volume features important results of this enterprise as well as a number of selected papers from the international conference on Word order variation and change, organized by Project B4 in October 2005 in Berlin.

2. Word order Variation and Word order Change in Germanic The modern Germanic languages display a great variety of different word orders. In German the object precedes the verb (OV), in English it obligatorily follows the verb, while Icelandic allows for both word orders (OV/VO). In German main clauses, the finite verb appears in second position (V2), Icelandic also shows V2-patterns in embedded clauses, while English did not develop a generalized V2-rule at all. The issue of how it came about that so closely related languages developed such diverse word order patterns was one of the main research questions in Germanic linguistics in the past hundred years. Research in this tradition has focussed on the following issues.

2.1. What is the basic word order in Germanic? One of the central issues was the question of what counts as the base position of the verb in Germanic. In the oldest Germanic texts, the verb appears in a great number of different positions: besides V2 and verb final orders, V1 and medial positions of the verb are quite common as well. Moreover, all these different orders can be found in main as well as in embedded clauses. While Braune (1894) held that verb placement is free in Germanic, Erdmann (1886) and Tomanetz (1879) assumed that due to its high frequency, V2 order must display the base position of the verb. At the end, Delbrück (1911), building on Wackernagel’s (1892) work, succeeded in arguing that the verb-final pattern should be treated as the basic pattern in Germanic. Newer investigations, however, indicate a high proportion of VO orders in the older stages of all Germanic languages (cf. Kavanagh (1970), who claims that 50% of embedded clauses in OHG display VO order). Given hat OV order is to be considered as the basic word order, word orders with

Introduction

3

postposed objects or preposed verbs were treated as marked word orders whose functions were described with the unanalysed notion of emphasis (cf. Näf 1979, Dittmer and Dittmer 1998). According to Fleischmann (1973) preposing of the verb (V1 or V2) in OHG served the purpose of putting the situation described by the verb into the foreground, while postposing of the verb signalled that the situation described by the verb is to be interpreted as part of the background. In this scenario, the differentiation of verb placement in main and embedded clauses in German can be analyzed as the grammaticalisation of different word order patterns that were originally motivated by information structure.

2.2. The development of V2 Following Delbrück, the question arises how V2 orders have developed from original verb final orders in Germanic. The traditional answer is that according to Wackernagel’s law (which has clitics move into second position in Indo-European languages), light finite verbs start to be placed after the accented first constituent in main clauses in Germanic; a movement operation that was later extended to all finite verbs. A very influential proposal about the development of V2 was put forward by Vennemann (1974). He assumes that Germanic had inherited the basic verb final order (SXV) from Indo-European and that the V2 pattern first initiated with the preposing of thematic stressed constituents. In this scenario, TVX (initial V2) constitutes an intermediate stage in the development of OV languages to VO languages. Fourquet (1974) proposes a completely different account of the development of V2. In his treatment, V2 results from a gradual process that places the verb in a sentence-medial position due to postposing of nonpronominal constituents. In his scenario, an intermediate pattern arises in which pronominal elements precede the verb in main clauses that is well documented in Old English (OE) texts of the earlier period. Until recently it was assumed that all Germanic languages, including English, had developed a generalized V2-rule. Newer studies of the development of verb placement in English have shown that OE only displayed a residual V2 pattern: V2 only appears if an operator (wh-word or negation) is preposed, while the preposing of constituents without operator status yields V3 word order. In order to account for residual V2, Kiparsky (1995, 1996) introduced structural positions for Focus and Topic in the C-domain. In his account,

4 Roland Hinterhölzl and Svetlana Petrova only movement of an operator into the Focus position is accompanied by movement of the verb, while with preposed topics, analyzed as occupying adjoined positions to the CP, the verb may stay in a lower position. This account raises the question of why OHG was rather different from Old English in this respect, since already in this stage of German, preposing of a non-operator constituent regularly triggered V2 as well.

2.3. The change from OV to VO in English Most studies about the development of English into a pure VO language assume that OE was an OV language of the modern German type and that VO structures are innovations that came about and spread in the Middle English (ME) period. It is important to note that at the same time in which OV structures disappear, English lost most of its nominal and verbal morphology. Therefore this development was characterized as a process in which due to the loss of morphological distinctions, verb placement serves to distinguish the grammatical functions of subject and objects (cf., among others, Vennemann 1974). Along these lines, Roberts (1997) tries to explain the change in word order with the loss of Case morphology in Early Middle English (EME). Note, however, that this account must remain unsatisfactory under a wider Germanic perspective, since it raises the question why Dutch had not also developed into a VO language. There is a growing number of works that take into account the great variety of word orders in early English. Most prominent among these are the accounts of Kroch (1989) and Pintzuk (1999) which relate the presence of OV and VO word orders to the availability of an OV grammar and a VO grammar (double base hypothesis). According to them the availability of two grammars is due to language contact between Anglo-Saxons (basic OV grammar) and Scandinavian settlers (basic VO grammar) in the 10th century. Also this account remains unsatisfactory since a) one of its presuppositions, namely that the Scandinavian settlers spoke a VO language, cannot be proven and b) it does not take into account that already OE displayed VO word orders. Since VO orders can also be found in OHG and in OI (cf. Hróarsdóttir 2000), it is reasonable to assume that VO orders in English are not due to language contact but are part of the common inheritance of the Germanic languages. So far it has not been investigated whether the availability of OV and VO orders in OE could not be due to information-structural factors and

Introduction

5

there are no comparative investigations of OV and VO orders in the oldest stages of Germanic, that is, in OE, OHG and OI.

2.4. The development of the sentence bracket in German As discussed above, the positioning of arguments and adjuncts after the verb constitutes a common word order pattern in the early stages of the Germanic languages. Given that the basic word order in these languages was verb final, these word orders were analysed as involving exbraciation or extraposition (cf. Behaghel 1932). In the German tradition, the problem of extraposition was discussed under the perspective of the development of the German sentence bracket and constituted the subject of a large number of empirical studies. Considering only the later stage of OHG, there are three studies that note a significant reduction of extraposed direct objects in this period (cf. Bolli 1975, Näf 1979 and Borter 1982). Unfortunately, these studies are purely descriptive and do not take into consideration the information-structural role of extraposed constituents. In general, most of these studies treat the development of the German sentence bracket as a purely stylistic phenomenon that is brought about by socio-cultural factors: the influence of Latin classical writers (Behaghel 1892) or the spreading of the written style of formal bureaucratic language (Ebert 1999). In all these studies, the role of grammar in yielding (different) stylistic preferences is not taken into account. It is remarkable that the development of English has been described as a change in grammar while the development of German has been described as a purely stylistic change in most formal studies, including the approach of Lenerz (1984), which only relates the development of V2 to a change in grammar, namely the introduction of an obligatory C-projection. Formal treatments of the development of word order regularities in Germanic have mostly ignored the role of information structure so far. Traditional and functional accounts of these developments have often pointed out the influence of information-structural parameters on these processes, but failed to show how stylistic change can be related to grammar change. Recent works in generative theory on restrictions on movement as well as on restrictions on word order in the left periphery have proposed to integrate pragmatic notions like topic and focus into the syntax by defining pragmatically determined syntactic nodes in the split C-domain (Müller & Sternefeld 1993, Rizzi 1997). These proposals about the information-struc-

6 Roland Hinterhölzl and Svetlana Petrova tural organisation of the C-domain, as discussed above, has been adopted by Kiparsky (1995, 1996) and applied to the analysis of the left periphery in Indo-European languages. Furthermore, several studies on the representation of focus have shown that focussing of a constituent involves changes in word order in many languages. In the following, Zubizarreta (1998) developed a general theory about the interaction between word order, prosody and information structure. The application of these theories and of insights on the interaction between core grammar and information-structure in the analysis of older stages of Germanic is still wanting. Finally, as already pointed out above, in the new model of change in which variation in language use is crucial in processes of grammar change, information structure acquires a central role in explaining word order variation and change: the expression of IS-categories in terms of different grammatical means and their more pronounced or reduced use by speakers in specific contexts and communicative situations can be taken as a major source of synchronic variation as well as a crucial link between grammatical form and stylistic form. This volume collects a number of papers that – by taking into account information-structural factors – try to relate formal and functional explanations in order to yield a more complete and unified account of the development of word order regularities in Germanic.

3. The structure of the book The volume consists of three sections, each dedicated to different aspects of the complex task of investigating the role of information structure in historic texts. First, there is the question of which information structural notions are relevant for explaining word order variation. Then, the question arises of how to analyse the interaction between core grammar and information-structural conditions. These questions are addressed in the theoretical contributions in Section 2. The authors discuss pertinent theoretical issues concerning word order variation and word order change and review current research on the role of information structure in the grammar. They agree that information structure is a factor responsible for the emergence of new word order patterns and constructions in the early stages of the Germanic languages, which provided the basis for the further development of word order regularities in the individual Germanic languages and their dialects.

Introduction

7

Second, there is the question of how to retrieve IS-categories in texts of historical corpus languages, i.e. in texts available in written form only, with no or very restricted access to prosodic information. It is well-known that the Germanic languages are languages in which information structure is reflected prominently in the intonation of constituents. These issues are addressed in detail in Section 3, where different philological and linguistic methods of finding secure cues to unmarked word order and information structure are presented. Section 4, finally, turns to the information-structural analysis of historical texts proper. The papers present individual analyses of the main texts of the Old German corpus and show to which degree an approach is successful which tries to make sense of word order variation in terms of IScategories and to which degree such an approach allows to make generalisations across different texts and genres.

3.1. Information structure and word order variation The contributions in Section 2 feature individual studies on word order variation in OHG, OE and OI, arguing that the consideration of information structure offers new and interesting ways of explaining the pertinent variation. The contribution by Katrin Axel focuses on word order variation in the left periphery in OHG and investigates the status of V2 in OHG texts from Isidor to Notker. Axel argues that the modern V2-system was already in place in OHG leading to the regular fronting of wh-operators and topics, while CP-expletives were still missing in OHG. According to Axel, this system existed parallel to an older system in which clause type was not signalled via word order but by a limited class of particles. Axel surmises that the older system disappeared when the innovative rule of stylistic fronting was extended, giving rise to a further consolidation of the dominant V2-pattern. The contribution by Roland Hinterhölzl focuses on word order variation in the right periphery in OHG and discusses the factors that led to the establishment of the famous sentence bracket in modern German. Hinterhölzl argues that word order variation within one grammar is due to the expression of different IS-categories and word order change involves a change in the mapping between syntactic structure and prosodic structure in which information structure plays a crucial role. The contribution by Thorbjorg Hróarsdóttir discusses aspects of word order variation in the right periphery in OI showing that this variation is

8 Roland Hinterhölzl and Svetlana Petrova due to a complex interaction between prosodic and information-structural factors. Hróarsdóttir also discusses the change to basic VO order in Icelandic, showing that traditional treatments fail to account for the slow gradual loss of OV orders, and instead argues for a change in language use, namely in the expression of IS-categories, which paved the way to the eventual change in grammar. The contribution by Ans van Kemenade addresses aspects of word order variation in the middle field in OE showing that a certain class of short adverbs separates the domain of discourse-given and discourse-new constituents in the clause, pointing out the similarities of these patterns with scrambling in German. Van Kemenade argues that the syntax, though constituting a tightly circumscribed system, offers a limited amount of options which defines the space in which language users can manipulate word order according to information-structural needs. According to van Kemenade, this system of IS-based ordering of constituents disappeared, when a specific type of referring expressions, namely simple demonstratives, were lost in ME.

3.2. How to identify unmarked word order and IS-categories in a corpus language The contributions in Section 3 address various methodological issues concerning a) the identification of the unmarked word order in a corpus language displaying mixed word orders and b) the identification of IScategories in historic texts that provide no or very little clues to prosody. The contribution by Thomas Krisch uses various cross-linguistic generalisations concerning deletions in coordination and in ellipsis to probe the unmarked word order in ancient Indo-European languages like Latin, Greek, Sanskrit and Hittite. Krisch concludes from evidence for backward gapping that these languages were underlyingly SOV languages, while forward gapping is interpreted as a reflex of V-to-C movement. Furthermore, Krisch argues that deletions in coordination provide insights into the information-structural value of the constituents in the clause. The contribution by Jürg Fleischer constitutes a careful philological study of different palaeographic phenomena, foremost accent diacritics and word separation, as potential clues to prosodic information in different manuscripts of the OHG period. Fleischer addresses the important issue of whether diacritics signal prosodic prominence on a single word in the text. His analysis shows, however, that each of the OHG manuscripts displays

Introduction

9

its own system of accentuation in which diacritics may serve more than one single function. While in Notker’s texts, accents are used to mark vowel quantity, the same diacritics are devices for different types of graphical and phonetic disambiguation in other texts. Nevertheless, Fleischer provides examples in which the use of an accent is related to prosodic prominence in the text. But any conclusion on the role of accentuation as a strategy of marking information structure has to take into account the remaining functions of diacritics in the particular record. This result leaves the historical linguist with a careful and stepwise procedure of text interpretation as the only reliable method of identifying IS-categories in historic texts. The contribution by Svetlana Petrova and Michael Solf addresses the different issues one is confronted with in investigating the role of information structure in word order variation in historic texts. The first issue is how to identify the structures that represent instances of authentic OHG word order, given the fact that the large texts of this period are either poetic or constitute translations from Latin. The second issue concerns the determination of the information-structural role of a constituent in a text and presupposes an analysis that breaks up the rather controversial informationstructural notions, like topic, focus and background, into their constitutive features that can be used to make the assignment of IS-categories in a given text sequence on the basis of contextual interpretation as objective and tractable as possible. The exposition of this methological procedure is embedded in a thorough discussion of the analysis of the basic IS-notions in the literature and concludes with some illustrative examples of how IScategories are to be assigned to a given constituent in the text of the OHG Tatian translation.

3.3. Information structure in texts: five case studies The contributions in Section 4 constitute detailed empirical studies addressing the question of whether the consideration of information-structural aspects can provide us with a deeper and coherent explanation of the great amount of word order variation in the main texts of the German inheritance. Thus, we selected contributions from researchers which combine the theoretical linguistic knowledge with the necessary philological expertise on the individual text. If not indicated by the author itself, the contributions of Section 4 use the same IS-notions as defined by Petrova and Solf in Section 3, in order to guarantee that the individual results in this section can be evaluated and compared with each other.

10

Roland Hinterhölzl and Svetlana Petrova

The contribution by Eva Schlachter discusses word order variation in the OHG Isidor translation which is among the earliest texts of the OHG corpus. She examines the position of the finite verb in subordinated clauses introduced by dhazs. She observes that an extraordinarily high percentage of ‘light’ verbs, i.e. copula and auxiliary verbs, end up in a higher clause position yielding structural variation in the right periphery. Schlachter also demonstrates that variation in the positional realisation of main vs. light verbs in the Isidor text is associated with different strategies to highlight new and contrastive information in the clause. The contribution by Svetlana Petrova addresses the role of information structure in word order variation in the right periphery in the largest and oldest prosa text of the German inheritance, namely the Tatian translation. Based on a thorough investigation of word order in embedded clauses, she shows that constituents are mapped according to their informationstructural contribution into three different fields of the clause: a) discourse given nominal constituents are realized adjacent to C, b) constituents that are narrowly (contrastively) focussed are realized left-adjacent to the finite verb, while c) constituents that represent new information focus are realized after the finite verb. While the placement of discourse given material is identical to that in modern German, OHG seems to distinguish two types of focus positions. Petrova proposes that the OHG-system in this way allows to avoid focus ambiguities that pervade the NHG-system: if the most deeply embedded argument is stressed in NHG, this pattern can express either narrow focus (on the argument) or wide focus. Word order and information structure in the Gospel Harmony by Otfrid von Weissenburg are addressed in the contribution by Andreas Lötscher. Lötscher argues that word order in this text is largely determined by the same rules and principles that govern modern German clause structure, rejecting the standard view that word order variation is entirely due to the poetic nature of this text. In particular, Lötscher investigates the placement of constituents within or outside of the sentence bracket formed by the finite and non-finite verb in the clause. Lötscher argues that this major characteristic of German word order is well-established in the syntax of Otfrid, with the distinction that arguments can optionally be placed in the Mittelfeld or in the Nachfeld. Lötscher argues that their distribution is not random but is governed by information-structural factors. The contribution by Rosemarie Lühr examines a late-OHG text, namely the translation of Boethius’s De Consolatione Philosophiae from Latin to OHG by Notker of St. Gall. The main issue is to investigate the way in which the OHG translation reflects information-structural phenom-

Introduction

11

ena of the Latin original. Lühr’s careful study shows that Notker was aware of the language-specific means to express IS-categories in the original and investigates the means by which these categories are expressed in the OHG text. The most interesting cases are those in which Notker considerably transformed the word order determined by the original, in order to provide a clear identification of the IS-categories in his translation. Lühr also argues that these changes arguably respect the rules of a genuine grammatical system and therefore provide an insight into the effects of information structure on OHG word order, by pointing out that constructions like the Latin hyperbaton, which serve to mark focus but are ungrammatical in Germanic, are translated without splitting up constituents, as it is typical for Latin. The contribution by Sonja Linde addresses word order and information structure in OS, a language that is almost unexplored from a syntactic viewpoint. She argues that the syntactic variation in the left periphery in the poetic text of Heliand, which constitutes the major part of the OS corpus, is best understood if it is regarded as the result of two competing systems, one involving verb fronting to C and optional movement to SpecC, and one representing an archaic stage in which, in line with Kiparsky (1995), the functional category C has not developed yet. Linde shows that in the latter system, the syntactic realisation of sentence constituents with respect to the finite verb was determined by pragmatic, i.e. informationstructural and discourse-related factors.

References Behaghel, Otto 1892 Zur deutschen Wortstellung. Zeitschrift für den deutschen Unterricht 6: 262–267. 1932 Deutsche Syntax. Eine geschichtliche Darstellung. Band IV. Periodenbau. Heidelberg: Carl Winter. Bolli, Ernst 1975 Die verbale Klammer bei Notker. Untersuchungen zur Wortstellung in der Boethius-Übersetzung. Berlin/New York: Walter de Gruyter. Borter, Alfred 1982 Syntaktische Klammerbildung in Notkers Psalter. Berlin/New York: Walter de Gruyter.

12

Roland Hinterhölzl and Svetlana Petrova

Braune, Wilhelm 1894 Zur Lehre von der deutschen Wortstellung. In Forschungen zur deutschen Philologie. Festgabe für Rudolf Hildebrand zum 13. März 1894, ed. Rudolf Kögel, 34–51. Leipzig: Veit & Co. Delbrück, Berthold 1911 Germanische Syntax II. Zur Stellung des Verbums. Leipzig: Teubner. Dittmer, Arne and Dittmer, Ernst 1998 Studien zur Wortstellung – Satzgliedstellung in der althochdeutschen Tatianübersetzung. Göttingen: Vandenhoeck & Ruprecht. Ebert, Robert Peter 1999 Historische Syntax des Deutschen II: 1300–1750. 2d ed. Bern: Lang. Erdmann, Oskar 1886 Grundzüge der deutschen Syntax. Erste Abteilung. Gebrauch der Wortklassen. Die Formationen des Verbums in einfachen Sätzen und in Satzverbindungen. Stuttgart: Cotta. Fleischmann, Klaus 1973 Verbstellung und Relieftheorie. Ein Versuch zur Geschichte des deutschen Nebensatzes. München: Wilhelm Fink. Fourquet, Jean 1974 Genetische Betrachtungen über den deutschen Satzbau. In Studien zur deutschen Literatur und Sprache des Mittelalters. Festschrift für Hugo Moser zum 65. Geburtstag, eds. Werner Besch, Günther Jungbluth, Gerhard Meissburger and Eberhard Nellmann, 314–323. Berlin: Erich Schmidt. Hinterhölzl, Roland 2004 Language Change versus Grammar Change: What diachronic data reveal about the distinction between core grammar and periphery. In Diachronic Clues to Synchronic Grammar, eds. Carola Trips and Eric Fuß, 131–160. Amsterdam: John Benjamins. Hróarsdóttir, Þorbjörg 2000 Word Order Change in Icelandic: From OV to VO. Amsterdam/New York: John Benjamins. 2003 Language Change and Language Acquisition. Nordlyd 31 (1): 116–131. Kavanagh, Joseph Hayes 1970 The Verb-Final Rule in Old High German: A Diachronic Analysis of Surface Structure Constraints. Dissertation, University of Michigan. Kiparsky, Paul 1995 Indo-European Origins of Germanic Syntax. In Clause Structure and Language Change, eds. Adrian Battye and Ian Roberts, 140–169. New York/Oxford.

Introduction 1996

13

The Shift to Head-Initial VP in Germanic. In Studies in Comparative Germanic Syntax, eds. Höskuldur Thráinsson, Samuel David Epstein and Steve Peter, 140–179. Dordrecht/Boston/London: Kluwer Academic Publishers. Kroch, Anthony 1989 Reflexes of Grammar in Patterns of Language Change. Journal of Language Variation and Change 1 (3): 199–244. Lenerz, Jürgen 1984 Syntaktischer Wandel und Grammatiktheorie. Eine Untersuchung an Beispielen aus der Sprachgeschichte des Deutschen. Tübingen: Niemeyer. Lightfood, David 1999 The Development of Language. Oxford: Blackwell. Müller, Gereon and Wolfgang Sternefeld 1993 Improper Movement and Unambiguous Binding. Linguistic Inquiry 24: 461–507. Näf, Anton 1979 Die Wortstellung in Notkers Consolatio. Berlin/New York: Mouton de Gruyter. Pintzuk, Susan 1999 Phrase Structures in Competition: Variation and Change in Old English Word Order. New York and London: Garland. Rizzi, Luigi 1997 The Fine Structure of the Left Periphery. In Elements of Grammar. Handbook in Generative Syntax, ed. Liliane Haegeman, 281–337. Dordrecht: Kluwer. Roberts, Ian 1997 Directionality and word order change in the history of English. In Parameters of morphosyntactic change, eds. Ans van Kemenade and Nigel Vincent, 398–426. Cambridge: Cambridge University Press. Tomanetz, Karl 1879 Die Relativsätze bei den ahd. Uebersetzern des 8. u. 9. Jahrh. Wien: Carl Gerold's Sohn. Vennemann, Theo 1974 Topics, Subjects, and Word Order: From SXV to SVX via TVX. In Historical Linguistics, eds. John M. Anderson and Charles Johns, 339–376. Amsterdam: North-Holland Publ. Co. Wackernagel, Wilhelm 1892 Über ein Gesetz der indogermanischen Wortstellung. Indogermanische Forschungen. Zeitschrift für indogermanische Sprach- und Altertumskunde 1: 333–436. Zubizaretta, Maria Luisa 1998 Prosody, Focus and Word Order. Cambridge: MIT Press.

The role of information structure in the grammar

The verb-second property in Old High German: Different ways of filling the prefield Katrin Axel

1. Introduction* This paper deals with the question of how the prefield was filled in Old High German (OHG). This is an important issue if we want to find out how well the verb-second (= V2) property was already established at this earliest attested stage of the German language (ca. 750–1050 A.D.). Descriptively speaking, the V2 property involves the three components in (1): (1)

a. obligatory placement of some XP in front of the finite verb b. the XP-movement in (a) is restricted to one XP c. obligatory preposing of the finite verb

In this paper, I will only consider the V2 component in (1a). According to standard assumptions in generative syntax, the landing site of obligatory XP-fronting is the specifier position of CP (SpecCP). Regarding V-to-Cmovement (cf. 1c), it may suffice to say that it was already very much generalized in OHG (see Axel 2007: Chapter 2 for details). As for (1b), it should be mentioned that we still find a certain amount of V3 order in the 8th and 9th century prose texts, most notably with personal pronouns and with certain adverbials (Lippert 1974; Tomaselli 1995; Axel 2007: Chapter 4). For the purposes of this paper, we will neglect this issue and simply assume that there was only one specifier position (= SpecC) in front of the preposed finite verb. The present study is based on an investigation of the following texts:1 First, the major prose texts from the 8th and 9th centuries, i.e. the Isidor translation (ca. 790), the so-called Monsee Fragments (ca. 790–800) and the OHG version of Tatian’s Gospel Harmony (ca. 830). Second, the translation of Boethius’ Consolatio Philosophiæ by the late OHG writer Notker Labeo (first half 11th century). Unfortunately, all these texts are translations of theological or philosophical works. This means that we have to take into account the word order in the Latin source sentences in order to decide which word order patterns were real reflexes of the native OHG grammar.

18

Katrin Axel

2. The prefield in Old High German: A short overview A typical property of V2 languages is that an XP can be moved to the SpecC position regardless of its syntactic category or syntactic function. This can already be shown to hold for 8th and 9th century OHG. As can be seen in the examples in (2), the clause-initial XP-position is not always occupied by the subject noun phrase as in (2a), but may also host a pronominal, (2b), or full object, a temporal adverb, (2c), an adjectival predicate, (2d), or the infinite part of the verb, such as the past participle in (2e). Note that in (2), as in the following examples, the fronted XP is given in square brackets and the finite verb is underlined. (2)

a. [Druhtin] suuor dauite in uuaarnissu (I 610) in truth Lord swore Daviddat ‘the Lord swore to David in truth’ Iurauit dominus dauid in ueritate thó thie scribara/ 2 (T 227,10) b. / [ín] bihieltun PARTICLE the scribes him watched ‘and the scribes watched him’ /obseruabant autem scribe‫ ݸ‬/ ein kneht …/ (T 251,7) c. /[hier] ist here is one boy ‘there is one little boy here …’ /est puer unus hic … / her./ (T 313,14) d. /[toot] ist dead is he ‘he is dead’ /quia mortuus est/ e. [Araugit] ist in dhes aldin uuizssodes boohhum dhazs … revealed is in the old testament’s books that ‘it is revealed in the books of the Old Testament that …’ Pateat ueteris testamenti apicibus (I 252)

In the Isidor and Tatian examples (2a) to (2c) the fillings of the first position clearly differ from those in the corresponding Latin sentences, which suggests that these word orders were licensed by the native OHG grammar. In (2d) the relative order of the adjectival phrase and the finite verb is maintained in the OHG translation, but the clause has been converted from a subordinate clause into a main clause and the subject pronoun her has been inserted in postfinite position. In (2e) it can be argued that the word orders are parallel in both languages since the participle araugit is

The verb-second property in Old High German

19

the translation of the sentence-initial finite verb in the Latin source. Note, however, that the example is from the OHG Isidor translation, which is generally considered to be of outstanding quality (e.g. Lippert 1974: 188). So already in the earlier prose texts practically any maximal category is attested in the OHG prefield. The same could, of course, easily be demonstrated with material from late OHG texts (see Näf 1979: 302–313). However, some fillings of the prefield occur much more frequently than others. A subject noun phrase or an adverbial (in the form of an AdvP or PP) is found there more often in SpecC than, for example, a direct-object noun phrase or an infinite verb in sentences which deviate from the Latin.3 For modern German it can easily be shown that even though virtually any constituent can move to the prefield, these different fillings of the prefield have an effect on the information structure of the sentence. Recently, Fanselow (2002, 2003, 2004) and Frey (2004) have put forward syntactic accounts couched in the minimalist framework which distinguish between three ways to fill the prefield: (i) operator movement, (ii) Stylistic Fronting (or Formal Movement), and (iii) Merge of expletive es. In the following, I will discuss which of these different ways of filling the prefield had already developed in OHG. Accordingly this paper is organized as follows: In the next section, I will consider different types of operator movement. Sections 4 and 5 deal with the questions of whether ‘Stylistic Fronting’ and the precursor of our expletive es had already developed in OHG. In these two sections I will also discuss the different types of V1 declaratives that are attested in the OHG prose texts. Finally, in section 6 I will sketch a hypothesis concerning the relation between XPfronting and sentence particles in OHG.

3. Operator movement Operator movement is the type of XP-fronting which arguably has the longest history: It has been shown to even have Proto-Indo-European roots (e.g. Delbrück 1888: 24; Kiparsky 1995). Note also that certain types of operator-initial sentences have been proposed to have played a central role in the development of V-to-C-movement in the early Germanic languages (see Kiparsky 1995 on Old English; and Eythórsson 1995; Ferraresi 1997, 2005 and Roberts 1996 on Gothic). Not surprisingly, already in the 8th and 9th century OHG prose texts, whphrases are positioned at the left sentence periphery in front of the pre-

20

Katrin Axel

posed finite verb, cf. (3a). We even find evidence for long wh-extraction from subordinate thaz-clauses in these early texts, cf. (3b).4 (3)

thú ti héra/ (T 257,12) a. /… [uuanne]i quami when came you here ‘when did you come here?’ /… quando huc uenisti./ ti garauuemes] …/ b. /[uuár]i uuili [CP t'i thaz uuir that we prepare ...? where will2.sg ‘where do you want us to prepare ...?’ /ubi uis paremus … / (T 555,8)

Furthermore, there is evidence that focused constituents could be moved to SpecC in OHG. Behaghel (1932: 51, 255ff.) already observed marked word orders in declarative clauses where the initial position was occupied by constituents that usually occurred postfinitely. If we investigate written corpus data, in particular historical corpus data, it is, of course, difficult to analyse the information-structural or semanto-pragmatic status of the preposed XPs. We do not have direct access to prosodic information and we cannot ask informants which interpretation a particular construction has. The so-called question test is, however, one relatively straightforward way to identify a focused element that can also be applied to corpus data. It is illustrated in (4) with a modern German example from Fanselow (2004: 4–5). According to Fanselow, the fronting of the direct object ein Buch is triggered by a focus feature in C, cf. (4). (4)

a. (– –

Was hat er gelesen? ‘what has he read’?) modern German gelesen [Ein BUCH] hat er a book has he read ‘he has read a book’ Ein BUCH [[C hat] [IP er gelesen ]]] b. [CP [+foc]-feature in C (cf. Fanselow 2004: 5)

In the OHG prose texts we can also find examples with such questionanswer sequences. The Tatian sentences in (5a), for example, constitute a question-answer sequence and we may therefore assume that the pronoun ein in SpecC is focused. Note that ein here is a cataphoric pronoun which denotes the propositional content of the following complex sentence (= /obathu uuolles thuruhthigan uuésan./ fár forcoufi thiu thu habes./ ‘if you want to be perfect, go, sell what you have’ [T 357,16f.]).

The verb-second property in Old High German

(5)

21

a. (/uuaz ist mir nohnu uuan./ ‘What do I still lack?’ [T 357,12]) thir uuan./ (T 357,15) /[ein] ist youdat missing onenom is ‘one thing you lack’ /unum tibi deest./ [ thir uuan ]]] b. [CP ein [[C ist ]

We also find cases where the context suggests a contrastive interpretation. So in the second sentence in example (6a) from Tatian, the dative object (i.e. the pronoun anderemo) has been fronted to SpecC and is contrasted with the dative object (thesemo) in the first sentence.5 Similarly, in the late OHG example in (6b) from the Consolatio, the fronted dative object (i.e. the DP herbeste) is contrasted with temo lenzen ‘the spring’ in the following sentence, thus we can suspect that it was contrastively focused. (6)

a. /Inti ih quidu zi thesemo far/… and I say to this.one go (T 181,22) /... [anderemo] quidu quim/ say come otherdat ‘and I say to this one, ‘Go’, … and to another, ‘Come’’ /& dico huic uade . /... & alio ueni./ êrâ . náls temo lénzen b. [Hérbeste]gáb kót tîe not thedat springdat autumndat gave God the honour ‘God has given this award to the autumn, but not to the spring’ (N Boethius I 36,3)

Furthermore, the prefield/SpecC position in V2 languages can, of course, host topicalized XPs. This is illustrated in (7), again with a modern German example adapted from Fanselow (2004: 5). (7)

a. ( – Soll ich was über Ede erzählen?/ ‘do you want me to say something about Ede?’) – Diesen Verbrecher hat man endlich verhaftet criminal has one finally arrested thisacc ‘that criminal has finally been arrested’ b. [CP Diesen Verbrecher [[C hat] [ man endlich verhaftet ]]] [+top]-feature in C (cf. Fanselow 2004: 5)

Many linguists subscribe to the so-called ‘aboutness concept’ of topics, which was proposed by Reinhart (1981). A controversial issue is whether an aboutness topic has to already be given in the preceding context as in example (7) or whether it can denote a new referent (see Frey 2004 for an

22

Katrin Axel

overview of the relevant literature). Here, it may suffice to say that in the OHG texts we find many examples for the uncontroversial case where the fronted XPs denote a given referent. So in the Tatian example in (8) the object pronoun ín, which has an anaphoric antecedent (the noun phrase man) in the preceding sentence, has been fronted to SpecC. In the Isidor example in (9) the preposition phrase umbi dhiz refers back to the content of the dhazs-clause in the preceding sentence. (8)

(/… Inti uuas thár man/ thes zesua hant thurri uuas/ ‘and there was a man whose right hand was withered’ [T 227,8]) thie scribara/ (T 227,10) /[ín] bihieltun thó watched PARTICLE the scribes himacc ‘and the scribes watched him’ /obseruabant autem scribe/‫ݸ‬

(9)

(… dhazs dher aerloso man … christ, got endi druhtin uurdi chinemnit ‘that the impious man … was named Christ, God and Lord’ [I 168]) auh so chiscriban (I 171) [Umbi dhiz] nist NEG-is also so written about this in dhero sibunzo tradungum in the Septuagint ‘about this is not written thus in the Septuagint’ Unde et in translatione LXX non habet

Fanselow claims that the fronting of a topic in Present-Day German is an instance of operator/A-bar movement just as the fronting of a focused constituent. Frey (2004) argues that operator movement, i.e. real A-bar movement, always results in a contrastive interpretation. In his approach (noncontrastive) topics as in the modern German example in (7) move to the prefield via A-movement, more precisely via ‘Formal Movement’, which always affects the highest constituent in the middle field just as ‘Stylistic Fronting’ in Fanselow’s account (see section 4.1). This is possible because Frey assumes that there is a topic position at the left edge of the middle field into which topics can be scrambled via A-movement.6 Since it is very difficult, if not impossible, to find out which analysis is more convincing for OHG solely on the basis of written corpus data, I will not try to answer this question. I shall merely draw the descriptive conclusion that constituents functioning as topics are frequently attested in the OHG prefield. To sum up, already in the earlier OHG prose texts we find different types of operators in the SpecC position: wh-phrases and focused XPs. Furthermore, the SpecC position could host topicalized constituents.

The verb-second property in Old High German

23

4. Stylistic Fronting 4.1. Stylistic Fronting in Present-Day German In most modern German clauses the element fronted to SpecC is simply the subject. It is clear that subject-initial sentences follow conditions of use other than the sentences discussed above. They can be thetic, uttered in out-of-the-blue contexts or as answers to a question such as ‘what happened?’. However, it has been known for quite some time that there are other elements that can show up in clause-initial position without special pragmatic force. For example, temporal adverbs and sentence level adverbs are like subjects (see Fanselow 2002, 2003, 2004). They can be clauseinitial without being a topic or focus. So the sentences in (10) and (11) would also be perfect answers to the question ‘what happened?’. (10) a. [CP [ Gestern]i [[hat]k [IP ti ein Bär yesterday has anom bear ‘yesterday, a bear attacked a man’ b. dass gestern ein Bär einen that yesterday anom bear aacc ‘that yesterday a bear attacked a man’

einen Mann attackiert tk ]]] aacc man attacked Mann attackiert man attacked

hat has

(11) a. [CP [ Glücklicherweise]i [[ hat]k [ ti Peter angerufen tk ]]] fortunately has Peter called ‘fortunately, Peter has called’ b. dass glücklicherweise Peter angerufen hat that fortunately Peter called has ‘that fortunately Peter has called’ Note that sentence adverbials such as glücklicherweise in (11) cannot by their very nature be topics. The same is true for quantified XPs such as niemand ‘nobody’ in (12). (12) [Niemand] hilft ihnen umzuziehen nobody helps them to.move ‘nobody is helping them move house’ What temporal adverbs and sentential adverbs have in common is that they can be the structurally highest phrases in the middle field of an unmarked subordinate clause. This is illustrated in the b-examples in (10) and (11). When there are no high adverbials present, the highest category in the middle field is usually just the highest argument, which is normally the

24

Katrin Axel

nominative argument (i.e. the subject). It has been pointed out, however, by Lenerz (1977) that there are predicate classes (e.g. unaccusative predicates, psychological predicates, and passives of ditransitive verbs) with which the oblique argument (most often a dative) precedes the nominative one in the middle field of pragmatically unmarked subordinate clauses (DAT > NOM). Accordingly, in the corresponding main clauses these oblique arguments are fronted in the unmarked case: Fahrrad gestohlen (13) a. [Einem Kind] wurde das child became thenom bike stolen adat ‘the bike was stolen from a child’ (cf. Fanselow 2002: 231) der Text entfallen b. [Einem Schauspieler]ist actor is thenom text forgotten adat ‘an actor has forgotten the text’ (cf. Fanselow 2002: 231) Fanselow (2002, 2003, 2004) argues that the phrases in (10), (11), (12) and (13) have been fronted by a German version of Stylistic Fronting at the CPlevel which is reminiscent of Stylistic Fronting in modern Icelandic. In a pragmatically unmarked clause the highest category in the middle field (= IP in Fanselow’s account) is fronted to SpecC. The exact definition is as follows: (14) When C has no semantic or pragmatic feature, its EPP [= Extended Projection Principle, K.A.] feature attracts the closest (= highest) category in IP. (Fanselow 2004: 7) This type of Stylistic Fronting at the CP-level is, of course, an essential ingredient of the V2 property: the SpecC position has to be filled even when there is no semantic or pragmatic trigger.7

4.2.

Stylistic Fronting in Old High German

We have ample evidence that there was already XP-movement which was semanto-pragmatically vacuous in OHG. Already in the prose texts from the 8th and 9th centuries it is easy to find elements in SpecC which are neither focused nor topics. The sentences in (15), for example, arguably have neutral word order, i.e. a broad focus reading. The fronted elements in these sentences can also not be topics: Negative indefinite subjects as in (15a), indefinites with a non-specific interpretation as in (15b), and correlatives of infinitival or gerundival clauses as in (15c) cannot by their very

The verb-second property in Old High German

25

nature be topics. The same is true for sentence adverbials as in (15d). Note that in all these examples – except (15a) – the relevant XPs have been fronted contrary to the Latin. guot nibi ein got./ (T 355,30) (15) a. /[neoman] nist NEG-is good except the.one God no one ‘no one is good except God alone’ /nemo bonus nisi unus deus/ inan …/ b. /uuanta8 [manage diuuala] giengun/ in since many demons went in him ‘since many demons had gone into him …’ /quia intrauerunt demonia multa/ in eum …/ (T 189,31) sus zi uuesanne/ (T 513,5) c. /[iz] gilimpfit it behoves so to be ‘it behoves to be so’ /oport& enim haec fieri/ d. endi [chiuuisso] ist christus in dheru selbun salbidhu and certainly is Christ in this same salve chimeinit meant ‘and certainly Christ is meant in this same salve’ et utique christus ipsa unctione monstratur (I 144) So just like in modern German, the OHG SpecC position had to be filled even when there was no semantic or pragmatic trigger.

4.3.

Stylistic Fronting with non-canonical predicates

As was already mentioned above, in Present-Day German sentences with non-canonical predicates the filling of the SpecC position is somewhat special. With impersonal psych-predicates, for example, it is either the quasiargument es which is fronted to SpecC, (16b), or the (highest ranked) oblique argument, (16a). In the latter case the non-obligatory es may or may not be be present in the middle field. In the former case, however, it is not possible to drop the es, (16b). (16) a. [Dem Jungen] graut boydat terrifies thedat ‘the boy is dreading the exam’

(es) it

vor by

der the

Prüfung exam

26

Katrin Axel

b. *(Es) graut dem Jungen it terrifies thedat boydat ‘the boy is dreading the exam’

vor by

der the

Prüfung exam

The quasi-argument es also occurs with intransitive impersonal verbs (meteorological verbs and the like), cf. (17), but in contrast to the impersonal psych-verbs it cannot be dropped in such cases, cf. (17a).9 (17) a. dass *(es) regnet that it rains ‘that it is raining’ b. *(Es) regnet it rains ‘it is raining’ Similarly, impersonal predicates with sentential complements are often realized with the correlative es, (18a). Again it is only possible to drop the correlative es in the middle field and not in SpecC, (18b). (18) a. weil (es) mich wundert, dass Peter raucht because it meacc astonishes that Peter smokes ‘because I am astonished that Peter smokes’ mich, dass Peter raucht b. *(Es) wundert it astonishes meacc that Peter smokes ‘I am astonished that Peter smokes’ One could argue that in transitive and intransitive impersonal constructions as in (16) and (17) a nominative argument is always present and that there is variation whether it is realized overtly (as es) or whether it is realized as an empty pro (Safir 1985; Cardinaletti 1990; Sternefeld 2006, II: 560). Similarly in the special case of impersonals with sentential complements one could argue that the correlative es, which is coindexed with the extraposed complement clause, can be left out due to pro-drop (Hentschel 2003). The ungrammaticality of V1 declaratives as in (16b), (17b) and in (18b) shows that in contrast to the overt subject pronoun es, pro cannot move to SpecC to fulfil the V2 requirement. The SpecC position has to be filled by overt material. In OHG, neither the quasi-argument iٕ nor the correlative iٕ10 were very common. These elements only began to spread during Middle High German times (Held 1903; Abraham 1993). The only exceptions are intransitive predicates denoting meteorological phenomena. These fall into two groups: Non-complex, i.e. verbal predi-

The verb-second property in Old High German

27

cates such as âbandên ‘to become evening’ or nahtên ‘to become night’, already often occur with iٕ, whereas constructions with nominal, adjectival or adverbial predicates and copula verbs are mostly attested without iٕ (see Held 1903: 28; Haiman 1974: 103; Lenerz 1985a). The latter type of predicates could show surface V1 order in OHG: zit / nah sehsta. .../ (19) a. /... uuas tho time near sixth was PARTICLE ‘it was about the sixth hour’ /... hora erat / quasi sexta; .../ (N Boethius IV 205,10) b. Íst álso chád si. is so said she ‘“so it is,” she said’ Ita est inquit illa.

(T 275,29)

Sentences with transitive impersonal predicates predominantly exhibited V2 order in OHG. It was usually the (highest ranked) oblique argument that was fronted. This argument normally referred to the logical subject (= experiencer). Already Behaghel (1932: 70) observed that the oblique argument (most often a personal pronoun) was generally fronted in order to fulfil the V2 requirement in those cases where the quasi-argument iٕ was not present. The same is true for modern German. Here as well, the oblique arguments of impersonal verbs occur in SpecC in unmarked clauses: (20) /[mih] hungrita …/ meacc hungered ‘I was hungry …’ /esuriui enim …/

(T 543,26)

By contrast, in sentences with impersonal predicates that subcategorize for a propositional argument in the form of a ‘subject’ clause we find V1 order both in the 8th and 9th century texts, cf. the Tatian example in (21), and in the late OHG texts by Notker, cf. (22). thaz arfuoron fon In/ thie engila …/ (21) /uuard thô became PARTICLE that departed from them the angels ‘it came to pass that the angels departed from them’ /& factum ut discesserunt ab eis/ angeli …/ (T 87,17) (22) a. Íst áleuuâr [dáz man chît …] is very.true that one says ‘it is a true sentence that one says …’ Nam uera sententia est .

(N Boethius V 234,4)

28

Katrin Axel

b. Keskíhet óuh ófto . (N Boethius IV 220,10) happens also often [dáz kûotên geuuált ze_hánden chóme] . to hands comes that the.goodpl power ‘it also often happens that the chief command is given to good men’ Fit autem sepe . uti bonis summa rervm regenda deferatur . The V1 pattern is also attested with transitive impersonal predicates that subcategorize both for an accusative or dative experiencer argument and an infinitival/gerundival or finite clause, cf. (23). We also find examples with V2 order where the dative argument has been fronted contrary to the Latin source, cf. (24). As was mentioned above, fronting of the oblique argument was also possible in the case of impersonals without sentential complements, cf. (20). [christinheidi chilaupnissa chihoran] (23) a. lustida sie hear desired themacc Christianity’s belief ‘they wanted to hear the belief of Christianity’ christi fidem delectantur audire (I 693) [uuáz iz méine] b. Íst sie vuúnder what it means is sheacc wonder ‘she wonders what it means’ (N Boethius IV 210,29) [abur (24) /[íu] gilimphit behoves again youdat.pl ‘you must be born again’ /oport& uos nasci denuo/

giboran born

uuerdan]/ become

(T 405,26)

Note that ‘subject’ clauses, in particular finite ‘subject’ clauses, are generally extraposed in OHG as in modern German. This can be seen in (23b), where the clause occurs after the predicate vuúnder. Again, one could argue that all the OHG impersonal constructions without the expletive iٕ involve instead a pro-element which is either an empty quasi-argument or a correlative coindexed with the extraposed clause. For OHG this hypothesis seems to be even better motivated than for modern German since in the earlier OHG texts there is also referential prodrop attested in postfinite environments (Axel 2005, 2007: Chapter 6). This is illustrated in (25).

The verb-second property in Old High German

29

pro in cruci (25) a. [Sume] hahet pro in cross someacc hang2.pl ‘some of them you will crucify’ et ex illis occidetis et crucifigetis (MF XVIII,17 [St. Matthew 23:34]) b. [In dhemu druhtines nemin] archennemes ... pro fater in the Lord’s name recognize1.pl ... pro father ‘in the name of the Lord we recognize ... the Father’ In persona enim domini patrem accipimus (I 279) Given the assumption that there is an expletive pro in OHG impersonal constructions, it could be possible that in sentences as in (19) to (23), pro has been moved to SpecC as indicated in (26) for the sentences (19a) and (22a). The V1 order would then only be a surface phenomenon; the underlying order would be the regular V2 structure. (26) a. [CP proi [[C uuas] [tho ti zit nah sehsta. ...]]] áleuuâr]]] [ dáz man chît]i ] b. [CP proi [[C Íst ] [[ti This analysis, however, faces one serious problem: Also in Present-Day German there are impersonal predicates with which the quasi-argument es or correlative es are not obligatory and it can be argued that in the cases whithout es pro is present. In contrast to OHG, however, sentences with impersonal predicates cannot be realized as V1 sentences in Present-Day German. In other words, we would have to argue that a putative pro-element cannot move to SpecC in modern German to fulfil the V2 requirement (see also Cardinaletti 1990; Sternefeld 2006 I, 170). Viewed from a diachronic perspective it is then very unlikely that pro should have been licensed in SpecC in OHG times. If so, what would account for the diachronic change which resulted in a ban on fronting pro to SpecC? This is why I will stick to the idea that there is a general ban on empty expletives in SpecC in V2 languages. In other words, it may well be that we should analyse impersonal predicates as subcategorizing for pro, but this pro presumably stayed in the middle field and did not move to SpecC via Stylistic Fronting. So we may conclude that in the context of impersonal predicates, Stylistic Fronting had already developed in OHG times: However, there are still some residues where it sometimes failed to take place.11 A further type of non-canonical predicate is the passives of ditransitive verbs. In the last section it was shown that in modern German in the unmarked order the oblique argument precedes the nominative in such cases and consequently it is the oblique argument which is moved to SpecC via

30

Katrin Axel

Stylistic Fronting in main declarative clauses. In the OHG prose texts we find some examples with passivized ditransitives. In contrast to the impersonal psych-predicates, the oblique argument is usually not fronted and we get V1 order: furlazano (27) /sint thir forgiven are youdat.sg ‘your sins are forgiven’ /dimittuntur tibi peccata/

sunta/ sins

(T 193,24)

However, we have to be careful in drawing this conclusion since in all the relevant examples the oblique argument also occurs postverbally in the corresponding Latin sentences. To sum up: Sentences with impersonal predicates fall into two groups: In sentences with transitive impersonals the oblique experiencer was generally fronted to SpecC. Sentences with intransitive impersonals or intransitive sentences with subject clauses, on the other hand, were often realized as V1 sentences. This is presumably due to the fact that there was no overt argument present which could be fronted since the use of the overt quasiargument iٕ or the correlative iٕ was not yet obligatory. So the V1 orders with impersonals constituted exceptions to the Stylistic Fronting Rule in OHG, but they were exceptions that proved the rule. One factor that led to the consolidation of the V2 property in the history of German was the spread of overt quasi-arguments and correlatives.

5. An expletive in SpecC? In Present-Day German there are also constructions where Stylistic Fronting fails to apply and where there are also no fronted focused XPs or topics in SpecC either. In this case, the expletive es is merged in this position. One example would be existential or presentational constructions as in (28). As can be seen in (28b), the expletive es can also occur with definite subjects. This shows that it is different from the subject expletive in languages such as English (there) or Icelandic (það). German es is not a subject expletive, but a placeholder which satisfies the requirement that the SpecC position be filled. (28) a. [Es] braut sich ein Gewitter thunderstorm it brews REFL a ‘a thunderstorm is brewing’

zusammen together

The verb-second property in Old High German

31

Philharmoniker b. [Es] spielen die Wiener the Vienna Philharmonic.Orchestrapl it playpl ‘the Vienna Philharmonic Orchestra is playing’ In OHG this kind of prefield (or SpecC) expletive had not yet developed. In our corpus there are no such examples. This fits in with observations in the secondary literature (Brugmann 1917; Behaghel 1923: 319–320, 1928: 470–471; Lenerz 1985a; Abraham 1993) that the development of es as a prefield expletive took place in the Middle High German period. Now the question arises of how existential/presentational constructions looked in OHG. The answer is that they were simply realized with V1 order. Like in the modern language, the subject was not fronted but remained in the postfinite domain even in those cases where the corresponding Latin sentence had a preverbal subject, (29). Instead of merging an expletive, the SpecC position remained empty:12 (29) a. See13

quimit der bruti gomo ·(MF XX,8 [St. Matthew 25:6]) comes the bridegroom ‘behold, the bridegroom is coming’ Ecce sponsus uenit, hirta In thero lantskeffi.‘/ b. /uuarun thô PARTICLE shepherds in that country were ‘there were shepherds in the same country’ /Et pastores erant In regione eadem.‘/ (T 85,29) thar ouh sum uuitua/ In thero burgi .../ c. /uuas was there also some widow in that city ‘there was a certain widow in the same city...’ (T 415,2) /vidua autem quædam erat/ In ciuitate illa .../ in/ moises …/ (T 305,21) d. /arougta sih tho appeared REFL PARTICLE them Moses ‘Moses appeared unto them …’ /et ecce apparuit illis/ moises …/ INTERJ

In some examples with presentational constructions and V1 order we have evidence that the subject is not in its base position, but is extraposed. In (30), for example, the subject noun phrase gotes engila is found after the verb particle zuo. In the examples in (29) the so-called right-sentence bracket is not occupied by overt material (i.e. a non-finite verb or verbal particle). This is why the examples are in fact ambiguous with respect to the position of the subjects.

32

Katrin Axel

(30) /giengun thô t i zuo PARTICLE VERB.PART. went ‘there came God’s angels’ /Et ecce angeli accesserunt/

[gotes God’s

engila]i / angels

(T 115,30)

Most examples with existential/presentational constructions contain unaccusative predicates (most notably verbs of existence or appearance). In OHG texts, subject extraposition occurs quite frequently with such predicates not only in main clauses, but also in subordinate clauses.14 The frequent occurrence of ‘subject’ extraposition in unaccusative environments could be related to the fact that the subject is the underlying object in such cases. Again one could speculate that such constructions involve an expletive pro in the canonical subject position which is coindexed with the nominative noun phrase. Like the empty quasi-argument pro in truly subjectless constructions, this expletive pro has a nominative case feature, but in contrast to quasi-argument pro it is not restricted to the third person singular, but must also be able to bear a plural feature (e.g. in (29b) and (30) the associate noun phrase is in the plural). The question as to which position was the subject position in OHG is not really relevant for the issue at hand. We may simply assume that it was SpecVP as argued by e.g. Haider (1997) and Sternefeld (2006, II: 537), who propose a minimal model for modern German sentence structure with no I-projection. Again one could now argue that the V1 orders in (29) and (30) only pertain at the surface and that there are structurally V2 orders where the expletive pro-element has been fronted to SpecC. However, this analysis would face the same diachronic problem as was discussed above: If expletive pro could be moved to SpecC in OHG, why is this no longer possible today? For modern German Sternefeld argues that in sentences where nominative DPs are merged in direct object position there must be a silent category in SpecVP c-commanding the finite verb so that the required checking configuration for subject-verb agreement and the nominative case feature can be established, cf. (31a). This silent category differs from the pro that occurs in truly subjectless sentences in that it is not restricted to the third person singular, but may also bear a plural feature. ereignet] haben]] (31) a. dass sichj [VP Ø [V’ [VP Wunder tj miracles happened have that REFL ‘that miracles have happened’ ereignet b. [Es] haben sich Wunder REFL miracles happened it have ‘miracles have happened’

The verb-second property in Old High German

33

ereignet c. *Haben sich Wunder REFL miracles happened have ‘have miracles happened?’ In contrast to OHG, however, the corresponding main clauses cannot be realized with V1 order as is illustrated in (31b,c). Sternefeld (2006, II: 592) therefore concludes that there must be a general ban on moving invisible material to SpecC. So we may conclude that one factor that led to the consolidation of the V2 property was the development of an expletive element, a placeholder that was merged in the SpecC position when no XP was moved there. This development took place in the Middle High German period. Interestingly, in modern Bavarian the use of es as a prefield expletive is not very common either, at least it is more restricted than in the Standard language. As Weiß (1998: 102–103) shows, the adverb do is often used where es would occur in the Standard: Ungligg (32) [Do] is a there is an accident ‘an accident has happened’

bassierd happened

(Weiß 1998: 102)

One could now hypothesize that the modern and historical dialects in which the pronoun es is not common resort to an adverbial expletive instead. In the OHG Tatian we do indeed find examples where the adverb thô occurs in presentational constructions (see also Brugmann 1917 on Otfrid): uuas man In INTERJ PARTICLE was man in ‘behold, there was a man in Jerusalem’ /& ecce homo erat In hierusalem‘/

(33) /senonu

[tho]

hierusalem.‘/ Jerusalem’ (T 89,23)

We also have some evidence that thô only had a weak semantics in OHG: It is often used as a translation of Latin et ‘and’ or it is simply inserted into the OHG where the Latin does not even have a corresponding element (cf. Ruhfus 1897: 12 et passim; Betten 1987). If thô was really the OHG counterpart of our modern prefield expletive es, it could be speculated that in the presentational/existential constructions with V1 order that we discussed above thô was dropped as the result of topic drop: (34) [CP thô/ø [[C Vi]

[VP ... ti ]]]

The hypothesis of an elliptical thô is, however, not really convincing since this element was overtly present in many examples, albeit in postfinite

34

Katrin Axel

position, cf. (29a, d) and (30). In the following section its role will be discussed in more detail. To sum up: In OHG, the prefield expletive es had not yet developed. Existential/presentational constructions were realized as V1 structures. The hypothesis that there was topic drop of the adverb thô in SpecC in these cases is not convincing.

6. XP-movement to SpecC and the role of particles In the previous sections we discussed V1 order in the context of impersonal predicates and existential/presentational constructions. These are, however, not the only classes of examples with this word order. As (35) shows, we also find V1 order in declaratives with pronominal subjects.15 As was observed already by Erdmann (1886: 167) such examples are clearly not the precursors of our modern German es-V2 sentences since the latter are incompatible with pronoun subjects (see also Brugmann 1917: 35). úf/ (T 183,14) (35) a. /... árstuont siu tho up arose she PARTICLE ‘she rose’ /... & surrexit/ inti gifultun/ beidu thiu skef …/ b. /quamun sie thó both the boats came they PARTICLE and filled ‘they came and filled both of the boats’ /& uenerunt & impleuerunt/ ambas nauiculas .../ (T 125,27) Lenerz (1984: 131–132, 152–153; 1985b) proposes that the OHG V1 declaratives are the precursors of our modern narrative V1 declaratives that occur at the beginning of certain types of narrations, such as jokes (see also Maurer 1924 and Önnerfors 1997: 53ff. for many more examples): (36) Kommt da ein Kerl herein und fragt mich, … comes then a guy in and asks me ‘a guy comes in and asks me …’ (Lenerz 1985b: 104) According to Lenerz, V1 order in declaratives indicates that the sentence does not contain a thematic element, i.e. the whole proposition is rhematic. This also explains, he argues, why this word order often occurs at the beginning of texts or passages, in emphatic narration or where a change of the discourse topic occurs.

The verb-second property in Old High German

35

While it is true that there are many V1 declaratives which only contain rhematic material attested in the OHG prose texts, notably in Tatian, we also find a considerable number of examples where this is not the case: ûz sine scalca/ In uuega .../ (37) a. /giengun thô PARTICLE out his servants in streets went ‘his servants went out into the streets’ /& egressi sunt serui eius/ In uias .../ (T 425,16) thie hirta heimuuartes.‘/ b. /uuvrbun thô turned PARTICLE the shepherds homeward ‘the shepherds returned’ /& reuersi sunt pastores.‘/ (T 89,2) giheilit ther kneht in thero ziti./ c. /uuard tho in that time became PARTICLE healed the servant ‘his servant was healed in that hour’ /& sanatus est puer in illa hora;/ (T 183,7) The V1 declarative in (37a) from Tatian’s Gospel Harmony, for example, occurs in the middle of a bible story (the Parable of the Wedding Banquet [Matthew 22: 1–14]) and its subject sine scalca ‘his servants’ clearly refers to thematic material as the servants have been mentioned several times in the preceding discourse. The same is true for the examples in (35), which have pronoun subjects that are anaphorically linked to a preceding discourse referent. The V1 declaratives in (37b,c) even occur at the absolute end of bible stories (The Shepherds and the Angels [Luke 2: 8–20] from the Christmas Story and The Faith of the Centurion [Matthew 8: 5–13]). Here as well, the respective subjects thie hirta ‘the shepherds’ and ther kneht ‘the servant’ refer to thematic material.16 Note also that we cannot render OHG V1 declaratives as in (37) by our so-called narrative V1 declaratives in modern German.17 For example, it would be impossible to close the Christmas Story by the sentence Kehrten die Hirten um (lit. ‘turned the shepherds back’). This fits in well with claims in the older philological literature (e.g. Maurer 1924; Behaghel 1932: 37) that there is no diachronic connection between the OHG V1 declaratives and the type that arose in the late Middle High German period, which are still common in Present-Day German. If there are V1 declaratives attested which are neither the precursors of the later narrative type nor the historical counterparts of our modern es-V2 construction, what else could account for the fact that there is no XPmovement to SpecC (i.e. no Stylistic Fronting) in these cases? Interestingly, the vast majority of examples contain a characteristic feature: the par-

36

Katrin Axel

ticle thô/tho in postfinite position. As I have shown elsewhere in Axel (2007: Chapter 2), there are residues of a system of sentence particles in the early OHG texts (the interrogative particle inu/eno, the affirmative particle jâ/ja and the imperative particle nû/nu). It may have been the case that thô – besides its use as a normal temporal adverbial – belonged to this system as well and served as a kind of narrative-emphatic marker. Sentence particles encoding sentence-type, mood or aspects of information structure, however, are not really a typical feature of V2 languages. In V2 languages sentence types are primarily signalled by word order (V1 vs. V2). Viewed from this perspective, it would not be surprising that there was no Stylistic Fronting in sentences with postfinite thô. If thô was a narrative-declarative marker, it was not necessary to front an XP via Stylistic Fronting. We may thus hypothesize that it was the loss of sentence particles that contributed to the generalization of XP-fronting, i.e. the development of Stylistic Fronting. The OHG V1 declaratives with thô are probably vestiges of this older system.18

7. Summary In OHG, the SpecC position was already targeted by different types of XPmovement. For example, there are uncontroversial cases of operator movement attested, i.e. fronted wh-phrases and fronted focused XPs. Furthermore, the SpecC position often hosted topicalized constituents. Moreover, we have ample evidence that the phenomenon of Stylistic Fronting had developed as we can easily find XPs in SpecC which cannot be operators. However, there were certain environments where Stylistic Fronting did not yet apply in OHG. For example, in sentences with impersonal constructions, there was often no overt quasi-argument (or correlative) present that could be fronted. Furthermore, there was often no XP-fronting in sentences that contained the particle thô. I suggested that these are residues of an older system where Stylistic Fronting had not developed and sentence type was not signalled by word order but by the use of lexical elements in the form of sentence particles. Finally, the third strategy to fill SpecC – merge of expletive es – was totally absent from OHG. Consequently, existential or presentational constructions usually exhibited V1 order. We may thus conclude that it was a conspiracy of factors that shaped the prefield as we know it today. The rise of overt quasi-arguments and the development of a SpecC expletive were factors that contributed to the further consolidation of V2. One older and much more decisive step towards the development of

The verb-second property in Old High German

37

V2, however, was the innovation of Stylistic Fronting. I hypothesize that the lexical loss of certain particles may eventually be responsible for this innovation. Even though this hypothesis could not be investigated more thoroughly within the scope of this paper, it offers a new perspective for future research.

Notes *

1.

2.

3.

4.

5.

This work was supported by the Deutsche Forschungsgemeinschaft within the SFB 441 ‘Linguistische Datenstrukturen’. Some parts of this paper have been published in Axel (2007: Chapter 4). I would like to thank the audience at the ‘Workshop on Information Structure and Language Change’, an anonymous reviewer and the editors for their helpful comments. The Monsee Fragments are cited by section and line numbers, the Isidor by line numbers and the Tatian by page and line numbers (of the whole edition, not of the manuscript) according to the editions given in the reference section. The line numbers only indicate where the OHG sentence begins. If present, the corresponding Latin sentence is given (without line numbers) in a separate line. In many examples, underlining of the finite verb, boldface and/or bracketing have been added. The Present-Day English translations of the Isidor examples have largely been adopted from Robinson (1997). The slashes in the Tatian examples indicate where the line breaks are. In the Tatian manuscript (Cod. Sang. 56) the bilingual text is presented in two separate columns, Latin on the left and OHG on the right, a feature which has been reproduced in the edition by Masser (1994). The translators tried to achieve a translation which was as loyal as possible to the line breaks in the Latin parts, a technique which could result in non-German word orders (Dittmer & Dittmer 1998: 23). Bernhardt and Davis (1997: 55) observe that in the OHG Tatian, the prefield is most often occupied by the subject or an initial adverbial. Näf (1979: 323) mentions that in Notker’s Consolatio the subject occurs in the prefield in ca. 58% of V2 clauses. Note that also Latin wh-phrases are usually found at the left periphery. This is why there are hardly any OHG examples attested where the wh-phrase is fronted contrary to the Latin in the earlier texts. Näf (1979: 164) mentions, however, that in the late OHG Consolatio almost one-third of the examples with fronted wh-phrases do not have a direct counterpart. It could be objected that the placement of the dative object is parallel to the Latin as here the object alio also appears sentence-initially. Note, however, that the translator has inserted the finite verb (quidu) contrary to the Latin in this case. Instead of putting the verb in front of the dative object (i.e. quidu

38

6.

7.

8.

9.

10. 11.

12.

13.

Katrin Axel anderemo) parallel to the first sentence, the translator has decided to place anderemo in front of the verb. Frey’s account for modern German differs from Fanselow’s not only with respect to the question of how topicalization is analysed. He also assumes that there are different C-projections (FinP, KontrP and CP from right to left) which bear specific types of EPP-features: the EPP-feature in Fin is a ‘bare’ EPP-feature, and the one in Kontr is associated with the feature [KONTRAST]. In C there may be a feature which encodes reference to the discourse structure and is checked by base-generated ‘discourse adverbials’ such as kein Wunder ‘no wonder’. The V2 effect results from the constraint that there can only be one EPP-feature present in the C-domain. Such theoretical details, however, are not essential for the following argumentation, which primarily aims at giving a more descriptive overview of the grammar of the OHG prefield. For the purposes of this paper I have chosen to use Fanselow’s proposal as a theoretical background since it is somewhat simpler. Not surprisingly, this phenomenon is of younger origin than operator movement. It seems to be absent from Gothic, for instance. As Eythórsson (1995: 79–83) and Ferraresi (1997: 132) demonstrate for the Gothic bible translation, all subjects that have been moved to the C-domain are either definite or indefinite with a specific interpretation. The absence of non-specific indefinites comes as no surprise if we assume that in Gothic all XP-movement to the Cdomain had a semanto-pragmatic effect, i.e. all fronted XPs were either topics or focused. This would also explain why there was a fair amount of V1 order in Gothic, as has been observed by Eythórsson (1995: 58). Note that the causal conjunction uuanta is attested both with V2 order (as in the present example) and with verb-final order in OHG. So it could function both as a coordinating and as a subordinating conjunction just as Present-Day German weil ‘because’. The only exceptions are complex time/season expressions consisting of a copular verb and a nominal predicate. In such cases es is not obligatory (Haiman 1974: 103), but again it is only possible to drop the es in the middle field and not in SpecC. But see (15c) for an example with correlative iٕ. This fits in well with observations by Diels (1906: 98), who claimed that originally V1 was the dominant pattern in the context of impersonal constructions in Germanic. A further context where es occurs in modern German is the passives of intransitive verbs (e.g. es wird getanzt, ‘there was dancing (lit.: it was danced)’. In OHG, passives of truly transitive predicates are attested only very rarely (see Bishop 1977). The letters given in roman in the OHG and Latin examples (first and fourth line) indicate material that is unreadible in the manuscript and has been reconstructed by Hench (1890).

The verb-second property in Old High German

39

14. This word order pattern is also realized contrary to the Latin (see Axel 2007: Chapter 2). 15. In the examples in (35a,b) there is an et (&) in the Latin sentence followed by a verb form. Regardless of the question of whether et is a conjunction or an adverb, it cannot simply be argued that the V1 order in OHG is due to the Latin since there are many different translations attested for such et-sentences in Tatian. For example, et is also rendered by thô in prefinite position (Betten 1987). 16. Petrova (2006) also observes that there are V1 declaratives attested in Tatian that contain thematic material. She studies the relationship between word order and discourse relations and points out that V1 order generally occurs in sentences belonging to the main story line of the narrative, which means that they are coordinated in text structure. V2 clauses, on the other hand, can be argued to be subordinated in text structure as they appear in contexts providing additional, for example explanatory or descriptive information on a contextually given entity. 17. Önnerfors (1997: 71–76) argues against the view that V1 declaratives in modern German only contain rhematic material. In his approach the lack of a preverbal position results in a clause which is unpartitioned at the level of topiccomment structure. However, the problem remains that there are many V1 declaratives in OHG which could not be rendered by a narrative V1 declarative in modern German. 18. Note also that the particle thô is also often present in the V1 sentences with different types of impersonal constructions and in the existential/presentational constructions. In these cases the development of V2 was delayed simply because there often was no XP present which could be fronted to or merged in SpecC: The quasi-argument/correlative iٕ was not yet very widespread and the prefield expletive had yet to be ‘innovated’. We could thus hypothesize that thô was retained in these environments where the preconditions for the alternative strategy (i.e. XP-fronting) were not yet fulfilled.

References Primary Sources [I] [MF]

[N]

Der althochdeutsche Isidor. Nach der Pariser Handschrift und den Monseer Fragmenten. Neu hrsg. von H. Eggers. Tübingen, 1964. The Monsee Fragments. Newly collocated text. Introduction, notes, grammatical treatise and exhaustive glossary and a photolithographic fac-simile. Edited by G. A. Hench. Straßburg 1890. Notker der Deutsche. Boethius, »De consolatione Philosophiæ«. Buch I–V. Hrsg. von Petrus W. Tax. Tübingen, 1986, 1988, 1990.

40 [T]

Katrin Axel Die lateinisch-althochdeutsche Tatianbilingue Stiftsbibliothek St. Gallen Cod. 56. Unter Mitarbeit von Elisabeth De Felip-Jaud hrsg. von Achim Masser. Göttingen, 1994.

Secondary Sources Abraham, Werner 1993 Null subjects in the history of German: From IP to CP. Lingua 89: 117–142. Axel, Katrin 2005 Null subjects and verb placement in Old High German. In Linguistic Evidence: Empirical, Theoretical and Computational Perspectives, eds. Marga Reis and Stephan Kepser, 27–48. Berlin/New York: Mouton de Gruyter. 2007 Studies on Old High German Syntax: Left Sentence Periphery, Verb Placement and Verb-Second. Amsterdam/Philadelphia: John Benjamins Publishing Company. Behaghel, Otto 1923 Deutsche Syntax. Eine geschichtliche Darstellung. Volume 1: Die Wortklassen und Wortformen. Heidelberg: Winter. 1928 Deutsche Syntax. Eine geschichtliche Darstellung. Volume 3: Die Satzgebilde. Heidelberg: Winter. 1932 Deutsche Syntax. Eine geschichtliche Darstellung. Volume IV: Wortstellung, Periodenbau. Heidelberg: Winter. Bernhardt, Karl A. and Graeme Davis 1997 The Word Order of Old High German. Lewiston/Queenston/Lampeter: The Edwin Mellen Press. Betten, Anne 1987 Zur Satzverknüpfung im althochdeutschen Tatian. Textsyntaktische Betrachtungen zum Konnektor ‘thô’ und seinen lateinischen Entsprechungen. In Althochdeutsch, Volume I, eds. Rolf Bergmann, Heinrich Tiefenbach and Lothar Voetz, 395–407. Heidelberg: Winter. Bishop, Harry M. 1977 The ‘subjectless’ sentence of Old High German. Ph.D. dissertation, University of California, Berkeley. Brugmann, Karl 1917 Der Ursprung des Scheinsubjekts ‘es’ in den germanischen und den romanischen Sprachen. (Berichte über die Verhandlungen der Königl. Sächsischen Gesellschaft für Wissenschaften, Phil.-hist. Klasse, 69: 5.) Leipzig: B. G. Treubner. Cardinaletti, Anna 1990 Impersonal Constructions and Sentential Arguments in German. Padova: Padova Unipress.

The verb-second property in Old High German

41

Delbrück, Berthold 1968 Reprint. Altindische Syntax. Darmstadt: Wissenschaftliche Buchgesellschaft. Original edition, Halle/S.: Verlag der Buchhandlung des Waisenhauses, 1888. Diels, Paul 1906 Die Stellung des Verbums in der älteren althochdeutschen Prosa. Berlin: Mayer & Müller. Dittmer, Arne and Ernst Dittmer 1998 Studien zur Wortstellung – Satzgliedstellung in der althochdeutschen Tatianübersetzung. Für den Druck bearbeitet von M. Flör und J. Klempt. Göttingen: Vanderhoeck und Ruprecht. Erdmann, Oskar 1985 Reprint. Grundzüge der deutschen Syntax nach ihrer geschichtlichen Entwicklung, Volume I. Hildesheim: Georg Olms Verlag. Original edition, Stuttgart: Verlag der J.G. Cotta’schen Buchhandlung, 1886. Eythórsson, Thórhallur 1995 Verbal syntax in the early Germanic languages. Ph.D. dissertation, University of Cornell. Fanselow, Gisbert 2002 Quirky ‘subjects’ and other specifiers. In More than Words, eds. Barbara Stiebels and Ingrid Kaufmann, 227–250. Berlin: Akademie Verlag. 2003 Münchhausen-style head movement and the analysis of verb-second. In Syntax at Sunset: Head Movement and Syntactic Theory. UCLA Working Papers in Linguistics 10, ed. Anoop Mahajan, 40–76. 2004 Cyclic phonology-syntax-interaction: Movement to first position in German. In Working Papers of the SFB 632, Interdisciplinary Studies on Information Structure (ISIS) 1, eds. Shinichiro Ishihara, Michaela Schmitz and Anne Schwarz, 1–42. Potsdam: Universitätsverlag. Ferraresi, Gisella 1997 Word order and phrase structure in Gothic: A comparative study. Doctoral dissertation, University of Stuttgart. 2005 Word and Phrase Structure in Gothic. Leuven: Peeters. Frey, Werner 2004 The grammar-pragmatics interface and the German prefield. Sprache und Pragmatik 52: 1–39. Haider, Hubert 1997 Projective economy: On the minimal functional structure of the German clause. In German: Syntactic Problems – Problematic Syntax, eds. Werner Abraham and Elly van Gelderen, 83–103. Tübingen: Niemeyer. Haiman, John 1974 Targets and Syntactic Change. The Hague/Paris: Mouton.

42

Katrin Axel

Held, Karl 1903

Das Verbum ohne pronominales Subjekt in der älteren deutschen Sprache. Berlin: Mayer & Müller. Hentschel, Elke 2003 Es war einmal ein Subjekt. Linguistik online 13(1). http://www.linguistik-online.de/13_01/hentschel.html Kiparsky, Paul 1995 Indo-European origins of Germanic syntax. In Clause Structure and Language Change, eds. Adrian Battye and Ian Roberts, 140–169. Oxford: Oxford University Press. Lenerz, Jürgen 1977 Zur Abfolge nominaler Satzglieder im Deutschen. Tübingen: Narr. 1984 Syntaktischer Wandel und Grammatiktheorie. Eine Untersuchung an Beispielen aus der Sprachgeschichte des Deutschen. Tübingen: Niemeyer. 1985a Zur Theorie des syntaktischen Wandels: Das expletive ‘es’ in der Geschichte des Deutschen. In Erklärende Syntax des Deutschen, ed. Werner Abraham, 99–136. Tübingen: Narr. 1985b Diachronic syntax: Verb position and COMP in German. In Studies in German Grammar, ed. JindӚich Toman, 103–132. Dordrecht: Foris. Lippert, Jörg 1974 Beiträge zur Technik und Syntax althochdeutscher Übersetzungen unter besonderer Berücksichtigung der Isidorgruppe und des althochdeutschen Tatian. München: Fink. Maurer, Friedrich 1924 Zur Anfangsstellung des Verbs im Deutschen. In Beiträge zur germanischen Sprachwissenschaft. Festschrift für O. Behaghel, ed. Wilhelm Horn, 141–184. Heidelberg: Winter. Näf, Anton 1979 Die Wortstellung in Notkers Consolatio. Untersuchungen zur Syntax und Übersetzungstechnik. Berlin/New York: de Gruyter. Önnerfors, Olaf 1997 Verb-erst-Deklarativsätze. Grammatik und Pragmatik. Stockholm: Almqvist & Wiksell International. Petrova, Svetlana 2006 A discourse-based approach to verb placement in Early WestGermanic. In Interdisciplinary Studies on Information Structure (ISIS) 5, eds. Shinichiro Ishihara, Michaela Schmitz and Anne Schwarz, 153–185. Potsdam: Universitätsverlag. Reinhart, Tanya 1981 Pragmatics and linguistics: An analysis of sentence topics. Philosophica 27: 53–94.

The verb-second property in Old High German

43

Roberts, Ian 1996 Remarks on the Old English C-system and the diachrony of V2. In Language Change and Generative Grammar. Special Issue 7 (1995– 96) of Linguistische Berichte, eds. Ellen Brandner and Gisella Ferraresi, 154–167. Opladen: Westdeutscher Verlag. Robinson, Orrin W. 1997 Clause Subordination and Verb Placement in the Old High German Isidor Translation. Heidelberg: Winter. Ruhfus, Wilhelm 1897 Die Stellung des Verbums im ahd. Tatian. Doctoral dissertation, University of Heidelberg. Safir, Kenneth 1985 Missing subjects in German. In Studies in German Grammar, ed. JindӚich Toman, 231–287. Dordrecht: Foris. Schrodt, Richard 2004 Althochdeutsche Grammatik II: Syntax. Tübingen: Niemeyer. Sternefeld, Wolfgang 2006 Syntax: Eine merkmalbasierte generative Analyse des Deutschen. Vol. 1–2. Tübingen: Stauffenburg. Tomaselli, Alessandra 1995 Cases of verb third in Old High German. In Clause Structure and Language Change, eds. Allan Battye and Ian Roberts, 345–369. Oxford: Oxford University Press. Weiß, Helmut 1998 Syntax des Bairischen. Studien zur Grammatik einer natürlichen Sprache. Tübingen: Niemeyer.

The role of information structure in word order variation and word order change Roland Hinterhölzl

1. Introduction In this paper, I propose a novel account of word order variation and word order change in terms of competition between prosodically more or less marked forms within one grammar (contrary to the double base hypothesis). It is argued that variation within one grammar is due to the expression of different information-structural categories and word order change involves a change in the mapping between syntactic structure and prosodic structure in which Information Structure (IS) plays a crucial role. 1.1. Word order change in German One of the most intriguing developments in the history of the Germanic languages, next to the grammaticalization of Verb Second (V2), is the change in basic word order in English and Scandinavian. This development involved a change from the presumed Indo-European basic OV order to the basic VO order in these languages. In this scenario, German (and Dutch) retained (modulo some changes in the application of extraposition) the inherited base order. The traditional explanation of this phenomenon is to assume that the loss of Case led to a positional marking of grammatical functions. However, this account faces serious difficulties if we consider the development of Dutch and Icelandic, since Dutch has also lost its Case distinctions but retained OV order, while Icelandic has preserved its rich Case morphology, but nevertheless changed to basic VO order (cf. Hróarsdóttir 1998 for additional discussion of this issue). Recently an alternative approach for the change in word order in English was proposed that assumes that the change from OV to VO is due to language contact and grammar competition in Early Middle English (EME) (cf. Pintzuk 1999, Kroch and Taylor 2000). This approach is based on the so-called double base hypothesis (Pintzuk 1999) according to which word

46

Roland Hinterhölzl

order variation follows from the co-existence of competing grammars that differ with respect to the head parameter of VP and IP. One of the central assumptions of the double hypothesis is that VO orders are an EMEinnovation that was brought about by language contact between AngloSaxons and the Scandinavian settlers in the 10th century. 1.2. Word order variation in Older Germanic If we look at word order regularities in the older stages of the Germanic languages, then we find that both OV- and VO-properties already existed in Old English (OE), rendering the contact scenario from above less plausible. Furthermore, if we look at Old High German (OHG) and Old Icelandic (OI) (cf. Hróarsdóttir 1998), we find a similar kind of variation in word order suggesting that these mixed word order properties should not be treated in terms of language contact but may simply be part of the common Germanic inheritance. In the following I will restrict myself to a discussion of word order variation in OE and OHG. The examples in (1) illustrate typical OV-properties in OE. For example, in (1a) the direct object and the verb particle precede the finite verb in final position within an embedded clause and in (1b) the non-finite verb precedes the finite auxiliary in sentence final position, as is typical in OVlanguages. The same state of affairs, maybe less surprisingly, also holds in OHG, as is illustrated in (2). (1)

a. þæt he his stefne up ahof (Pintzuk 1991:71) that he his voice up raised ‘that he raised up his voice’ b. forþon of Breotone nædran on because from Britain adders on scippe lædde wæron (Pintzuk 1991:117) ships brought were ‘because adders were brought on the ships from Britain’

(2)

a. soso zi In gisprochan uuas how to them spoken was ‘how it was spoken to them’ b. thaz then alton giqu&an what to-the old ones said ‘what was said to the old ones’

(T 37, 5) uúas was

(T 64, 13a)

Information structure, word order and word order change 47

However, in both languages, we also find ample evidence of properties that one would associate with VO-languages, as is illustrated in (3) for OE and in (4) for OHG. For instance in (3a), the direct object follows the selecting verb in an embedded clause and in (3d) the manner adverb follows the verb it modifies. Both properties are typical of VO-languages. (3)

a. þæt ænig mon atellan mæge ealne one demm that any man relate can all the misery (Pintzuk 1991:36) ‘that any man can relate to all the misery’ b. forðam ðe he licettað hi because that they pretended themselves unscyldige (van Kemenade 1987:35) innocent ‘because they themselves pretended to be innocent’ c. he ahof þæt cild up geedcucod and he raised the child up quickened and ansund (van Kemenade 1987:36) healthy ‘he raised the quickened and healthy child up’ d. þæt he þæt unaliefede dod aliefedlice that he the unlawful did lawfully (van Kemenade 1987:36) that he did the unlawful lawfully’

Similar examples can be found in OHG. For example in (4a), the subject follows the selecting verb in an embedded clause and in (4b), the participle follows the selecting auxiliary, an order that is ungrammatical in modern standard German (though there are dialects that allow for right-branching verb clusters), but typical for a VO-language. (4)

a. thaz gibrieuit uuvrdi al these that listed was-SUB all this ‘that all this mankind was listed’ b. thaz sie uuvrdin gitoufit that they were baptized ‘that they were baptized’

umbiuuerft (T 35, 9) mankind (T 46, 25)

While it has been argued for English that the variation illustrated in (3) should not be accounted for in terms of assuming a basic OV-grammar plus extraposition of heavy material and really calls for the assumption of an

48

Roland Hinterhölzl

additional VO-grammar (cf. Pintzuk 1999), the case has never been made for OHG. The standard treatment of OHG is that of an OV-language that allowed for a greater amount of extraposition than modern German (cf. Lenerz 1984). In this connection, it is interesting to note that there is convincing evidence that OHG had a right-peripheral focus position, providing an explanation for why and when constituents follow the selecting verb in embedded clauses in OHG. Some examples of this are given in (5)–(7) from the Tatian translation. These examples are of particular interest, since they deviate in word order from the Latin original. Let us have a closer look at (5), which illustrates very well how OHG typically differs from Latin. In (5), the discourse-given DP thin ouga is preposed, while the constituent that carries the new important or relevant information, namely luttar, is put in post-finite position, to occupy the default position for a focussed constituent. (6) illustrates a case of a postverbal nominal predicate that constitutes the new information in the given context, while (7) shows the same property for a direct object. (5)

liohtfaz thes lihhamen ist ouga / oba thin ouga uuirdit luttar / the light of the body is the eye. if your eye becomes light, thanne ist al thin lihhamo liohter (T 69, 21ff.) then is all your body brighter ‘The light of the body is the eye. If your eye becomes light, then all your body is brighter.’ Lucerna corporis. est oculus. / si fuerit oculus tuus simplex. / totum corpus tuum lucidum erit.

(6)

ther giheizan who named ‘who is named Petrus’ qui vocatur p&rus

ist is

p&rus (T 54, 15) Petrus

(7)

Inti bráhtun imo/ alle ubil habante / […]]/ Inti thie thár hab&un and brought him all sick ones and those there had diuual (T 59,1) devil ‘and they brought him all the sick ones and those that had the devil’ &obtulerunt ei/ omnes male habentes/ […]/ &qui demonia habebant

Information structure, word order and word order change 49

Another important aspect of the data in (3) and (4) is that we do find not only sentences with pure OV-properties and sentences with pure VOproperties in these languages, but very often we find sentences with mixed word orders. For instance, in (3a) it can be seen that, while the direct object follows the selecting verb (and the modal), a typical VO-property, the infinitive precedes the selecting modal which is rather typical of an OVlanguage. In (4d), the direct object precedes the selecting verb, a typical property of an OV-language, while the manner adverb follows the verb that it modifies, which is typical of VO-languages and ungrammatical in OVlanguages. Likewise in (4), while the subject follows the selecting verb (and the auxiliary), the participle precedes the auxiliary, as is typical in OV-languages. Given this state of affairs, at least two questions arise at this point. First, we must address the issue of how to account for mixed word orders in the older stages in these languages. This question will be dealt with in the following section. Second, we would like to know which factors led German and English to develop from a common OV/VO basis into pure OV and pure VO languages, respectively. This question will be addressed in section 3.

2. Word order variation and IS Our approach is to assume that OV/VO orders do not signal the presence of two grammars, contrary to the double base hypothesis (cf. Pintzuk 1991), but that the variation illustrated in (1)–(7) is due to the expression of different IS-categories within one grammar. The advantage of this approach is clear when it comes to the characterization of mixed word orders. Proponents of the double base hypothesis not only have to assume that a speaker possesses two grammars or two settings of the head complement parameter, but also that he can switch between the settings of the head complement parameter within one sentence. To give a concrete example, in order to derive the sequence relate can all the misery in (3a) proponents of the double base hypothesis have to assume that the speaker can switch from an OV grammar – in order to derive the verbal complex Infinitive > finite Modal – to a VO grammar, to account for the object in postverbal position. Note specifically, that the assumption of an OV grammar plus extraposition in this case will not be sufficient, since well-behaved OV languages like German and Dutch do not allow for extraposition of DP-arguments.

50

Roland Hinterhölzl

I have argued in Hinterhölzl (2004a) that word order change should not be explained as a change in the head complement parameter, since OV and VO languages differ in properties that cannot be subsumed under the head complement parameter. These properties involve the availability of heavy adjuncts in the middle field and the order and the preverbal versus postverbal positioning of event-related adverbs. Instead I have argued that word order properties are defined by prosodic properties, namely the headedness of phonological phrases. Therefore, I will dispense with the head complement parameter and adopt the Universal Base Hypothesis (UBH) (cf. Kayne 1994), according to which all syntactic structure is head-initial. In such an approach, it is necessary to distinguish between the base order, which is universally defined as Specifier – Head – Complement, and the unmarked word order of a language. For instance, the unmarked word order in German (OV) cannot be taken to be basic property (to be identified with the base order) anymore, but has to be derived from other properties in the language. Nespor, Guasti and Christophe (1996) propose that the head complement parameter is determined by the predominant, that is, unmarked prosodic patterns in an early phase during language acquisition (the rhythmic activation principle). More specifically, they argue that the decisive information for the child is the placement of main prominence within the phonological phrase. In Hinterhölzl (2004a), I have adopted this approach and proposed that the unmarked word order in the phrases of a language are determined by the predominant, that is, unmarked prosodic patterns in that language. That a language can have several unmarked prosodic patterns is shown by German. While with DPs and PPs the unmarked prosodic pattern is (weak strong) ((w s)), the unmarked prosodic pattern with VPs is (s w). It is interesting to note that APs show both types of prosodic patterns with a strong preference for the verbal pattern (s w), as is illustrated in (8). (8a) is the neutral order, while (8b) is rather marked and can only be used for specific communicative purposes. (8)

a. weil Hans [[auf die Maria] stolz] ist (unmarked) b. weil Hans [[stolz [auf die Maria]] ist (marked) since Hans (of the Maria) proud (of the Maria) is ‘since Hans is proud of Maria’

In conclusion, I would like to propose that syntactic structures are not marked per se (say, in terms of complexity), but count as marked or un-

Information structure, word order and word order change 51

marked if they realize marked or unmarked prosodic patterns. Since the unmarked word order in a language is defined by the predominant, that is to say, the most frequent prosodic pattern in a language, a change in frequency of use of a prosodic pattern can lead to a change in unmarked word order. This is the approach to word order change that I would like to pursue in the following section. Since the expression of IS-categories influences the default mapping between syntactic structure and prosodic structures, as I will show below, IS plays an important role in processes of word order change. But before we take a look at issues of word order change, let me first discuss which factors determine the word order regularities in OE and OHG. There is the observation by Behaghel (1932) that pronouns and unmodified nouns tend to precede the verb, while modified nouns, PPs and other heavy material tend to follow the verb that gave rise to the generalization in (9). The question arises from which principle of grammar this tendency derives from. (9)

Light elements precede heavy elements in OE, OI and OHG. (Behaghel 1932: Das Gesetz der wachsenden Glieder)

There is another generalization that emerged from our IS-analysis of the Tatian translation (to be revised below) that derives (9) as a mere corollary. Given that discourse-given elements are typically realized as light elements, while foccussed constituents may count as prosodically heavy elements, since they receive stress, (10) derives the tendency expressed in (9). (10) C

background

V

focus

The generalisation in (10) allows us to account for word order variation within one grammar by taking into account the information-structural contribution of a constituent in the discourse: for instance, a direct object will precede the verb (in embedded clauses) if it is discourse-given but will follow it, if it is discourse-new. According to the generalisation in (9) a direct object will be placed preverbally if it is realized as a pronoun or single noun, but postverbally if it is made heavy by modification. As I have indicated above, the two conditions are not independent of each other, but it would be interesting to see which one of them is more basic (see also Hroarsdottir this volume for similar observations in OI).

52

Roland Hinterhölzl

2.1. Prosodic constraints and word order In this section, I will argue that Behaghel’s law can be derived from a (violable) interface condition that applies in the mapping between syntactic structure and prosodic structure. First, however, let us take note of the fact that Behaghel’s law can be overwritten by syntactic requirements. The generalisation in (10) needs to be refined for OHG in as much as contrastive foci are concerned. While new information focus is typically realized in postverbal position, contrastive focus is realized in preverbal position and seems to involve leftward movement of the finite verb, as is illustrated in (11). (11) C

background

contrastive focus

V

presentational focus

In this respect, OHG seems to pattern with Yiddish, which is a West Germanic language that has preserved mixed word orders. According to the description of Diesing (1997), constrastive foci pattern with background elements in occupying a preverbal position. As I have argued in Hinterhölzl (2004a), the pattern in (11) can be derived from the following assumptions about the syntax of focus. Assuming the UBH, arguments move out of the VP to be licensed in (Case-) Agreement positions. A structural focus position is located above these licensing positions. The word order facts in (11) then follow from the following assumptions: A) The verb moves into the Focus head. B) A contrastively focused phrase moves into [Spec,FocP]. C) A constituent that represents new information focus just stays in scope of the Focus head, while D) background elements move out of the scope domain of the Focus head. This is illustrated in (12). (12) Assumptions about the syntax of focus (Hinterhölzl 2004a) [C background [FocP ContrastF V [AgrP PresentationF [VP]]]] Coming back to the placement of constrastive foci, an illustrative example is given in (13). Note that in example (13) – which needs to receive a contrastive interpretation, since the contrast is made explicit in the context – the contrastive element is placed preverbally against the order in the Latin origin, signifying that we are dealing with an independent requirement of OHG.

Information structure, word order and word order change 53

(13) thane thu fastes/ salbo thin houbit/ Inti thin annuzi thuah/ zithiu thaz when you fast, anoint your head, and your face wash, so that thu mannon nisís gisehan/ fastenti. úzouh thimeno fater you men not appear fasting. but to your Father (T 68, 28–32) ‘[When you fast, do not be like the hypocrites…] When you fast, anoint your head and wash your face so that you do not appear to men to be fasting but to your Father’ tu autem cum ieiunas/ unge caput tuum/ & faciem tuam laua/ ne uideatis hominibus/ ieiunans. Sed patri tuo In this context, it is interesting to note that PPs that are placed predominantly postverbally in accordance to Behaghel’s law (due to their heaviness) appear preverbally when contrastively focussed, as is illustrated in (14). (14) b&onte nicur& filu sprehan/ sósó thie heidanon mán/ sie uuanen Praying be careful not to much speak, as the heathen men. they think thaz sie in iro filusprahhi / sín gihórte (T 67, 23–26) that they in their many words are heard ‘And when you pray, do not use vain repetitions as the heathens do. For they think that they will be heard for their many words’ orantes autem. nolite multum loqui/ sicut &hnici.’/ putant enim quia in multiloquio/ exaudiantur. The PP in (14) comprising three words is rather heavy but nevertheless appears preverbally to express that the constituent is contrastively focussed. Despite of cases like (14), we can nevertheless assume that the restriction behind Behaghel’s rule is real and we will see that it is still operative in Modern English and Modern German, albeit in different domains, as is shown in the following section. 2.2. Prosodic constraints and phases The first question that arises with Behaghel’s law is the issue of when a constituent counts as heavy. In this paper, I propose that heaviness should be identified with a branching prosodic constituent and that Behaghel’s law should be properly understood as flowing out of a condition that defines the best match between syntactic structure and prosodic structure.

54

Roland Hinterhölzl

But first, let us look at some other instances in which heaviness has been argued to play a role in cases of word order differences between languages. I think it was Haider (2000), who first observed that adverbs in the English middle field are subject to conditions which are absent in a typical OV-language like German, as is illustrated in (15) and (16) (cf. also Hinterhölzl 2001). (15) a. John (more) often (* than Peter) read the book b. Hans hat öfter (als der Peter) das Buch gelesen (16) a. John read the book more often than Peter b.* Hans hat das Buch gelesen öfter (als Peter) (15a) shows that the English middle field does not tolerate heavy constituents, while no problem arises in the German middle field (15b). More specifically, we can make the following empirical generalization: the head of the adjunct may not contain material on its right. Such adjuncts must be postponed as in (16a), while heaviness alone does not constitute a license for postposition in German (16b). A prosodic restriction can also be used to account for the difference in the placement of event-related adverbs in German and English in general. While Time, Place and Manner adverbs appear in the order T>P>M preverbally in OV-languages, they appear in the exact mirror order postverbally in VOlanguages (cf. Haider 2000, Hinterhölzl 2002), as is illustrated in (17). (17) a. b.

C C

T P M-V V- M P T

OV-languages VO-languages

In Hinterhölzl (2002; 2004a), I have argued that the order found in German is basic and that the English order is to be derived from the German word order in terms of successive cyclic VP-intrapositon that pied-pipes the adjunct at each step1. Furthermore, I have argued that VP-intraposition came about due to a stylistic rule of light predicate raising that was operative in OE and affected typically event-related adjuncts, since they were primarily realized as rather heavy NPs and PPs. In this respect, it is interesting to note that a similar restriction also applies in German verb clusters. German verb clusters are predominantly leftbranching, but right-branching verb clusters are possible as long as the most deeply embedded cluster is left-branching (cf. Hinterhölzl 1999). A case in question is given in (18a). However, once a right-branching verb cluster is

Information structure, word order and word order change 55

introduced, the verb cluster must be also right-branching at the next level up, as is illustrated by the contrast in (18b) and (18c). (18) a. b.?? c.

weil er den Text muß lesen können since he the text must read can ‘since he must be able to read the text’ weil er den Text [[müssen [lesen können]] wird] since he the text must read can will ‘since he will have to be able to read the text’ weil er den Text [wird [müssen [lesen können]]] since he the text will must read can ‘since he will have to be able to read the text’

The formation of verb clusters is motivated by the following two licensing requirements (cf. Hinterhölzl 2006): dependent verbs move into dedicated positions in the V-domain to be temporally linked and to check the subcategorisation of the selecting verb. F20 is responsible for temporal linking and Asp0, the highest head in the V-domain, is responsible for checking the subcategorisation of the matrix verb. In this approach, left-branching verb clusters are derived if the dependent verbs are spelled out in the highest Specifier, as is illustrated in (19a), while right-branching verb clusters are derived if the dependent verbs are spelled out in the lower Specifier (19b). The only exception to this rule are verb clusters comprising an IPP-infinitive (an infinitive that replaces a part participle). These verb clusters are obligatorily right-branching, since in this case, there is a null morpheme that moves to the highest head in the Vdomain, stranding the dependent verbs in the lower Specifier, as is illustrated in (19c). (19) a. [AspP [lesen können] muss [F2P [lesen können] [VP]]] b. [AspP [lesen können] muss [F2P [lesen können] [VP]]] c. [AspP 0-hat [F2P [lesen können ] [VP ]]] The interesting question now is what the generalisation illustrated in (18bc) results from. In any event, the generalisation cannot be derived from a hard syntactic condition in West Germanic, since we can find many instances that violate it, as is the case with IPP-infinitives in West Flemish. (20) illustrates a right-branching verb cluster headed by willen which itself sits on a left-branch with respect to the selecting auxiliary een.

56

Roland Hinterhölzl

(20) dan ze toch kosten [[willen [dienen then they but could want-IPP that boek kuopen]] een] book buy have ‘if they only could have wanted to buy that book’ A possible solution is to relegate the contrast in (18bc) to a violable interface condition that determines the best match between a given syntactic structure and a prosodic output structure. Given that left- and rightbranching verb clusters are mapped onto left- and right-headed phonological phrases, a possible candidate for such an interface condition is (21). (21) Mapping Condition between syntactic structure and prosodic structure: A right-headed phonological phrase (in a verb cluster) must sit on a right branch with respect to the syntactic head that is to become its prosodic sister (21) can probably be formulated in a more elegant way. But it is meant to account for the patterns in (22). The patterns in (22) indicate the deeper reason that probably lies behind the condition in (21): once fixed, stress tends to stay in a peripheral position within a certain domain. (22b) is okay, since stress remains left-peripheral and (22c) is okay, since stress remains right-peripheral, while stress in (22a) is neither leftnor right-peripheral, unless H itself is stressed. (18b) violates the condition in (21). The violation can be circumvented, if the (already) right-branching verb cluster is spelled out in the lower Specifier, as is the case in (18c). (22) a.* 3 b. ok 3 c. ok 3 2 2 2 2 2 ((A B) H) ((A B) H) (H (A B )) Note, however, that condition (21) is a condition that can be taken to apply in the V-domain and in the I-domain (excluding the subject) in VOlanguages. In fact, (21) can be used to rule out (15a), since more often than Peter is a right-headed prosodic constituent that sits on left branch with respect to the verb phrase that is to become its prosodic sister. Since German tolerates heavy adjuncts and arguments in the middle field, this condition cannot be taken to apply in the German middle field, but – as we have seen above – is still active in the German V-domain. Why should that be so?

Information structure, word order and word order change 57

In this respect, it is interesting to note that prosodic constraints can apply phasewise and can thus hold in one phase while not being operative in another phase. The crucial evidence comes from restrictions on VPtopicalization in German. As is illustrated in (23), topicalized rightbranching verb clusters exhibit an interesting contrast. In general, the topicalization of a right-branching verb cluster leads to ungrammaticality, unless it comprises an IPP-infinitive, in which case the topicalization is rather marked but grammatical. (23) a.?* b.?

[müssen [lesen können]] wird er den Text (ok lesen können müssen wird) must read can will he the text [haben [lesen wollen]] wird er den Text (* lesen wollen haben wird) have read want-IPP will he the text ‘he will have wanted to read the text’

The violation in (23a) cannot be accrued in the C-domain, since the Cdomain in German does tolerate right-branching prosodic constituents. Thus the violation must have been induced in the V-domain. A violation of (21) can then be taken to be induced in a phase-based derivation, if we make the assumption that the Aspect phrase (and not the vP as in Chomsky 2001) constitutes the edge of the strong VP-phase. Since we can assume that topicalization involves movement from one strong phase (the VP) into the next strong phase (the CP), it must be taken to be subject to the Phase Impenetrability Condition, given in (24b). Phase Condition (Chomsky 2001) (24) a. Evaluation for a phase is done at the level of the next highest strong phase b. Phase Impenetrability Condition (PIC) The complement of a strong phase a is not accessible to operations at the level of the next highest strong phase b, but only the head and the edge of a are Given the PIC, extraction of the right-branching verb cluster must take place via the highest Specifier in the V-domain, namely [Spec,AspP]. In this position, however, the verb cluster will induce a violation of the mapping condition in (21), which will result in ungrammaticality if there is an alternative that does not violate it. Since the verb cluster could have been spelled-out as a left-branching verb cluster, (23a) is ungrammatical. Since

58

Roland Hinterhölzl

IPP-infinitives only allow for right-branching verb clusters, there is no derivational alternative in the case of (23b), thus the resulting topicalization is prosodically marked but grammatical. This solution, however, is only feasible in a framework in which Spellout applies cyclically, including the cyclic, that is, phasewise application of conditions that map syntactic structure onto prosodic structure. More specifically, we have seen above that interface conditions can apply phasewise in the sense that they may hold in one phase (the V-domain in German) without holding in another phase (the I-domain, or middle field, in German). In conclusion, word order variation can be accounted for by the differential expression of IS-categories within one grammar and word order preferences (Behaghel’s law) are due to violable interface conditions that define the ideal mapping between syntactic structure and prosodic structure in the course of the derivation. As we have seen above in the case of contrastive foci, information-structural restrictions play a crucial role in this mapping.

3. Word order change and IS Given the discussion of the previous section, the development of German is to be described as one in which certain factors led to the non-application of the prosodic constraint in the German middle field allowing for the Spellout of branching constituents in the preverbal domain, while the development of English is to be described as one in which certain (other) factors not only led to the retention of the prosodic constraint in the middle field but also caused light (non-branching) constituents like pronouns to be spelled out in the postverbal domain. I will concentrate here on discussing the potential factors that can help us understand the development of German. As far as the development of English is concerned, I will only discuss it in so far as to make plausible why the pertinent factors working in German did not have the same effect in the history of English. Given the presence of both a preverbal and a postverbal focus position, we can assume that OHG had two unmarked word order patters: due to focus restructuring (cf. Hinterhölzl 2004a) preverbal focus will give rise to prosodic phrases of the type (s w) and postverbal focus will give rise to prosodic phrases of the type (w s). Hence, if we can identify a factor that led to the occurrence of more and more focussed (stressed) preverbal constituents, then we can envisage a slow development in which certain word

Information structure, word order and word order change 59

order patterns become more and more marginalized to a point where postverbal stressed constituents are prosodically highly marked such that they can only be used for specific communicative purposes. In the following, I would like to argue that this factor was the grammaticalization of the definite determiner. Following this line, we have to consider a development from the system in (12) above, in which noun phrase interpretation was signalled by syntactic position to a determiner system which signals noun phrase interpretation with morphological means. First, note, that the morphological system that became grammaticalized is not congruent with the distinction between given and new discourse referents (that was embodied in the syntactic system in (12)), since the definite determiner came to signal that the discourse referent – whether given in the discourse or discourse-new – is uniquely identifiable in the context. Thus, the development of this morphological determiner system has blurred the original information-structural distinction of (nominal) arguments. The orginal distinction between definites, however, is still visible in the scrambling rule of modern German. Haider and Rosengren (1998) note that neither definiteness nor specificity can be taken to be triggers of scrambling, citing examples like (25), where a specific indefinite NP in (25a) and a definite NP in (25b) seemingly occur in their base position. (25) a. wenn wer eine rothaarige Frau sucht if someone a red-haired woman seeks dann ist das Maria then is it Maria ‘if someone seeks a red-haired woman, then it is Maria’ b. dass er wem ihr Kleid gezeigt hat, that he someone her dress shown has, Maria nicht gefallen Maria not pleased ‘that he has shown someone her dress, hasn’t pleased Maria’

hat has

Note, however, that their examples only involve NPs that must be characterized as discourse-new. While they are right about definiteness not being a trigger for scrambling, they are wrong about specificity. If specificity is understood as memberhood in a set that is given in the discourse (cf. Enç 1991) it can be taken as a valid trigger of scrambling. In conclusion, only discourse-given DPs may scramble; DPs that are discourse-new, even when

60

Roland Hinterhölzl

they are uniquely identifiable (by binding or bridging, for instance) stay in the focus-domain and receive stress (cf. Hinterhölzl 2004b). But why should the grammaticalization of a definite determiner have an influence on the unmarked word order? There are three observations that fit nicely into an explanatory pattern of this kind. First, recent studies of the grammaticialization of the definite determiner (cf. Oubouzar 1992, Leiss 2000, Demske 2001) agree that the definite determiner – derived from the demonstrative pronoun – first appears in contexts with pragmatic definite interpretations in OHG. The pertinent distinction between pragmatic and semantic definites stems from Löbner (1985). While semantic definites are uniquely identifiable on the basis of their lexical meaning (functional concepts), pragmatic definites are uniquely identifiable in the context by being discourse given. The grammaticalization of the definite determiner starts in the earliest OHG-texts and is concluded in Notker (early 11th century), in which text all semantic groups of nouns (including abstract nouns and uniquely referring expressions) appear with the definite determiner. The slow stepwise process of grammaticalization can be examplified by the occurrences of the definite determiner in Otfrid. In this text, the definite determiner regularly appears with pragmatic definites, that is, in discourse-anaphoric uses, as is illustrated in (26), while with semantic definites, the determiner is still missing, as is illustrated in (27). Examples are taken from Demske (2001). (26) a. ein burg ist thar in lante… (O.I.11.23) a town is there in country ‘there is a town in the country’ b. zi theru steti fuart er thia druhtines to this city leads he the Lord’s muater (O.I.11.26) mother ‘he leads the Lord’s mother to this city’ (27) a. tho ward himil offan (O.I.25.15) there became sky open ‘there the sky opened’ b. inti iz hera in worolt sante (O.I.13.5) and it here into world sent ‘and sent it here into the world’ c. in ira barm si sazta [barno bezista] (O.I.13.10) in her lap she placed child most-beloved ‘she placed the most-beloved child in her lap’

Information structure, word order and word order change 61

The second observation concerns another statement by Behaghel (1932), given in (28), who notes that noun phrases with a determiner are introduced preverbally. (28) Behaghel (p. 79): ‘Substantiva mit Pronomen stehen auf der Seite der einfachen Wörter; zum Teil mag das daher rühren, dass ihnen der Artikel früher fehlte.’ Behaghel’s observation is in line with the stepwise grammaticalization from pragmatic definites to semantic definites, since discourse anaphoric NPs are typically realized preverbally in OHG. We now can envisage a development in two phases, as sketched in (29). (29) phase 1: the determiner is introduced preverbally for discourse-given referents phase 2: the use of the determiner is extended to all uniquely identfiable NPs, which are placed according to the pattern in phase 1 preverbally There are two consequences of interest in such a scenario. A) more DPs appear preverbally, which due to focus restructuring strengthen the prosodic pattern (s w) and B) preverbal DPs have a profound effect on the prosodic make-up of the language, since they introduce right-branching constituents on a left branch. We can assume that when the determiner is first introduced in discourse anaphoric contexts, it will carry stress (with the discourse anaphoric noun being destressed), such that the initial introduction of determiners in the preverbal domain does not violate the prosodic condition in (21). However, as the determiner is grammaticalized it becomes deaccented and stress will be placed on the noun per default (on the basis that lexical heads are stronger than functional heads) giving rise to more and more rightbranching constituents in the preverbal domain. The scenario sketched in (29) makes the following prediction: there should be a stage where an increased number of focused DPs appear preverbally but focused predicates (predicative adjective, nouns and participles) still appear predominantly in postverbal position. This prediction is borne out since the verbal cluster including predicative elements predominantly remained right-branching till the Early New High German period (cf. Ebert 1986). The third observation that supports this scenario concerns the time frame of these two developments. Bolli (1975), Borter (1982) and Näf

62

Roland Hinterhölzl

(1979), investigating the development of the German sentence bracket, note that in the late OHG Notker text a larger number of Accusative objects start to be placed in preverbal position. This is exactly the same period in which the definite determiner has become fully grammaticalized. If this scenario for German is correct, then the question arises why the grammaticalization of the definite determiner did not have the same effect in the history of English and in the development of the Scandinavian languages. As far as Scandinavian is concerned, Leiss (2000) in her interesting study on the relation between (verbal) aspect and (nominal) definitness notes that in OI the definite determiner first appears with discourse-new NPs that are not to be read as indefinites, hence in postverbal position. Thus we can conclude that – while the introduction of the definite determiner lead to the strengthening of the prosodic pattern (s w) in OHG – the very same process lead to the strengthening of the prosodic pattern (w s) in OI, explaining why the two languages have developed in opposite directions. The case of the development in English is more difficult to deal with. Philippi (1997) notes that the literature on OE suggests that the language did not have a definite or indefinite article, making use of demonstratives and numerals instead. Also van Kemenade (p.c.) reports that the definite determiner was grammaticalized in English at least 100 years later than in German. So, while the grammaticalization of the definite determiner in German was completed at the end of the OHG-period, the grammaticalization of the determiner in English must have started in the beginning of the ME-period. OE had two demonstrative pronouns se (‘that’) and þes (‘this’). Hroarsdottir (2006) reports that the demonstrative se which had different Case forms in OE was split into two invariant forms, the (article) and that (demonstrative) at the turn of the OE to the EME-period. Thus, it is possible that the English definite determiner took a different path of grammaticalization than the German one. We can assume that the was unaccented and therefore was placed with its containing noun phrase postverbally in accordance with the condition in (21). More research on the development of the definite determiner in English is necessary to evaluate the validity of this scenario, but I think it has enough initial plausibility to provide an interesting alternative to the improbable standard scenario that tries to relate word order change with the loss of Case morphology in the history of English. Above, I have provided a plausible scenario for why the prosodic condition in (21) was not restricted to the V-domain in the history of English (as

Information structure, word order and word order change 63

it was in the history of German), but what is still missing at this point is an explanation for why English has also started to postpone light elements, like pronouns. At this point I can only speculate about the origin of this development. But it stands to reason that language contact was at issue in this case, since at the end of the OE-period the Anglo-Saxon pronoun system was partially replaced with Scandinavian pronouns. Weak Scandinavian pronouns (nowadays), as is evidenced by the phenomenon of object shift, are enclitic elements that are spelled out in the smallest domain that contains a suitable host, with a suitable host being the head of the phase that the pronoun belongs to, that is, the verb in the VP and the preposition in the PP (cf. Hinterhölzl (to appear)). This explanation is supported by the fact that V2-second (before its loss) is generalized in the northern contact area in the EME-period in that the pattern Topic pronoun finite verb, in which the Old Saxon pronoun can be analyzed as proclitic element on the verb, is replaced with the pattern Topic finite verb pronoun, in which the EME-pronoun can be analyzsed as enclitic element on the verb (see Kroch and Taylor 2000 for more details on this development).

4. Conclusions The paper presents a number of novel ideas about the relation between grammar and prosody and the role that information-structure plays in language change. First it is argued that unmarked word order is determined by prosodic properties. Since information-structural constraints play a crucial role in the mapping between syntactic structure and prosodic structure, IS can help us to gain new insights into processes of word order change. For instance, we have seen that what counts as unmarked word order in a language may be influenced by focus articulation, since the order focus > verb can strengthen the prosodic pattern (s w) in the middle field. Furthermore, I have argued that interface conditions, like the condition in (21) are violable or best match conditions, leading to marked and unmarked syntactic structures. Since speakers, on the one hand, generally prefer unmarked structures and, on the other hand, tend to use marked structures to achieve particular communicative goals in specific communicative situations, we derive that in most stages there will be some sort of competition of marked and unmarked forms within one grammar.

64

Roland Hinterhölzl

While the strive towards prosodic unmarkedness might be the source for word order tendencies, the strive towards prosodic markedness might be one explanation for why we find variation in the domain of syntax, particularly in terms of word order and prosodic phrasing. In general, a marked form a will be weeded out by an unmarked form b if it does not give rise to a different interpretation from b. It follows that a marked syntactic form will only be grammaticalized if it can be identified with a distinct pragmatic interpretation. Possibly, this is the reason for why IS-categories are imported into the syntax in grammaticalization processes. In conclusion, if these general observations are on the right track, then the conception of narrow syntax as a computational system that only (or predominantly) operates on formal features (Case and phi-features) seems misguided to me und must be rejected on the basis of the above observations on word order change.

Note 1. In this approach a modern English sentence, like (ia), is derived as is illustrated in (ib-e). (i) a. John visited them in Vienna on Friday b. […[on Friday [in Vienna [John visited them]]]] c. […[on Friday [[John visited them] in Vienna tVP]]] d. […[[John visited them] in Vienna ] on Friday ] e. [IP Johni [[[[VP ti visited them] k in Vienna t k]j on Friday tj]]

References Behaghel, Otto 1932 Deutsche Syntax. Band 4. Heidelberg: Carl Winters Universitätsbuchhandlung. Bolli, Ernst 1975 Die verbale Klammer bei Notker. Untersuchungen zur Wortstellung in der Boethius-Übersetzung. Berlin/New York: Mouton de Gruyter. Borter, Alfred 1982 Syntaktische Klammerbildung in Notkers Psalter. Berlin/New York: Mouton de Gruyter. Chomsky, Noam 2001 Derivation by Phase. In Ken Hale: A life in language, ed. Micheal J. Kenstowicz, 1–52. Cambridge, M.A.: The MIT Press.

Information structure, word order and word order change 65 Demske, Ulrike 2001 Merkmale und Relationen: Diachrone Studien zur Nominalphrase des Deutschen. Berlin/New York: Walter de Gruyter. Diesing, Molly 1997 Yiddish VP order and the Typology of Object Movement in Germanic. Natural Language and Linguistic Theory 15: 369–427. Ebert, Rober Peter 1986 Historische Syntax des Deutschen II. 1300–1750. Bern: Peter Lang. Enç, Mürvet 1991 The semantics of specificity. Linguistic Inquiry 22 (1): 1–25. Haider, Hubert 2000 Adverb Placement-Convergence of Structure and Licensing. Theoretical Linguistics 26: 95–134. Haider, Hubert and Inger Rosengren 1998 Scrambling. Sprache & Pragmatik 49. Hinterhölzl, Roland 1999 Restructuring Infinitives and the Theory of Complementation. Doctoral dissertation, Department of Linguistics, University of Southern California. 2001 Event-related Adjuncts and the OV/VO Distinction. In WCCFL 20 Proceedings, eds. Karine Megerdoomian and Leora Bar-el, 276–289. Somerville: Cascadilla Press. 2002 Parametric Variation and Scrambling in English. In Studies in Comparative Germanic Syntax (Proceedings of CGSW 15 in Groningen, May 2000), eds. Werner Abraham and Jan-Wouter Zwart, 131–150. Amsterdam & Philadelphia: John Benjamins. 2004a Language Change versus Grammar Change: What diachronic data reveal about the distinction between core grammar and periphery. In Diachronic Clues to Synchronic Grammar, eds. Eric Fuss and Carola Trips, 131–160. (Linguistik Aktuell 72.) Amsterdam: John Benjamins. 2004b Scrambling, Optionality and Non-lexical Triggers. In Triggers, eds. Anne Breitbarth and Henk van Riemsdijk, 173–203. (Studies in Generative Grammar 75.) Berlin: Mouton de Gruyter. 2006 Scrambling, Remnant Movement and Restructuring. Oxford: Oxford University Press. to appear A phase-based comparative approach to modification and word order in Germanic. To appear in: Syntax. Hroarsdottir, Þorbjörg 1998 Verb Phrase Syntax in the History of Icelandic. Doctoral dissertation, Universitetet i Tromsö. 2006 Loss of OV and the role of morphology. Ms., Universitetet i Tromsö.

66

Roland Hinterhölzl

Kayne, Richard 1994 The Antisymmetry of Syntax. (Linguistic Inquiry Monograph 25.) Cambridge, M.A.: MIT Press. Kemenade, Ans van 1987 Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris. Kroch, Anthony and Ann Taylor 2000 Verb-Object Order in Early Middle English. In Diachronic Syntax. Models and Mechanisms, eds. Susan Pintzuk, George Tsoulas and Anthony Warner, 132–163. Oxford: University Press. Leiss, Elisabeth 2000 Artikel und Aspekt: die grammatischen Muster von Definitheit. Berlin: Mouton de Gruyter. Lenerz, Jürgen 1984 Syntaktischer Wandel und Grammatiktheorie. Tübingen: Niemeyer. Löbner, Sebastian 1985 Definites. Journal of Semantics 4: 279–326. Näf, Anton 1979 Die Wortstellung in Notkers Consolation. Berlin/New York: Mouton de Gruyter. Nespor, Marina, Maria Teresa Guasti and Anne Christophe 1996 Selecting Word Order: The Rhythmic Activation Principle. In Interfaces in Phonology, ed. Ursula Kleinhenz, 1–26. Berlin: Akademie Verlag. Oubouzar, Erika 1992 Zur Ausbildung des bestimmten Artikels im Althochdeutschen. In Althochdeutsch: Syntax und Semantik, ed. Yvon Desportes, 71–87. Lyon: Universite Lyon III Jean Molin. Philippi, Julia 1997 The rise of the article in the Germanic languages. In Parameters of Morphosyntactic Change, eds. Ans van Kemenade and Nigel Vincent, 62–93. Cambridge: Cambridge University Press. Pintzuk, Susan 1991 Phrase structures in competition: variation and change in Old English word order. Doctoral dissertation, University of Pennsylvania. 1999 Phrase Structures in Competition: variation and change in Old English word order. New York: Garland.

OV languages: Expressions of cues Þorbjörg Hróarsdóttir

1. Introduction The aim of this chapter is to argue that the cue for the OV/VO parameter was expressed through information structure in Older Icelandic (OI). I will show how a change in encoding of information structure may lead to a change in the basic order of verb and its complements. The claim is that a gradual change in language use, namely the increased postverbal positioning of focused material led to a sudden change in the grammar, where the option of VP-extraction disappeared. Roberts (1997) argues that English lost overt object-movement due to the loss of morphological case in Middle English (ME). As seen from the English viewpoint, low-level facts of inflectional morphology may express the relevant cue for parameters and so the loss of inflection may (but does not have to) lead to a grammar change. This analysis does not carry over to Icelandic as the loss of OV in Icelandic took place despite rich case morphology. I assume that learners must watch out for a universal cue for the positive value of the OV/VO parameter. However, this cue may be expressed differently among the languages: while it may have been expressed through morphology in Old English (OE), it was expressed through information structure in OI. In both cases, external effects led to fewer expressions of the relevant cue and a grammar change took place. I will illustrate how the grammar change in OI was due to a change in information structure. The loss of OV in the grammar was caused by a prior language change that involved objects that express new information: These objects gradually gained a higher frequency in postverbal position. The OV/VO word order patterns are more or less stable until the seventeenth century where there is a gradual loss of the OV word order patterns, or, a gradual increase in language use, placing focused elements in postverbal (VO) position (a shift in discourse property). Morphological triggers, or the absence thereof, have been related to clustering of syntactic properties in both diachronic change and acquisition. The correlation between overt verb movement and ‘rich’ agreement paradigm has, for instance, been proposed by several researchers in recent

68 Þorbjörg Hróarsdóttir years (see Kosmeijer 1986; Roberts 1993; Rohrbacher 1999; Holmberg and Platzack 1995; Vikner 1997; Koeneman 2000).1 If I am correct, then the relationship between strong morphology and OV order has been overestimated. I claim that this relation is only onedirectional, in the way that languages that lose strong morphology also tend to lose overt OV order (English), because of structure sandwiching, but not vice versa. Therefore, languages may have overt OV orders despite weak morphology, and languages may lose overt OV structure regardless of whether there have been any changes in the morphology or not (Icelandic). Moreover, many modern languages have rigid word order despite their case morphology, and vice versa (cf. Siewierska 1998).

2. Loss of OV in Icelandic grammar 2.1. Introduction Let us now look at the loss of OV word order in the history of Icelandic. Modern Icelandic has pure VO-order within the VP, as shown in (1). The word order in (1) with auxiliary verb – main verb – object is the only possible order of these elements in Modern Icelandic (abstracting away from topicalization and stylistic fronting). (1)

Þeir munu aldrei hafa lesið bókina they will never have read book.the ‘They will never have read the book’

Unlike in Modern Icelandic, several orders of the VP-internal arguments were possible at earlier stages in the history of Icelandic, including both OV and VO word order patterns (cf. Sigurðsson 1988; Rögnvaldsson 1996; Indriðason 1987; Hróarsdóttir 1996, 2000). The attested OV word order patterns were lost at the beginning of the nineteenth century. Two examples of OV word order in OI are shown in (2) below. (2)

Pure OV word order a. að hann hafi hana drepið (Álf) that he had her killed ‘that he had killed her’ b. að þú ... hafir það bréf fengið (letters) that you ... have that letter received ‘that you have received that letter’

OV languages 69

Icelandic has had a rich subject-verb agreement morphology and case morphology throughout its history. It is also generally assumed that both Old and Modern Icelandic have an obligatory overt movement of the finite verb out of the VP (to IP/CP), in both main and subordinate clauses.

2.2. The diachronic aspect Although it has been shown that it is possible to derive all the attested word order patterns in OI by assuming SVO word order base with overt versus covert object movements (cf. Hróarsdóttir 1996; 2000: Ch.5), this hypothesis does not easily handle the diachronic aspect. With regard to the order of objects and non-finite verbs, OI differs from Modern Icelandic in three major aspects, as shown in (3). (3)

a. b. c.

Vfin … object – Vaux – Vmain Vfin … Vaux – object – Vmain Vfin … (object) – Vmain – (*object) – Vaux – (object)

Some OI examples are shown below. (4)

Vfin … object – Vaux – Vmain a. að hon mundi eigi barn mega eiga (Finn) that she would not child may own ‘that she would not be allowed to have a child.’ b. ef hann hefði það viljað fága (Guðm) if he had it wanted clean ‘if he had wanted to clean it.’

(5)

Vfin … Vaux – object – Vmain a. að hann skyldi aldrei mega sól sjá (Árm) that he should never be-allowed sun to.see ‘that he should never be allowed to see the sun’ b. og ekki skal faðir minn geta þér hjálpað (Álf) and not shall father mine can you helped ‘And my father will not be able to help you.’

70 Þorbjörg Hróarsdóttir (6)

Vfin … Vmain – Vaux – object a. at hann mun raða vilia ferðum sínum (Finn) that he will decide want journeys his ‘that he wants to decide his own journeys’ b. þu munt haft hafa harða landtoku (Finn) you will had have hard landing ‘You have had a rough landing.’

(7)

Vfin … object – Vmain – Vaux a. og enginn þóttist þvílíkan veikleika séð hafa (Álf) and no-one pretended such weakness seen have ‘And no one pretended to have seen such weakness.’ b. að eg mundi hann sigrað geta (Árm) that I would him defeat could ‘that I would be able to defeat him’

All these word order patterns are ungrammatical in Modern Icelandic (abstracting away from negative and quantified phrase constructions), that is, the short object movement (within the VP), the long object movement (scrambling out of the VP, to the left of non-finite auxiliary verbs and sentential adverbs), and the possibility of placing the non-finite main verb to the left of the non-finite auxiliary (V-raising). The striking fact is that all these three leftward movement processes disappeared from the language at the same time in the history of Icelandic. Furthermore, the possibility of placing DPs, on the one hand, and PPs and small clause predicates, on the other hand, to the left of the main verb also seems to have disappeared from the language at the same time.

3. Theoretical implications A central question in the comparison of OV- and VO-languages is whether the difference in word order results from having more object movements in OV-languages, or more verb movements in VO-languages. Here we will argue, following Hróarsdóttir (2000), that although a uniform VO-base hypothesis with overt versus covert leftward movement of objects yields positive results in the synchronic analysis of the Older Icelandic VP, it cannot easily handle the diachronic aspect. Hence, we will agree with the original proposal of Haider (1992) that there are good reasons to assume that the verb moves more in VO-languages. Haider thinks of this verb

OV languages 71

movement as head movement. However, this could also be a VPmovement, provided that the complements have first moved out of this VP. This is the path that Hinterhölzl (2006), Kayne (1998) and Hróarsdóttir (2000) take. They suggest that VO-order might in some cases result from shifting a remnant VP containing the verb across complements extracted from the VP. Hence, the displacement of the verb could be a consequence of either (i) head movement of the verb, or (ii) movement of a larger constituent containing the verb, e.g. VP. We will assume the latter option here. Hróarsdóttir (2000) argues that the morphological features relevant for the triggering of leftward object movements in OI were indeed not weakened. Therefore, the objects still have to move to the left of the main verb (to [Spec, AgrOP]) in overt syntax in Modern Icelandic. Hróarsdóttir (2000) makes use of remnant-movement of various kinds of predicative phrases, and the long movement associated with ‘restructuring’ phenomena, to provide an analysis of OV orders, and correspondingly, a proposal as to which aspect of Icelandic syntax must have changed when VO word order became the norm: the essential change is loss of VP-extraction from VP. The claim is that the crucial difference between OV and VO languages is simply that OV-languages lack the VP-preposing Modern English and other VO languages have. The picture is slightly more complicated for mixed OV/VO languages such as OE, OI, Afrikaans, and Dutch. However, it is possible to construct a theory with a universal base that derives all the attested OV and VO word order patterns, by means of three transformations: – – –

optional extraction of the embedded VP from the matrix VP in Older Icelandic. obligatory and universal movement of the direct object out of the VP (to [SpecAgrOP] in the functional domain). obligatory preposing of the remnant VP, containing the finite auxiliary verb in all VO-languages, including all stages of Icelandic.

The preposing of the remnant VP will always mask the object movement, deriving VO word order only, as long as the option of extracting the embedded VP from the matrix VP has not been chosen. This hypothesis is able to account for all the various attested pure and mixed word order patterns in Icelandic, besides providing an account for the diachronic development; explaining the grammar change in question in terms of a single

72 Þorbjörg Hróarsdóttir parameter change. Given the remnant VP-hypothesis, all the various OV word order would disappear as soon as the possibility of extracting VPs out of VPs (PredPs into Spec of PredPs) was lost. In other words, the remnant finite VP will always necessarily contain all the non-finite verbs in addition to the finite verb, and the preposing of this remnant VP will always mask the object movement, deriving VO word order only. However, this proposal does not provide an explanation for why the third leftward movement process, the extraction of VPs out of VPs, was lost. The remaining question is what the trigger for VP-extraction was and why it was lost, or became inoperative, in the language. I argue that the strong feature that must have led to the additional (overt) movement of VP-extraction in OI entails a more complex representation for the child-learner, and hence, that it is marked. These conclusions are in accordance with Roberts’ (2001) claim that simplicity is a function of movement relations, since the old language here has an additional movement over the modern language. Hence, the old grammar is more complex for the child-learner than the new grammar not involving this additional movement. At a certain point in the history of Icelandic, the marked value of this parameter, that is, the strong feature leading to the VP-extraction, was no longer expressed (highly enough) in the PLD and the unmarked value became to be triggered by default.

4. Information structure and OV order in Older Icelandic It has been claimed that in various mixed word order languages such as Hungarian, Finnish, Northern Sámi, Russian and Yiddish, the existing OVand VO-orders correlate with properties such as specificity, or ‘familiarity’ of the object, or in other languages, topic-focus structure (see e.g. Holmberg 1998; Kiss 1998; Corver and Delfitto 1995). It has further been proposed that old information tends to precede new information and that light elements precede heavy elements. Hróarsdóttir (2000) studied the various OV and VO word order patterns in OI and argued that there is a parallelism between the factors that have been proposed to derive scrambling, and the factors that induce OV rather than VO word order in OI. The three most important factors are heaviness, definiteness, and information structure. The different behavior of pronouns and full nominal objects within the Germanic languages in particular has also led to a fruitful discussion in recent years, where pronouns in the West Germanic languages (including OE, Early Yiddish, and Early New High German) are assumed not to be

OV languages 73

able to move rightward; “they do not postpone beyond an otherwise final verb in the way that NPs can” (Kroch and Taylor 1994: 8). This is consistent with the prediction of ‘heaviness’; that is, pronouns, especially personal pronouns, possessive pronouns and demonstrative pronouns, have a tendency to occur higher (further to the left) in the sentence than phrases containing longer or heavier words (full DPs). Pintzuk (1996) argues that the position of complements in the history of English (except in the earliest stage) was determined by weight; the heavier the complement, the more likely it was to appear postverbally. Light constituents, particles, pronominal objects, and monosyllabic adverb could appear either before or after the non-finite main verb in OE sentences with the auxiliary verb before the main verb. In sentences with the auxiliary verb after the main verb, however, these light elements, unlike heavier constituents, appeared before the main verb (cf. Pintzuk 1996: 6). Another well-known tendency across languages is that indefinite nominal objects usually do not move to the left (scramble), they are not fronted; thus, they usually do not occur in a high position (cf. e.g. de Hoop 1992; Diesing 1992). It is generally assumed that only definite objects undergo scrambling (movement out of the VP). The division between definite and indefinite objects is consistent with the prediction that old information (definite objects) precede newer information. Indefinite NPs are strongly associated with new information; the occurrence of an indefinite NP indicates that it is introduced into the discourse for the first time. While old (given) information is recoverable from the preceding discourse, new information is unrecoverable or unpredictable from the preceding discourse. However, old information in a broader sense may also include information that has not been mentioned in the preceding discourse, but is inferable from the context or from shared knowledge. Related to the division between new and old information is focus, more exactly, focused (non-presupposed) categories can be fronted in many languages. It has, for instance, been suggested that the VP-internal position of preverbal objects (OV word order to the right of sentential adverbs) is a ‘focus position’ in various languages, such as Hungarian (cf. Horváth 1986; Farkas 1986; Kiss 1987), the Chadic languages (cf. Tuller 1992), and Basque (cf. Laka and Uriagereka 1987; Urbina 1989). Finnish and Russian are among the languages that exhibit very free word order. However, Holmberg (1998, 2000a) has shown that the permitted sequences of the subject, object and the verb in these languages are not in totally free variation, but that each variant is associated with a particular interpretation, such as topic, focus, contrast, specificity, and so forth. Thus, the object in

74 Þorbjörg Hróarsdóttir SOV constructions is (optionally) focused in Russian and Hungarian. In Finnish and Northern Sámi, on the other hand, the SOV construction has a non-focused object. As Holmberg (1998) notes, Northern Sámi appears to be ‘semi-SOV language’, with short object movement in non-finite VPs, where the movement is sensitive to focus. Let us now turn to the role of information structure as a cue for the OV/VO parameter in Older Icelandic. My basic claim is that the loss of OV word order in the grammar was caused by a prior language change that involved objects expressing new information: These objects gained in frequency in postverbal position. As mentioned, previous studies on the loss of OV word order in the history of Icelandic have focused on the abrupt disappearance of the OV orders in the beginning of the nineteenth century, that is, the grammar change. However, the cause of the grammar change must lie in some alteration to the PLD. Here, I will try to address the variation in the PLD, which led to the parameter change. I will illustrate the gradual loss of the OV word order patterns that took place in the centuries prior to the grammar change, focusing on the changes in the PLD that must have paved the way for the parameter change when OV word order was lost. I propose that the word order patterns in OI were derived by a peripheral rule that applied optionally according to prosodic output conditions at PF, thus accounting for the relatively free word order at that stage of the language. If we now focus only on the nominal objects in pre- and postverbal positions in the attested OI texts, then we see a gradual increase of VO word orders with nominal objects from the 17th century. Examples (8) through (11) show some simple cases of pre- and postverbal nominal objects in OI. (8)

Full DPs: OV word order a. efftir þad þeir høffdu eplid eted (Dín) after that they had apple.the eaten ‘after they had eaten the apple’ b. þä skilldu þeir lijffed missa (Dín) then should they live.the lose ‘Then they should die’

(9)

Full DPs: VO word order a. að hann hefði etið kjötið (Munn) that he had eaten meat.the ‘that he had eaten the meat’

OV languages 75

b.

hvört hann vilji ei kaupa þræla (Árm) whether he wanted not buy slaves ‘whether he didn’t want to buy slaves’

(10) Pronouns: OV word order a. að eg hafi hana beðið (letters) that I have her asked ‘that I have asked her’ b. at hann mundi þat eigi gera (Finn) that he would it not do ‘that he would not do that’ (11) Pronouns: VO word order a. að ég hafi aldrei borgað þér (letters) that I have never paid you ‘that I have never paid you’ b. að guð vilji brúka þig lengi (letters) that God wants use you long ‘that God wants to use you for a long time’ Consider Table 1, illustrating the total frequency for the texts from each century. Pronouns are preferred in preverbal position more often than full DPs in OI. The overall difference is not great, but it is clearest in the earliest texts. Table 1. Full DPs versus pronouns Texts 14th century 15th century 16th century 17th century 18th century 19th century

OV 75 66 69 94 25 99

VO 186 110 112 245 125 1603

Full DPs % OV 28.7% 37.5% 38.1% 27.7% 16.7% 5.8%

OV 92 54 34 162 27 258

VO 33 28 7 120 58 1050

Pronouns % OV 73.6% 75.0% 82.9% 57.4% 31.7% 19.7%

There is also a grammatical ordering with respect to old information preceding new information in the OI texts, as shown in Table 2. This means that objects previously mentioned within the five preceding sentences have a tendency to occur in preverbal position, more than objects introduced for the first time in the context.2

76 Þorbjörg Hróarsdóttir Table 2. Old versus new information Texts 14th century 15th century 16th century 17th century 18th century 19th century

OV 75 48 50 79 20 84

New information VO % OV 204 26.9% 136 26.1% 132 27.5% 314 20.1% 98 16.9% 1312 6.0%

OV 156 110 77 215 39 261

Old information VO % OV 36 81.3% 24 82.1% 17 82.0% 104 67.4% 43 47.6% 1072 20.0%

The reason for why we do not see a clear rule here, with OV structures always involving old information and VO structures always involving new information, is that the word order is also subject to a stylistic rule in terms of weight. This means that light DPs (including pronouns) are preferred in preverbal position and heavy DPs are preferred in postverbal position. Hence, a relatively heavy DP with old information would normally occur in postverbal position. The heaviness factor is illustrated in Table 3. Table 3. Number of words within full DPs

14th century 15th century 16th century 17th century 18th century 19th century

1 word % OV 72.5% 66.3% 60.0% 46.0% 43.8% 13.7%

2 words % OV 36.8% 32.7% 31.8% 23.5% 20.6% 4.6%

3+ words % OV 10.9% 10.3% 8.0% 2.8% 1.7% 1.6%

Some examples of the division of full DPs according to their heaviness are shown below. Only sentences with OV word order are exemplified. (12) DPs with one word a. Þar er kaleikur sem álfafólk hefur kirkjunni gefið (Álf) there is chalice that elves have church.the given ‘There is a chalice there, that some elves have given to the church’ b. at ek mun sæmd af þier hliota (Vikt) that I will honor of you get ‘that I will get honor from you’

OV languages 77

(13) DPs with two words a. hafer þu þinu lidi jatat þeim (Vikt) have you your assistance promised them ‘if you have promised them your assistance’ b. at hann muni eitthvert ráð til leggja (Guðm) that he would some advice to put ‘that he would offer some advice’ (14) DPs with three words a. að prestur þessi hefði sína fyrri konu misst (Álf) that priest this had his first wife lost ‘that this priest had lost his first wife’ b. því hann vildi þann saklausa mann til dauða dæma (Morð) because he wanted this innocent man to death sentence ‘because he wanted to sentence this innocent man to death’ No preverbal DPs in the corpus contained more than three words. On the other hand, postverbal DPs containing more than three words were common (163 examples found). Some of these examples are shown below. (15) DPs with more than three words: VO word order a. Ungbarna veiki hefur sópað burt miklum þorra þessa ungviðis (letters) baby sickness have swept away big majority these children ‘Diphtheria has taken most of these children’ b. eg heffe feinged eina fräbæra edur faheyrda sött (Dín) I have got one distinguished or outrageous sickness ‘I have got a distinguished or outrageous sickness’ c. Vil eg nú upphéðan heita yður minni fullri og fastri vináttu (Munn) want I now from-now-on promise you my complete and solid friendship ‘From now on I will promise you my complete and con stant friendship’ Single DPs (DPs containing only one word) occurred as frequently in preverbal position as in postverbal position in the earliest texts (dating from the fourteenth through the sixteenth centuries). In texts dating from the next two centuries (the seventeenth and the eighteenth centuries), single DPs started to occur in postverbal position more often, as illustrated in

78 Þorbjörg Hróarsdóttir Table 3 above. I take this to reflect a language change; that is, change in usage during the seventeenth and eighteenth centuries. Finally, in the nineteenth century texts and letters, preverbal single DPs have become very rare, as a consequence of an abrupt grammar change in the beginning of the nineteenth century. Single DPs thus lost their tendency to occur to the left of the main verb. DPs containing two words clearly are preferred in postverbal position, throughout the period studied. However, these DPs could also occur preverbally. DPs consisting of three or more words were very rare in preverbal position in all the texts and letters studied. It is therefore clear that light DPs more often occurred in preverbal position than heavy DPs in OI. The heaviness factor seems to have been stronger than the information factor: Light objects with old information are always preverbal and heavy objects with new information are always postverbal. Light objects with new information are also usually preverbal, while heavy objects with old information are usually postverbal. DPs with two words seem to have been more or less neutral with regard to the heaviness factor; hence, unfocused (old information) DPs with two words are usually preverbal, while focused (new information) DPs with two words are usually postverbal. Hence, pronouns and other light categories are preferred in preverbal position, while (heavy) full DPs and objects with new information are preferred in postverbal position. In other words, there is a prosodic ordering associated with syntactic positions with respect to: – – – –

full DPs are more likely to occur in VO structures than pronouns new information is more likely to occur in VO structures than old information heavy DPs are more likely to occur in VO structures than light DPs quantified and negated objects occur preverbally (OV) 3

When we take all these prosodic orderings into consideration, we can account for the OV/VO word order pattern in the OI corpus. In other words, it is more or less predictable which objects occur preverbally and which occur postverbally.

OV languages 79

5. Language change leading to grammar change As indicated in the tables in the preceding section, the OV/VO word order patterns are stable until the seventeenth century, where there is a gradual loss of the OV word order patterns for two centuries, or, in other words, a gradual increase in language use to have focused elements in postverbal (VO) position (the signaling of discourse properties). During the seventeenth and eighteenth centuries, a gradual change took place in the history of Icelandic, paving the way for the parameter change in the beginning of the nineteenth century, when we have a change from a grammar allowing the variation of both OV and VO word order patterns, to a grammar allowing only (pure) VO word order. Hence, DPs expressing new information were more often put in postverbal position, with the consequences that there was a gradual drop in the frequency of OV word order patterns until OV had dropped below a certain threshold to be useful as cues (Lightfoot 1999). In other words, this increased frequency of focused DPs in postverbal position gradually led to fewer expressions of the relevant cue. Hence, there is a(n abrupt) parameter change within the next generation acquiring the language (in the beginning of the nineteenth century), where this new generation has a new parameter setting that does not allow the variation of both OV and VO word order patterns, and instead only has the pure VO word order setting. In sum, this means that a shift in discourse property, or simply a change in language use, can lead to a grammar change; a change in the basic word order in the I-language. It is often claimed that this is connected with the placement of heavy (old information is light, new information is heavier) and stressed material (an element that is stressed is part of the focus of the clause) (cf. e.g. Hinterhölzl 2006). Of course, it is still unclear what brought about this change in language use. Maybe there is no explanation in structural terms, but only in sociolinguistic terms: a certain type of word order became unfashionable. The eighteenth century marked the most tragic age in Iceland’s history. In 1703, when the first complete census was taken, the population was approximately 50,000. From 1707–1709 the population sank to about 35,000 because of a devastating smallpox epidemic. Twice again the population declined catastropically, both during the years 1752–1757 and 1783–1785, owing to a series of famines and natural disasters. The summer of 1782 marked the beginning of the presumably most tragic period ever, the so-called Móðuharðindi, where a tremendous natural catastrophe took place with a great volcanic eruption that led to increased cold weather and hard times, including starvation and different diseases. It is usually assumed that 1/5 of the

80 Þorbjörg Hróarsdóttir population, or 10,000 people, died because of the Móðuharðindi. When the population was severely decimated by these epidemics, this placed the surviving population in a new situation. First, the age distribution changed, and the average life expectancy decreased. Second, the language would be more sensitive to other sources of noise, for instance, language contact. Historical/social changes in the form of increased intervention, administration, and the power of the Danish led to an increase in foreign influence in Iceland at a time when the stability of the language was less secure, due to the fact that quite a lot of the older bearers of the language had perished. Note that we expect the child-learner to be sensitive to changes in the surrounding language environment, since that language environment is what she uses as a model. The language environment providing the PLD had changed so that there was more insecurity about the correct word order patterns, due to an increased influence from Danish and a loss of older individuals, followed by an increase in the proportion of children. It is obvious that we can expect the language to change when the transmission of the tradition is disturbed. When a severe plague hits a population causing a sudden death of older speakers, much of the conservatism of the language (use) may be lost. If we take VO word order to be the more innovative structure, it is not surprising that the frequency of the use of the old OV structure drops when the language environment is affected in this way. We can assume that the devastating smallpox epidemics in 1707–1709 and other natural disasters in Iceland throughout the eighteenth century are a plausible example of the kind of population changes that can cause a language change, which may then lead to a parameter change; the loss of OV word order patterns. Hence, the case we are studying here is a language that had a variation (OV and VO word order) before the epidemic, and lost it after the epidemical years during the eighteenth century. As can be seen in the tables in section 4, the proportion of OV word order had remained surprisingly stable from the fourteenth to the seventeenth century. The first notable drop in frequency takes place during the eighteenth century, during the epidemical period. This reflects a language change, where constructions with OV word order came to be used less and less, until the frequency of OV had dropped below a certain threshold to be useful as cues (Lightfoot 1999). This in turn lead to a parameter change in the beginning of the nineteenth century where OV orders suddenly disappeared. To summarize, after the smallpox epidemics in the very beginning of the eighteenth century and the natural catastrophe Móðuharðindin in the middle of the century, the age distribution became different due to a loss of older individuals, followed by an increase in the proportion of children.

OV languages 81

Hence, the language environment had changed and we assume this to have led to more insecurity about the correct word order patterns. At the same time, there was an increase of foreign influence from Danish, a language that already had pure VO word order at the time. Finally, another social/historical change in Icelandic at the time of the language change supports our findings. Note that the frequency of OV/VO word order patterns showed a remarkable stability from the earliest texts until the eighteenth century. More interestingly, so did the Icelandic society: From the time of the settlement until the eighteenth century, it was a rather stable (old-fashioned) country of farmers, without any important formation of cities or towns. It was a society of big families living together. The elderly were not sent away to old people’s homes and the children were not sent away to nursery schools or kindergartens. The big family stayed together, and worked together, on the farm. Older children took care of their younger siblings, and most noticeable perhaps, the children grew up with their parents and their grandparents and their language (including language use and tradition). It was not until the eighteenth century that we have the formation of big towns. During 1760–1770, a great disease hit the Icelandic livestock and as much as half of the sheep were killed. As a consequence, people had to flee the countryside and move to the seaside to survive. The fishing industry increased and fishing towns thrived. Reykjavík gradually became the capital of Iceland. This also meant that the big families were split up and children went to school with other children and hence their language acquisition was now, to a much greater extent than before, affected by the language of other children. This is another example of a change in language environment that has had consequences for language use, leading to a parameter change. Together these historical/social changes during the eighteenth century in Iceland explain why the language change took place – there was a change in the language environment, affecting the use of the language – paving the way for the grammar change. There is also some direct evidence that OV and VO word order might have correlated with style; the frequency of OV word order can sometimes be associated with a more elevated or formal style (and formal style is typically more conservative). For instance, not only personal letters were studied dating from the nineteenth century, but also letters of a more formal style, written by Icelandic priests at a request from a special committee for archaeology in Copenhagen. When the proportions of OV orders in these two types of letters are compared, it appears that on average, the rate of OV word order is somewhat higher in the more formal letters than in the personal letters.

82 Þorbjörg Hróarsdóttir 6. Object shift I will conclude by illustrating the difference between OV word order and Modern Icelandic Object Shift. Note that although Modern Icelandic is generally assumed to have SVO surface word order, as already mentioned, there still exist three constructions in Modern Icelandic exhibiting some sort of SOV word order: Negative (and quantified) phrase construction, Object Shift, and Stylistic Fronting. We will leave Stylistic Fronting aside here, and take a look at a correlation between OV word order patterns in OI and Object Shift in the modern language. 4 Consider example (16). (16) Object Shift a. María las ekki bókina Mary read not book.the ‘Mary did not read the book’ b. María las bókina ekki Mary read book.the not ‘Mary did not read the book’ c. *María las ekki hana Mary read not it ‘Mary did not read it’ d. María las hana ekki Mary read it not ‘Mary did not read it’ Object Shift in Icelandic is generally assumed to involve a leftward movement of definite objects and pronouns across a sentence adverb, in clauses containing no non-finite verbs, that is, in sentences where verb movement has occurred (cf. Holmberg’s Generalization 1986). Thus, the employment of Object Shift does not lead to an OV word order, in a strict sense, but a word order where the object has moved out of the VP, since it shows up to the left of the sentence adverb, marking the left periphery of the VP. An important element in connection to OV orders in OI is the analysis of these possible OV orders still found in Modern Icelandic. This has been the subject of much recent work, and it would indeed be interesting to see what the empirical and analytical differences between the earlier stages and the present-day stage are. In particular, it is interesting to compare obligatory Object Shift of pronouns and optional Object Shift of definite DPs in Modern Icelandic. For instance, how did whatever changes that have taken place give rise to the differences between pronouns and non-pronouns that

OV languages 83

we observe today? It is very likely that English has undergone a parallel development; Late Middle English probably had optional Object Shift for non-pronouns and obligatory Object Shift for pronouns, whereas 16th century English was like the Modern Mainland Scandinavian languages; pronouns moved whenever the verb moved but no other complements moved (see Roberts 1995). Bobaljik and Tráinsson (1998) discuss the connection between Object Shift in Modern Icelandic, the remains of OV orders still existing, and information structure, giving independent support to our claim that the cue for OV order in OI indeed was expressed through information structure. They give the following example: (17) a. b.

Ég las þrjár bækur ekki I read three books not Ég las ekki þrjár bækur I read not three books ‘I didn’t read three books’

They claim that the alternation between (a) and (b) here depends on information structure, with a contrast between specific and non-specific, or given versus new information. Hence, in order to be able to shift, the DP object must represent old information, while a DP object representing new information cannot shift and must remain internal to the VP. They conclude that AgrOP has thus nothing to do with case theory, but instead with information structure, the mapping from syntax to semantics. Although the various attested OV orders in OI might at first sight seem to apply to all kinds of nominal objects (both pronouns and full DPs, definite objects as well as indefinite) there was a division or restriction, similar to the one of Object Shift in Modern Icelandic, which only applies to pronominal objects and definite DPs. It is interesting to note that in the OI data, where a sentential adverb intervenes between the preverbal complement and the non-finite main verb, there are no instances of indefinite objects. That is, the pattern [DPindefinite - sentence adverb - Vmain] seems to have been absent. Pronouns also had a much stronger tendency to occur to the left of an adverb than full DPs or other types of arguments, as illustrated in Table 4.

84 Þorbjörg Hróarsdóttir Table 4. Objects occurring higher than sentence adverbs

All centuries

Pronouns 27

DPs 5

PPs 3

The distinction is very striking: 27 pronouns occurred to the left of the sentence adverb, only 5 full DPs, all definite, and three arguments of another type (PP or a small clause predicate). Some of the OI sentences are shown below. (18) Pronoun to the left of a sentence adverb a. og vér fáum það ei þekkt (Álf) and we get it not recognized ‘And we cannot recognize it’ b. at hann mundi þat eigi gera (Finn) that he would it not do ‘that he wouldn’t do it’ c. að þeir hefðu þau aldrei út gefið (Morð) that they had them never out given ‘that they had never published them’ (19) Full DP to the left of a sentence adverb a. þui ath hann willdi leoninum giarna ekki grand giora (Sig) because that he wanted lion.the readily not damage do ‘because he didn’t want to harm the lion’ b. Enn hier med a ofan hef eg greind Kugillde alldrei teked (Afs) but here with in addition have I mentioned cow-values never taken ‘But, in addition, I have never taken these cows’ c. hefi eg þessa kapla aldri aptur fengið (Skál) have I these cables never back got ‘I haven’t got these cables back’ (20) Complemental adverbial/PP to the left of a sentence adverb a. þegar hann hefur þar ei lengi verið (Álf) when he has there not long been ‘when he hasn’t stayed there for a long time’

OV languages 85

b.

c.

því þeir hafa hér aldrei fyrri komið (Árm) because they have here never before come ‘because they have never been here before’ hefr síðan um þann Haug ekki grenslast verid (letters) has since about that grave-mound not inquired been ‘Since then, no one has inquired about that grave mound’

Furthermore, at most one argument could precede the sentence adverb at a time. In VO word order, on the other hand, two or three arguments frequently followed the sentence adverb. The existence of complements to the left of a sentence adverb, thus, gives strong support to leftward movements of the objects. In the patterns [object – Vaux – Vmain] and [object – Vmain – Vaux], the preverbal argument could either be a DP, a PP or a small clause predicate. However, only the DP occurred with any significant frequency. In almost all the examples found, the DP here is a pronoun. Two examples are shown in (21) and (21). (21) [DP – Vaux – Vmain] ef hann hefði þat viljað fága (Guðm) if he had it wanted clean ‘if he had wanted to clean it’ (22) [DP – Vmain – Vaux] að eg mundi hann sigrað geta (Guðm) that I would him defeat could ‘that I would be able to defeat him’ The pattern [Vmain – Vaux – object], on the other hand, occurred with full DPs only, but never pronouns. The sentences in (23) show some examples of this pattern. (23) [Vmain – Vaux – DP] a. at hann mun raða vilia ferðum sínum (Finn) that he will determine want journeys his ‘that he will want to determine his own journeys’ b. þu munt haft hafa harða landtoku (Finn) you will had have hard landing ‘You have had a rough landing’

86 Þorbjörg Hróarsdóttir Note also that the construction [Vmain – Vaux – object] is comparable to Object Shift in the modern language; when the main verb moves, the object moves as well, obligatorily if it is an unstressed pronoun. Consequently, we do indeed not expect to find a pronoun in the postverbal position after the main verb has moved in OI. Definite DPs and pronouns do not exit the VP, so that they move along within the preposed VP (containing the finite verb). Indefinite DPs (and CPs), on the other hand, would always have to move out of the VP, to a separate Focus-position, and therefore are stranded by the remnant VP-preposing. This correctly predicts the distinction between [Vmain – Vaux – DP] and *[Vmain – Vaux – pronoun], together with the fact that the object in the word order patterns [object – Vmain – Vaux] and [object – Vaux – Vmain] is usually a personal pronoun or a definite noun. Moreover, heavy XP-shift in general could be seen as an independent dimension, in line with Kayne’s (1998) proposal for Modern English. Here, we arrive at the analytical difference between OV word order and Modern Icelandic Object Shift: Object Shift can now be regarded as a movement of the object along with the VP, in front of the negative adverbial. While full DPs must move out of the VP (to get their case checked in [Spec, AgrOP] in the functional domain), unstressed pronouns do not have to exit the VP (because they do not have strong enough features).5 In terms of focus/heaviness, we claim that focused/heavy objects must move out of the VP, prior to the remnant VP-movement, and hence end up postverbally, while light objects do not have to exit the VP, and therefore they may move along within the preposed VP (containing the finite verb) and thereby end up preverbally.

7. Conclusions In this chapter, we have studied the change from object-verb order to verbobject order. A number of European languages have undergone such a change and we have focused here on the change in Icelandic, arguing that while the structural change was very similar in English and Icelandic, it took place at different times and different ways in the two languages, triggered by different E-language changes. Morphological triggers, or the absence thereof, have often been related to clustering of syntactic properties in both diachronic change and acquisition. As seen from the English viewpoint, low-level facts of inflectional morphology may express the relevant cue for parameters, and hence the loss of inflection may lead to a grammar change. I have shown that this analysis does not carry over to Icelandic, as

OV languages 87

the loss of OV in Icelandic took place despite rich case morphology. Instead, I have argued that the relevant cue may be expressed differently among languages: While it may have been expressed through morphology in English, it was expressed through information structure in Icelandic. In both cases, external effects led to fewer expressions of the relevant (universal) cue and a grammar change took place. I have claimed that the language change in OI reflects an increased frequency of focused elements in postverbal position – in the form of increased number of objects moving out of the VP (to an external Focusposition) prior to the remnant VP-movement – which again increases the overall number of VO word order – with the consequences that there was a gradual drop in the frequency of OV word order during the eighteenth century until the frequency of OV had dropped below a certain threshold to be useful as cues, causing a parameter change in the beginning of the nineteenth century, where the VP-out-of-VP option disappeared. Finally, I have shown how Object Shift in Modern Icelandic supports the claim that the cue for OV in OI was expressed through information structure.

Notes 1.

2. 3.

4.

5.

Hróarsdóttir (2000) studied the frequency of OV and VO patterns in various texts dating from the fourteenth to the nineteenth centuries, in addition to personal letters dating from throughout the nineteenth century. Here, I have built on this database, making it approximately double in size from that used earlier, based on a wider extraction of the same texts. Objects with old information here mean personal pronouns, and full DPs already mentioned in the preceding sentences. Quantified and negated objects always occurred preverbally in OI, and they still do in the modern language. These are the remaining OV orders in Modern Icelandic. Stylistic Fronting rarely leads to OV orders, but often to an order where a complement, or part of a complement, precedes the finite verb (see Maling 1990 for the first explicit discussion of Stylistic Fronting in Icelandic, together with the discussion in Jónsson 1991, 1996; Holmberg 1997, 2000b, and Hrafnbjargarson 2004). This proposal goes against Cardinaletti and Starke (1999).

88 Þorbjörg Hróarsdóttir References Bobaljik, Jonathan D. and Höskuldur Thráinsson 1998 Two heads aren’t always better than one. Syntax 1: 37–71. Cardinaletti, Anna and Michal Starke 1999 The typology of structural deficiency: A case study of the three classes of pronouns. In Clitics in the Languages of Europe, ed. Henk van Riemsdijk, 145–233. Berlin: Mouton de Gruyter. Corver, Norbert and Denis Delfitto 1995 Feature primitives and the syntax of specificity. GLOW Newsletter 34, spring 1995. Diesing, Molly 1992 Indefinites. Cambridge, Mass: MIT Press. Farkas, Donka 1986 On the syntactic position of focus in Hungarian. Natural Language and Linguistic Theory 4 (1): 77–96. Haider, Hubert 1992 Branching and discharge. Arbeitspapiere des Sonderforschungsbereichs 340. Vol. 23. University Stuttgart. Hinterhölzl, Roland 2006 Scrambling, Remnant Movement, and Restructuring in West Germanic. New York: Oxford University Press. Holmberg, Anders 1986 Word Order and Syntactic Feature in the Scandinavian Languages and English. Doctoral dissertation, University of Stockholm. 1997 Scandinavian Stylistic Fronting: Movement of phonological features in the syntax. Working Papers in Scandinavian Syntax 60: 81–124. 1998 Word order variation in some European SVO languages: A parametric approach. In Constituent Order in the Languages of Europe, ed. Anna Siewierska, 553–598. Berlin: Mouton de Gruyter. 2000a OV in Finnish. In The Derivation of VO and OV, ed. Peter Svenonius, 123–152. Amsterdam: John Benjamins. 2000b Scandinavian Stylistic Fronting: How any category can become an expletive. Linguistic Inquiry 31 (3): 445–483. Holmberg, Anders and Christer Platzack. 1995 The Role of Inflection in Scandinavian Syntax. New York/Oxford: Oxford University Press. Hoop, Helen de 1992 Case Configuration and Noun Phrase Interpretation. Doctoral dissertation, University of Groningen. Horváth, Julia 1986 Focus in the Theory of Grammar and the Syntax of Hungarian. Dordrecht: Foris.

OV languages 89 Hrafnbjargarson, Gunnar Hrafn 2004 Stylistic fronting. Studia Linguistica 58 (2): 88–134. Hróarsdóttir, Þorbjörg 1996 The decline of OV word order in the Icelandic VP: A diachronic study. Working Papers in Scandinavian Syntax 57: 92–141 2000 Word Order Change in Icelandic. From OV to VO. (Linguistik Aktuell/Linguistics Today 35). Amsterdam: John Benjamins. Indriðason, Þorsteinn 1987 Skýrsla um orðaröð í sagnlið. Ms., University of Iceland, Reykjavík. Jónsson, Jóhannes Gísli 1991 Stylistic Fronting in Icelandic. Working Papers in Scandinavian Syntax 48: 1–43. 1996 Clausal Architecture and Case in Icelandic. Doctoral dissertation, University of Massachusetts, Amherst. Kayne, Richard S. 1998 Overt vs. covert movement. Syntax 1: 128–191. Kiss, Katalin É. 1987 Configurationality in Hungarian. Dordrecht: Kluwer. 1998 Discourse-configurationality in the languages of Europe. In Constituent Order in the Languages of Europe, ed. Anna Siewierska, 681– 727. Berlin: Mouton de Gruyter. Koeneman, Olaf 2000 The Flexible Nature of Verb Movement. Doctoral dissertation, University of Utrecht. Kosmeijer, Wim 1986 The status of the finite inflection in Icelandic and Swedish. Working Papers in Scandinavian Syntax 26. Kroch, Anthony S. and Ann Taylor 1994 Remarks on the XV/VX alternation in Early Middle English. Ms., University of Pennsylvania, Philadelphia. Laka, Itziar and Juan Uriagereka 1987 Barriers for Basque and vice-versa. NELS 17. Amherst: University of Massachusetts. Lightfoot, David W. 1999 The Development of Language: Acquisition, Change, and Evolution. Oxford and Malden, Mass: Blackwell Publishers. Maling, Joan 1990 Inversion in embedded clauses in Modern Icelandic. In Modern Icelandic Syntax, eds. Joan Maling and Annie Zaenen, 71–91. [Syntax and Semantics 24]. San Diego: Academic Press.

90 Þorbjörg Hróarsdóttir Pintzuk, Susan 1996 From OV to VO in the history of English. Ms., University of York. [Paper presented at the 4th Diachronic Generative Syntax Conference, October 31-November 2, 1995. Université du Québec à Montréal] Roberts, Ian G. 1993 Verbs and Diachronic Syntax. Dordrecht: Kluwer. 1995 Object movement and verb movement in Early Modern English. In Studies in Comparative Germanic Syntax, eds. Hubert Haider, Susan Olsen and Sten Vikner, 269–284. Dordrecht: Kluwer. 1997 Directionality and word order change in the history of English. In Parameters of Morphosyntactic Change, eds. Ans van Kemenade and Nigel Vincent, 396–426. Cambridge: Cambridge University Press. 2001 Language change and learnability. In Language Acquisition and Learnability, ed. Stefano Bertolo, 81–125. Cambridge: Cambridge University Press. Rohrbacher, Bernhard W. 1999 Morphology-Driven Syntax: A theory of V to I raising and pro-drop. Amsterdam: John Benjamins. Rögnvaldsson, Eiríkur 1996 Word order variation in the VP in Old Icelandic. Working Papers in Scandinavian Syntax 58: 55–86. Siewierska, Anna 1998 Variation in major constituent order: A global and a European perspective. In Constituent Order in the Languages of Europe, ed. Anna Siewierska, 475–551. Berlin: Mouton de Gruyter. Sigurðsson, Halldór Á. 1988 From OV to VO: Evidence from Old Icelandic. Working Papers in Scandinavian Syntax 34: 1–41. Tuller, Laurice 1992 The syntax of postverbal focus constructions in Chadic. Natural Language and Linguistic Theory 10 (2): 303–334. Urbina, Jon Ortiz de 1989 Parameters in the Grammar of Basque. Dordrecht: Foris. Vikner, Sten 1997 V-to-I movement and inflection for person in all tenses. In The New Comparative Syntax, ed. Liliane Hageman, 189–213. London: Longman.

Discourse relations and word order change Ans van Kemenade

1.

Introduction

I explore, for Old English, against the broader backdrop of Old (West) Germanic, the relation between discourse, clause structure, and inflectional morphology.1 I will address a word order feature that has proved a knotty problem in the syntactic literature on older English, and propose an alternative approach that attempts to integrate the insights from syntactic approaches with those emerging from the literature on discourse. In section 2, I will first position the Old English dataset that prompted the approach developed in this article, and outline the issues it raises for syntactic analyses that have been proposed so far. Then, in section 3, I will present the basic tenets of an approach that attempts to unify insights from discourse analysis and information packaging with a syntactic approach. I start from the assumption that there is such a thing as a tightly circumscribed syntax, which makes available a limited set of options which define the limits to the space that language users have to manipulate word order according to information-structural needs. Specific claims will be: 1. Old English has a class of short adverbs which I will call discourse partitioners as they serve to partition the clause into domains for given information (referring back to a prominent discourse referent) and non-given information. Their use as discourse partitioners was later lost; 2. Old English has a type of referring expression lost in Middle English, the paradigm of simple demonstratives, which allow specific anaphoric reference to a discourse antecedent; this ensures that, in addition to the well-known range of pronouns, DP subjects with a demonstrative pronoun allow reference to a discourse antecedent and can be marked and positioned as given information. 3. the interplay of the first two strategies serves to allow a number of word order options in the higher part of the clause that were later lost. In section 4, I will discuss a number of issues concerning the modelling of these findings.

92

Ans van Kemenade

2.

An analytical problem in Old English

Before going on to discuss an analysis of Old English word order in discourse terms, let’s briefly position the dataset as it has been analysed in the literature so far. Its core consists of arguments that occur in a relatively high position in the clause, and includes personal pronouns (subject or object), but, rather problematically for a purely syntactic account, various other types of NP as well. A well-known feature of Old English word order is that personal pronouns may occur in positions higher than their DP counterparts. This is true almost categorically for subject pronouns, and optionally for object pronouns, as exemplified (1) for main clauses and by (2) for sub-clauses2: (1)

a. Be ðæm we magon suiðe swutule oncnawan ðæt ... By that, we may very clearly perceive that ... CP 26.181.16 ‘By that, we may perceive very clearly that ...’ b. þa axodon hine Pharisei & þa boceras hwi ne gað and then asked him Pharisees and the learned men, why not go þine leorningcnihtas æfter ure yldrena gesetnysse. ac your disciples after our forefathers' law, but besmitenum handum hyra hlaf þicgað with defiled hands their bread eat?

Mk (WSCp) 7.5

‘Then the Pharisees and scribes asked him, why walk not thy disciples according to the tradition of their elders, but eat bread with unwashed hands?’ c. Hwæt tacnað us ðonne Saul buton yfle hlafurdas? What betokens us then Saul except evil lords? CP 28.197.22 ‘What is signified to us by Saul but bad masters?’ (2)

a. Gif hire ðonne se wiðsace, ðonne is cynn ðæt him spiwe if it then he refuse, then is proper that him spits ðæt wif on ðæt nebb the woman in the face

CP 5.45.2

‘But if he refuse it, it is proper for the woman to spit in his face’

Discourse relations and word order change

93

b. gif hine ðonne ðæt fleah mid ealle ofergæð, ðonne ne mæg he if him then the albugo with all covers, then not-can he noht geseon. naught see

CP 11.69.18

‘if it [the pupil of the eye] is entirely covered with albugo, he cannot see anything.’ This fact has been taken to indicate that personal pronouns are syntactically special and exhibit a form of syntactic cliticization (van Kemenade 1987; Pintzuk 1991). Alternative analyses are those in terms of Germanic style weak pronouns in the sense of Cardinaletti and Starke (1996). Cardinaletti and Roberts (2002). The positional evidence was made more precise in Haeberli (1999); van Kemenade (1999; 2000), which is illustrated by the examples (3), reflecting a word order template as in (4): (3)

a. Ne het he us na leornian heofonas to wyrcenne not ordered he us not learn heavens to make ‘He did not bid us learn to make the heavens’

ÆLS.XVI.127

b. Ne sæde na ure Drihten þæt he mid cynehelme oððe mid not said not our Lord that he with diadem or with purple purpuran gescryd, cumin wolde to us clothed, come wanted to us

ÆLS.XXXI.762

‘Our Lord said not that He would come to us with a diadem or clothed with purple’ (4)

personal pronoun – secondary negator – nominal subject

Very similar syntactic analyses of this word order template are Haeberli (1999) and van Kemenade (1999; 2000), following up Bobaljik and Jonas (1996); Bobaljik and Thrainsson (1998) whose structures I adapt for the time being as (5)3: (5)

[CP [XP] C [AgrP PronounSubj/Obj Agr [NegP Neg Adv Neg [TP DPSubj T ... ]]]

Personal pronouns occur in a designated pronoun position on the left of a NegP in Spec,AgrP position, while the position for DP subjects is lower than NegP, in SpecTP. It turns out then, that the secondary negator has

94

Ans van Kemenade

considerable diagnostic value in pinpointing the position of elements in the higher part of the clause. This is (as) true for a number of other adverbs as well, and is illustrated here for ðonne: (6)

a. Hu mæg he ðonne ðæt lof & ðone gilp fleon how may he then praise and vainglory avoid ‘How can he avoid praise and vainglory...?’

CP. 9.57.18

b. Hu gerades mæg ðonne se biscep brucan ðære hirdelican are. how properly may then the bishop enjoy the pastoral dignity CP. 18.133.3 ‘How, then, can the bishop properly enjoy the pastoral dignity?’ By way of example, findings for root clause questions in Cura Pastoralis, based on a lexical search of the whole text, are as follows: Table 1. Relative position of adverb ðonne and various types of subject in Cura Pastoralis

Nominal subject subject left of ðonne subject right of ðonne

0 17

Personal pronoun subject or object 10 0

This yields a clear picture for main clauses. However, the position is considerably more complicated when we consider sub-clauses. (see also Haeberli and Ingham 2007 for early Middle English4). Examination of all the subclauses containing þa/þonne shows the distribution for subjects in table 2. Throughout the article, the data are based on extensive searches in the York Corpus of Old English (Taylor et al. 2003) and the Penn Helsinki parsed Corpus of Middle English 2 (Taylor 2000). Table 2. Relative position of adverb ðonne and various types of subject in YCOE

Old English subclauses Left of þa/þonne Right of þa/þonne

Pronominal subjects 1250 5

DP subjects 293 419

Table 2 shows that subject pronouns almost categorically occur on the left of the adverb. An example of this, also including object pronouns, is (7).

Discourse relations and word order change

(7)

95

on Salomonnes bocum, hit is awrieten ðæt mon ne scyle in Solomon’s books, it is written that that one not shall cweðan to his frind: Ga, cum to morgen, ðonne selle ic ðe say to one’s friend: “Go, come tomorrow, then give I you hwæthwugu, gif he hit him ðonne sellan mæge something, if he it him then give may.”

CP.44.323.24

‘in the books of Solomon, it is written that we are not to say to our friend: “Go, and come tomorrow, then I will give you something,” if we can give it him then.’ However, the figures for DP subjects in table 2 show that we are not dealing merely with a designated position for personal pronouns. The relatively high frequency of DP subjects left of the adverb renders this particularly problematic, and there is no readily discernible generalization with respect to the choice of DP subject on the left or right of the adverb; higher and lower DP subjects, for instance, include definite DP’s, as the following representative examples show: (8)

Forðæm bið se sige micle mara ðe man mid geðylde gewinð, therefore is the victory much greater which one with patience wins, forðæm sio gesceadwisnes ðonne hæfð ofercumen ðæt mod & because this wisdom then has overcome the mind and gewielð, swelce he self hæbbe hiene selfne gewildne, & sio geðyld subdued, as if he self have himself conquered, and the patience hæbbe ðæt mod geðreatod & gecafstrod. have the mind intimidated and curbed.

CP[Cotton]:218.19.42

‘Therefore the victory which is won with patience is much greater, because this wisdom (patience) has overcome and subdued the mind, as if he himself had conquered himself, and patience had intimidated and curbed the mind.’ (9)

Gif ðonne se sacerd bið ungerad ðæs lareowdomes, hwæt forstent If then the teacher is unskilled in instruction, what avails ðonne his gehlyd? then his cry?

CP.15.91.25

‘then if the teacher is unskilled in instruction, what avails his cry?’

96

Ans van Kemenade

In the next section, I will pursue a different perspective on this problem, incorporating insights from both discourse and syntax.

3.

Information packaging strategies

In this section, I will consider issues pertaining to the relation between discourse and syntax in Old and Middle English. The claim is that, to varying degrees, Old and Middle English syntax allowed the language user greater flexibility to incorporate information structural considerations by allowing a wider range of possibilities for positioning given information before new information. In contrast to present-day English, in which the subject typically marks given information (e.g. Birner and Ward 1998), earlier English differentiates between two positions which may both host various types of subject, or personal pronoun objects, the differentiation being keyed to the distinction ‘given’ vs. ‘new’. Here, I follow up previous work, in particular van Kemenade and Los (2006); van Kemenade and Miliüev (2005). When looking at the literature on information packaging, we usually come across at least five major types of strategies to mark information structure: (10) a. b. c. d. e.

Intonation Ellipsis Particles Types of referring expressions Word order

I will defer from discussing the first two strategies: intonation is a phenomenon which is particularly hard to track in any detail in historical text material; the state of our knowledge of ellipsis in the history of English is as yet limited, and instructive exploration is beyond the scope of this article. In this section, I discuss the interplay of strategies in Old English that mark its more prominent discourse character relative to Middle English and especially Modern English. More particularly, I claim that the Old Germanic languages show evidence for a clausal organization that is defined more clearly in terms of discourse domains. In particular, I claim that there is an area relatively high in the clause whose right edge is marked by a a short adverb that functions as a discourse partitioner. The evidence discussed here is from two such adverbs, þa and þonne, as representatives of a larger set. This area hosts specific types of referring expressions, viz.

Discourse relations and word order change

97

various types of pronouns and definite DP’s containing a demonstrative pronoun. The function of these referring expressions in this area is to refer back to a prominent discourse antecedent. Thus, the conclusion will be that Old English possesses the minimal morphology necessary to partition the clause in terms of an area for given/presupposed information (referring back to a (prominent) discourse antecedent) and non-given information. In the transition to Middle English, this morphology is lost: the paradigm for demonstrative pronouns is levelled to one definite determiner the; discourse partitioners (or perhaps rather the discourse partitioner use of a set of adverbs) are lost, and this leads to a clausal organization that is more rigidly syntactically defined.

3.1. Particles It is well-known that languages whose syntactic organization is more tailored to discourse needs, make use, in varying degrees, of particles that mark discourse-relevant domains in word order. Thus, for instance, Aboh (2004, 2006) shows that the word order of Gungbe, a Kwa language, is determined by Topic and Focus. Topics occur on the immediate left of a topic particle yà as in (11a), focussed NPs on the immediate left of a focus particle wȑ, as in (11b): (11) a. Ùn sè ǚƥ dàn lƥ yà Kòfí hù ì 1sg hear that snake Det Top Kofi kill 3sg ‘I heard that, as for the snake, Kofi killed it’ b. Ùn sè ǚƥ dàn lƥ wȑ Kòfí hù 1sg hear that snake Det Foc Kofi kill ‘I heard that Kofi killed THE SNAKE’ Such particles serve as discourse partitioners then, and require that the relevant NP is on their immediate left. Particles do not need to be designated elements; they can also be adverbs, for instance. Adverbs are routinely used in syntactic work as word order diagnostics, even more particularly in the literature on the Germanic languages. We saw in the previous section that secondary negators (negative adverbs) have strong diagnostic value, and the same goes for adverbs like þa and þonne ‘then’. A similar diagnostic value is ascribed to the interjection la ‘lo’ in Kato (1995). In a very different vein, Brinton (2006) claims that the temporal adverb þa ‘then’ is

98

Ans van Kemenade

a discourse partitioner. The claim made here follows up van Kemenade and Los (2006) and van Kemenade and Miliüev (2005) and goes a step further: Old English þa and þonne, both literally meaning ‘then’, are prominent members of a set of adverbs and particles that take discourse-linked material on their left. The term ‘discourse-linked’ here corresponds with what is known in the literature as given information. This set of adverbs and particles forms no obvious natural class, as it includes þa and þonne both literally meaning ‘then’; and nu (literally ‘now’); Furthermore, it includes na, whose primary use is that of negator; la ‘lo’, which is an exclamative element; and eac ’also’, which is rather hard to classify. What unites these elements, beside their fixed position, is that they are often used as rhetorical devices, serving to admonish or exhort, and, as claimed, to separate given from non-given material in the clause.5 This is particularly clearly the case when they are used in clause-internal position, as discussed in van Kemenade and Los (2006). I illustrate briefly here with the following examples: (12) a. Hwæt ða la ongunnon þa godes cempan hnexian and what then lo began then God’s champions yield and heora mod awendon to hyre maga sarnysse their mood turn to their kinsmen’s anguish.

ÆLS(Sebastian) 48

‘Well, then, behold! God’s champions began to yield, and to turn their thought on their kinsmen’s anguish.’ b. Hwæt þa se biscop hine bliþelice gefullode, and ealle his What then the bishop him blithely baptized, and all his hiwan on þæs Hælendes naman. household in the Saviour’s name

ÆLS(Basil) 167

‘and indeed the bishop blithely baptized him and all his household in the Saviour’s name.’ In (12), the exclamatory part of the clause in each case primarily serves to draw particular attention to the event following. The temporal effect of þa is minimal at best. This particular use of þa illustrates very well the function of adverbs/particles as viewed here: the focus part of the clause is “introduced” by the particle in the sense that it immediately precedes the focus part. The discussion here is restricted to þa and þonne, as they are the most consistent members of this set.

Discourse relations and word order change

99

3.2. Types of referring expressions A crucial ingredient of the account of Old and Middle English morphology in relation to discourse and syntax here is that Old English had a specific class of discourse-linking expressions that was lost in the later history of English: the weak demonstrative pronouns. This class of pronouns seems to have been one of the properties of Old English that sustained its discourse-configurational character. Before I go on to discuss their properties, let me list the paradigm, found in any grammar of Old English. Table 3. Old English demonstrative pronouns

Nom Acc Gen Dat

singular masculine se þone þæs þĺm

plural feminine sƝo þƗ þĺre þĺre

neuter þæt þæt þæs þĺm

þƗ þƗ þƗra þĺm

The demonstrative pronouns are often called definite determiners, as one of their primary known uses in Old English is to mark definiteness in a DP, thus se monn ‘the man’; þĺre fæmnan fæder; ‘the woman’s father’ and so on, see recently Denison (2006). However, the term ‘determiner’ is, I claim, a singular misnomer, in particular because it is often used when comparing the Old English demonstratives to the present-day English definite determiners, which are clearly morphologically invariant and cannot be used independently. In comparison, Old English demonstratives are a rather versatile class. While they mark definiteness like the present-day English determiners do, they can also be used as independent pronouns: (13) a. Swa fela swa hine underfengon, þam he forgeaf anweald So many so him received, them he gave authority Godes bearn to beonne God’s child to be ÆHom 1.42.26 ‘To all those who received him, he gave authority to be God’s children’ b. Be þam awrat Moyses se mæra heretoga, In principio fecit By them wrote Moses the great general, In principio fecit

100

Ans van Kemenade

Deus celum et terram Deus celum et terram

ÆHom I, 70, 46

‘About those, Moyses the great general wrote: In principio fecit Deus celum et terram’ This in itself shows that this set of pronouns has independent pronominal function, and serves to mark a type of pronominal reference. This is reinforced by the fact that they are also used as relative pronouns, which are anaphoric elements par excellence, as they must refer back to the antecedent of the relative clause they introduce, under identity of features for number and gender, and, optionally, of case as in (14b). (14) a. Ure Drihten arærde anes ealdormonnes dohtor, seo ðe læg dead Our Lord raised an alderman’s daughter (A), who (N) that lay dead ÆHom, VI, 176 ‘Our Lord brought to life an alderman’s daughter who lay dead’ b. Ic wat witodlice ðæt ge secað ðone hælend ðone ðe on I know truly that you eek the Lord (A) whom (A) that on rode ahangen wæs cross hung was

Mt. 1766

‘I know truly that you seek the Lord, who was hung on the cross’ My claim here is that demonstrative pronouns mark a type of (pronominal) reference that is both definite and specific. I would argue that the paradigm of weak demonstrative pronouns, which carried inflection for case, number and gender and were used as independent demonstratives, relative pronouns and definite “determiners”, provided a versatile extra class of referential expressions that played an important role in discourse reference. The fact that a demonstrative can also be used as definiteness marker on NP, yields a more versatile form of discourse linking, as it allows a definite NP to have specific discourse reference, and to be positioned accordingly: on the left of a discourse partitioner. I will elaborate this in section 4. 3.3. Word order Present-day English is characterized by a rigid SVO syntax. In the unmarked case, the subject corresponds with given information, the object with new information (Birner and Ward 1998). It is well-known that earlier

Discourse relations and word order change

101

English had a larger number of syntactic options available beside SVO word orders. Consider in this respect the vast literature on OV word order and its loss, and on the Old English version of the Verb Second constraint. For summaries of the issues playing a role in the syntactic debate on these word orders, the reader is referred to Fischer et. al (2000) and references cited there, and to Kroch and Taylor (2000), Pintzuk and Taylor (2006), and to the articles in Taylor and van der Wurff (2005). Likewise, there is a considerable literature on Old English word order in discourse terms, including Brinton (2006), Bech (2001), to name but a few. A general point about the literature is that purely syntactic treatments have some difficulty coming to grips with such essentially relational notions as given/new information, discourse reference, closeness of the antecedent and so on, which are all relevant to a proper treatment of Old English word order. The literature so far in terms of discourse, on the other hand, does scant justice to insights emerging from the syntactic literature. As yet, there is little or no literature on the interplay between syntax and discourse. In this article, I start from the assumption that there is such a thing as a tightly circumscribed syntax, which makes available a limited set of options which define the limits to the space that language users have to manipulate word order according to information-structural needs. Viewed from this perspective, Old English syntax allows the language user more leeway than present-day English does. This relative freedom has often been ascribed (exclusively and rather impressionistically) to the presence of inflectional morphology, in particular with respect to the relative positional freedom of NPs. No doubt, the presence of case morphology is important for syntactic purposes in identifying the argument status and grammatical role of NPs; I would argue, however, that it is not only case marking per se that yields word order flexibility, but rather the fact that an inherent part of this case system was the paradigm of weak demonstrative pronouns, which carried inflection for case, number and gender and were used as independent demonstratives, relative pronouns and definite “determiners”.

4.

Discourse and syntax in Old English

4.1. The position of pronouns The focus in this treatment is then the position/area that in more purely syntactic analyses has been associated with high personal pronouns or the Spec,AgrP position in the structure (5) above. We have now seen that it hosts

102

Ans van Kemenade

a subset of subject DP’s as well. From a syntactic viewpoint, this yields a problem primarily for subclauses as we saw in discussion in section 2. We will therefore first consider sub-clauses, and look at the discourse conditions and syntactic conditions that various types of high elements are subject to. The hypothesis, following van Kemenade and Los (2006) and van Kemenade and Miliüev (2005), is that material appearing on the left of þa/þonne is discourse-linked, i.e. any material occurring there refers specifically to a discourse antecedent. Material in the discourse-linked area in effect comprises continued topics in the discourse sense: subjects and a range of pronouns, essentially given information. At this point, it may be useful to emphasize that this is a notion crucially different from ‘topicalization’ as it is known in the syntactic literature, which involves movement to Spec,CP in root clauses, and whose discourse effect is more typically associated with contrast, comparison and scene-setting. Returning to our discourse-linked area, it should come as no surprise that a range of pronoun types occur here. First and foremost, this applies to personal pronouns. Subject pronouns near-categorically appear in this area, and if they occur in conjunction with object pronoun(s), they always precede the object pronoun(s). (15) illustrates a double object example where all three arguments are pronouns; the order of the pronouns is fixed as in (15): nominative – accusative – dative. This is a crosslinguistically widely attested order. (15) on Salomonnes bocum, hit is awrieten ðæt mon ne scyle cweðan in Solomon’s books, it is written that one not shall say to his frind: Ga,cum to morgen, ðonne selle ic ðe hwæthwugu, to his friend: “Go,come tomorrow, then give I you something, gif he hit him ðonne sellan mæge. if he it him then give may.”

CP.44.323.24

‘in the books of Solomon, it is written that we are not to say to our friend: “Go, and come tomorrow, then I will give you something,” if we can give it him then.’ An object pronoun may appear alone in the high position, as in (16): (16) gif hine ðonne ðæt fleah mid ealle ofergæð, ðonne ne mæg he if him then the albugo with all covers, then not-can he noht geseon. naught see.

CP 11.69.17

Discourse relations and word order change

103

‘if it [the pupil of the eye] is entirely covered with albugo, he can not see anything.’ Object pronouns can occur in lower positions as well. Miliüev (in prep, chapter 2) shows that object pronouns in the high position have a prominent discourse antecedent. DP objects are not attested in the discourse-linked area. There is one substantial set of apparent counterexamples to this: preposed dative experiencers, as illustrated in (17)6. (17) þa þæm hearpere ða ðuhte ðæt hine þa nanes then the harp-player then seemed that him then no þinges ne lyste on ðisse worulde thing not pleased in this world

Bo 35.102.9.1978

‘Then it seemed to the harp-player that nothing in this world pleased him anymore’ We can account for this in terms of Allen (1995), who shows that preposed dative experiencers behave as subjects with respect to coordinate subject deletion. The least we can say is that these constructions involve unaccusative verbs, i.e. verbs without an external argument; the subject position in such clauses can host one of the internal arguments (van Kemenade 1997). The high position optionally hosts a range of indefinite pronoun subjects as well.7 This may well raise a problem for the idea that material in the high position is discourse-linked. However, Miliüev (in prep) shows that, while indefinites in the lower subject position are genuinely indefinite, high indefinites in fact do have a presuppositional reading: choice of referent from a discourse-given set. To illustrate, consider (18) (from Miliüev (in prep)): (18) a. ða gebroþra, þe for hwylcere neode utfarað and þy ilcum the brothers, who for which need go-out and the same daege to mynstre cuman maegen, ne dyrstlaecen hi þaet hy day to monastery come may, not dare they that they ahwaer ut of mynstre etan, þeah hy hwa bidde, anywhere out of monastery eat even-if them anyone invites,

104

Ans van Kemenade

butan hit him from heora abbode geboden sy unless it them from their abbot allowed be BenR:51.79.15.927 ‘the brothers, who for some reason need to go out and are able to come back to the monastery the same day, they are not allowed to eat anywhere outside the monastery, even if someone invites them, unless it has been allowed by the abbot’ b. Gif hwa hit þonne butan his leafe gedyrstlaece, beo he on if anyone it then out his permission dares, be he in amansumunge excommunication BenR:51.79.18.928 ‘if anyone then dares to ignore his permission, he should be excommunicated’ The relevant hwa is in (18b). (18a) provides the context, which makes it clear that hwa ‘whoever’ in (18b) picks a referent from the set of brothers mentioned in a): those brothers who for some reason need to go out and are able to come back the same day. Miliüev notes that hwa can occur in the lower subject position as well. Interestingly, it occurs there typically when there is no set of possible referents as in (18a), as illustrated in (19): (19) and næfre se læweda man ne healed ealdorscipe ofer gehadodum. and never the lay man not hold eldership over clergy. Gif hit ðonne hwa deð, wite he þæt he deð ongean Cristes gesetnysse If it then anyone does, know he that he does against Christ’s ordinance ÆCHom_II_45:345.308.7714 ‘and a layman should never hold supremacy over clergy. If anyone then does, he should know that he acts against Christ’s ordi nance…’ The potential set of referents in the context is that of laymen, but this is cancelled by the negation. Let’s now turn to non-pronominal subjects in the high position. DP’s in the high position comprise possessive DP’s and DP’s with a definite demonstrative pronoun. In the lower position on the right of þa/þonne, we find proper names, DP’s with a definite demonstrative pronoun, and a range of indefinite and generic expressions. The focus here will be on what determines the choice between higher position and lower subject position.

Discourse relations and word order change

105

Let’s consider here again the examples (8) and (9) above, which are repeated here for convenience as (20) and (21), with some more context for (21): (20) Forðæm bið se sige micle mara ðe man mid geðylde gewinð, therefore is the victory much greater which one with patience wins, forðæm sio gesceadwisnes ðonne hæfð ofercumen ðæt mod & because this wisdom then has overcome the mind and gewielð, swelce he self hæbbe hiene selfne gewildne, & sio geðyld subdued, as if he self have himself conquered, and the patience hæbbe ðæt mod geðreatod & gecafstrod. have the mind intimidated and curbed.

CP[Cotton]:218.19.42

‘Therefore the victory which is won with patience is much greater, because this wisdom (patience) has overcome and subdued the mind, as if he himself had conquered himself, and patience had intimidated and curbed the mind.’ In (20), the DP sio gesceadwisnes lit. ‘the wisdom’, refers back to the PP mid geðylde in the previous clause and should therefore more properly read as ‘this wisdom’. This is a good illustration of the referential properties of the demonstrative pronoun sio: it is this that renders the reference to the antecedent specific.With this in mind let us now look at (21): There are three discourse particles in this text fragment. In the first case, the DP ða sacerdas precedes the particle nu, and has a specific reading, signalling a specific group of priests, those who have undertaken the priesthood and proclaim it loudly. In the next sentence, the subject se sacerd occurs on the right of the particle, and has a generic reading, referring to ‘priest’ in his office of priest. ‘The priest’ here does not refer to a specific person. (21) Forðæm se se ðe ðonne sacerdhad onfehð, he onfehð Because he who that then priesthood undertakes, he undertakes friccan scire & foreryneles ða her iernað beforan kyningum, & herald’s office footmen who here run before kings, & bodigeað hira færelt & hiera willan hlydende. Sua sculun proclaim their journey and their will loudly. So shall ða sacerdas nu faran hlydende & bodiende beforan ðæm the priests now run proclaiming and preaching before the

106

Ans van Kemenade

egeslican deman ðe him suiðe andrysnlic æfter gæð. Gif awful judge who them very majestically after goes. If ðonne se sacerd bið ðæs ungerad lareowdomes, hwæt forstent then the teacher is unskilled in instruction, what avails ðonne his gehlyd? then his cry?

CP, 15,91,25

‘For he who undertakes the priesthood undertakes the office of herald and footmen who run before kings, loudly proclaiming their journey and will. Thus priests ought to run and proclaim loudly be fore the awful judge, who follows them in great majesty. But if the priest is unskilled in instruction, what avails his cry?’ These examples give us a perspective on the choice between the higher and lower subject position: the higher position is for those definite DP-subjects that refer back specifically to a discourse antecedent. This is possible because the demonstrative pronoun allows specific reference. The lower subject position is for those DP-subjects which are definite but have a generic or otherwise non-specific reading. Van Kemenade, Miliüev and Baayen (2008) offer sophisticated statistical support for the analysis here, in an approach much inspired by Bresnan et.al. (2007). A database was created including all the clauses containing þa/þonne. For the present analysis, it is primarily subclauses that are relevant, as the articulation of the ‘high’ area in main clauses is further complicated by various types of movement to CP. To each NP, numerical values were assigned for discourse-relevant properties, including: NP function (subject, various types of object); NP Type (personal pronoun, weak demonstrative, strong demonstrative, definite NP, indefinite NP etc.); NP position (left periphery, high (left of particle), mid (immediate right of particle), low); specificity (non/specific); distance of antecedent (counted in clauses). The interaction between these factors was calculated in the R open statistical environment (R Development Core team 2004). Summarizing for DP subjects, it is shown that the combination of the four values ‘definite NP’, ‘specific NP’, ‘high position’ and ‘antecedent is present in the discourse’ (less than 3 clauses away) is of the statistically highly significant. Moreover, each of these four values contributes highly significantly to this combination. I take this as strong evidence for the approach outlined here. This allows a number of conclusions. The first and foremost conclusion is that the ‘high’ area, as marked off by the adverb, is reserved for

Discourse relations and word order change

107

discourse-linked material: personal pronouns (subject and object), presuppositional indefinite pronouns, specific definite DP’s. The second conclusion is that this additional layer in the clause is licensed by the presence of the adverb. The third conclusion is that definite DP can establish a discourse link in this high position by virtue of the anaphoric character of the demonstrative pronoun. The final conclusion is that, even in subclauses, the adverb has strong diagnostic value. The general point emerging from this discussion then is that the variety of elements occurring in the high position has been unified under one principled non-syntactically defined term: discourse-linked material. Let us now try to define with some more precision the syntactic properties and structure of this ‘high area’: in particular, let’s consider where we stand now with respect to the partial clausal architecture in (5) above, repeated here for convenience as (22): (22) [CP [XP] C [AgrP PronounSubj/Obj Agr [NegP Neg Adv Neg [TP DPSubj T ... ]]] The discourse-linked domain we have identified here is positionally comparable with AgrP in this structure; the position of the particle þa/þonne corresponds with NegP; the lower subject position corresponds with spec,TP, leaving aside further lower subject positions. I will dub the discourse-linking area ȈP, following Nilsen (2003); van Kemenade and Miliüev (to appear). Ȉ stands for ‘scrambling’, which seems an appropriate characterization of this phenomenon. ȈP then is a high scrambling position, distinct from and higher than the position usually associated with object scrambling in the Germanic languages. Given the generalizations that have been made above concerning discourse linking, the conclusion seems inevitable that the position of þa/þonne is higher than TP: only discourselinked material appears on its left; transitive DP subjects may appear on its right. Next, let’s consider the internal structure of ȈP. It was observed above that up to three personal pronouns may appear here, in a fixed order. A DP subject may also combine with an object pronoun in ȈP, but here, too, the order is: subject DP - object pronoun, as illustrated by (23), showing that within ȈP, the order of subjects and objects is fixed. (23) Gif he wif self hæbbe, gange hio ut mid him. Gif se if he woman self have, go she out with him. If the

108

Ans van Kemenade

hlaford him þonne wif sealde sie hio & hire bearn þæs hlafordes lord him then woman gave, be she and her children the lord’s. LAW2,28.11.2-3 ‘If he will have the woman himself, she will go with him. Now if the lord gave him the woman, she and her children will belong to the lord.’ If an approach in terms of ‘high’ scrambling is correct, the facts clearly show that it is discourse reference that triggers scrambling. This view is controversial: scrambling is treated by many as syntactic optionality (e.g. Haider 2005). While it is no doubt correct that scrambling is optional in the sense that the syntax allows two positions that can both be chosen, the generalizations about discourse reference here, considered for each example in its broader context, show that scrambling is strongly discoursesensitive. This is in line with Hinterhoelzl (2004) on present-day German. He argues that there are two triggers for scrambling in present-day German: scope and specificity, where specificity is taken to mean that the scrambled constituent is part of a set of referents given in the context. The only circumstance in which an object may appear on the left of a subject then is when an object pronoun is in ȈP, and the (DP)-subject is in a lower position. We have seen several examples of this above, and I repeat (16) here for convenience: (24) gif hine ðonne ðæt fleah mid ealle ofergæð, ðonne ne mæg he noht if him then the albugo with all covers, then not-can he naught geseon. see CP:11.69.17 ‘if it [the pupil of the eye] is entirely covered with albugo, he can not see anything.’ The ordering restrictions summarized here suggest that there are tight syntactic restrictions on movement to ȈP. The fact that a DP-object cannot move to ȈP suggests a minimality effect: movement is blocked by an intervening subject in spec,TP, cf. Miliüev (in prep). This also explains why dative experiencers (as exemplified in (17) above) can prepose to the discourse-linked area: the constructions in which they feature involve unaccusative verbs, so that there is no nominative subject in Spec,TP. This raises the question why an object pronoun can cross a subject in Spec,TP. I assume, following Miliüev (in prep) that object pronouns undergo syntactic

Discourse relations and word order change

109

cliticization (adjunction) to the head of ȈP. This enables us to revise (22) as (25): (25) [ȈP d-linked subject [Ȉ (obj pro) [Ȉ (obj pro) [TP]]] The structure (25) does not specify what the status of the adverb is with respect to ȈP. One possibility here, in line with Aboh’s (2004; 2006) treatment of Gungbe, is to say that the adverb is the head of ȈP, and thus licenses an extra position available for scrambling. Another possibility is that the adverb is itself a scrambled element, moving to ȈP in order to anchor the clause by referring back to a point in time pre-established in the discourse. I would for the moment like to remain neutral on these possibilities, leaving the answer for further research into the discourse context in Old English, and the role of discourse-markers cross-linguistically. 4.2. The position of the finite verb A question left unaddressed so far is how the various positions under discussion here relate to the position of the finite verb in sub-clauses. We can safely say that appearance of the finite verb in ȈP is rare. Starting from table 2 above, repeated here as table 4, occurrence of the finite verb on the left of þa/þonne is given in table 5 (only subject-initial patterns have been included, as the finite verb never appears on the left of the particle when the subject is lower). For the sake of comparison, I provide Table 6 with the same patterns in root clauses (restricting the search to subject-initial root clauses, although the position of the subject in subject-initial root clauses is almost certainly higher than ȈP in at least a number of cases). Table 4. Subject - þa/þonne order in Old English sub-clauses

Old English subclauses Left of þa/þonne Right of þa/þonne

Pronominal subjects 1250 5

DP subjects 293 419

Table 5. Subject - finite verb - þa/þonne order in Old English sub-clauses

Old English subclauses Left of þa/þonne Subject-Vf- þa/þonne

Pronominal subjects 1250 53 (4.2%)

DP subjects 293 46 (15.7%)

110

Ans van Kemenade

Table 6. Subject - finite verb - þa/þonne order in Old English main clauses

Old English root clauses Left of þa/þonne Subject-Vf- þa/þonne

Pronominal subjects

DP subjects

2533 629 (34.6%)

1817 392 (21.6%)

It should be noted that the numbers for incidence of the finite verb in Ȉ in table 5, low as they are, are still flattered. Many of the examples included in these numbers occur in embedded root environments (complements of bridge verbs, non-restrictive relatives, clauses introduced by ‘because’ as exemplified in (26))8: (26) Ĉæt is swelce he hæbbe befæsð his hond oðrum menn, ðonne that is as if he have committed his hand other men, when he gebint hine selfne to him mid his wordum ðæt he sceal he binds himself to him with his words that he shall niede ða giemenne & ða geornfulnesse ymb ðone habban ðe necessarily the care and the zeal about him have whom he ær ne ðorfte forðæm he hine hæfð ðonne gehæftne he before not needed, because he him has then held mid his agnum wordum with his own words

CP:28.193.7.1283

‘It is like committing his hand to another, when he binds himself to him with his words that he will necessarily be careful and zealous about him for whom it was not necessary before, because he holds him with his own words, … When embedded root environments are left out of account, the figures are very low indeed. Another piece of evidence that the finite verb does not feature in ȈP is that the presence of a high object pronoun is apparently incompatible with finite verb placement. In the 51 cases in which an object pronoun appears on the left of the particle, either alone or in combination with a subject, the finite verb does not occur on the left of the particle. The evidence suggests then that our scrambling area is reserved for arguments. If our scrambling area is for discourse-linked material as demon-

Discourse relations and word order change

111

strated, an interesting way of putting it may be that verbs are not discourselinked. It will be intriguing to pursue this perspective, especially because it potentially provides a novel perspective on the discussion of verb placement, but I will resist that temptation, which would take me far beyond the scope of this article, and will have to be left for further research, including cross-linguistic research. The numbers in tables 5 and 6 also raise interesting questions about finite verb movement in root clauses. Although the incidence of finite verb movement as evident from table 6 is higher than in non-root clauses, the number is not nearly a high as we might expect on the basis of the overall picture we have of finite verb movement in root clauses, suggesting that here, too, it will be highly relevant to consider discourse-linking in more detail. However, our findings for ȈP in sub-clauses cannot be straightforwardly extended to root clauses, as the left periphery of root clauses is obscured by a range of clause typing strategies as in e.g. questions and non-subject-initial clauses, to name but a few. The picture is fraught with evidential problems and this, too, must be left for further research.

5.

Diachronic developments

In this section, I will attempt to put the findings so far in a diachronic perspective. I will first outline this perspective as a hypothesis, and then fill in the evidence as far as it will take us at this stage. The starting point is, on the one hand, the analysis of Old English as presented here, and, on the other hand, a consensus view of present-day English syntax. The latter consists simply of the observation that present-day English is SVO, has a designated functional position for auxiliaries, and shows only some relic features in main clauses of a stage where it had more than one subject position. In discourse terms, the rigid SVO structure of present-day English encodes the subject as the unmarked topic. The evidence presented below shows the following surface changes: (27)

a. b. c. d.

the loss of the discourse partitioning function of þa/þonne the loss of high object pronouns the loss of the demonstrative pronoun paradigm increase of Vf-adverb order

The changes in (27) evidence a loss of the elements that determined the discourse flexibility of Old English. Let’s consider them briefly in turn.

112

Ans van Kemenade

The loss of the discourse particle function of þa/þonne and the loss of high object pronouns: the transition to Middle English shows a very marked decrease in the use of þa/þonne in absolute numbers. Discounting clauseinitial and conjunction uses of the adverbs, the Old English search file contains some 2.500 instances of þa/þonne in sub-clauses. The Middle English search file has about 180. Given this sharp drop in frequency, it is questionable whether the discourse marker function of the adverb would have been represented in the language environment robustly enough to remain learnable as a discourse marker. Apart from this, as discussed in van Kemenade and Los (2006), the evidence for the empirical generalizations established above for Old English has dwindled in Middle English. There are only two examples of object pronouns on the left of þa/þonne, both in combination with a subject, and both in one of the earliest Middle English texts. These facts seem to warrant the conclusion that, at least in sub-clauses, the subject has become the unmarked topic. Going back from Middle English to Old English, and assuming that object pronouns on the left of the particle are a feature of the discourse organization discussed here for Old English, it is interesting to note that there appears to be some change in the Old English period already. When comparing the corpus of Alfredian texts (9th century) with that of the texts by Ælfric and Wulfstan (10th century), it turns out that of the total number, about 75% of the high object pronouns is in the earlier texts. The loss of the demonstrative pronoun paradigm: as discussed above, Old English demonstrative pronouns were an interclausal tracking device par excellence. They functioned as relative pronouns, had independent pronominal reference, and in their use as definiteness markers also allowed specific reference to a discourse antecedent. Demonstrative pronouns were a relatively early victim of case loss, earliest in the East and North. Smith (1996) discusses the collapse of the paradigm in the first and final continuations of the Peterborough Chronicle (North-east Midlands, 1130– 1154), and notes speakers’ attempts to restructure the system, a certain sign of its demise. The increase of Vf-adverb order: it was noted above that the finite verb is not a welcome guest in ȈP: the numbers of Vf-adverb order in the subclause are quite low. Here again, there is already some development within the Old English period. About two thirds of the Vf/þa/þonne word orders are from 10th century, which means that between the late 9th and late 10th centuries, the incidence of Vf-adverb order has been doubled. This devel-

Discourse relations and word order change

113

opment is continued in the transition to Middle English. In table 7, the figures are presented for Vf-Adverb order with nominal and pronominal subjects, for the first two sub-periods of Middle English, M1 (AD 11501250) and M2 (AD 1250-1350). Given the low numbers for the Middle English variants of þa/þonne, I expanded the search to all temporal adverbs. Table 7. Subject-finite verb-temporal adverb order in early Middle English

ME subclauses M1 M2

S pro-Vf-Adv 121/484 25.5% 42/75 56 %

S DP-Vf-Adv 54/140 38,5 % 14/24 58,3 %

Table 7 shows a very striking fact: by the M2 period, there are no patterns to suggest that nominal and pronominal subjects are different in subclauses (though there still are major asymmetries in main clauses). Put differently: while the evidence for the discourse particle status of þa/þonne had already dwindled in the M1 period, by the M2 period, all evidence for ȈP in sub-clauses had disappeared. On the evidence from sub-clauses, there would be no reason to assume an analysis other than one in which the subject is licensed in Spec, TP, the finite verb in T, with the adverb merged below T. In this respect, the left periphery of sub-clauses is simplified earlier than that of main clauses. I interpret this development as the loss of high scambling, which took place earlier in sub-clauses than in main clauses. The result of this loss is that the adverb comes to be merged below T. Note that this does not require a very drastic reanalysis: the patterns in the Old English discourse-flexible grammar with ȈP on the left of an adverb functioning as a discourse partitioner, are not incompatible with the more rigid syntax of M2, in which the position of subject and verb is well on the way to becoming regulated in TP. The positions for subjects do not change drastically either: in the Old English ȈP, subjects could readily occur as continued topics. What is lost in the transition to Middle English is the ingredients necessary for a discourse-flexible grammar: the discourse partitioning function of a number of adverbs, and the articulate paradigm of demonstrative pronouns which yields an inter-clausal tracking device for a wider range of subjects. Discourse takes what is made available by the syntax.

114

Ans van Kemenade

Notes 1.

2.

3. 4.

5.

6.

7.

8.

I thank Tanja Miliüev and Elizabeth Traugott for helpful discussion of the issues in this paper. Thanks also to the editors of this volume for their patience, constructive comments, and help. The references follow the system of short titles as employed in Healey and Venezky (1985 [1980]) (in turn based on the system of Mitchell, Ball and Cameron (1975, 1979)) Haeberli (1999) dubs SpecAgrP a subject1 position and Spec,TP a Subject2 position. But the “PronounSubj/Obj” position hosts object pronouns as well. Haeberli and Ingham (2007) rightly draw attention to the problem for subclauses. But most of their analysis is devoted to showing that in early Middle English, the position of the adverb is lower than T, drawing on evidence from the relative positioning of adverb and finite verb. I will refrain from taking their detailed argument into account, first, because their account has nothing to say on the motivation for the choice of subject position, except to stipulate that pronouns are high; second, they ignore the fact that nominal subjects readily appear on the right of the adverb; third, their word order evidence is based broadly on adverbs generally, which renders a comparison with two specific adverbs in Old English problematic; fourth, their word order evidence on finite verb position cannot be replicated for Old English in relation to þa/þonne. Finally, their evidence for early Middle English implies a discontinuity between Old and Middle English that is of a magnitude hard to imagine. Note that the set of elements identified here suggest a considerable overlap with their present-day German cousins, which in the literature are called modal particles. The meanings described in the literature for German (e.g. Thurmair 1991) are in some respects akin. A detailed semantic study of Old English particles is beyond our scope here and is left for further research; note that the claim made here goes no further than to say that these adverbs separate discourse domains. As far as I have been able to trace in the literature, this makes their behaviour different from that of modal particles in present-day German. There is a further and rather intriguing set of counterexamples, all involving the impersonal subject pronoun man. This is discussed in detail in Miliüev (in prep). I refrain here from discussion of the properties of the impersonal pronoun man. The reader is referred to van Kemenade and Miliüev (2005) and Miliüev (in prep). I leave it for further research to work out the precise numbers, which requires hand-correction of the search files from YCOE.

Discourse relations and word order change

115

References Aboh, Enoch Oladé 2004 The Morphosyntax of Complement-head Sequences. Clause structure and word order pattern in Kwa. New York: Oxford University Press. 2006 Complementation in Saramaccan and Gungbe: The case of c-type modal particles. Natural Language and Linguistic Theory 24 (1): 1– 55. Allen, Cynthia 1995 Case Marking and Reanalysis. Oxford: Clarendon Press. Bech, Kristin 2001 Word order patterns in Old and Middle English: A syntactic and pragmatic study. Doctoral dissertation, University of Bergen. Birner, Betty J. and Gregory Ward 1998 Information Status and Canonical Word Order in English. Amsterdam/Philadelphia: John Benjamins. Bobaljik, Jonathan and Dianne Jonas 1996 Subject positions and the roles of TP. Linguistic Inquiry 27 (2): 195– 236. Bobaljik, Jonathan, and Höskuldur Thráinsson 1998 Two heads aren’t always better than one. Syntax 1 (1): 37–71. Bresnan, Joan, Anna Cueni, Tatiana Nikitina and Harald Baayen 2007 Predicting the dative alternation. In Cognitive Foundations of Interpretation, eds. Gerlof Bouma, Irene Kraemer and Joost Zwarts, 69– 94. Royal Netherlands Academy of Arts and Sciences. Brinton, Laurel 2006 Pathways in the development of pragmatic markers in English. In The Handbook of the History of English, eds. Ans van Kemenade and Bettelou Los, 307–334. Oxford and Malden, MA: Blackwell. Cardinaletti, Anna and Ian Roberts 2002 Reprint. Clause structure and X-second. In Functional Structure in DP and IP: The Cartography of Syntactic Structures. Vol. I, ed. Guglielmo Cinque, 123–166. Oxford: Oxford University Press. Original Edition, Oxford: Oxford University Press, 1991. Cardinaletti, Anna and Michal Starke 1996 Deficient pronouns: A view from Germanic. A study in the unified description of Germanic and Romance. In Studies in Comparative Germanic Syntax. Vol. II, eds. Höskuldur Thráinsson, Samuel David Epstein and Steve Peter, 21–65. Dordrecht: Kluwer. Denison, David 2006 Category change and gradience in the determiner system. In The Handbook of the History of English, eds. Ans van Kemenade and Bettelou Los, 279–304. Oxford and Malden, MA: Blackwell.

116

Ans van Kemenade

Fischer, Olga, Ans van Kemenade, Willem Koopman and Wim van der Wurff 2000 The Syntax of Early English. Cambridge: Cambridge University Press. Haeberli, Eric 1999 Features, categories and the syntax of A-positions. Synchronic and diachronic variation in the Germanic languages. Doctoral dissertation, University of Geneva. Haeberli, Eric and Richard Ingham 2007 The position of negation and adverbs in Early Middle English. Lingua 117: 1–25. Haider, Hubert 2005 How to turn German into Icelandic – and derive the OV-VO contrasts. The Journal of Comparative Germanic Linguistics 8 (1-2): 1–53. Healey, Antoinette Dipaolo and Richard Venezky 1985 A Microfiche Concordance to Old English. Toronto: The Pontifical Institute of Mediaeval Studies. [1980] Hinterhölzl, Roland 2004 Scrambling, optionality and non-lexical triggers. In Triggers, eds. Anne Breitbarth and Henk van Riemsdijk, 173–204. Berlin: Mouton de Gruyter. Kato, Kozo 1995 The interjection la and subject pronouns in Old English. Linguistics and Philology 15: 23–40. Kemenade, Ans van 1987 Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris. 1997 V2 and Embedded Topicalization in Old and Middle English. In Parameters of Morphosyntactic Change, eds. Ans van Kemenade and Nigel Vincent, 326–352. Cambridge: Cambridge University Press. 1999 Sentential negation and clause structure in Old English. In Negation in the History of English, eds. Ingrid Tieken-Boon van Ostade, Gunnel Tottie and Wim van der Wurff, 147–165. Berlin: Mouton de Gruyter. 2000 Jespersen’s cycle revisited: Formal properties of grammaticalization. In Diachronic Syntax: Models and Mechanisms, eds. Susan Pintzuk, George Tsoulas and Anthony Warner, 51–74. Oxford/New York: Oxford University Press. Kemenade, Ans van and Tanja Miliüev 2005 Syntax and discourse in Old English and Middle English word order. In Articles from the 8th Diachronic Generative Syntax Conference, eds. Steven Anderson and Dianne Jonas. To appear with Oxford University Press.

Discourse relations and word order change

117

Kemenade, Ans van and Bettelou Los 2006 Discourse Adverbs and Clausal Syntax in Old and Middle English. In The Handbook of the History of English, eds. Ans van Kemenade and Bettelou Los, 224–248. Oxford: Blackwell. Kemenade, Ans van, Tanja Miliüev and R. Harald Baayen 2008 The balance between discourse and syntax in Old and Middle English. In English Historical Linguistics 2006. Vol. I: Syntax and Morphology, eds. Marina Dossena and Maurizio Gotti. 3–22. Amsterdam: John Benjamins. Kroch, Anthony and Ann Taylor 2000 Verb-object order in Early Middle English. In Diachronic Syntax: Models and Mechanisms, eds. Susan Pintzuk, George Tsoulas and Anthony Warner, 132–163. Oxford/New York: Oxford University Press. Miliüev, Tanja in prep. Discourse linking in Old English. Doctoral dissertation, Radboud University Nijmegen. Mitchell, Bruce, Catherine Ball and Angus Cameron (eds.) 1975 Short titles of Old English texts. Anglo-Saxon England 4: 207–221. 1979 Addenda and corrigenda. Anglo-Saxon England 8: 331–333. Nilsen, Øystein 2003 Eliminating Positions: Syntax and Semantics of Sentential Modification. LOT dissertations 73. Pintzuk, Susan 1991 Phrase structures in competition: variation and change in Old English word order. Doctoral dissertation, University of Pennsylvania. Pintzuk, Susan and Ann Taylor 2006 The loss of OV order in the history of English. In The Handbook of the History of English, eds. Ans van Kemenade and Bettelou Los, 249–279. Oxford and Malden, MA: Blackwell. R Development Core Team 2004 R: a language and environment for statistical computing. http://R-project.org. Rizzi, Luigi 1997 The fine structure of the left periphery. In Elements of Grammar: Handbook in Generative Syntax, ed. Liliane Haegeman, 281–337. Dordrecht: Kluwer. Taylor, Ann 2000 The Penn-Helsinki Parsed Corpus of Middle English 2. University of Pennsylvania, Dept. of Linguistics.

118

Ans van Kemenade

Taylor, Ann, Anthony Warner, Susan Pintzuk and Frank Beths 2003 The York-Toronto-Helsinki Parsed Corpus of Old English Prose. York: Department of Language and Linguistic Science, University of York. Available through the Oxford Text Archive. Taylor, Ann and Wim van der Wurff (eds.) 2005 Aspects of OV and VO word order in the history of English. English Language and Linguistics 9 (1). Cambridge: Cambridge University Press. Thurmair, Maria 1991 Kombinieren Sie doch nur ruhig auch mal Modalpartikeln. Combinatorial regularities for modal particles and their use as an instrument of analysis. Multilingua 10: 19–42.

Methodological problems of the information-structural analysis of data from historical text corpora

On the methods of information-structural analysis in historical texts: A case study on Old High German Svetlana Petrova and Michael Solf

1. Introduction It is an undisputable fact that information-structural research made considerable advance in the past few decades by exploring a wide range of phenomena in the systems of present-day languages spoken worldwide. Historical linguistics has also been interested in the matter of how old languages attempt to achieve the needs of communicative explicitness and stylistic expressivity. Leaving aside a huge variety of accounts within the ancient and medieval grammar tradition, even more recent descriptive works on the grammar of the ancient languages give direct or at least implicit accounts of grammatical phenomena which correlate with properties of contextual reference and pragmatics in these languages. It is most revealing that one of the earliest works that influenced the emergency of what we call ‘Functional Sentence Perspective’ or ‘Information Structure’ today, the seminal work of Henri Weil ([1844] 31879), was dedicated to the sentence structure of ancient Greek and Latin which he regarded as a reflection of the natural flow of thoughts both in conditions of neutral as well as emphatic speech. In more recent times, information-structural features are still a key to the proper understanding of historical text sources. Taking as an example the role of information structure for the interpretation of Biblical Hebrew texts and Old Hebrew narratives, only within a short period of time a series of monographs like those by Disse (1998), Heimerdinger (1999) and Floor (2004) appeared. As for the early Germanic languages, properties of the information-structural organization of the utterance have regularly been addressed in earlier studies aiming at explaining word order variation and verb placement. Accounts like those of Bean (1983) and Stockwell (1984) on Old English syntax have drawn a connection between the pragmatic features of sentence constituents and their placement in the structure of the clause, while in Hopper’s (1979a and b) typological account, different word order patterns in Old English are directly associated with properties of discourse organization, namely with the function of distinguishing parts

122

Svetlana Petrova and Michael Solf

of foregrounding vs. backgrounding in narratives. Similar attempts are found in the syntactic exploration of the remaining early Germanic languages as well. As early as in the accounts on the syntax of Old Saxon by Ries (1880: 1–11) or Old High German (OHG) by Behaghel (1932, IV: 3– 9), among many others, the distribution of different word order patterns involves terms and concepts which from the perspective of modern linguistics clearly pertain to the field of information structure and discourse analysis. Meanwhile, information structure has started to be taken into account as a factor responsible for the emergency of novel word order patterns and constructions and thus introducing variation as a precondition of language change in the course of language development, cf. Hinterhölzl (2004, in this volume). This view opens new perspectives on the explanation of word order variation and syntactic change in language history. Given the considerable advance in recent studies on information structure, there is a demand for a more systematic, large-scale description of the information-structural properties of sentence constituents and their syntactic realization in early Germanic. As promising as it seems, such an enterprise is concerned with a series of methodological problems and questions which have to be addressed prior to any information-structural analysis of a historical text. The present paper aims at providing a method of dealing with information-structural questions on the basis of evidence from OHG and reflects the experience of a research project investigating the role of information structure for the development of the word order regularities in the Germanic languages.1 Problems arising in dealing with information structure in a text from a historic corpus can broadly be classified into two major groups. The first one concerns the quality of the written data available to us, i.e. the reliability of the written text sources as well as the authenticity of the constructions and patterns provided in them. The second problem concerns the information-structural analysis proper, i.e. how to handle the huge variety of theoretical notions and terminological items used in present day’s research and how to apply them to data from historic corpora.

2. The philological issue Any research on syntax and information structure in a historical text has to address the reliability of its data base, i.e. the question whether the word order patterns and syntactic constructions provided in the written texts

Information-structural analysis in historical texts

123

sources are representative for the system of the language under investigation. In this respect, OHG displays an especially difficult ground for any conclusions on syntax, as was also recently shown by Fleischer (2006). The major part of the Old High German corpus is made up by two types of texts: i) vernacular translations of Latin biblical or other religious sources (Isidor (late 8th century), Tatian (9th century), Notker (11th century)) as well as ii) metrical poetry (the poetic Gospel Harmony of Otfrid von Weißenburg, 9th century). In both types of texts, it cannot be excluded that the ordering of sentence constituents is subject to influences imposed either by the syntax of the Latin original or by the properties of the poetic form like rhyme and metre. Therefore, we have to conclude that none of the Old High German sources of a considerable size may be viewed as a good example of authentic prose representative for the system of the dialects spoken at the period of time. For this reason, prior to the syntactic investigation of OHG, some methodological solutions are required. Let us analyse in more detail the specific problems that the individual sources of the Old High German corpus display and consider texts that can be assumed to be more appropriate for the purpose of the intended study than others. E.g., Otfrid’s Gospel harmony, the only poetic record among the more extensive ones, has been viewed to display structures and other grammatical phenomena explainable only with respect to the purposes of end rhyme and metre (cf. Ingenbleek 1880, Fleischer 2006: 35–37). Looking at the translations, there are above all two records of a considerable size: the so called Old High German Tatian and Notker’s extensive work. Both expose the problem of Latin influence, and in addition to that, Notker’s work is simply too young to draw conclusions from it on the earliest stages of the syntax of OHG. The remaining text, the Old High German Tatian translation, is a record of a most disputed quality with respect to syntactic analysis, while in the fields of phonology and morphology, it has helped tremendously to shape our view on OHG. This Gospel harmony, which only survived as one single manuscript (St. Gallen Cod. 56), represents the largest Old High German text written before 850, a translation from the Latin, a bilingual source which provides the original and its translation in two juxtaposed columns. Due to the fact that a comparison of the Old High German text with the Latin source is immediately possible, an overwhelming syntactic similarity between the Latin source and its Old High German translation is visible at once. Consider the example in (1) in which the Old High German sentence follows the structure of its Latin original without any exception:

124 (1)

Svetlana Petrova and Michael Solf

inti quamun fugala inti frazun thiu (T 108, 2) and came birds and ate them ‘and birds came and ate them’ & uenerunt uolucres & comederunt ea

Given this, many philologists including Sievers (1895:LXX) classified Tatian as a “slavish imitation of the original” which is worthless for any account on syntax. This general skepticism is furthermore fed by the fact that we discover a number of syntactic loan constructions all over the text of Tatian. This becomes particularly clear when it comes to the translation of participial constructions typical for the Latin like the ablativus absolutus. The scribes quite often imitate it by an Old High German dativus absolutus, a construction which is viewed to be a syntactic loan occurring in Old High German translations only, rather than in native writings (cf. Lühr 2005): (2)

fon erdu skeffun / forlazenen Inti árleitten to earthdat shipsdat forsaken and broughtpart allen folg&un imo (T 56, 11–12) all followed him & subductis a terra nauibus / relictis omnibus.’ secuti sunt eum; ‘And when they had brought their ships to land, they forsook all, and followed him.’

This might sound discouraging for any further work on the syntax of Tatian. However, investigations on other aspects of the language of Tatian, as the use of subject pronouns (cf. Eggenberger 1961) or the rendering of AcI-construction of the Latin original (Dentschewa 1987), show that the Old High German text does not in each case imitate the original but systematically implements different means of expression, therefore displaying some genuine features constitutive for OHG. Accordingly, Eggenberger (1961: 87) classifies the Tatian as a kind of ‘mixed form’ placing it between the type of interlinear translations like the Old High German Benedictine Rule and free translations like Isidor. As far as word order is concerned, as early as in the investigation by Ruhfus (1897) it has been noted, that there are numerous instances in which the Old High German text deviates from the structure of the underlying original. Ruhfus explicitly bases his analysis on such cases pointing out that these are appropriate for research on syntax:

Information-structural analysis in historical texts

125

“dass eine untersuchung der wortstellung des ahd. Tatian […] bei all seiner sklavischen abhängigkeit von der lateinischen vorlage ebenso beachtenswerte ergebnisse liefern muss, wie gleiche untersuchungen an der freien und gewandten übersetzung des Isidor oder an der dichtung Otfrids, wenn man sich nur auf eine zusammenstellung der abweichungen des ahd. vom lat. beschränkt und daraus schlüsse zieht.” ‘that an investigation of the word order of the OHG Tatian [...] despite all his slavish dependency on the Latin original ought to give remarkable results as much as similar investigations of the more free and elegant translation of Isidor or of the poetic work of Otfrid, if one only considers the collection of deviations of the OHG from the Lat. and draws conclusions from it.’ (Ruhfus 1897: 1)

In more recent times, Dittmer and Dittmer (1998) also provide a syntactic analysis of the Old High German Tatian based on instances of word order differences to the original. As a methodological pre-condition, they assume that such instances are to be taken as evidence for authentic Old High German syntax: “In the first place, only those instances which differ from the original can be taken as evidence for genuine Old High German word order.”(“Beweiskräftig für genuin althochdeutsche Wortstellung sind in erster Linie nur die von der Vorlage abweichenden Belege.”) (Dittmer and Dittmer 1998: 36). The idea that research on word order in OHG is possible on the basis of examples showing a word order which differs with respect to the original is supported by new insights into the translation technique of this text provided by Masser (1994). He discovers that each line in the Old High German text translates exactly the same material found in the corresponding Latin line. Among thousands of lines we only find relatively few cases neglecting this principle, and the majority of them are found on the few pages contributed by scribe İ2. Astonishingly, this point seems to have passed unnoticed until recently: the former authoritative edition of Sievers (1892) at least and others did not pay the necessary attention to this. Studies on word order ignoring the role of this translation principle threaten to base their conclusions on false premises. The translation technique implemented in Tatian certainly imposes restrictions on the possibility of rendering a genuine word order pattern to the structures found in the original, cf. Masser (1997a and b). But at the same time, deviations from the word order of the Latin text within a line can be valued as evidence of genuine syntax and thus be made an object of separate investigation. Under these conditions, the text of the Old High German Tatian may be considered as one of the most abundant data collections for research on

126

Svetlana Petrova and Michael Solf

Old High German syntax, as recently shown by Fleischer, Hinterhölzl and Solf (2008). The fact that we can make direct comparison with the Latin part is an especially helpful device to take control over the influence of the Latin structure. These considerations motivate our decision to base our data collection and initial investigations on syntactically deviating structures of Tatian. This should not imply that we simply assume that the remaining texts are unsuitable for an account on syntax. We are aware that data from other sources is needed for comparison, and we strongly appreciate the efforts of other specialists to provide methods of dealing with these texts as well, cf. the contributions by Lühr, Lötscher and Schlachter in this volume. In the following, we want to provide a classification of the different types of deviations encountered in Tatian according to the edition of Masser (1994).3Altogether we find several thousands of relevant instances, an amazingly high number of significant divergences from the word order of the Latin original. We first consider the frequent changes within the boundaries of syntactic constituents and then turn to changes which surpass the constituent boundary.

2.1. Changes within the constituents An outstandingly great number of divergences take place within the boundaries of syntactic constituents. This is the case when different types of modifiers, e.g. adjectives, possessive pronouns or genitive attributes, which are post-nominal in Latin are realized pre-nominally in OHG, cf. (3): (3)

uuard gifullit heilages geistes (T 31, 9–10) was filled holygen spiritgen ‘was filled by the Holy Spirit’ Impl&us est spiritu sancto

Concerning issues of information structure and discourse organization, the regular use of the demonstrative pronoun as an article appears to be a notable feature, cf. (4): (4)

kneht uuvohs (T 32, 6) ther the boy grew ‘the boy grew’ puer autem crescebat

Information-structural analysis in historical texts

127

2.2. Changes involving the order of constituents in the sentence Changes affecting the relative order of sentence constituents are especially important for issues of word order and information structure. Here we can roughly distinguish the following cases: i) a constituent missing in the original is inserted in the translation; ii) the relative order of constituents is changed with respect to the original; iii) the translation differs from the original in providing an idiomatic, more precise expression, or it even provides supplementary information. An important subgroup is constituted by instances of Latin synthetic forms which lack a synthetic counterpart in OHG. Although the process is confined to the verb phrase, and strictly speaking not surpassing the constituent boundary, too, the translator(s) had to dissolve the form into an analytic construction and had to decide on the placement of the finite auxiliary with respect to the non-finite verb. As examples (5) and (6) show, both orders, i.e. Aux – Participle and Participle – Aux, occur in Tatian: (5)

thaz uuari gifullit (T 34, 27) that became fulfilled ‘that it might be fulfilled’ ut adimpler&ur

(6)

thiu dar giquetan ist which there called is ‘which is called Ephrem quæ dicitur ephrem

ephrem (T 234, 27) Ephrem

2.2.1. Insertion Different types of constituents lacking a lexical equivalent in the original are inserted in the vernacular texts, e.g. subject pronouns (7)4, object pronouns (8), adverbials (9), and finite verb forms (10): (7)

hab& giloubon (T 200, 12) oba ir if you have faith ‘if you were true believers’ si habueritis fidem

128

Svetlana Petrova and Michael Solf

(8)

thó antlingota imo then answered him ‘then the Saviour answered’ Respondit ihesus

ther the

heilant (T 52, 30) Saviour

(9)

Inti só sie thó ubar and when they there over ‘and when they crossed there’ Et cum transfr&assent

feritun (T 20, 17) went

(10)

Inti Ira namo uuas elisab&h (T 26, 2) and her name was Elisabeth ‘and her name was Elisabeth’ & nomen eius elisab&h

In addition, we also find instances where Latin constituents, especially particles, are not given an equivalent in the translation, cf. lat. ergo in (11): (11)

thane thú tuos elimosinam (T 66, 29) when you do charity ‘when you do charity’ cum facies elimosinam ergo

2.2.2. Transposition For investigations of the conditions on word order the regular transposition of constituents are of the greatest importance. Foremost, the placement of the finite verb is of special interest in this connection. The finite verb may be shifted to the first (12), second (13) or a later position (14) within the sentence: (12)

hant In then heilant anan (T 297, 20) Inti legitum iro and laid their hands in the Saviour on ‘and laid their hands on the Saviour’ & manus iniecerunt In ihesum

(13)

elisab&h uuas unberenti (T 26, 7) Elisabeth was barren ‘Elisabeth was barren’ ess& elisab&h sterilis

Information-structural analysis in historical texts

(14)

129

quamun / zi themo heilante came to the Saviour ‘and when the apostles came to the Saviour’ Et cum uenissent apostoli / ad ihesum (T 102, 29–30)

mitthiu

thie postoli

when

these messengers

Good evidence for authentic structures is especially given in the following two cases. First, multiple changes may indicate that the translator paid more attention to the stylistic quality of his translation. The more is changed, the more we ought to assume, that an authentic structure is expressed. Cf. (15) where the transposition of the copula sint is being accompanied by the insertion of the subject pronoun sie: (15)

sie sint blinte inti blintero they are blind and blind-GenPl ‘They are blind and leaders of the blind’ caeci sunt duces cecorum

leitidon (T 128,14) leaders

The second group contains the cases of transpositions violating the line boundary: (16)

ih ni / haben diuual (T 219, 8–9) I Neg-have devil ‘I don’t have the devil’ ego demonium / non habeo

The authenticity of the transposition in this example is being confirmed by the fact that the scribe neglects the line boundary, which strongly violates the main translational principle of this text. Otherwise, whereas this example prima facie indeed points at a certain freedom in translating the Latin, there are other cases where the violation of the line boundary is not an intended means: it might have been for reasons of space or simple carelessness5 as well. The mere consistency of the translation with respect to the line-for-line principle might lead to the conclusion that every single case of its violation has to be of particular value. However, this is not always the case, for instance, when the Latin word is divided by the line boundary and the scribe was forced to decide where to chain up the translation: (17)

Intj / uûizagon sint ouh and prophets are also ‘and the prophets are dead as well’ & pro / ph&tæ mortui sunt

tote (T 219, 23–24) dead

130

Svetlana Petrova and Michael Solf

A higher degree of certainty is only possible by comparing similar instances.

2.2.3. Free translations Considering the limited freedom of the translation some regularly occurring instances are especially remarkable which show constructions with a sense more or less diverging from the original. We may assume, that the Old High German expression grants a better understanding and is therefore preferred to a word for word translation. As evidence for authentic OHG they are of a particular high value. Example (18) shows, that the range of changes may go as far as to forming of a relative clause without a Latin counterpart: (18)

hiez fon theru burg thiu of the city which is-named ‘out of the city, which was called Nazareth’ de ciuitate nazar&h

nazar&h (T 35, 15) Nazareth

In (19), the subject gotes engil is related to a new finite verb quam, which the translator preferred to the possible literal equivalent inti senonu engil gotes gistuont nah in: (19)

Inti gistuont nâh in quam thara gotes engil came there gods angel and stood near him ‘there the angel of god came and stood near them’ & ecce angelus domini st&it Iuxta illos (T 35, 32)

Finally, example (20) demonstrates a Latin participle construction formed into a temporal subordinate clause: (20)

thaz gihorta herodes ther cuning (T 39, 17) thô when that heard Herod the king ‘when Herod the king heard that’ audiens autem herodes rex

Information-structural analysis in historical texts

131

2.3. Some doubtful cases The different instances of divergences require an interpretation within the respective context. It is sure, for example, that single divergences do not always guarantee the authenticity of the whole sentence. Example (21) provides a transposition of constituents within the absolute construction. As we stated above, this kind of construction represents a syntactic loan, and therefore it is misleading to assume an authentic Old High German example in this case: (21)

In Inti Inphanganemo antuuvrte answerdat in and recieveddat ‘and having received the answer in a dream’ & responso accepto In somnis

troume (T 40, 20) dream

The structure in (22) provides us with another interesting example of the difficult approach to Latin participle constructions. Despite of the transposition of the object pronoun and the dissolution of the Latin passive infinitive the translation here follows the original accusativum cum infinitivo by risking to make the whole sentence incomprehensible: (22)

uuenan her uuolti Inan genemnitan uuesan (T 30, 29) wantedsubj heakk namedakk be whoakk he ‘*whom he would him have called’ / ‘how he would have him called’ quem uell& uocari eum

As a source for Old High German syntax this example is at least of a doubtful quality. It mixes up features of a different origin: while the subject pronoun is being inserted and the object pronoun is being transposed, oddly enough the accusative of the interrogative pronoun is being kept, and the accusativum cum infinitivo that is maintained transpires the character of a Latin sentence. To conclude, superficial decisions about the authenticity of syntactic structures in Old High German sources risk to ignore the multiple problems connected with the way the corpus is handed down to us. As has been demonstrated for the case of the Old High German Tatian, the possibility of comparing the translation with its original definitively supports the search for authentic syntactic structures.

132

Svetlana Petrova and Michael Solf

3. The theoretical issue The second type of methodological problems in need of clarification concerns the information-structural analysis in texts from historical corpora. This question itself involves two aspects to be discussed separately. The first one concerns the conceptual and terminological determination of the information-structural categories. As information structure involves some of the most controversial issues in today’s linguistic research, we have to cope with a number of different notions and terminological expressions for the main categories like topic and focus. Second, we have to look for a way of identifying these categories in texts which are available in written form only, thus offering little clues to the prosodic realization of the utterance.

3.1. The definition of information-structural categories Most researchers agree on the notion that information structure concerns the formal realization of an utterance according to the context as well as to the state of informedness of the interlocutors (cf. Lambrecht 1994: 6). It is also well known that information structure manifests itself in a variety of domains, such as phonology, morphology and syntax, though the utilization of formal means marking information-structural properties strongly differs from language to language. On a par with the great diversity of formal means used to express information-structural properties on different levels of language structure, current research on information structure is characterized by a great number of theoretical approaches to the explanation of the central notions in question which has led to a considerable amount of conceptual and terminological divergence and confusion, cf. Musan (2002: 202–208). Literally every notion or feature that has been proposed as relevant to information structure is subject to debate and exhaustive discussion in the literature, and there is no agreement on the definition of any category that has been proposed to capture information-structural properties in language. In fact, this situation points at the immense complexity of information structure as a linguistic phenomenon. An approach that promises to yield good results in capturing all relevant aspects of information packaging should therefore aim at isolating different functional dimensions of information structure into which a more precisely defined inventory of categories may be applied. Such an approach was proposed by Molnár (1993) and adopted for the information-structural analysis of both modern and extinct

Information-structural analysis in historical texts

133

languages, e.g. by Zybatow (1999) for Russian or by Disse (1998) for Biblical Hebrew. Adopting this approach, we assume that there are at least three functional layers on which properties related to information structure have to be distinguished: i.

ii.

iii.

the pragmatic status of sentence constituents, i.e. the distinction between given vs. new information in discourse as seen in the classical opposition of theme vs. rheme the predicational structure of the utterance allowing for a bipartite division of the sentence into a starting point, or topic of the utterance and a predication, or comment on this topic the distinction of focus vs. background in terms of communicative weight or relevance for the development of the discourse.

These layers of information structure are viewed to function independently in language but to interact with one another thus yielding the full picture of the information-structural shaping of an utterance. Decomposing information structure and creating a multi-layered model in the way proposed by Molnár (1993) and Krifka (2007) bears substantial consequences for the understanding of the information-structural categories and their proper description in theory. First, and contrary to a common assumption in previous literature, topic and focus need not to be mutually exclusive but are allowed to overlap as they pertain to different layers of informationstructural segmentation. Furthermore, a model like this offers a way to avoid the common practice of identifying the topic of an utterance with the contextually given information as well as of restricting focus to new information only. In this sense, the segmental approach provides the basic scheme for a more fine-grained analysis of the information-structural value of individual parts of an utterance. It allows for the assignment of a broad variety of information-structural features as well as for the evaluation of how different combinations of features influence the positioning of constituents in the clause. As shall become clear later, such a cumulative approach is of inestimable value for the information-structural analysis of sentence constituents in text from historical corpora. Nevertheless, the explanation and the catalogue of categories that make up the functional dimensions of information structure in the model given above are far from being unproblematic and unquestionable. The following elaboration gives an overview over the most relevant concepts concerning the information-structural categories as well as over some main points in the discussion on them in current research.

134

Svetlana Petrova and Michael Solf

3.1.1. Given vs. New Starting with the distinction between given vs. new information in a sentence, it may appear that the assignment of these categories is the most unproblematic one in analysing information structure in a running text of the historic corpus of a language. However, the discussion provided in Prince (1981: 225–232), among many others, points at the different aspects these notions have acquired in literature as well as at some considerable problems in applying them to natural data: – given vs. new in the sense of shared knowledge: given is the information that the speaker believes the listener already knows and accepts as true, whereas new is that information which the speaker believes that the listener does not yet know, cf. Clark and Haviland (1977: 4) – given vs. new in the sense of cognitive activation/salience: “Given (or old) information is that knowledge that the speaker assumes to be in the consciousness of the addressee at the time of the utterance. So-called new information is what the speaker assumes he is introducing into the addressee’s consciousness by what he says.”, cf. Chafe (1976: 30) – given vs. new in the sense of predictability/recoverability: “An element in a sentence represents old, predictable information if it is recoverable from the preceding context; if it is not recoverable, it represents new, unpredictable information.”, cf. Kuno (1978: 282–283) Prince (1981) shows that the definitions of ‘given’ vs. ‘new’ proposed in the literature – though being not completely independent – bear significant differences in various important points. In comparing the concept of Clark and Haviland (1977) with the one of Chafe (1976) it becomes obvious that both notions explain the given/new-distinction in terms of cognition since both ‘activation’ and ‘knowledge’ concern the cognitive state of the interlocutors. Nevertheless, a basic difference appears: an entity that is not explicitly pre-mentioned but inferable from the common knowledge of the interlocutors counts as given in the sense of Clark and Haviland (1977) but as new, i.e. known but not necessarily activated, in the concept of Chafe (1976). Accordingly, this yields two different interpretations of one and the same constituent (your father) in a sentence like (23)6:

Information-structural analysis in historical texts

135

(23) I saw your father yesterday. a. I saw your fathergiv yesterday given=known, i.e. inferable according to Clark and Haviland (1977) b. I saw your fathernew yesterday new=not activated according to Chafe (1976) At the same time, as Prince (1981: 229) also remarks, the expression a twoheaded man in (24) should gain – according to Chafe’s definition – the same status as the expression your father in (23b), a situation that is clearly counter-intuitive: (24) I saw a two-headed mannew yesterday new=not activated according to Chafe (1976) Another problem arises with respect to the uniformity of formal means expressing givenness in language. According to Clark and Haviland (1977) the referent of the definite expression the beer both in (25a) and (26a) is given by virtue of its status as being known via contextual preestablishment in (25) and inferentiality in (26), respectively. Nevertheless, the ‘given’ referent can be pronominalized only in the case of (25b), cf. the ungrammaticality in (26b): (25) a. We got some beer out of the trunk. The beer was warm. b. We got some beer out of the trunk. It was warm. (26) a. We got some picnic supplies out of the trunk. The beer was warm. b. We got some picnic supplies out of the trunk. *It was warm. Following this argumentation, Prince arrives at the conclusion that the traditional dichotomy of ‘given’ vs. ‘new’ is evidently too narrow to capture the fine but nevertheless significant differences regarding the activation state of referential expressions in natural discourse. As a result, she proposes to suspend the old dichotomy and exchange it for a model of a scalar representation of features in which ‘given’ and ‘new’ are only the two endpoints in a wide range of subcategories. Prince’s proposal involves a triple of notions under the core idea of “assumed familiarity”: i) new, ii) inferable and iii) evoked. Each of these pragmatic states is subdivided into deeper categories. So i) the notion of

136

Svetlana Petrova and Michael Solf

‘newness’ falls into the subcategories of i-i) ‘brand-new’ items whose existence in the mental world of the listener is being created at the time of the utterance, and i-ii) ‘unused’ items which are ‘known’ to the hearer but not activated at the time of utterance. The introduction of brand-new referents is itself divided into i-i’) ‘anchored’ vs. i-i’’)‘unanchored’ due to the fact that a brand-new item may be established in the context as linked to another entity – an anchor – that is not brand-new itself. This occurs e.g. in the case of modified NPs like a person I know or a girl I work with. On the other pole of the scale, Prince considers iii) the category of entities already ‘evoked’ in the discourse model. These categories are either iii-i) evoked textually, i.e. they are explicitly mentioned in previous context, or they are iii-ii) evoked situationally, e.g. when they refer to the interlocutors themselves or to referents which are salient in the communicative situation. In between, Prince distinguishes an additional, and more complex, category of ii) ‘inferable’ referents whose mentioning in an utterance may logically be inferred by the listener according to the entities already evoked in the context. Here, one subclass is represented by the so-called ii-i) ‘containing inferables’, i.e. by referents staying in a set-member relationship (picnic supplies-beer), and another one is formed by the so-called ii-ii) ‘noncontaining inferables’, i.e. by referents staying in an analogy relation to one another (bus-driver; party-music etc.). Building upon this new taxonomy of the informational state of referents, Lambrecht (1994: 74–113) accounts for some formal means typically correlating with them: i) presence vs. absence of accent; ii) pronominal vs. lexical coding; ii) definite vs. indefinite marking. According to the different kinds of combination of features in the actual realization of referents of these pragmatic classes, Lambrecht maps these categories onto an ordered relation, a so-called scale of identifiability of referents presented with some supplementary terminology in (27), cf. Lambrecht (1994: 109): (27)

i) ii) iii) iv) v) vi) vii)

unidentifiable/brand-new unindentifiable anchored/brand-new anchored inactive/unused textually accessible situationally accessible inferentially accessible active/given

This scale of identifiability correlates with the formal explicitness of referents in discourse. The lower a referent is to be classified according to the

Information-structural analysis in historical texts

137

scale in (27), the less explicit its formal representation is in real communication. Turned the other way round, the growth of formal explicitness in the representation of a referent points at a higher need to activate or reactivate the identification state of the referent. Furthermore, Lambrecht (1994: 113–115 and 165) refers to a substantial cross-relation between the identification state of a referent and its appropriateness to provide the topic of an utterance. He argues that the lower a referent is mapped on the scale of identification, the more it is likely to act as the topic of an utterance. In doing so, Lambrecht tangles one of the most problematic and difficult questions in information-structural research, the relationship between givenness and topicality in language, a question to be discussed in the following section.

3.1.2. Topic vs. comment According to the multi-layer approach on information structure proposed by Molnár (1993) and advocated in section 3.1 above, the topic vs. comment distinction refers to the possibility to identify a bipartite division of the predication in terms of an item used as the starting point, or topic and another one providing the predication, i.e. the comment on this topic. A classical form of this kind of predicational separation occurs in copular constructions or generic utterances of the kind given in (28) and (29) for which the basic definition provided by Hockett (1958: 201) applies: “the speaker announces a topic and then says something about it”: (28)

[The house]T [is green.]C

(29)

[Trees]T [have green leaves.]C

Apart from this very basic notion originally intended to describe the constituent structure of predicative constructions, a range of competing concepts have been proposed making topic one of the most problematic terms to work with in information-structural analysis. First of all, different syntactic strategies have been discussed as a means of marking the topic of the sentence. Constructions like Left dislocation (Frey 2005), Hanging topic or ‘As for…’ are among these. The Prague School of Functional Sentence Perspective applied a more general view by assuming the topic to be the first element in a sentence (Firbas 1966). This notion was also adopted by Halliday (1967) to cover all sorts of ‘topical-

138

Svetlana Petrova and Michael Solf

ized’ elements in a clause. For German, moving a constituent to the ‘Vorfeld’, i.e. to the single position before the verb in root clauses has also been discussed with respect to topicality. And although this type of dislocation has been shown to apply for other information-structural categories as well (Axel in this volume), studies like Molnár (1993) observe a strong tendency of topical material to appear clause-initally in German. Arguments against this purely configurational account on topics come from alternative approaches to the explanation of this notion, e.g. from the aboutness-concept of Reinhart (1981) which is one of the most widely accepted ones in the literature today. Reinhart (1981) adopts the basic notion of pragmatic ‘aboutness’ as a distinctive feature of topichood and develops tests triggering typical topic readings. According to these, the item identified as the topic of the sentence acts as the referent X to be inserted in a proposition of the kind A says about X that X…or as for X, X is… Analysing the appropriateness of utterances gained from the insertion of different sorts of NPs in these tests, Reinhart shows that e.g. quantified NPs “are often hard, and sometimes impossible, to interpret as topics” (1981: 65). This behaviour she explains by virtue of the fact that the quantified NPs, failing to pass the topic tests, do not allow for an interpretation as referents. This is in clear contrast to universally quantified NPs denoting sets of referents and being available to act as topics. According to this, Reinhart restricts topichood to referentiality, and compares sentence topics with the referents acting as the entries of a subject catalogue in a library under which the propositions made about them are stored. Topics are in the sense of Reinhart “referential entries under which we classify propositions“ (1981: 80). In this way, a sentence topic is allowed to appear anywhere in the sentence. Or put in different terms, a sentence-initial constituent which is not referential fails to classify as a topic. Another widely accepted account on topichood is the one proposed by Gundel (1988) who relates pragmatic ‘aboutness’ to the conditions of givenness and accessibility. Departing from a definition of the topic-comment structure as presented in (30): (30)

Topic definition: An entity, E, is the topic of a sentence, S, if in using S the speaker intends to increase the addressee’s knowledge about, request information about, or otherwise get the addressee to act with respect to E. (1988: 210)

Information-structural analysis in historical texts

139

Comment definition: A predication, P, is the comment of a sentence, S, if in using S the speaker intends P to be assessed relative to the topic of S. Gundel arrives at the conclusion that in terms of felicitous communication, shared familiarity on the topic expression appears to be a necessary precondition of topichood: „An entitiy E can […] serve as a topic […] if, both speaker and addressee have previous knowledge of or familiarity with E“ (1988: 212). According to this definition, new referents are banned from acting as sentence topics, an option held open by Molnár (1993). Furthermore, the concept of Gundel aims at identifying givenness and accessibility with topichood though no identification of newness with comment is implied; cf. Gundel (1988: 212). This simply means that given referents are equally allowed to be part of the topic or the comment of a sentence, while no operational methods are proposed to decide which of more than one given or accessible referents in a clause should be taken as the topic of the sentence. Still another account on topic is promoted by Chafe (1976) who argues for a basic differentiation between topics in topic-prominent languages like Mandarin and topics in languages like English. For the latter, he criticizes the definition of Halliday (1967) by pointing to apparent contrastive effects paired with the function of topicalization in English in sentences of the type This play John saw yesterday (Chafe 1976: 49). Consequently, Chafe proposes to suspend the term topic for such constructions in English. Instead, he pays more attention to the first type of topics, i.e. the so-called Chinese-style topic which he assumes to represent the prototypical case of topichood. However, analysing the interpretation of sentences containing typical Chinese-style topics, as the one in (31): (31)

nèi-xie shùmu shù-shƝn those tree tree-trunk

dà. big (Chafe 1976: 50)

Chafe concludes that the role of the topic is not so much to say what the sentence is about, but to “limit the applicability of the main predication to a certain restricted domain” (Chafe 1976: 50). As not only nominal referents but also adverbials of time and location are allowed in this function, Chafe extends the notion of topic to the overall concept of ‘frame-setting’ used to subsume all sorts of elements providing the “spatial, temporal, or individual framework within which the main predication holds“ (1976, 50).

140

Svetlana Petrova and Michael Solf

Comparing Chafe‘s notion on topic as a set of different frame-setting elements with the aboutness-concept of Reinhart (1981) and the familiarity-concepts of Gundel (1988) described above, the question arises whether both frame-setting topics and aboutness/familiarity-topic represent different parts of one and the same phenomenon or whether these are completely different phenomena to be held apart in information-structural analysis. In current research, the former option seems to prevail. Analysing different aspects of topic-comment constructions in German, Jacobs (2001) promotes the view that frame-setting and ‘usual’ topics might represent “two different but similar prototypes of T[opic]C[omment]” (2001: 658) thus making topicality a polysemous category. In line with this view, Maienborn et al.7 explore aboutness-topics and frame-setting as two different notions of sentence topichood which share some important syntactic, semantic and pragmatic properties with respect to the rest of the sentence as well as to the larger discourse.

3.1.3. Focus vs. background The third functional layer within the model of information packaging, the layer of the focus vs. background distinction, is subject to no less controversy than the other two domains discussed above. There is some common agreement on the explanation of focus in terms of informational relevance and emphasis. This consideration implies two main aspects, one related to the contents of the utterance and one related to its formal realization. According to the first one, it is intuitively appealing to suggest that not all parts of the information exchanged between the interlocutors are equally important or relevant with respect to achieving the communicative goals of the utterance. Secondly, the notion of focus in terms of emphasis relates to the formal prominence of certain parts over the rest of the utterance, most commonly manifested in terms of prosodic prominence, e.g. main stress, phrasing and intonation in so-called intonational languages. According to many standard assumptions, the informational weight of focused constituents is explained in terms of the distinction between presupposed, or given vs. newly added, or asserted information in discourse (cf. the overview in SzendrĘi 2004: 230). The difference between the informational relevance of parts of the utterance is best illustrated by question-answer pairs where the newly added, asserted information may be precisely identified as that part of the answer which corresponds to the whphrase in the preceding question. On the formal side, the newly added in-

Information-structural analysis in historical texts

141

formation is marked by means of prosodic prominence illustrated by capitalizing the word carrying sentence stress, cf. (32): (32)

Q: What did John eat? A: John ate PIZZA. (cf. SzendrĘi 2004: 230)

According to the high operational value of question-answer pairs for focus detection, most focus theories stick to exploring this phenomenon on the basis of such discourse examples only. However, it is misleading to identify focus with ‘newness’ in terms of the informational status of the referents. As many examples in previous literature have shown, information provided in the answer to a preceding wh-question prosodically behaves the same way as the new one in (32) even if it refers to entities which are contextually given or inferable, cf. (33): (33)

Q: Who did Felixi praise? A: a. Hei praised HIMSELFi. b. Hei praised YOU. c. Hei praised his BROTHER. (cf. Lambrecht 1994: 260)

With respect to this property, already Halliday (1967) warns against identifying focus with pragmatic newness: “What is focal is “new” information; not in the sense that it cannot have been previously mentioned, although it is often the case that it has not been, but in the sense that the speaker presents it as not being recoverable from the preceding discourse (Halliday 1967: 204). The same kind of correlation between focus and given information regularly emerges in cases of contrastive focus. Here, a part of the utterance gains informational relevance over the rest of the sentence by virtue of the fact that it is placed in a relation of contrast to another item which is either explicitly given or implied as part of the context. Consider (34) in which the answer picks up referents selected from a set of alternatives explicitly mentioned in the question: (34)

Q: Who did she invite to her birthday party, Johni or Peterj? A: She invited JOHNi. PETERj she cannot stand.

Lambrecht (1994: 209–218) provides some more considerations showing that the identification of focus with that portion of the answer which con-

142

Svetlana Petrova and Michael Solf

veys the ‘new information’ is too simplifying and inaccurate. In his analysis of the example in (35): (35)

Q: Where did you go last night? A: I went to the MOVIES. (cf. Lambrecht 1994: 209)

he argues that the new information in the answer is neither identical with the referent of the expression the movies nor to be restricted to any missing referent actually. What is new is in fact the relationship between the denotatum of the expression the movies and the proposition made explicit in an answer of the kind in (35’): (35’)

the place where I went last night was the movies.

Following this argumentation, focus is the device marking the establishment of a new kind of relation between a denotatum and the rest of the proposition, thus creating a “new state of information in the addressee’s mind” cf. Lambrecht (1994: 210). In this sense, a focus relation is guaranteed even in sentences operating with given material only, as is the case in (36c) of the famous syllogism or in (34) above: (36)

a. Every human is mortal. b. Socrates is a human. c. Socrates is mortal.

Another approach avoiding the definition of focus in terms of newness is found in the model of Rooths (1985). His meanwhile broadly accepted method involves a definition of focus in which the assignment of a formal focus feature on a constituent signals that it has been chosen from a set of alternatives and that is inserted in an open proposition.

3.2. Access to prosodic information in OHG? So far, the problems of the overall definition of information-structural categories have been addressed. Relating this issue to the informationstructural analysis of data from the historic corpus of a language, still more difficulties arise. One major problem comes from the fact that we have to deal with a quantitatively restricted amount of data. Accordingly, no additional data elicitation is possible in this case, i.e. no tests for checking the

Information-structural analysis in historical texts

143

appropriateness of sentences as well as no negative evidence is available. This simply means that the information-structural analysis of corpus data has to do without some basic methods most widely applied in synchronic research on information structure. Another problem arises from the fact that the data to be analysed here is attested in written form only and thus provides no reliable access to prosodic information which is crucial to the detection of a number of phenomena related to information packaging as accent/de-accenting, phrasing and intonation. The previous literature gives sporadic accounts on indications drawn from the graphical representation of the manuscripts, though in sum, these basically concern the macro-structural dimension of text organization and episode subdivision, and less the level of prosody. For instance, we have quite reliable access to the separation of chapters (so-called ‘fits’) in some texts of the early Germanic corpus since these are either marked by Roman numbers as in the manuscripts of the Old English Beowulf and the Old Saxon Heliand, cf. Bästlein (1991), or introduced by individual chapter titles as in the Old High German Gospel Harmony by Otfrid. Simmler (1998) also shows that in the Old High German Tatian translation, numerous concordance notes attached to the Latin (and sometimes to the Old High German) part of the text give evidence on text organization and episode-division, next to the use of capital letters at the beginning of new episodes. Such clues are not available on the micro-structural level. In fact, the most representative of the Old High German manuscripts contain a number of accent diacritics. According to recent observations, certain information on prosody can be inferred in the texts of Otfrid and Notker, cf. Kleiber (2004: 119–142) and Fleischer (in this volume). However, the functional explanation of accent placement in Tatian is more than questionable. Analysing the numerous accent diacritics found here, Fleischer (in this volume) arrives at the conclusion that these may be assigned different functions which strongly interfere with each other thus leaving little ground for any substantial conclusions on prosodic prominence. Fleischer (in this volume) also points at spacing between words as another potential clue to prosodic information in Old High German manuscripts. The absence of a blank space between words may be interpreted as a reflection of the fact that these words were pronounced together, without a phonological break between them. Consider the different representation of the personal pronoun in (37):

144 (37)

Svetlana Petrova and Michael Solf

eno thu bistu mera / unsaremo fater iacobe father Jacob verily you are-you more ourdat ‘And you, are you more than our father Jacob?’ numquid tu maior es / patre nostro iacob. (T 131, 18–19)

The graphical representation of cases like the second occurrence of the personal pronoun in (37) allows for the conclusion that a functional word is de-accented and cliticized to a main word as well as that no prosodic phrase boundary intervenes between these words. On the contrary, large spaces between words have been analysed as signalling a speech pause or a phrasing boundary, cf. also Masser (1997: 58). However, this is a catalogue of potential indications whose relation to prosody cannot be claimed without any doubt. Moreover, the proper evaluation and systematic analysis of this scattered evidence on the textorganization and the prosodic realization in corpus data depends on the availability of precise diplomatic editions reflecting these properties of the graphical representation of the texts. In numerous cases, a comparison with or an exploration of the manuscripts is necessary. These problems clearly point at the fact that for the proper informationstructural analysis of sentences from historical corpora, context interpretation and discourse organization are the most reliable source to clues for the information-structural value of sentence constituents. These aspects shall be taken as a basis for a model of information-structural analysis on sentences from the Old High German Tatian proposed in the following section.

4. A proposal for the information-structural analysis of corpus data 4.1. The main scheme Due to the lack of a proper definition of the information-structural categories which can be applied to the data immediately, and given the difficulties arising in texts of the historic corpus of a language, we propose a novel approach based on the multi-layered model on information structure presented in section 3.1 above. We maintain the notion that informational status, topic-comment and focus-background are basic distinctions which interact with one another but nevertheless reflect different aspects of information packaging. Therefore, the attributes belonging to these different distinctions have to be kept apart in real data analysis. However, we dispense with the idea of assigning the main categories of topic and focus

Information-structural analysis in historical texts

145

directly; rather, we aim at assigning to the sentence constituents a possibly wide range of features which have been connected with information structurae in the previous literature. In doing so, we are able to draw conclusions about the interdependence between certain features or combinations of features on individual constituents and their positioning in the clause. Departing from the analysis of the basic notions in informationstructural research, we distinguish the following attributes and features considered as constitutive for the main information-structural categories.

4.1.1. Informational status of discourse referents: given, new and accessible Given the argumentation in Prince (1981) discussed in section 3.1.1 above and some further considerations in Dik (1989: 268–270) and Lambrecht (1994), we agree on a scalar representation of features attested to the informational status of discourse referents. The notions ‘given’ and ‘new’ are viewed as the endpoints of the scale. However, for the sake of unambiguous applicability, these two notions are bound to textual pre-establishment: ‘given’ is restricted to expressions referring to explicitly pre-mentioned referents, while ‘new’ covers referents introduced to context for the first time. The anaphoric expressions that serve as linguistic correlates of givenness can be represented on a scale of explicitness starting with zeroanaphors, personal and demonstrative pronouns and ending up with full expressions like full NPs, proper names and epithets. Consider the following text-example adapted from Dik (1989: 271): (38)

Yesterday I got a phone call from the tax inspectori. Hei/ the mani/ the jockeri wanted me to come to hisi office and hei /øi gave me the impression that I was in for some trouble.

Contrary to that, the term ‘new’ covers notions that are not explicitly preestablished in previous context. Typically, a new discourse referent is introduced explicitly following one of the strategies outlined by Dik (1989: 268): – meta-linguistic information: I’m going to tell you a story about X – as an object or second argument in sentences with transitive verbs: In …, we saw X

146

Svetlana Petrova and Michael Solf

– as the subject of an existential or presentation construction of the type Once upon a time, there lived/was … which is typical for textopening sentences – with verbs of motion, denoting “appearing on the scene” (Dik 1989:268): Suddenly, right before our eyes, X appeared … Between the states of ‘given’ and ‘new’ as polar values, a whole range of categories is situated sharing the property of not being previously mentioned but nevertheless staying in a certain relation of relevance either to the communicative situation as a whole or to entities already established in discourse. These categories are subsumed under the label of ‘accessible’ entities due to the fact that they need not be introduced explicitly but are semi-active at the time of the utterance and are thus available for proper reference. We distinguish the following cases of accessibility: – expressions referring to the interlocutors, e.g. the deictic pronouns of the first and second person singular/plural as well as full NP, proper names etc. that are coreferential with them – anchoring: entities that are introduced in a certain relation to already activated referents; a friend of mine …, a person I work with … – bridging: entities which stay in part-whole relationship or in a relation of analogy to an already pre-established referent; John gave a party last week, but the music was awful … – shared familiarity: entities belonging to the common knowledge of the interlocutors like real names, proper names, etc.

4.1.2. Predicational separation: topic vs. comment Bearing in mind the different accounts on topicality and the lack of a commonly accepted definition of this term, we decompose this notion with respect to all features – both conceptual and formal ones – viewed to be constitutive for topicality in language: – givenness/accessibility of referents: to be acquired from the layer of informational status, see above – aboutness: whether or not any constitutent of the sentence fits to replace X in the topic-test A says about X that X … – definiteness: whether the referential expressions involve a definite or indefinite reading

Information-structural analysis in historical texts

147

– syntactic realization in the clause: whether any of the constituents in the sentence is involved in a syntactic construction associated with topicality (e.g. Left dislocation LD, hanging topic HD, ‘As for …’, ‘Vorfeld’ etc.) The co-occurrence of all features on one and the same constituent would yield an optimal topic candidate; the less features apply on a constituent the less it fits to be identified as the topic of the sentence. This method facilitates a clear decision on the assignment of the topic category in examples like the following small discourse: (39)

a. senonu tho uuas behold there was

man man

In hierusalem.’ / in Jerusalem

b. thes namo thisgen name

uuas giheizzan simeon, / c. Inti ther man uuas reht Inti gotforht.’ was named Simon and this man was just and devout ‘a. And behold, there was a man in Jerusalem b. whose name was Simeon, c. and this man was just and devout’ (Luke 2, 25) a. & ecce homo erat In hierusalem.’/ b. cui nomen simeon,/ c. & homo iste iustus & timoratus.’ (T 37, 23–25) Here, the expressions thes namo ‘whose name’ in (39b) and ther man ‘this man’ in (39c) refer back to the discourse referent man ‘a man’ introduced in the initial conjunct (39a). The referent namo in (39b) is an instance of anchoring since it is introduced in relation to an already mentioned entity, and ther man in (39c) is a full anaphor to ‘a man’. Both anaphoric expressions are referential, definite and realized clause-initially in the corresponding conjuncts. Therefore, these constituents can be identified as topics, while the remaining part of the utterances provides the comment on these topics, respectively. This interpretation satisfies the intuition that the sentences in (39b–c) predicate on a given entity by saying something about it, e.g. by assigning a property to it. The first conjunct (39a), however, is a case of a presentational construction establishing the topic of the subsequent utterances but displaying no topic-comment structure itself (cf. Sasse 1995: 4–5). Moreover, the referent ‘a man’ in (39a) doesn’t bear any significant topic features: it is referential but new, indefinite and occurs in a late position in the clause. For these reasons, no topic assignment applies to ‘a man’ in (39a). However, the method described above does not allow for an unambiguous solution in each case. Consider instances like (40) in which the criteria

148

Svetlana Petrova and Michael Solf

announced above equally apply to more than one discourse referent in the clause: (40)

quad Iru ther engil (T 28, 3−15) said herdat this angel ‘Then the angel said to her’ (Luke 1, 30) & ait angelus ei

The sentence is nested within a dialogue sequence in which the expressions Iru and the angel relate to explicitly pre-established referents whose relative order in the sentence is shifted in OHG against the Latin original. Although crucial criteria for topicality like givenness and definitness apply for both of the referents involved, it is difficult to decide which of them qualifies as the aboutnesss-topic of the utterance. Broadly speaking, the sentence is about both of them, though in fact the application of the topictest A says about X that X … does not yield a proper paraphrase of the contents of the sentence. A more appealing interpretation is the one considering the sentence to predicate on the event itself, i.e. on the fact that the conversation between the referents is carried further. In this case, the given referents are involved in a new state of affairs but are not the topic of the clause. Consequently, the most decisive issue in assigning the category topic to a constituent bearing any of the features relevant for topics involves the presence of a topic-comment division in the utterance: only if the sentence allows for a categorical reading (Sasse 1995: 4–5), i.e. if it announces a unit as the starting point of an utterance and then makes a statement on this unit, is it reasonable to assign a topic category to any constituent bearing any of the features relevant for topics.

4.1.3. Informational relevance: focus vs. background In section 3.1.3 we argued for a pragmatically rather than prosodically defined notion of focus. Following the discussion in current research, we defined focus as that part of an utterance that provides the most relevant information in a particular context as opposed to the (not so relevant) rest of information making up the background. There are at least two ways for a part of an utterance to gain information-structural relevance over the rest of the sentence:

Information-structural analysis in historical texts

149

– it provides ‘new’ information in the sense that it is either requested in a preceding wh-question or necessary to develop the discourse; or – it stands in a relation of contrast to another constituent in the discourse. According to this, we distinguish the types of new-information focus (nif) and contrastive focus (cf), respectively. We also assume that newinformation focus and contrastive focus are not mutually exclusive but may apply within one and the same domain. As we also referred to in the above discussion, the term ‘new-information focus’ is somewhat misleading as it should not equate focus with ‘newness’ but stand for the new type of relation that emerges when an open proposition is saturated by inserting the missing material out of a set of possible alternatives, regardless of the fact whether this material is known, given, inferable, etc. to the interlocutors. Related to referents, ‘new’ is only a category on the level of informational status and not a constitutive feature of focus. The assignment of the category focus to a particular part of the proposition points to the material which promotes the establishment of this new relation between referents in the addressee’s mind. In this respect, the literature refers to the ‘focus domain’ of a sentence, i.e. to the different extension of the focus material in the utterance. Typically, the following cases are distinguished: – the focus domain spreads over the entire proposition, e.g. in all-new sentences opening a text discourse in so-called thetic sentences as the one in (41), cf. Sasse (1995: 4–7): (41)

Q: What’s that smell? A: [The KITCHEN is burning.]F

– the focus domain comprises the whole VP (42a), a single constituent within the VP (42b), or a part of a constituent (42c) respectively. The following examples show a different assignment of the focus domain for one and the same structure depending on the preceding context: (42)

a. b.

Q: What is Peter doing? A: He [is reading a PAPER.]F Q: What is Peter reading? A: He is reading [a PAPER.]F

VP-focus XP-focus

150

Svetlana Petrova and Michael Solf

c.

Q: What sort of paper is Peter reading? A: He is reading a paper [on SCIENCE]F

focus on a part of XP

– focus spreads over a sub-constituent, i.e. a prefix or a morphological part of a word which is not necessarily identical with the syllable carrying lexical stress, cf. (43): (43)

Q: Is she Chinese? A: No, she is [JA]Fpanese. (cf. Lambrecht 1994: 241)

While contrastive focus is located on the constituent viewed to bear the contrastive relation to another constituent, the determination of the domain of new-information focus reveals to be more problematic in our analysis. As was shown in the discussion on focus in section 3.1.3 as well as in the examples in this section, the identification of the focus domain of the utterance in current literature is based on (constructed) question-answer pairs. In striking contrast to that, no such methods are established for the identification of focus in sentences from natural running discourse as is the one in our analysis on Old High German data. However, we consider it possible to adopt the operational methods of focus detection established for dialogue sequences to texts from narratives, reports, etc. Let us assume that in line with the quaestio-theory by Klein and von Stutterheim (1992) each sentence of a running discourse is the adequate answer to an implicit question. In defining this question on the basis of the preceding discourse, one can identify those parts of the sentence that carry further the discourse and thus constitute the focus domain in sentences of running narrative texts.

4.2. Examples The model proposed to capture the multitude of information-structurally relevant features of sentences constituents in data from the historic corpus of a language is demonstrated on the basis of sentences from the Old High German Tatian translation based on the edition of Masser (1994). For reasons described in section 2.1 above, the investigation primarily addresses sentences in which the Old High German text deviates from the sentence structure of the corresponding Latin original.

Information-structural analysis in historical texts

151

Below, some examples concerning the application of the model proposed as well as some comments to the assignment of features to the constituents are given. Example 1 Lat. Fuit in diebus herodis regis / iudeĊ quidam sacerdos / […] / & uxor illi de filiabus aaron/& nomen eius elisab&h, / erant autem iusti ambo ante deum OHG uuar [sic!] In tagun herodes thes cuninges / Iudeno sumer biscof /[…] / Inti quena Imo fon aarones tohterun / Inti ira namo uuas elisab&h / siu uuarun rehtiu beidu fora gote (T 25, 2) ‘There was in the days of Herod, the king of Judea, a certain priest […] and his wife of the daughters of Aaron, and her name was Elizabeth. And they were both righteous before God’ Sentence siu uuarun rehtiu beidu fora gote they were righteous both before God Lat. OHG informational status TCS aboutness definiteness syntactic realization FBS nif cf

siu giv top ab def init bgr

erant uuarun

autem

iusti rehtiu

ambo beidu

ante fora

deum gote acc

comment def noninit focus nif

Analysis: In the OHG sentence, the subject pronoun siu-3pl ‘they’ is added against the Latin original and placed at the beginning of the clause, in front of the finite copula. The added constituent refers to entities already preestablished in the discourse. The application of the topic test A says about X that X… fits well to paraphrase the contents of the sentence and supports the interpretation of the constituent siu as the aboutness-topic of the sentence. All other features of topicality apply for siu as well, and we assign the category topic to it. The focus of the sentence covers the VP as the new information answering the implicit question “What about Zacharias and Elizabeth?”

152

Svetlana Petrova and Michael Solf

Example 2 Lat. Et pastores erant in regione eadem.‘ / uigilantes & custodientes uigilias noctis / supra gregem suum OHG uuarun thô hirta In thero lantskeffi.‘ / uuahhante Inti bihaltante nahtuuahta / ubar ero euuit, (T 35, 29−31) ‘There were in the same country shepherds living out in the fields, keeping watch over their flock’ Sentence uuarun thô hirta In thero lantskeffi were then shepherds in that region Lat. OHG informational status TCS aboutness definiteness syntactic realization FBS nif cf

Et uuarun

pastores thô

erant hirta new

in in

indef noninit focus nif

regione thero

eadem landskeffi giv

def nononit

Analysis: The sentence opens a new paragraph marked by concordance notes on the margins of the Latin part of the text as well as by capitalizing of the initial letter, cf. Masser (1994: 85). A new referent hirta ‘shepherds’ is introduced to the context; the expression denoting the new referent is placed in front of the finite copula in the Latin part and shifted after the finite verb in the Old High German sentence. Although the sentence contains the given and definite expression in thero landskeffi ‘in the same country’, the general understanding of the sentence in this context does not support the intuition that it is meant not to provide further information on the region itself. The insertion of this referent into the topic-test A says about X that X … does not yield a proper paraphrase of the sentence either. Following these considerations, we decide not to assign topic to in thero landskeffi. Furthermore, the instance is a typical case of a presentational, all-focus sentence in which no topic-comment division applies. Thus, no topic category is assigned at all. As an introductory sentence of a new sub-episode, the entire sentence is focal.

Information-structural analysis in historical texts

153

Example 3 Lat. Haec locutus sum uobis / ut In me pacem habeatis / In mundo presuram habebitis OHG thisiu sprahih íu / thaz in mir habet sibba / In theru uueralti habet ir thrucnessi (T 290, 7–9) ‘These things I have spoken to you, that in Me you may have peace. In the world, you will have turbulation’ Sentence thaz in mir habet sibba that in me have-2Pl peace Lat. OHG informational status TCS aboutness definiteness syntactic realization FBS nif cf

ut thaz

In in

me mir giv

def

cf_1

pacem habet

habeatis sibba new

def

cf_2

Analysis: In this adverbial clause the direct object sibba ‘peace’ is shifted to the postverbal position in Old High German against the corresponding Latin structure. Two contrastive pairs are involved in this small discourse: one between in me and In theru uueralti, and another one between sibba and thrucnessi. For this reason, in me and sibba are analysed as two different instances of contrastive focus. The precise identification of the new information focus is difficult, though the most general implicit question fitting to the context is “Why have I spoken to you these words?” pointing at the entire subordinate clause as focal. Further subdivisions within this wide focus domain, e.g. taking the role of the finite verb as re-structuring device, are also possible but will be left out here.

5. Summary and further discussion Pursuing the new perspectives which information structure reveals for the explanation of word order variation in early Germanic languages, we face the demand for a precise and systematic description of the informationstructural value of sentence constituents and its impact on the syntactic

154

Svetlana Petrova and Michael Solf

realization of these constituents in texts of the earliest Germanic records. Two main problems connected with this issue have been discussed: the possibility of isolating authentic material allowing for generalizations on true Germanic syntax as well as the methodology of applying criteria for information-structural description to data available in written form only. Concerning the first question in the light of investigating the role of information structure in OHG, it has been proposed to utilize the OHG Tatian translation as one of the earliest and largest data collections from this period and to restrict the analysis to a corpus of sentences deviating from the structure of the Latin original. Thus, the possible influence of the syntax of the Latin original on the structure of the Old High German translation is minimized. Furthermore, we reflected the theoretical difficulties which informationstructural research on historical data involves. We provided an overview of the controversial treatment of any information-structural category in current research and addressed the problems of assigning informationstructural categories in written texts. It turned out that it is impossible to start from a clearly defined notion on the main categories in informationstructural research and that contextual information and discourse organization are the most reliable sources for pragmatic features in historic data. According to these considerations we proposed a model based on the collection of prototypical features relevant for the information-structurally categories as shown on some selected instances of the Old High German Tatian translation. The implementation of this model to a database and its statistical evaluation can help to come to grips with the enormous variation in the domain of word order and verb placement in early Germanic. It also allows for the detection of different combinations of features favoring special word-order patterns, and provides a way of looking for the distribution of patterns according to discourse-structural or other organizational principles. Furthermore, significant differences in the quantitative relations of competing patterns can be revealed, e.g. concerning different dialects, genres, or scribes. In this way, the statistical evaluation of a corpus enriched with the features of the proposed information-structural scheme is of enormous value for detecting some ordering principles in an apparently unordered system and for localizing domains in which the establishment of general rules first applied.

Information-structural analysis in historical texts

155

Acknowledgments The present paper is a revised version of a talk given in the Workshop ‘Information Structure and Language Change’ in Berlin (Germany), September 30th–Oktober 1st, 2005 organized by Project B4 of Collaborate Research Center 632 ‘Information Structure: the linguisitc means for structuring utterances, sentences and texts’ of Potsdam University and Humboldt University Berlin. We thank all participants of the workshop for questions and discussions, and an anonymous reviewer for comments on an earlier version of the paper.

Notes 1.

2.

3. 4. 5. 6. 7.

We refer to the work of project B4 (principal investigators Karin Donhauser and Roland Hinterhölzl) as part of Collaborative Research Center 632 “Information Structure” at Humboldt University Berlin launched in July 2003 and funded by the German Research Foundation (DFG). For a more detailed description on the issues of research see http://www.linguistik.hu-berlin.designato.de/sprachgeschichte/ forschung/ informationsstruktur/index. php or http://www.ling.uni-potsdam.de/sfb/. Here we find on four pages alone eleven violatons of the translation principle described above. A number of seven instances, as mentioned by Dittmer and Dittmer (1998: 24–25), who did not examine the whole text, is proven to be too low, cf. Fleischer et al. (2008: 220), who account for 46 instances in total. Sonderegger (1965) provides a similar attempt to classify the different kinds of syntactic divergences in Old High German translations. For the distribution of the Old High German subject pronoun cf. Eggenberger (1961). As Masser (1991) seems to suppose for scribe İ. The example is taken from Chafe (1976: 30). Cf. the homepage of Project 9 at the Centre for General Linguistics, Typology and Universals Research (ZAS) Berlin http://www.zas.gwz-berlin.de [20.07.2006].

156

Svetlana Petrova and Michael Solf

References Bästlein, Ulf Christian 1991 Gliederungsinitialen in frühmittelalterlichen Epenhandschriften. Studie zur Problematik ihres Auftretens, ihrer Entwicklung und Funktion in lateinischen und volkssprachlichen Texten der Karolinger- und Ottonenzeit. Europäische Hochschulschriften. Reihe I. Deutsche Sprache und Literatur; 1167. Frankfurt a. M.: Peter Lang. Bean, Marian C. 1983 The Development of Word Order Patterns in Old English. Croom Helm Linguistic Series. London: Croom Helm. Behaghel, Otto 1932 Deutsche Syntax. Eine geschichtliche Darstellung. Band IV. Wortstellung. Periodenbau. Heidelberg: Carl Winter. Chafe, Wallace L. 1976 Givenness, Contrastiveness, Definiteness, Subjects, Topics, and Point of View. In Subject and Topic, ed. Charles N. Li, 25–55. New York: Academic Press. Clark, Herb H. and Haviland, Susan E. 1977 Comprehension and the given-new contrast. In Discourse Production and Comprehension, ed. Roy O. Freedle, 1–40. Hillsdale, N.J.: Lawrence Erlbaum Associates. Dentschewa, Emilia 1987 Zur sprachlichen Eigenständigkeit der althochdeutschen TatianÜbersetzung in bezug auf den Gebrauch des Infinitivs. Beiträge zur Erforschung der deutschen Sprache 7: 207–232. Dik, Simon C. 1989 The Theory of Functional Grammar. Part I: The Structure of the Clause. Dordrecht: Foris Publications. Disse, Andreas 1998 Informationsstruktur im biblischen Hebräisch. Sprachwissenschaftliche Grundlagen und exegetische Konsequenzen einer Korpusuntersuchung zu den Büchern Deuteronomium, Richter und 2 Könige. Arbeiten zu Text und Sprache im Alten Testament; 56/Teil 1. St. Ottilien: EOS Verlag. Dittmer, Arne and Dittmer, Ernst 1998 Studien zur Wortstellung – Satzgliedstellung in der althochdeutschen Tatianübersetzung. Göttingen: Vandenhoeck & Ruprecht. Eggenberger, Jakob 1961 Das Subjektspronomen im Althochdeutschen. Chur: Sulser.

Information-structural analysis in historical texts Firbas, Jan 1966

157

On defining the ‘theme’ in functional sentence analysis. In Travaux Linguistiques de Prague, vol. I, 267–280. University of Alabama Press. Fleischer, Jürg 2006 Zur Methodologie althochdeutscher Syntaxforschung. Beiträge zur Geschichte der deutschen Sprache und Literatur 128: 25–69. this vol. Paleographic clues to Old High German prosody? – Accents, word separation, and related phenomena in Old High German manuscripts In New Approaches to Word Order Variation and Change in the Germanic Languages, eds. Roland Hinterhölzl and Svetlana Petrova. Berlin: Mouton de Gruyter. Fleischer, Jürg, Hinterhölzl, Roland and Solf, Michael 2008 Zum Quellenwert des althochdeutschen Tatian für die Syntaxforschung. Zeitschrift für germanistische Linguistik 36: 210–239. Floor, Sebastian Jonathan 2004 From Information Structure, Topic and Focus, to Theme in Biblical Hebrew Narrative. Ph.D. University of Stellenbosch. Frey, Werner 2005 Notes on the syntax and the pragmatics of German Left Dislocation. In The Syntax and the Semantics of the Left Sentence Periphery, eds. Horst Lohnstein and Susanne Trissler, 203–233. Berlin: Mouton de Gruyter. Gundel, Jeanette K. 1988 Universals of topic-comment structure. In Studies in Syntactic Typology, eds. Michael Hammond, Edith A. Moravcsik and Jessica R. Wirth. Amsterdam: John Benjamins. Halliday, Michael A. K. 1967 Notes on Transitivity and Theme in English, II. Journal of Linguistics 3: 199–244. Heimerdinger, Jean-Marc 1999 Topic, Focus and Foreground in Ancient Hebrew Narratives. Sheffield: Sheffield Academic Press. Hinterhölzl, Roland 2004 Language Change versus Grammar Change: What diachronic data reveal about the distinction between core grammar and periphery. In Diachronic Clues to Synchronic Grammar, eds. Eric Fuß and Carola Trips, 131–160. Amsterdam and Philadelphia: John Benjamins. this vol. The role of Information Structure in word Order Variation and word Order Change. In New Approaches to Word Order Variation and Change in the Germanic Languages, eds. Roland Hinterhölzl and Svetlana Petrova. Berlin: Mouton de Gruyter.

158

Svetlana Petrova and Michael Solf

Hockett, Charles F. 1958 A Course in Modern Linguistics. New York: Macmillan. Hopper, Paul J. 1979a Some Observations on the Typology of Focus and Aspect in Narrative Language. Studies in Language 3.1: 37–64. 1979b Aspect and Foregrounding in Discourse. In Syntax and Semantics, ed. Talmy Givón, 213–241. San Diego et al.: Academic Press, INC. Ingenbleek, Theodor. 1880 Über den Einfluss des Reimes auf die Sprache Otfrids besonders in Bezug auf die Laut- und Formenlehre. Mit einem Reimlexikon zu Otfrid. Straßburg: Karl J. Trübner. Jacobs, Joachim 2001 The dimensions of topic-comment. Linguistics. An interdisciplinary journal of the language sciences 39/4: 641–681. Kleiber, Wolfgang 2004 Otfrid von Weißenburg. Evangelienbuch. Band I: Edition nach dem Wiener Codex 2687. Teil 2: Einleitung und Apparat. Tübingen: Niemeyer Klein, Wolfgang, and Stutterheim, Christiane von 1992 Textstruktur und referentielle Bewegung. Zeitschrift für Literaturwissenschaft und Linguistik 86: 67–92. Krifka, Manfred 2007 Basic notions on Information Structure. In The notions of Information Structure, eds. Caroline Féry et al., 13–56. Potsdam: Universitätsverlag. Kuno, Susumo 1972 Functional Sentence Perspective: a Case Study from Japanese and English. Linguistic Inquiry 3: 269–320. 1978 Generative Discourse Analysis in America. In Current Trends in Textlinguistics, ed. Wolfgang U. Dressler, 275–294. Berlin and New York: de Gruyter. Lambrecht, Knud 1994 Information structure and sentence form:Topic, focus and the mental representations of discourse referents. Cambridge: Cambridge University Press. Lühr, Rosemarie 2005 Der Einfluß der klassischen Sprachen auf die germanische Grammatik. In Sprachkontakt und Sprachwandel. Akten der XI. Fachtagung der Indogermanischen Gesellschaft, eds. Gerhard Meiser and Olav Hackstein, 341–362. Wiesbaden: Reichert.

Information-structural analysis in historical texts

159

Masser, Achim 1991 Die lateinisch-althochdeutsche Tatianbilingue des Cod. Sang. 56. Nachrichten der Akademie der Wissenschaften in Göttingen, I. Philologisch-historische Klasse, 1991, 3. Göttingen: Vandenhoeck & Ruprecht. 1997a Syntaxprobleme im althochdeutschen Tatian. In Semantik der syntaktischen Beziehungen. Akten des Pariser Kolloquiums zur Erforschung des Althochdeutschen 1994, ed. Yvon Desportes, 123–140. Heidelberg: Carl Winter. 1997b Wege zu gesprochenem Althochdeutsch. In Grammatica Ianua Artium. Festschrift für Rolf Bergmann zum 60. Geburtstag, eds. Elvira Glaser and Michael Schlaefer, 49–70. Heidelberg: Carl Winter. Masser, Achim ed. 1994 Die lateinisch-althochdeutsche Tatianbilingue Stiftsbibliothek St. Gallen Cod. 56. Göttingen: Vandenhoeck & Ruprecht. Molnár, Valéria 1993 Zur Pragmatik und Grammatik des TOPIK-Begriffs. In Wortstellung und Informationsstruktur, ed. Marga Reis, 155–202. Tübingen: Max Niemeyer. Musan, Renate 2002 Informationsstrukturelle Dimensionen im Deutschen: Zur Variation der Wortstellung im Mittelfeld. Zeitschrift für Germanistische Linguistik 30: 198–221. Prince, Ellen F. 1981 Toward a Taxonomy of Given-New Information. In Radical Pragmatics, ed. Peter Cole, 223–255. New York: Academic Press. Reinhart, Tanya 1981 Pragmatics and Linguistics: An Analysis of Sentence Topics. Philosophica 27: 53–94. Ries, John 1880 Die Stellung von Subject und Prädicatsverbum im Heliand: Nebst einem Anhang metrischer Excurse. Straßburg/London: Karl J. Trübner. Rooth, Mats 1985 Association with Focus. Ph.D. University of Massachusets at Amherst. Ruhfus, Wilhelm 1897 Die Stellung des Verbums im althochdeutschen Tatian. Dortmund: Ruhfus. Sasse, Hans-Jürgen 1995 “Theticity” and VS order: a case study. Sprachtypologie und Universalienforschung 48, 3–31.

160

Svetlana Petrova and Michael Solf

Sievers, Eduard 1960 Reprint. Tatian. 2. neubearb. Aufl. Paderborn: Schöningh. Original edition, Paderborn: Schöningh, 1892. Simmler, Franz 1998 Makrostrukturen in der lateinisch-althochdeutschen Tatianbilingue. In Deutsche Grammatik: Thema in Variationen, eds. Karin Donhauser and Ludwig M. Eichinger, 299–335. Heidelberg: Carl Winter. Sonderegger, Stefan 1965 Frühe Übersetzungsschichten im Althochdeutschen. In Philologia Deutsch, eds. Werner Kohlschmitt and Paul Zinsli, 101–114. Bern: Francke. Stockwell, Robert P. 1984 On the history of the verb-second rule in English. In Historical Syntax, ed. Jacek Fisiak, 575–592. Berlin / New York / Amsterdam: Mouton. SzendrĘi, Kriszta 2004 Focus and the interaction syntax and pragmatics. Lingua 114: 229– 254. Weil, Henri 1879 L'ordre des mots dans les langues anciennes comparées aux langues modernes. Question de grammaire générale. 3rd edition. Paris: F. Vieweg. [1844] Zybatow, Gerhild 1999 Informationsstruktur im Russischen. In Sprachspezifische Aspekte der Informationsverteilung, ed. Monika Doherty, 61–89. Berlin: Akademie Verlag.

Paleographic clues to prosody? – Accents, word separation, and other phenomena in Old High German manuscripts Jürg Fleischer

1. Introduction At least since Saussure, language has been viewed as an oral system of communication. Written language is only secondary vis-à-vis the primary system, i.e., spoken language. When we look at historical stages of languages we are normally confined to written records, that is, we are forced to deal with the secondary system. It is only comparatively recently that it has become possible to record sound and to thus gain access to the primary linguistic system of past times. Obviously, only written evidence is extant for Old High German (usually dated c. 750–1050). Old High German texts make use of an alphabetical writing system based on the Latin alphabet. Simplifying considerably, the primary task of Old High German script is to encode the segments of the spoken language using (combinations of) the letters of the Latin alphabet; suprasegmental information is usually not provided in a consistent manner. However, some prosodic information can possibly be deduced from paleographic phenomena such as accents, (absence of) word separation, punctuation, and others. These will be discussed with respect to their potential value for prosodic research. Since most existing work relevant to our purposes deals with accents and since most is probably known about the (different) functions of accents, these are treated most extensively. Word separation and the other paleographic phenomena are discussed in a more synoptic fashion. I will primarily be concerned with Old High German manuscripts; however, as a point of reference I will often refer to (Medieval) Latin. Many phenomena to be observed in Old High German script have parallels in Latin, which is only natural if we consider the situation of Old High German: in Carolingian Europe, vernacular writing had at best a marginal status vis-à-vis Latin literacy (see e.g. McKitterick 1989: 21–22, Green 1994: 270–271).

162

Jürg Fleischer

2. Written language and reading in the early Middle Ages According to the ancient conception, the written word was a record of and dependent upon the spoken word (this is close to modern structuralist views). This changed in late Antiquity and in the early Middle Ages, when the written word acquired “the position of a medium in its own right, whose function was no longer seen to be wholly dependent upon the spoken word.” (Parkes 1993: 20). The conception of written language as an independent entity fully holds for the Carolingian period: “By the ninth century readers and scribes had come to perceive the written medium as an autonomous manifestation of language, which was apprehended as much by the eye as by the ear.” (Parkes 1993: 34). One of the earliest representatives of the new attitude towards written language is Isidore of Seville (c. 560–636): “Isidore regarded letters of the alphabet as signs without sounds […] Writing was no longer merely the record of the spoken word but could also signal directly to the mind through the eye.” (Parkes 1993: 21). It is for this reason that “Isidore preferred silent reading to ensure better comprehension of the text” (Parkes 1993: 21). However, interestingly, even silent reading is still partially conceived of as a bodily activity (cf. below): Acceptabilior est sensibus lectio tacita quam aperta; amplius enim intellectus instruitur quando uox legentis quiescit et sub silentio lingua mouetur. ‘Silent reading is more acceptable to the senses than reading aloud; for the understanding is instructed more fully when the voice of the reader is silent and the tongue is moved silently.’ (Isidore, Sententiae III, 14, 9 [Cazier 1998 ed.]; translation Parkes 1993: 21 and mine)1

The development of silent reading is to some extent parallel to the changes concerning the concept of the written word just discussed. In the ancient world written texts were mostly read aloud; “a reader always heard the words of a text in his mind” (Parkes 1993: 9). Reading habits began to change in the early Middle Ages. According to Saenger (1997) the development of word separation (see section 4) is closely connected to these changes: “In the West, the ability to read silently and rapidly is a result of the historical evolution of word separation that, beginning in the seventh century, changed the format of the written page […]” (Saenger 1997: 6). As to actual Carolingian reading habits, however, the situation is unclear (this, incidentally, still holds for the high Middle Ages; see Scholz 1980: 103–111). According to many sources, reading aloud was the usual practice. Illich (1993a: 54; see also Illich 1993b, especially 23–24) emphasizes that medieval reading was a bodily activity:

Prosodic information in Old High German manuscripts

163

The reader’s ears pay attention, and strain to catch what the reader’s mouth gives forth. In this manner the sequence of letters translates directly into body movements and patterns nerve impulses. The lines are a sound track picked up by the mouth and voiced by the reader for his own ear. By reading, the page is literally embodied, incorporated.

For this reason, medieval monasteries “are described to us in various sources as the dwelling places of mumblers and munchers.” (Illich 1993a: 54). With respect to devotional reading of private prayer books Saenger (1997: 39) states: The nomenclature for devotional reading in early medieval Continental monastic rules and customaries is too ambiguous for us to be certain that silent reading was in fact practiced by Continental monks of the Carolingian age. However, many literate sources suggest that reading was oral.

Nevertheless, given the emerging concept of written language as an entity in its own right, the question must be asked whether written texts were perceived as having a prosodic structure at all by those who used them. Although there is naturally some uncertainty, this question can be answered positively. As discussed above, if silent reading was practised at all in Carolingian times, it was practised only rarely. On the other hand, there is clear evidence that some texts must have had a prosodic structure, for example, texts that were regularly recited in the liturgy or that were used for semi-liturgical purposes (such as readings at table, etc.). For nonliturgical texts, however, this might not hold: complicated philosophical or theological treatises for instance, which were probably sometimes read silently (see above), may not have been perceived by their users to have had a prosodic structure. The arguments discussed so far apply to written language and its relation to oral language in general. In Carolingian Europe, Latin was, of course, the most commonly written language: as in Antiquity, “the status of Latin as the primary means of communication in writing was maintained” (McKitterick 1989: 21). The written vernacular could at best play a marginal role compared to Latin. On the other hand, Latin was nobody’s native language, at least in the Germanic-speaking parts of the empire.2 To be sure, there was a Latin oral tradition in the early Middle Ages, but in most cases Latin texts were read or recited, as in the liturgy, for instance (see Stotz 2002: 149–153). Although we have records describing certain people freely expressing themselves in spoken Latin, in disputations for example, it is also clear that it could never be taken for granted that people could

164

Jürg Fleischer

hold conversations in Latin (see Stotz 2002: 152). Although Latin dominated monastic life, for Carolingian monks in the Germanic-speaking lands it was a second language that had to be learned in a process of formal instruction. Compared to the vernacular, Latin must have been perceived as a more abstract entity. Even if Latin texts were possibly read silently on some occasions, this might not have been the case for written vernacular texts. The reception mode of medieval literature has been a stimulating field of debate, following the seminal work of Scholz (1980). Scholz (1980) claims that the primary reception of the “classical” Middle High German literature often happened through individual reading rather than public performances (see also Green 1994). In the case of individual reception, reading might have been silent (although this is by no means sure; see above), but in the case of recital in front of an audience, reading was surely aloud; in that case the texts performed also displayed a prosodic structure. For at least some of the Old High German texts it is clear that they were intended not only for individual reading, but also for recital in front of an audience. In dedicating the Evangelienbuch to King Louis the German, Otfrid expresses his hope that the king would let the work be read out (‘thaz er sa lésan heiziT’, Lud. 88); Green (1994: 181) comments that this expression “generally refers to public recital rather than private reading”. All existing manuscripts of Otfrid’s Evangelienbuch display a system of accentuation that would have been useful for a reciting reader (see section 3.2.). The fact that in one manuscript neumes have been added to a line (see Klaper 2004 and the literature cited there) is clear evidence that this text was also intended to be intoned by at least some of its users. As to another Old High German author Green (1994: 185) convincingly argues that with the translations of Notker (see section 3.1), whose works belong to a school context, at least in some occasions “we have a typical oral situation: a text was made available by the teacher reading it out.” In the remainder of this article, I take it as given that the Old High German texts discussed were at least sometimes read aloud. If this is correct, then looking for clues to prosodic information may not be completely in vain, even though we have to stick to written language. The task then becomes a consideration of whether prosodic information can be inferred from Old High German writing in any systematic way.

Prosodic information in Old High German manuscripts

165

3. Accent diacritics If we disregard some Insular script remnants from the 8th and early 9th century (see Bischoff 1986: 126–128), all (vernacular and Latin) records from before the end of the Old High German period are written in the Carolingian minuscule (see Bischoff 1986: 151–171). Many Old High German manuscripts display diacritic accents, which have been treated to some extent by Sievers (1909) and Gabriel (1969: 44–101); see also the summarizing account by Schneider (2007). Accents in Old (and Middle) High German manuscripts are usually used in an unsystematic way (Schneider 2007: 17), but there are important exceptions. The habit of accentuating texts was doubtlessly taken over from Latin writing; Latin itself had adopted this from Greek. Many Latin grammarians provide a chapter De accentibus (see Sievers 1909: 9–11, Gabriel 1969: 45–49). Unfortunately “the use of accent marks in Northern Europe […] has been almost totally ignored by students of Latin paleography” (Saenger 1990: 59). Bruckner (1980), Bischoff (1986: 226) and Saenger (1997: 68– 69) give only brief accounts. Loew (1914: 274–277) and Newton (1999: 179–181, 193–194) provide some information with respect to the Beneventan script, in use in southern Italy when large parts of continental Europe adhered to the Carolingian minuscule (see Bischoff 1986: 160).3 According to Bruckner (1980: 260), the use of accents spread across the West in the 10th century and became frequent in the 11th. If this is correct, it would seem that accents in Old High German (or other vernacular languages, e.g., Old English) appeared to some extent prior to their appearance in Latin writing. However, according to other sporadic observations, accents were being used in Latin writing earlier than indicated by Bruckner (see, e.g., Gabriel 1969: 49), albeit not yet in a consequent and systematized manner (Bruckner 1980 possibly refers to this). There are two basic accents in Old High German script: the acute (e.g., ) and the circumflex (e.g., ). The third classic accent, the grave (e.g., ), which is used in Greek writing and described by Latin grammarians as well, plays no role in Old High German; some Latin grammarians already judge it superfluous (see Sievers 1909: 10). The two different accent forms are carefully distinguished in the works of Notker, but many manuscripts use only one type of accent (mostly the acute; see Sievers 1909: 123). If the two types of accents are distinguished (and if they are to be interpreted as signs indicating emphasis) there is a tendency for the circumflex to be associated with long vowels. In some cases scribes produce forms of accents not clearly identifiable as either acute or circumflex

166

Jürg Fleischer

(see Sievers 1909: 5): the acute is usually written as a slash, which may have a downward hook at the upper right. If this hook is relatively prominent, a form not distinguishable from a circumflex may emerge (see, e.g., Masser 1994: 15 or Gabriel 1969: 51 with respect to the St. Gall Tatian manuscript; example (7), rendered as acute in the edition by Masser 1994, provides a relevant example). Regarding their position in script it has often been observed that accents are written before or after the vowel letter that is expected to bear the accent or that the accents take the length of two or even three letters (see Gabriel 1969: 51–52 with respect to the St. Gall Tatian manuscript). In some cases, the fact that the position of an accent cannot be unequivocally ascribed to one letter can lead to multiple interpretations of its function. While accents in Medieval German manuscripts are usually used in an unsystematic way, the works of Otfrid and Notker, both dating from our period of interest, exhibit important attempts to systematize accentuation (see Schneider 2007: 18). However, with the exceptions of Otfrid and Notker, no systematization or consistency in Old High German accentuation can be discerned. These exceptions aside, it is virtually impossible to assign an unequivocal function to Old High German accents; indeed, in many individual cases an accent can be attributed more than one function, as will be illustrated in section 3.3. Ornamental usage of accents and scribal errors cannot be excluded from the discussion either (see Sievers 1909: 121). I will therefore begin my discussion of accents with Notker (section 3.1.) and Otfrid (section 3.2.), the two systems that are most consistent and best understood.4 Only after examining these two cases will I discuss accentuation in the St. Gall Tatian manuscript, where it seems to perform many different functions (section 3.3.). The works discussed represent a sizeable proportion of the Old High German records.

3.1. Accents in the works of Notker of St. Gall The most elaborate Old High German accentuation system can be found in the works of Notker of St. Gall (c. 950–1022). As emphasized by Sonderegger (1970: 54), Notker’s accentuation has some predecessors in St. Gall’s Old High German glosses and, thus, in the St. Gall scriptorium, but Notker is the only one to use accents in a systematic manner. Notker’s system has been described, sometimes with reference to one particular manuscript, by, among others, Fleischer (1882), Sievers (1909: 21–27), Gabriel (1969: 61–81), Zürcher (1978: 30–35, 80–83, 165–169) and King

Prosodic information in Old High German manuscripts

167

(1986). It is briefly discussed by Schneider (1999: 93; 2007: 18–19). We know explicitly from one of Notker’s writings that he attributed great importance to the accents when writing German. In a letter to Hugo Bishop of Sitten we find the following passage, which clearly demonstrates that accents are a matter of pronunciation for Notker: Oportet autem scire quia uerba theutonica sine accentv scribenda non sunt . prĊter articulos ipsi soli sine accentu pronuntiantur acuto et circumflexo ‘One needs to know that the German words must not be written without an accent, with the exception of the articles;5 they alone are pronounced without acute or circumflex accent’ (quoted from Tax 1996: 349; see also King 1986: XXXVIII; translation mine)

There is consensus that of all of the manuscripts that have come down to us the first two books of Notker’s translation of Boethius’ De Consolatione Philosophiæ in Cod. Sang. 825, 4–108 provide the most elaborate and consistent instantiation of Notker’s system (see Fleischer 1882: 130–131, King 1986: XXXVIII, Schneider 2007: 18). I will provide examples exclusively from this passage. The most important feature of Notker’s system is that acute and circumflex are neatly distinguished. According to a somewhat generalizing formula, every stressed long vowel displays a circumflex and every stressed short vowel displays an acute; in addition, long vowels in syllables not bearing primary stress often display a circumflex (see Braune / Reiffenstein 2004: 16). According to King (1986: XXXIX) this might be seen as an indication of secondary stress, whereas others have thought that the circumflex should be solely seen as an indication of vowel length (thus, for Sievers 1909: 27 this is the primary function; on the other hand, for Gabriel 1969: 64 the indication of stress is primary).6 Interestingly, the acute also occurs in such contexts, albeit less often. In the dative plural léidúngôn ‘prosecutions’ (Cod. Sang. 825, 14,26; Notker, Boethius 14,26 ed. Tax) all three vowels display an accent, only the first of which is likely to indicate primary stress; similarly, in the dative plural adjective únzálaháftên ‘countless’ (Cod. Sang. 825, 21,26; Notker, Boethius, 21,26 ed. Tax) not all acute accents can indicate primary stress. In Notker’s system almost every word is accentuated; it is thus obvious that it is based on word accent. The fact that a word displays accent(s) therefore does not mean that the respective item is especially prominent from a sentence prosodic perspective. Given this state of affairs, the fact that some words are not accentuated is much more telling. Since Notker’s accentuation is so consistent we can be quite sure that absence of an accent

168

Jürg Fleischer

is meaningful; as we have seen above, Notker himself states that some words are pronounced with no accent. For instance, nominative personal pronoun forms may or may not be accentuated, but in sentence-initial position they always display an accent (see Fleischer 1882: 150, King 1986: XL). It thus seems that in Notker’s Old High German unstressed subject pronouns are not possible in sentence-initial position. It would be a natural consequence to postulate that this position is reserved for elements bearing stress. In sentence-internal position, on the other hand, both forms are possible; there is thus a certain “freedom” with respect to accentuation (see Fleischer 1882: 151). This is illustrated by the following example, which displays the personal pronoun ih ‘I’ after the verb in two clauses that are parallel with respect to their construction; only in the second instance does an accent appear: (1)

Énes iího ih · tísses neiího íh· (Cod. Sang. 825, 24,20)7 that confess I this NEG-confess I ‘That I confess; this I do not confess.’8

Illustration 1. Cod. Sang. 825, 24,20; Stiftsbibliothek St. Gallen / Codices Electronici Sangallenses9

Notker’s accentuation undoubtedly provides rich prosodic information. At the level of the clause, paradoxical as it may seem, words not displaying accents are most interesting. Since Notker’s translations are prosaic (in contrast to Otfrid’s work; see section 3.2.), there is no interference from a meter. Of all accentuations systems, Notker’s is therefore the most valuable source with respect to Old High German prosody. Unfortunately from the point of view of sentence prosodic research, it primarily allows the identification of unstressed elements. Especially stressed constituents cannot be differentiated from normally stressed elements since both can be equally expected to display an accent in Notker’s system.

Prosodic information in Old High German manuscripts

169

3.2. Accents in Otfrid’s Evangelienbuch The Evangelienbuch of Otfrid of Weissenburg (c. 800–c. 870) provides another rather elaborate and quite consistently applied accentuation system, first described by Piper (1882) and Sobel (1882; see also Schneider 2007: 18). There are four different manuscripts extant, of which the Viennese Codex, which was partially written and corrected by Otfrid himself (see e.g. Kleiber 2004b: 53, 55), is undoubtedly the most precious (see Kleiber 2004b: 120). In the Viennese Codex (dated 863–871), as well as two other manuscripts of the Evangelienbuch, two main paleographic shapes of acute accents are carefully distinguished, the “phonetic” and the “rhythmic” accents (see, e.g., Sievers 1909: 7, Kleiber 2004b: 120).10 The phonetic accents are far smaller and often written higher above the letter than the rhythmic accents. They are in a different color of ink than the main text, which suggests that they where added separately, after the main text had been written (see Kleiber 2004b: 131).11 The phonetic accents usually consist of two fine slashes, distributed over two letters, mostly combinations of or with another vowel grapheme, for instance íó ‘ever; always’, íú ‘you (dat. pl.)’, íámer ‘always’, úáben ‘to arrange’ (see Kleiber 2004b: 131). According to 19th century Otfrid editors, the function of these accents was to avoid the pronunciation of and as semivowels. It has been suggested that the accents in question ensure vocalic realization of and in ambiguous contexts, that is, before a vowel grapheme (see Sievers 1909: 8, especially note 1), which explains the label “phonetic accent”. A few potential minimal pairs exist. For example, can stand for /iu/ ‘you’ or /ju/ ‘already’ (see Kleiber 2004b: 121). Disambiguation by means of accents would be especially useful here. Kleiber (2004b: 131) provides a slightly modified interpretation of Otfrid’s “phonetic accents”. According to him, in most instances they ensure the diphthongal realization of the graphemes accentuated. Indeed, in addition to the examples quoted above, óúgen ‘eyes’, with neither nor as its first letter, also bears phonetic accents, although the initial in can only be interpreted as a vowel. However, ensuring diphthongal realization of the accentuated letters can only be a partial explanation of the observed distribution. In examples such as úúorolt ‘world’ (IV 26,34 ed. Kleiber 2004a) the “phonetic” accents are used in a word that undoubtedly does not display a vowel or diphthong as its initial segment (see Sievers 1909: 8, note 1; Gabriel 1969: 55, note 35). There is thus a paradox: the phonetic accents seem to indicate

170

Jürg Fleischer

vocalic or diphthongal realization in some contexts, but cannot serve this function in others. Incidentally, there are also cases of possibly incorrect phonetic accents, for example íú standing for /ju/ ‘already’, not /iu/ ‘you (dat. pl.)’ in Sos ér gihiaz íú uuánne · themo drútmanne (I 10,13 ed. Kleiber 2004a). The so-called “rhythmic accents” in Otfrid’s Evangelienbuch are attributed a rather different function. They are used to identify the main accented syllables. It is most interesting that Otfrid himself, who, as discussed above, wrote some passages of the Viennese Codex and acted as a corrector of the whole manuscript, corrected many accents. For example the passage quoted below (I 1,111) was originally written with the following accents (see Kleiber 2004b: 123): (2)

a. Gidán ist es nu rédina · thaz síe sind guate thégana done is of_it now speech that they are good heroes ‘It is told now that they are good heroes’

With this accentuation the second hemistich displays exactly the same rhythmic pattern as the first one (namely: ` ^ ` ` ` ^ ` `). Both hemistichs have a very regular rhythmic structure in which the second and the sixth syllable are stressed most. However, it was apparently the author’s intention to emphasize the adjective guate ‘good’; he corrected the accentuation with the following result (see Kleiber 2004b: 123): (2)

b. Gidán ist es nu rédina · thaz sie sind gúate thegana (I 1,111 ed. Kleiber 2004a) done is of_it now speech that they are good heroes ‘It is told now that they are good heroes’

Obviously the adjective good is seen as most important in the dependent clause, and this is expressed through an accentuation that works against an easy but mechanical repetition of the accentuation pattern of the first hemistich. Instances such as this one clearly show the importance of Otfrid’s accents when it comes to questions of prosodic prominence. There can be no doubt that Otfrid’s rhythmic accents mark the words that the author wanted to be most prominent in the recitation of his work. From a sentence prosodic perspective they are therefore more interesting than Notker’s system, since they allow identification of the most emphasized elements. However, we have to keep in mind that we are dealing with a poetic text, whose prosody might differ from natural patterns. This potentially diminishes the value of Otfrid’s accentuation data.

Prosodic information in Old High German manuscripts

171

3.3. Other accentuation usages: Tatian What is so extraordinary about Notker and Otfrid is the consistency and systematicity of their accentuations. If consistency is lacking and a manuscript only displays some scattered instances of accentuation, as is the rule in Old and Middle High German manuscripts (see Schneider 2007: 18), it becomes much more difficult to interpret these instances in a functional light. This will become clear at various instances below when individual instances of accentuation in the St. Gall Tatian manuscript are discussed. The most common and important functions of accents in Old High German manuscripts outside the works of Notker and Otfrid seem to be (1) the indication of the pronunciation of certain letters; (2) the highlighting of monosyllabic words and weakly stressed or unstressed syllables; (3) the indication of special emphasis on the word accentuated. Of these different functions the first and the third are to some extent comparable to the systems encountered in Notker and Otfrid. Only the latter function is interesting from the point of view of prosodic research, but since considerable overlapping of the different functions plays a role in the actual occurrences of accents, the other cases must be looked at as well. The classification of accents advocated here is often blurred and additional criteria might play a role.12 Much room is left for interpretation. A particular danger lies in circular argumentation: if one is inclined to believe that an accent marks stress due to emphasis on the accentuated word, one will opt for an interpretation of the respective passage supporting emphasis of the respective element, referring to the accent supporting this interpretation, etc. An important paleographic problem posed by scattered accents is that in many cases there are doubts about whether they were written by the original scribes or added by later users. Since accents consist of just one or two slashes it is often difficult to identify their paleographic origin. Therefore, in this section I will only discuss data from the St. Gall Tatian manuscript: in this record six individual scribes can easily be differentiated (see Masser 1994: 31). As already noted by Sievers (1909: 14) and confirmed by the discussion in Gabriel (1969: 51–57), these scribes display clear differences in their accentuation patterns. There is, therefore, every reason to assume that the accentuation observable in the St. Gall Tatian manuscript is original. Of course, the possibility that some accents were added by later users cannot be excluded, but statistically this is relatively unlikely: It would be difficult to explain why accents that were added by later users display differences that correspond to those between the well-known portions by the different scribes.

172

Jürg Fleischer

According to Sievers (1909: 8), one original function of acute accents was to identify is and us in graphical contexts where many vertical strokes come together. In Carolingian minuscule (which has no dot in this script; one could therefore use in modern printing) and , as well as <m> and , consist of dominating vertical strokes, which makes it difficult in a word like mınımum to identify the end of one letter and the beginning of the next.13 In addition, Carolingian was difficult to differentiate from , even independent of the graphic context (see Sievers 1909: 8). Therefore, acute accents were added to and in certain graphic environments in order to differentiate them from other letters (as a matter of fact, the dot on our modern i, which originated considerably later, seems to have evolved from such usage). It is probably this function of accents which may be seen in the accentuated instances of the prefix int- (New High German ent-), in which the i occurs before an n. Such examples can be observed in scribe ȕ’s portion of the Tatian manuscript, e.g., ínlazu ‘I release’ (55,20) or íntphieng ‘he got’ (57,23). However, since this scribe’s work also displays instances of other accentuated prefixes (not necessarily featuring i or u in a graphically ‘difficult’ environment), the function of these accents can also (or additionally) be to mark an unstressed syllable (see below). It is possible that a usage according to which accents were only written above and gave rise to the “phonetic accents” known from Otfrid, among others (see section 3.2.): the shared element is that only and are accentuated. However, as we have seen in our discussion of Otfrid, it is difficult to pin down the exact function since “phonetic accents” are encountered on diphthongs as well as on semivowels. As discussed in section 3.2., Otfrid’s phonetic accents have two remarkable paleographic properties: they consist of two acute accents, and the phonetic accents are carefully distinguished in script from the “rhythmic” accents. As to the second property, I am not aware of any other Old High German manuscript outside the Otfrid family which distinguishes between two paleographic shapes of acutes. As to the first property, however, “double acutes” are also found in scribe į’s portion of the St. Gall Tatian manuscript, e.g., íú ‘you (dat. pl.)’ (171,29), íúuuariu ‘your’ (173,21), íógilichemo ‘each (dat. sg. m.)’ (177,16), etc. (see Sievers 1909: 19, Gabriel 1969: 55). As one can see from the words quoted, these double acutes seem to perform the same function as Otfrid’s “phonetic accents”. However, in scribe į’s portion we also find singleton acutes in such contexts, e.g., íu ‘you (dat. pl.)’ (183,21) or íuuuer ‘your’ (185,5).

Prosodic information in Old High German manuscripts

173

According to Sievers (1909: 9), the s standing for the velar semivowel (which can be rendered by , , and any combination of these letters in Old High German script, e.g., , , etc.) often display phonetic accents. Many examples can be found in scribe ȕ’s portion of the St. Gall Tatian manuscript, for example uúas ‘was’ (114,32), uúis ‘be!’ (96,1), etc. (see Sievers 1909: 18). This scribe even displays one example of an acute on in íar ‘years’ (114,29), which according to Sievers (1909: 18) ensures semivocalic realization of this letter. This usage can also be found in scribe į’s portion, e.g., in uúilis ‘want (2. sg.)’ (183,2), uúas ‘was (3. sg.)’ (188,1), or úuege ‘way (dat. sg.)’ (190,14) etc. A rather different function of accents seems to be the highlighting of monosyllabic words. This usage is known from Latin script (see Bruckner 1980: 260); according to Loew (1914: 276, note 1), this habit is attested in early Caroline manuscripts (Sievers 1909: 15 maintains the same of 8th/9th century manuscripts). According to Saenger (1990: 57–58, 1997: 54–55) this usage is already attested in ancient Latin books from Egypt. Sievers (1909: 15) states that its function is to highlight the small words, which were often written together with the following word (see section 4.). Scribe ȕ of the St. Gall Tatian manuscript has many acute accents on “small” words such as prepositions, auxiliaries, articles, pronouns etc. Examples are fón ‘from’ (52,16), só ‘so’ (58,1) íz ‘it’ (55,10), tház ‘that’ (54,5) etc. (see Sievers 1909: 15, Gabriel 1969: 53). It is especially interesting that this scribe uses a very similar pattern in the Latin text of his portion. Here it is most often pronouns that display an accent, such as mé ‘me’ (see below), but monosyllabic nouns such as ós ‘mouth’ (74,14),14 verbal forms such as dát ‘gives’ (122,6) and conjunctions such as ác ‘and; too’ (111,30) or án ‘whether’ (89,25) are also accentuated (see Gabriel 1969: 56). Some of these accents could also be interpreted as signs of stress (see below), but this is virtually impossible in the case of some of the conjunctions or prepositions quoted. In the following passage (Cod. Sang. 56, 85,31–32), there are accents on all of the monosyllabic words in the Latin and Old High German texts with the exception of mih ‘me’: (3)

domine.‘ permitte mé primum ire herro láz mih ér faran & sepelire patrem meum; Inti bigraban minan fater Lord let me before go and bury my father ‘Lord, suffer me first to go and bury my father.’

174

Jürg Fleischer

Illustration 2. Cod. Sang. 56, 85,31–32; Stiftsbibliothek St. Gallen / Codices Electronici Sangallenses15

As can be seen, the accentuation of the Latin and the Old High German texts does not correspond (which makes it unlikely, for example, that mé/mih should be interpreted as emphasized). Monosyllabicity seems to be the only reason for the accentuation. In the following passage, the purpose of the accentuation also seems to be to highlight the monosyllabic words (see Gabriel 1969: 53), i.e., to make them better visible to a reader. It is unlikely that all of the accentuated words are stressed (incidentally, the accents on uúaz and íu are most naturally interpreted as phonetic): (4)

só uúaz só hér íu quede só tuot ír (Cod. Sang. 56, 81,25) so what so he you say so do you ‘Whatsoever he saith unto you, do it’

Illustration 3. Cod. Sang. 56, 81,25; Stiftsbibliothek St. Gallen / Codices Electronici Sangallenses

Prosodic information in Old High German manuscripts

175

In addition to monosyllabic words there are examples of obviously unstressed or, at most, secondarily stressed syllables in polysyllabic words displaying an accent. Such cases can be found in scribe ȕ’s portion of the St. Gall Tatian manuscript, e.g., ubár ‘over’ (78,20), odá ‘or’ (71,6), ehír ‘ear (of corn)’ (112,2), the masculine accusative form of the personal pronoun Inán ‘him’ (57,32), or the proper noun genitives andreasés and p&rusés (both 52,1) (see Sievers 1909: 15–18, Gabriel 1969: 53).16 According to Sievers (1909: 18) and Gabriel (1969: 53), syllables that are especially prone to be eroded in the recital are highlighted in such instances. The examples of accentuation just discussed seem to indicate the exact reverse of the usage of accents in the works of Notker or of Otfrid’s rhythmic accents: elements likely to go unnoticed (monosyllabic words, unstressed or weakly stressed syllables) display an accent in order to be made more prominent for the reader, precisely because they are not especially prominent per se. However, the stress-indicating function known from Otfrid and Notker may possibly also be found in the Tatian manuscript. The following passage, from scribe ȕ’s portion, is extraordinary in that the accentuation of the Latin and Old High German texts is parallel (in contrast to the passage quoted above): (5)

ihesus autem ait illi. sequere mé. tho quad imo ther heilant. folge mír (Cod. Sang. 56, 86,1) then said him the savior follow me ‘But Jesus said unto him, Follow me.’

Illustration 4. Cod. Sang. 56, 86,1; Stiftsbibliothek St. Gallen / Codices Electronici Sangallenses

This passage corresponds to the beginning of the second sentence in “And another of his disciples said unto him, Lord, suffer me first to go and bury my father. But Jesus said unto him, Follow me; and let the dead bury their dead.” (Matt. 8:21–22 King James Version). Since the command to come with Jesus is opposed to another option, it seems reasonably justified to

176

Jürg Fleischer

assume contrastive accent on mé/mír and to take the accents as a support for this interpretation. However, given the accentuation habits of this scribe we ought not take the accents as decisive: As discussed above, this scribe very often puts accents on monosyllabic words. Were the accents on mé/mír to be taken as evidence that we are dealing with contrastive stress in this passage, the danger of circular argumentation mentioned above materializes. It is probably impossible to tell which function is being performed by the accents in the passage quoted. An intriguing and probably realistic interpretation would be that the scribe wanted the accents to perform both functions. As a matter of fact, in many instances different functional motivations can be attributed to a single accentuation (as just hypothesized it is even conceivable that a scribe wanted an accent to perform more than one of the possible functions). The “phonetic” function and that of marking monosyllabic words may co-occur, for instance, in examples like íu ‘you (dat. pl.)’, a word that is both diphthongal and monosyllabic; given that there are certain contexts in which this personal pronoun may be especially stressed, even the stress-indicating function could play a role in some instances. There are several passages where the stress-indicating function cannot be distinguished from other functions. Sievers (1909: 20) mentions that scribe ȗ of the Old High German Tatian manuscript has many instances of uú in which the accent is probably phonetic, but that in every instance the accent could also mean that the vowel following uú is stressed. Here, then, it is impossible to tell the “phonetic” from the stress-indicating function. There are other cases in which it is difficult to differentiate between the monosyllabicity-indicating and the stress-indicating function. For example, according to Masser (1997: 60), scribe ȗ of the St. Gall Tatian manuscript uses acute accents in words that are emphasized. He quotes the passage in which “Crucify him, crucify him” is shouted to Pilate. In the Latin text we find crucifige crucifige; in the Old High German text the imperatives have acute accents: háh háh (308,25). It is of course possible that these accents indicate stress. However, the words are monosyllabic, which Masser (1997) does not consider in his discussion. Hence, from a methodological point of view it is not justified to interpret these accents as indications of stress (alone). To conclude this section, it can be seen that accents may perform an array of different functions in the St. Gall Tatian manuscript. They may serve the “phonetic” function known from Otfrid, but they may also highlight monosyllabic words or unstressed syllables, and finally they may indicate stress and thus emphasize the words accentuated. This function is of fun-

Prosodic information in Old High German manuscripts

177

damental interest for prosodic research. What remains intriguing from a modern point of view is the multitude of different functions performed by one symbol, and the fact that it might be intrinsically impossible to tell the different functions apart in individual instances. Apparently, however, this posed no problem for Carolingian scribes.

4. Word separation Roman scribes would usually operate by “copying texts in scriptio continua – that is, without separating words or indicating any pauses within a major section of the text” (Parkes 1993: 10). Separation by spaces17 first appears in the Insular area (see Saenger 1997: 84–99). It is then used in the Carolingian period (see Müller 1964: 35, Bischoff 1986: 229, Saenger 1990); Parkes (1993: 31) discusses several cases in which Carolingian scribes copying older manuscripts introduced separation. However, spaces were not yet used in a consequent manner (see Saenger 1997: 100): in Carolingian minuscule manuscripts “space was present, but not consistently between every word” (Saenger 1997: 32). Saenger (1997: 32) refers to this kind of separation as “aerated script” (as opposed to “word separation” proper). In central and southern Germany “scribes, after the mid-ninth century, began to write Latin in intensely aerated script, sometimes approaching separation, and began to develop modes for writing the vernacular that were clearly stimulated by Anglo-Saxon models.” (Saenger 1997: 102). It is notable that “only at the Abbey of Saint Gall, one of the larger Continental colonies of Irish and Anglo-Saxon monks, did the evolution of text format go so far as to continue consistent emulation of Insular word separation.” (Saenger 1997: 103). The Vocabularius Sancti Galli, a Latin–Old High German dictionary, provides an early example of Continental word separation (see Saenger 1997: 103). If separation is followed in principle in a manuscript, the absence of spaces between units that are words from a syntactic point of view or the presence of unusually large spaces between syntactic words might be interpreted from a prosodic perspective.18 Most Old High German manuscripts display separation between words (or larger or smaller units), although entire lines may still give the impression of scriptio continua (see Nübling 1992: 337). One recurring pattern is that preposed small (usually functional) words are written together with the word that follows, a preposition or the negation particle, for instance (see Voetz 2006: 51, 56; Bischoff 1986: 229 reports the same of prepositions “and other short words”

178

Jürg Fleischer

in Latin script). Postpositioning of small words is also attested (see for instance Nübling 1992: 339). In some instances, however, units smaller than the word are separated. According to Frey (1988: 140) and Voetz (2006: 59), the syllable (rather than the morpheme) turns out to be the relevant unit in this respect (see also example [7], which features a blank space between two syllables of a word). We cannot find much uniformity with respect to word separation in Old High German manuscripts (Nübling 1993: 338, Voetz 2006: 59). Yet, according to Nübling (1992: 338), in autochthonous texts certain parts of speech are more regularly written together with the preceding or following word than others, namely, subject pronouns following verbs or conjunctions, prepositions directly before a personal pronoun, noun, or article, and the negation particle ni followed by a verb (Nübling 1992: 339–341). Since some of these combinations are known to be hosts and clitics in modern German dialects (Nübling 1993: 342), it seems reasonable to argue that clitic forms might be characterized by a lack of word separation in Old High German script. Absence of word separation might therefore be a valuable criterion by which to identify clitics. The examples discussed so far involve absence of word separation, but the presence of (extra large) spaces also might be interpretable. According to Masser (1997: 57), in scribe ȗ’s portion of the St. Gall Tatian manuscript larger spaces are used when a passage in direct speech is rendered; Masser (1997: 57) interprets the larger space as an indication of a pause before the direct speech begins. One relevant example is the following (Cod. Sang. 56, 274,6); here, a somewhat larger space separates ei/imo from quis/uuer, the first word of the quoted direct speech: (6)

& dicit ei quis est Inti quad imo uuer ist (Masser 1997: 58) and said him who is ‘… and said unto him: “Who is it?”’

Illustration 5. Cod. Sang. 56, 274,6; Stiftsbibliothek St. Gallen / Codices Electronici Sangallenses

Prosodic information in Old High German manuscripts

179

According to Masser (1997: 58), it might also be a function of spaces to render special emphasis. In the following example from scribe ȗ’s portion of the Tatian manuscript (Cod. Sang. 56, 224,13), two syllables of one word are separated from each other; additionally, the word also bears an accent: (7)

Iú uuer sunta uuon& (Masser 1997: 61) your sin remains ‘your sin remaineth’

Illustration 6. Cod. Sang. 56, 224,13; Stiftsbibliothek St. Gallen / Codices Electronici Sangallenses

According to Masser (1997: 61) Iú uuer is stressed in this passage, which provides the motivation for the accent. Masser explicitly rejects the possibility of a phonetic accent, pointing to the blank space which is seen as an additional element rendering emphasis (see Masser 1997: 61, note 37). As a matter of fact, I think that Masser is indeed right in interpreting Iú uuer as being emphasized in this passage. However, given that many phonetic accents in iu(wer) can be observed elsewhere, from a methodological point of view taking the accent as proof is not warranted. Quite to the contrary, one could just as well argue that the scribe inserted an extra space between the first and second syllables of the word because he did not regard the accent as an indicator of stress or, to put it more cautiously, because he did not think that the accent alone was enough to indicate stress. To sum up, there are clear cases of separation that can be interpreted from a prosodic perspective. However, much more systematic work is needed here. Detailed analyses of manuscripts or facsimile editions are a prerequisite, since editions are extremely unreliable with respect to word separation.19 Finally, one must bear in mind that trivial criteria such as lack of space in a line (possibly due to holes in the parchment or the like) or the desire to fill a line with a relatively small amount of text can lead to extraordinarily spaced units.

180

Jürg Fleischer

5. Other paleographic phenomena Other paleographic phenomena that might provide information relevant to prosody but which cannot be treated here in due detail, are punctuation, use of initial and capital letters, and certain assimilations. In Antiquity, punctuation symbols functioned to indicate pauses. Punctuation according to this principle is labeled rhetorical (see Müller 1964, Bischoff 1986: 224, Schneider 1999: 89, Simmler 2003: 2473). The period of the Carolingian Renovation is characterized by, among other things, an augmentation of the punctuation symbols used in texts.20 There is, on the one hand, a two-fold system of points that was already in use (see Parkes 1993: 31), but Carolingian scribes “often combined [this] with other marks” (Parkes 1993: 32; see also Bischoff 1986: 224). Alcuin (c. 730– 804), who “settled in Francia to assist Charlemagne in his endeavors to promote liturgical, educational and administrative reforms” (Parkes 1993: 30; see also Simmler 2003: 2473), promoted the use of punctuation (Parkes 1993: 32, Saenger 1997: 36). One invention of this period is the question mark (Bischoff 1986: 225, Saenger 1997: 73-74). Since this punctuation symbol can be traced back to the quilisma neume, indicating “trembling and mounting tone” (Bischoff 1986: 225; see also Kleiber 2004b: 137), it is natural to assume that questions at that time, as in contemporary German, displayed an intonation different to declarative clauses; this symbol therefore provides prosodic information. There are some analyses or at least symbol inventories of individual Old High German manuscripts, for instance that by Masser (1994: 15) on the Tatian, Simmler’s (1997) treatment of two minor texts, or that of Kleiber (2004b: 134–137) with respect to the Viennese Otfrid manuscript. Simmler (2003: 2472–2477) provides a summarizing review of the punctuation of various Old High German texts. In the bilingual Latin–Old High German Tatian manuscript, punctuation symbols are primarily, though not exclusively, encountered in the Latin part. Some Old High German passages also display punctuation, while others completely lack it (see Masser 1994: 15). This indicates that the Old High German text only has a subordinate function vis-à-vis the Latin, as can also be seen with respect to initials (see below). According to Simmler (2003: 2477), Old High German punctuation has to be viewed not as rhetorical, but rather syntactic punctuation, i.e., syntactic units not rhetorical ones are separated by punctuation. If this view is correct, the value of Old High German punctuation with respect to prosody is rather small. Clearly, more work is needed here.

Prosodic information in Old High German manuscripts

181

Capital and initial letters might also provide information relevant to prosody, although they are primarily a device for structuring text: capitals and initials indicating the beginning of a new paragraph might for example correspond to a longer pause in the lecture. Kleiber has shown that initial and capital letters of different heights and scripts play a fundamental role with respect to text structuring in the Viennese and Palatinate codices of Otfrid’s Evangelienbuch (Kleiber 1971: 53–63; see also Bästlein 1991: 233–242, Kleiber 2004b: 67). The same seems to hold for the different manuscripts of the Old Saxon Heliand (see Bästlein 1991: 218–226). In the bilingual Latin–Old High German Tatian manuscript, initials and capitals are only encountered in the Latin part (see Masser 1994: 14); this part clearly dominates in terms of text structuring features, many of which are present only in the Latin portion (see Simmler 1998: 307, 333). Finally, another class of phenomena that might provide information about prosodic properties are assimilations and other phonological processes. For instance, there are different forms of personal pronouns in various Old High German manuscripts, some of which do not form a syllable (e.g., the form ih ‘I’ also appears as a shortened -h). Instances of such forms can be interpreted as unstressed and provide solid evidence for viewing the respective forms as the written expression of clitics (see Nübling 1992: 339). Another potential source of information is Notker’s Anlautgesetz. Simplifying somewhat, according to this allophonic rule, /b/, /d/ < Germanic *þ, and /g/ are realized as

, , and respectively if they are at the beginning of a sentence or clause or if the preceding word ends in an obstruent, but otherwise as , , and respectively, i.e. after sonorants (Braune and Reiffenstein 2004: 105; see also Lahiri and Kraehenmann 2004: 3). Thus, if the unexpected grapheme occurs after a sonorant, this might indicate a boundary between two intonation units. Lahiri & Kraehenmann (2004: 11), discussing the St. Gall manuscript of De nuptiis Philologiae et Mercurii, observe that in a few instances in which

comes after a sonorant instead of , the sonorant is in fact followed by a punctuation sign.

6. Outlook This survey of potential clues to prosodic information in Old High German manuscripts is probably far from complete, and I have not been able to treat all the features exhaustively or in the required depth. Still, I hope to have shown that accents, word separation and other paleographical pheno-

182

Jürg Fleischer

mena potentially provide an interesting resource for prosodic research. The difficulties in interpreting such paleographic features and the danger of circular argumentation must be emphasized, however. So far very little systematic work exploring the paleographical phenomena discussed here and their potential value for prosodic studies has been carried out. Further research in this area, based on careful analyses of the manuscripts, could lead to interesting new insights. Acknowledgments The problems and ideas discussed in this paper first captivated me in December 2003, when I was invited by the project Die Rolle der Informationsstruktur bei der Herausbildung von Wortstellungsregularitäten im Germanischen (part of the Sonderforschungsbereich 632: Informationsstruktur: Die sprachlichen Mittel der Gliederung von Äußerung, Satz und Text), then in its initial phase, to discuss some of the methodological issues facing the project. Since then, the same questions have kept reappearing and have been discussed intensively on many occasions. For help and suggestions regarding the present paper I am grateful to Brigitte Bulitta, Oliver Ernst, Nikolaus Henkel, Roland Hinterhölzl, Astrid Kraehenmann, Thomas Krisch, Andreas Nievergelt, Svetlana Petrova, Oliver Schallert, Eva Schlachter, and Michael Solf. Needless to say, all remaining errors are my responsibility. Finally, many thanks are due to Mark Pennay for smoothing out my English and to the Stiftsbibliothek St. Gallen for allowing me to reproduce the illustrations. My work on Old High German syntax was financed by a Fellowship for Advanced Researchers granted by the Swiss National Science Foundation (PA001--105063), which allowed me to spend a very fruitful time at the Humboldt-Universität zu Berlin; it is herewith gratefully acknowledged. Notes 1.

2.

Part of this passage (namely, amplius enim intellectus instruitur quando uox legentis quiescit) is quoted and translated by Parkes (1993: 21); the translation of this passage is taken from there. This might have been different in the western, Romance-speaking parts of the empire, where Latin, according to McKitterick (1989: 21), “was in no sense a second or foreign language”, but “continued to be understood.” (see also Saenger 1997: 101).

Prosodic information in Old High German manuscripts 3.

183

Due to the apparent lack of work relating to accent usage in Carolingian manuscripts I will sometimes refer to the works relating to the Beneventana. Since especially the monastery of Montecassino, one of the leading scriptories using this script, influenced the whole monastic world of its time (Bischoff 1986: 151), this can to some extent be justified, although it must be stressed that most of the evidence quoted is younger than the Old High German records with which we are dealing. Still, some parallels can be observed, which makes comparison worthwhile, its anachronism and geographic distance notwithstanding. 4. Due to space limitations I cannot discuss the (early Middle High German) accentuation system found in the various Williram manuscripts, which seems to be an offspring of Notker’s system (see Kruse 1913: 8, Gärtner 1991: 47, Schneider 2007: 19). For a discussion of this system see Kruse (1913) and Gärtner (1991: 45–55). Most interestingly from a sentence-prosodic perspective, it seems to be clear for this system that monosyllabic function words displaying an accent bear sentence stress (see Gärtner 1991: 49–50). 5. In this passage Notker mentions only articulos, but in his actual usage it is not only articles that do not bear accents. One might wonder whether Notker’s Medieval Latin usage of articulos encompasses other parts of speech in addition to articles. Unfortunately King (1986) and Hellgardt (1979) do not discuss this question; in his New High German translation, Sonderegger (1970: 83) renders articulos as “Artikel”. 6. According to Gabriel (1969: 64–65) Notker in his diphthong accentuations éi, éu, óu and íu as opposed to ûo, îe, îa, and îo also intended to render pitch differences. This view is rejected by King (1986: XLI–XLII). 7. In this article texts are quoted from reliable editions (Otfrid: Kleiber 2004a) or from manuscript sources (Tatian: Cod. Sang. 56; Notker, Boethius: Cod. Sang. 825); in the latter case the page and line of the respective codices is quoted. This information easily allows to find the respective references in the editions by Masser (1994) and King (1986), which are diplomatic; they are therefore not quoted additionally. 8. Since the examples quoted are usually quite long, no gloss indicating the grammatical categories is provided; the only abbreviations used are NEG = “negative particle” and PREF = “verbal prefix”. The meaning of the examples is clarified by more idiomatic translations. If an Old High German example is phrased in close accordance to a biblical passage, the idiomatic translation is taken from the King James Version. 9. The illustrations of this article are all details from manuscripts hosted at St. Gall’s Abbey library (Stiftsbibliothek St. Gallen); facsimiles can be accessed via: http://www.cesg.unifr.ch/virt_bib/manuscripts.htm 10. In the youngest Otfrid manuscript the different accents are not distinguished any longer as far as their paleographical shape is concerned (see Sievers 1909: 8). As a matter of fact, in this manuscript the scribe put (rhythmic) ac-

184

11.

12.

13.

14.

15.

16. 17.

Jürg Fleischer cents nearly randomly, obviously not knowing what he was doing (see Pivernetz 2000: 121). Some systematization can be seen in his use of accents on íá, íó, and íú, however (see Pivernetz 2000: 122); thus, the principles of phonetic accentuation seem to have been followed to some extent. The different accents are distinguished in print in Kleiber’s (2004a) edition by use of two different acute accents symbols. In the present article this cannot be reproduced. In order to distinguish Otfrid’s different types of accents, the “rhythmic accents” (along with the letters above which they are placed) are printed in bold. Incidentally, there is a third type of accent of somewhat unclear function in the Viennese manuscript, disregarded here. It occurs only in a small portion of the text, while the phonetic and rhythmic accents are used throughout the codex (see Kleiber 2004b: 120). For instance, it seems that in direct speech, mentioned by Masser (1997: 60) with respect to scribe ȗ of the Tatian manuscript, more accents are encountered as compared to other passages. The same phenomenon seems to be observed in the Beneventana (see Newton 1999: 181). This example is taken from Loew (1914: 277, note 1), where the same is said with respect to the Beneventana. According to Saenger (1997: 56), this disambiguating function of the accent became especially prominent in the late tenth century; Schneider (1999: 93, 2007: 21), referring to German manuscripts, describes this function for the thirteenth century, too. Incidentally, the accent on ós ‘mouth’ might also be used to distinguish ǀs ‘mouth’ from ǂs ‘bone’; this function of the accent is noticeable in the writings of Isidore (Keller 1908: 103; see also Keller 1908: 106, Gabriel 1969: 49). According to Saenger (1997: 98), who refers to European vernacular languages in general, but particularly also mentions “Old German”, acute accents were employed “to supplement the deficiencies of the Roman alphabet and to avoid homographs for words distinguished by vowel length.” The use of an acutelike symbol (termed apex originally, but eventually lumped together with the acute sign) to avoid confusion with a homograph displaying a short vowel seems to be very old (see Keller 1908: 103–104 and Saenger 1997: 54, who quote a passage by Quintilian, c. 40–c. 118). Newton (1999: 181) mentions a similar function with respect to the Beneventana. Illustrations of text from the Tatian manuscript (Cod. Sang. 56) which contain both Latin and Old High German text (Illustrations 2, 4 and 5) are reproduced with the Latin text before the Old High German text. This differs from the layout in the manuscript, but allows the relevant paleographic details to be shown in a greater magnification. Apart from scribe ȕ this usage can only be found three times in the portion of scribe Ȗ in the St. Gall Tatian manuscript (see Sievers 1909: 15). By “separation” I am referring to separation by spaces only. An alternative and/or supplementary indication of word or syllable separation by points usually called prosodiae (see Saenger 1990: 55–56, 1997: 26, 53–58) is at-

Prosodic information in Old High German manuscripts

185

tested in some Old High German records (see Krotz 2002: 119, Voetz 2006: 57). 18. It seems that in glosses and interlinear versions spaces might perform a rather different function, namely, to indicate the morphological structure of the Latin words that were glossed (see Henkel 1996: 59–60). 19. This even holds for the otherwise very faithful edition of the Viennese Otfrid manuscript by Kleiber (2004a; see Voetz 2006: 60). 20. An equivalent to punctuation by means of punctuation symbols is an arrangement principle known as per cola et commata that goes back to Jerome (see Müller 1964: 28, 70, Parkes 1993: 35). It is frequently encountered in Vulgate manuscripts. In this arrangement, every constituent to be separated by pauses from other constituents is written on a new line; the needed information to recite a text with the appropriate pauses is thus easily provided (see Müller 1964: 70–71). Among Old High German records the Tatian manuscript displays this kind of arrangement (see Bischoff 1986: 48).

References Bästlein, Ulf Christian 1991 Gliederungsinitialen in frühmittelalterlichen Epenhandschriften: Studie zur Problematik ihres Auftretens, ihrer Entwicklung und Funktion in lateinischen und volkssprachlichen Texten der Karolinger- und Ottonenzeit. Frankfurt am Main etc.: Peter Lang. Bischoff, Bernhard 2004 Reprint. Paläographie des römischen Altertums und das abendländischen Mittelalters. 2., überarbeitete Auflage. Berlin: Erich Schmidt. Original edition, Berlin: Erich Schmidt, 1986. Braune, Wilhelm and Ingo Reiffenstein 2004 Althochdeutsche Grammatik I: Laut- und Formenlehre. 15. Auflage bearbeitet von Ingo Reiffenstein. Tübingen: Niemeyer. Bruckner, Albert 1980 Akzente. Lexikon des Mittelalters. Vol. 1, 259–260. München/Zürich: Artemis. Fleischer, Oskar 1882 Das accentuationssystem Notkers in seinem Boethius. Zeitschrift für deutsche Philologie 14: 129–172, 285–300. [also partially published as diss. 1882 Halle: Waisenhaus.] Frey, Evelyn 1988 Wortteilung und Silbenstruktur im Althochdeutschen. Mit einem Anhang zur mittelhochdeutschen „Speculum ecclesiae“-Handschrift. Diss. München: Kopier- und Druckzentrale.

186

Jürg Fleischer

Gabriel, Eugen 1969 Die Entwicklung der althochdeutschen Vokalquantitäten in den oberdeutschen Mundarten. Wien: Österreichische Akademie der Wissenschaften. Gärtner, Kurt 1991 Die Williram-Überlieferung als Quellengrundlage für eine neue Grammatik des Mittelhochdeutschen. Zeitschrift für deutsche Philologie 101, Sonderheft: Mittelhochdeutsche Grammatik als Aufgabe, ed. Klaus-Peter Wegerea, 23–55. Green, D. H. 1994 Medieval Listening and Reading: The Primary Reception of German Literature 800–1300. Cambridge: Cambridge University Press. Hellgardt, Ernst 1979 Notkers des Deutschen Brief an Bischof Hugo von Sitten. In Befund und Deutung. Zum Verhältnis von Empirie und Interpretation in Sprach- und Literaturwissenschaft, eds. Klaus Grubmüller, Ernst Hellgardt, Heinrich Jellissen and Marga Reis, 169–192. Tübingen: Niemeyer. Henkel, Nikolaus 1996 Die althochdeutschen Interlinearversionen: zum sprach- und literaturhistorischen Zeugniswert einer Quellengruppe. In Übersetzen im Mittelalter. Cambridger Kolloquium 1994, eds. Joachim Heinzle, L. Peter Johnson and Gisela Vollmann-Profe, 46–77. Berlin: Erich Schmidt. Illich, Ivan 1993a In the Vineyard of the Text: A Commentary to Hugh’s Didascalicon. Chicago and London: The University of Chicago Press. 1993b Lectio Divina. In Schriftlichkeit im frühen Mittelalter, ed. Ursula Schaefer, 19–35. Tübingen: Narr. Keller, Wolfgang 1908 Über die Akzente in den angelsächsischen Handschriften. In Untersuchungen und Quellen zur germanischen und romanischen Philologie. Johann von Kelle dargebracht von seinen Kollegen und Schülern. Erster Teil, 97–120. Prag: Bellmann. King, James C. 1986 Die Akzente bei Notker und in der Fachliteratur, besonders bei Sehrt und Starck. In Notker der Deutsche, Boethius, »De consolatione Philosophiæ«, ed. Petrus W. Tax, XXXVII–XLIII. Tübingen: Niemeyer. Klaper, Michael 2004 Musikhistorische Interpretation. In Kleiber 2004b: 148–153.

Prosodic information in Old High German manuscripts

187

Kleiber, Wolfgang 1971 Otfrid von Weißenburg: Untersuchungen zur handschriftlichen Überlieferung und Studien zum Aufbau des Evangelienbuches. Bern/München: Francke. 2004a Otfrid von Weißenburg: Evangelienbuch. Band I: Edition nach dem Wiener Codex 2687. Herausgegeben und bearbeitet von Wolfgang Kleiber unter Mitarbeit von Rita Heuser. Teil 1: Text. Tübingen: Niemeyer. 2004b Otfrid von Weißenburg: Evangelienbuch. Band I: Edition nach dem Wiener Codex 2687. Herausgegeben und bearbeitet von Wolfgang Kleiber unter Mitarbeit von Rita Heuser. Teil 2: Einleitung und Apparat. Mit Beiträgen von Wolfgang Haubrichs, Norbert Kössinger, Otto Mazal, Norbert H. Ott und Michael Klaper. Tübingen: Niemeyer. Krotz, Elke 2002 Auf den Spuren des althochdeutschen Isidor: Studien zur Pariser Handschrift, den Monseer Fragmenten und zum Codex Junius 25. Mit einer Neuedition des Glossars Jc. Heidelberg: Winter. Kruse, Heinrich 1913 Die Accente in den Handschriften von Willirams Übersetzung und Auslegung des hohen Liedes. Diss. Greifswald: Adler. Lahiri, Aditi and Astrid Kraehenmann 2004 On maintaining and extending contrasts: Notker’s Anlautgesetz. Transactions of the Philological Society 102: 1–55. Loew, Elias A. 1980 Reprint. The Beneventan Script. A History of the South Italian Minuscule. Second edition prepard and enlarged by Virginia Brown. Volume I: Text. Roma: Storia e Letteratura. Original edition, Oxford: Clarendon, 1914. Masser, Achim 1994 Die lateinisch-althochdeutsche Tatianbilingue Stiftsbibliothek St. Gallen Cod. 56. Göttingen: Vandenhoeck & Ruprecht. 1997 Wege zu gesprochenem Althochdeutsch. In Grammatica Ianua Artium. Festschrift für Rolf Bergman zum 60. Geburtstag, eds. Elvira Glaser and Michael Schlaefer, 49–70. Heidelberg: Winter. McKitterick, Rosamond 1989 The Carolingians and the Written Word. Cambridge: Cambridge University Press. Müller, Rudolf Wolfgang 1964 Rhetorische und syntaktische Interpunktion. Untersuchung zur Pausenbezeichnung im antiken Latein. Diss. Tübingen.

188

Jürg Fleischer

Newton, Francis 1999 The Scriptorium and Library at Monte Cassino, 1058–1105. Cambridge: Cambridge University Press. Nübling, Damaris 1992 Klitika im Deutschen: Schriftsprache, Umgangssprache, alemannische Dialekte. Tübingen: Narr. Parkes, Malcolm B. 1993 Pause and Effect. An Introduction to the History of Punctuation in the West. Berkeley/Los Angeles: University of California Press. Piper, Paul 1882 Zu Otfrid: 1. Otfrids accente. Beiträge zur Geschichte der deutschen Sprache und Litteratur 8: 225–244. Pivernetz, Karin 2000 Otfrid von Weißenburg. Das ‘Evangelienbuch’ in der Überlieferung der Freisinger Handschrift (Bayerische Staatsbibliothek München, cgm. 14). Edition und Untersuchungen. II: Untersuchungen. Göppingen: Kümmerle. Saenger, Paul 1990 The separation of words and the order of words: the genesis of medieval reading. Scrittura e Civilità 14: 49–74. 1997 Space between Word: The Origins of Silent Reading. Stanford: Stanford University Press. Scholz, Manfred Günter 1980 Hören und Lesen. Studien zur primären Rezeption der Literatur im 12. und 13. Jahrhundert. Wiesbaden: Steiner. Schneider, Karin 1999 Paläographie und Handschriftenkunde für Germanisten: eine Einführung. Tübingen: Niemeyer. 2007 Akzentuierung in mittelalterlichen deutschsprachigen Handschriften. In Edition und Sprachgeschichte. Baseler Fachtagung 2.–4. März 2005, eds. Michael Stolz, Robert Schöller and Gabriel Viehhauser, 17–24. Tübingen: Niemeyer. Sievers, Paul 1909 Die Accente in althochdeutschen und altsächsischen Handschriften. Berlin: Mayer & Müller. Simmler, Franz 1997 Interpungierungsmittel und ihre Funktionen in der Lorscher Beichte und im Weißenburger Katechismus des 9. Jahrhunderts. In Grammatica Ianua Artium. Festschrift für Rolf Bergman zum 60. Geburtstag, eds. Elvira Glaser and Michael Schlaefer, 93–114. Heidelberg: Winter.

Prosodic information in Old High German manuscripts 1998

189

Makrostrukturen in der lateinisch-althochdeutschen Tatianbilingue. In Deutsche Grammatik – Thema in Variationen. Festschrift für HansWerner Eroms zum 60. Geburtstag, eds. Karin Donhauser and Ludwig M. Eichinger, 299–335. Heidelberg: Winter. 2003 Geschichte der Interpunktionssysteme im Deutschen. In Sprachgeschichte. Ein Handbuch zur Geschichte der deutschen Sprache und ihrer Erforschung. 2., vollständig neu bearbeitete und erweiterte Auflage, 3. Teilband, eds. Werner Besch, Anne Betten, Oskar Reichmann and Stefan Sonderegger, 2472–2504. Berlin/New York: Walter de Gruyter. Sobel, Naphtali 1882 Die Accente in Otfrids Evangelienbuch: eine metrische Untersuchung. Strassburg: Trübner. Sonderegger, Stefan 1970 Althochdeutsch in St. Gallen: Ergebnisse und Probleme der althochdeutschen Sprachüberlieferung in St. Gallen vom 8. bis ins 12. Jahrhundert. St. Gallen: Ostschweiz, Sigmaringen: Thorbecke. Stotz, Peter 2002 Handbuch zur lateinischen Sprache des Mittelalters. Erster Band: Einleitung, lexikologische Praxis, Wörter und Sachen, Lehnwortgut. München: Beck. Tax, Petrus W. ed. 1986 Notker der Deutsche: Boethius, “De consolatione Philosophiæ” Buch I/II. Tübingen: Niemeyer. 1996 Notkers Brief. In Notker der Deutsche: Die kleineren Schriften, herausgegeben von James C. King und Petrus W. Tax, 347–349. Tübingen: Niemeyer. Voetz, Lothar 2006 Einige Beobachtungen zur Getrennt- und Zusammenschreibung im Althochdeutschen. In Neue Perspektiven der Sprachgeschichte: Internationales Kolloquium des Zentrums für Mittelalterstudien der Otto-Friedrich-Universität Bamberg 11. und 12. Februar 2005, eds. Ursula Götz and Stefanie Stricker, 51–64. Heidelberg: Winter. Zürcher, Josef 1978 Graphetik – Graphemik – Graphematik unter besonderer Berücksichtigung von Notkers Marcianus Capella. Diss. Zürich: akuFotodruck.

On the “syntax of silence” in Proto-Indo-European Thomas Krisch

Abstract Using material from the ancient Indo-European languages Latin, Greek, Sanskrit and Hittite and insights of modern linguistic theory, this paper discusses two phenomena of ellipsis: gapping and object ellipsis. Both kinds of ellipsis are shown to be operative in these languages and can be reconstructed for Proto-Indo-European. Gapping appears in the two variants of forward and backward gapping in the languages discussed. In accordance with observations in literature using generative theory, which tell us that backward gapping is only possible in SOV languages, we conclude that backward gapping is a further piece of evidence for an underlying SOV structure of these languages and of Proto-Indo-European. The fact that there exists forward gapping in all of the languages discussed is interpreted as a reflex of a V-to-C movement. Overt V-to-C-movement is only attested in Latin, Greek and Sanskrit and not in Hittite, though. This fact is interpreted as a Hittite innovation, forward gapping being a remnant of Proto-Indo-European V-to-C-movement in this language. Object ellipsis operates in forward direction and depends on factors of functional sentence perspective. 1. Introduction1 This paper deals with some aspects of the “syntax of silence” (ellipsis) in Indo-European. There are not many investigations into this phenomenon for ancient Indo-European languages and also for modern languages there is still much work to be done. During the last decades, Generative Grammar, though, has made some substantial contributions to our understanding of ellipsis where it is syntactically conditioned. The moderate aims of the present paper are to discuss some theoretical issues, to describe some phenomena appearing in ancient Indo-European languages (AIELs) and to reconstruct some elliptical constructions for Proto- Indo-European (PIE).

192

Thomas Krisch

In this paper, I leave aside the ubiquitous ellipsis of the verb “to be” in AIELs and refer to Karl Praust’s recent illuminating paper (Praust 2003), where he reconstructs a PIE (phonologically) silent injunctive of the verb “to be”, which he sees reflected in ancient IE verbless nominal sentences. I here also leave aside the question of an ellipsis of the verb in case of verbs of movement with preverbs (cf. e.g. Krisch 1984: 93-94) in AIELs. This paper also does not deal with the well known type of ellipsis in more or less lexicalized NPs of AIELs where the head (the N) is not expressed and an originally attributively used adjective is nominalized.2

2. Ellipses and implicatures Ellipses are a performance factor of natural languages. Speakers and (to a little lesser degree) also writers of texts often use them. Speaker and hearer have the tacit knowledge that certain parts of utterances may remain unexpressed. Modern relevance theory claims that speakers/writers do not express truth values in a direct way but rather suggest truth values (“schematic logical forms” cf. Blakemore 2002: 77) to the hearer/reader who has to complete the utterance by deriving inferences if necessary.3 Speaker and hearer follow the principle of maximal relevance of the utterance in the context and the speaker as well as the hearer expect implicatures. Implicatures are involved even in seemingly complete utterances as in (1): (1)

Mein Schwager wohnt in München, aber seine Tante ist Ärztin in Wien. “My brother-in-law lives in Munich but his aunt is a medical doctor in Vienna”

The meaning of this sentence contains the truth values of both sentences that form part of it. Thus, the sentence is only true iff the speaker’s brotherin-law lives in Munich and iff his aunt is a medical doctor in Vienna. But in addition to that, the hearer/reader can draw the implicature (activated by the conjunctive particle aber “but”) that there is some conflict between the utterance Mein Schwager wohnt in München “my brother-in-law lives in Munich” and seine Tante ist Ärztin in Wien “his aunt is a medical doctor in Vienna”, and the reader/hearer will worry about the reason of this conflict.4 This type of implicature triggered by the semantics of certain elements of speech has been called “conventional implicature” for the last 30 years.5 The other type of implicature (“conversational implicature”) is triggered by the context and by encyclopaedic knowledge. Cf. the much cited sentence in (2):

On the “syntax of silence” in Proto-Indo-European 193 (2)

Herr M. beherrscht seine Muttersprache und hat meine Lehrveranstaltungen regelmäßig besucht. “Mr. M. has a command of his mother tongue and regularly attended my lectures”.

If you utter this sentence in the context of an expert opinion to a scholarship this is a negative statement, namely Mr. M. does not have any other merits. If one sees the notions of conventional implicature vs. conversational implicature as poles of a continuum, our paper looks at examples which are nearer to the pole of conventional implicature.

3. Remarks on the history of generative research in ellipsis The most influential researcher in this field has been John Robert (‘Haj’) Ross (e.g. Ross 1970). Among other things he introduced the notion of “gapping” to denote ellipsis of the verb caused by coordinate constructions. I adopt Richards’ definition of gapping here: “Gapping involves ellipsis of a portion of the verb phrase, including the verb but excluding one or more VP-internal constituents” (Richards 1998: 158). The direction of gapping depends on the syntactic type. Backward gapping is only allowed in SOV languages (cf. (3c)), whereas forward gapping may appear in SVO (cf. the English examples in (3a)), SOV (cf. German subordinate clauses in (3b)) or VSO6 languages (cf. the Irish example in (3d)). Only SOV languages allow gapping in both directions (cf. (3b) vs. (3c)). In our examples the element(s) that can be gapped is/are put in deleted capital letters. (3)

a. English (SVO) John likes fish and Peter LIKES meat. Max seemed to be trying to begin to make love to Harriet and Fred SEEMED TO BE TRYING TO BEGIN TO MAKE LOVE to Sue. b. German (SOV) Ich weiß, dass die Kinder Fisch lieben und I know that the children fish love and die Eltern Fleisch LIEBEN the parents meat LOVE “I know that the children love fish and the parents (love) meat.” c. Ich weiß, dass die Kinder Fisch LIEBEN und and I know that the children fish LOVE die Eltern Fleisch lieben. the parents meat love “I know that the children (love) fish and the parents love meat.”

194

Thomas Krisch d. Modern Irish (VSO) (cf. Steedman 2000:177) Chonaic Eoghan Siobhán agus CHONAIC Saw Eoghan Siobhán and SAW Ciarán Ciarán “Eoghan saw Siobhán and Eoghnaí (saw) Ciarán”

Eoghnaí Eoghnaí

Gapping thus seems to correlate with the directionality of verbal government. In languages with government to the right (SVO, VSO) only forward gapping is allowed, whereas SOV languages with government to the left show backward gapping. A subclass of SOV languages like German allow a verb before the object in some clause types (in German especially in main declarative clauses), cf. (4): (4)

Die The

Kinder lieben children love

Fisch. fish

In generative grammar, this characteristic has been explained by a movement of the verb from original position after its object inside the verb phrase (VP) into a position called “C(omplementizer) position” (cf. also 4.3). The result may be a structure like (4), which superficially looks like an SVO structure.7 By syntactic analogy (abduction8), then, the same gapping pattern as in SVO languages, namely forward gapping can be applied to these SOV languages if the verb in the first sentence of the conjoined sentences appears in final position (cf. (3b)).

4. Gapping as a phenomenon of Proto-Indo-European syntax 4.1. Examples from ancient Indo-European Languages Gapping is, indeed, a type of ellipsis that can be traced back to the PIE language. Let us first look at some examples from AIELs. A sentence with backward gapping that reoccurs in the Old Hittite9 ritual corpus a number of times is (5a). There likewise exists forward gapping in Hittite (cf. (5b)):

On the “syntax of silence” in Proto-Indo-European 195 (5)

Hittite a. (backward ellipsis) StBoT 12 (Neu 1970; Hittite thunderstorm ritual), Rs. III, 21´ G]IŠ DINNANA.GAL.GAL LÚ.MEŠIal-li-ri-e[„ i„-]Ia-mi-an-zi “The big I„tar-instrument [probably a lyra, TK] (resounds), the priest-singers resound” LÚ.MEŠ GIŠ DINNANA GAL.GAL IŠ)AMAI Iallirie„ I„tar-instrument big.big RESOUNDS priest-singer.NOM.PL i„Iamianzi resound.PRS.3PL b. (forward ellipsis) StBoT 12 (Neu 1970; Hittite thunderstorm ritual), Rs. III, 38´: LUGAL-uš Iu-u-up-pa-ri ši-pa-an-ti MUNUS.LUGAL-ša n[a-at-]ta “The king libates into a tureen, and the queen (does) n[o]t (libate).” LUGAL-uš IVppari šipanti MUNUS.LUGAL-ša king.NOM.SG tureen.LOC.SG libate.3SG.PRS queen-and[CLT.] n[at]ta ŠIPANTI not LIBATES

In Latin, too, one can observe backward and forward ellipses, cf. the examples (6a) and (6b): (6)

Latin a. (backward ellipsis) (cf. Gaeta and Luraghi 2001: 95) Caes. Bell.Gall. 1,40,13 (AcI in indirect speech, a speech of Julius Cesar in a war assembly) suam innocentiam perpetua vita, felicitatem Helvetiorum bello esse perspectam. “(that) his [scil. Ceasar’s TK] selflessness (was to be seen) in (his) whole life, (his) success was to be seen in the war against the Helvetians” suam innocentiam perpetua selflessness.ACC.SG.F. continuous.ABL.SG his.ACC.SG.F.REFL felicitatem vita ESSE PERSPECTAM success.ACC.SG.F life.ABL.SG. TO BE SEEN Helvetiorum bello esse war.ABL.SG be.INF Helvetians.GEN.PL perspectam seen.PTCP.PRF.PASS.ACC.SG.F

196

Thomas Krisch b. (forward ellipsis) Plaut. Mil. 990 Viden tu illam oculis venaturam facere atque aucupium auribus? “Do you see her making a hunt with her eyes and (making) a bird-hunt with her ears?” Vide-n tu illam oculis venaturam facere atque eye.ABL.PL hunt.ACC.SG make.INF and See.2SG-Q you her auribus aucupium FACERE ear.ABL.PL bird-catching.ACC.SG. MAKE

Also Ancient Greek provides attestations for both, backward and forward gapping, as the following examples illustrate: (7)

Ancient Greek a. (backward ellipsis) Hdt. 2,56,1 (cf. Gaeta and Luraghi 2001: 105)

Ei alFth‚Ps hoi Phoínikes exÇgagon t…s hir…s gynaĩkas kaì t¼n mèn autéPn es LibýFn, t¼n dè es t¼n Hell@da apédonto, … “If the Phoenicians did in fact carry away the sacred women and (sell) one of them in Libya and sell one in Hellas, …”

Ei  alFth‚Ps hoi  Phoínikes  If

truly

exÇgagon 

the Phoenicians they.carried.away

gynaĩkas  kaì 

t¼n 

mèn 

autéPn

t…s  hir…s the sacred

es

women

and this.ACC.SG.F EMPH of.them [GEN.PL] to APÉDONTO t¼n  dè  es t¼n Libya.ACC SOLD this.ACC.SG.F but to ART.ACC.SG.F

LibýFn

Hell@da 

apédonto

Greece.ACC.SG.F sold.3PL.AOR.MID b. (forward ellipsis) Hdt. 2,180,2 (cf. Gaeta and Luraghi 2001: 100)

… Ámasis mèn gár sphi édPke chília styptFríFs tálanta, hoi dè en AigýptPi oikéontes HéllFnes eíkosi mnéas … for Amasis gave them a thousand talents’ weight of alum, but the Greek settlers in Egypt (gave them) twenty minae (of silver).” Ámasis mèn gár sphi édPke Amasis.NOM.SG EMPH for them.DAT gave.3SG.AOR tálanta hoi dè chília styptFríFs thousand alum.GEN.SG talent.ACC.PL ART.NOM.PL.M but oikéontes HéllFnes en AigýptPi in Egypt.DAT live.PTCP.PRS.NOM.PL.M Greek.NOM.PL.M mnéas SPHI ÉD0SAN eíkosi THEM GAVE twenty minae.ACC.PL.F

On the “syntax of silence” in Proto-Indo-European 197

In (7b), one may observe two types of elliptical expressions gapping of the verb édPsan including one of its arguments and also the usual ellipsis of “silver” with the unit of measurement “Mnéa”. Vedic Sanskrit, too, shows gapping in both directions, cf. (8a) and (8b): (8)

Vedic Sanskrit a. (backward ellipsis) (cf. Gaeta and Luraghi 2001: 96) RV 6,75,2 dh@nvanà gx dh@nvanBj¡ê jayema “with the bow (we want to win) the cows with the bow we want to win the battle.”

dh@nvanà

gx

JAYEMA

dh@nvanà

bow.INS.SG  cow.ACC.PL.F WE.WANT.TO.WIN  bow.INS.SG

Bj¡ê

jayema

battle.ACC.SG win.1PS.PL.OPT.PRS b. (forward ellipsis) RV 10,42,10

g¢bhiù ñaremxmatiê dur‚vàê y@vena kù£dham puruhåta v¡÷vàm “Through cows we want to overcome malignant helplessness, through corn (we want to overcome) all (kinds of) hunger, o much invoked one!”

g¢bhiù

ñarema

@matiê

cow.INS.PL overcome.1PL.OPT.PRS

dur‚vàê

y@vena

helplessness.ACC.SG.F

TAREMA

malignant.ACC.SG.F corn.INS.SG WE.WANT.TO.OVERCOME

kù£dham

puruhåta

hunger.ACC.SG.F

much.invoked.one.VOC.SG all.ACC.SG.F

v¡÷vàm

4.2. Sloppy identity in ancient Indo-European languages The Hittite example (5a) as well the Greek example (7b) show a phenomenon well attested in modern languages: the elliptical form does not need to be completely identical to its model. In (5a), the overt verbal form i„Iamianzi is a third person plural form but one has to recover a third person singular (išIamai) by conventional and conversational implicature. In (7b), the overt verbal form édPke by (conventional and conversational) implicature triggers a third plural form édPsan to be recovered. This phenomenon of only partial identity of the overt and the recovered element has been termed “sloppy identity” in the literature. (5a) and (7b) show that this property of ellipsis also existed in the AIELs.

198

Thomas Krisch

4.3. Backward and forward gapping as a PIE phenomenon and SOV The examples given above all belong to the four ancient Indo-European language families that are the most important ones for the reconstruction of PIE syntax10 (Anatolian, Italic, Greek and Indo-Aryan). They all show the same phenomena: backward and forward gapping. It is therefore safe to reconstruct both possibilities for the Proto-language. The question why PIE and the daughter languages allowed both directions of gapping may be answered by the predominantly SOV character of PIE (cf. also section 3. above)11 and its daughter languages Hittite, Latin, Greek and Vedic Sanskrit and by the fact that there exists V to C movement in all of these languages (as in the German example (4) above) except Hittite, cf. the examples (7b) and (8b) above and the examples in (9) (the moved verb is put in bold letters): (9)12

a. Latin Ter. Ph. 594–596 vixdum dimidium dixeram, intellexerat / gaudebat, me laudabat. quaerebat senem, / dis gratias agebat ... “I scarcely had said half (of it) he had understood. He was glad, praised me, looked for the old man, thanked the gods ...” me laudabat, quaerebat senem look.for.3SG.IPF old.man.ACC.SG me.ACC.SG praise.3SG.IPF b. Greek Il. 18, 476–477 (cf. also Krisch 2001: 169-171) [... th©ken en akmothétᯉ mégan ákmona, génto dè cheirì / raist©ra krater¼n,] hetérēphi dè génto pyrágrēn “[he [scil. Hephaistos, TK] set on the anvil-block a great anvil, he seized a massive hammer with one hand,] but with the other hand he seized the fire-tongs” dè génto pyrágrēn hetérēphi with.the.other but seize.3SG.MID.AOR fire.tongs.ACC.SG.F c. Vedic Sanskrit RV 1,85,7

v¡ùõur y@d dh„vad vÆùaõam madacy£taü v@yo n@ sãdann @dhi barh¡ùi priy‚ “when Viùõu supported the bull reeling with excitement [scil. god Indra, TK], they [scil. the Marut, the storm-gods, TK] sat down like birds on the dear Barhis [i.e. the sacrificial grass, TK].”

v¡ùõur

y@d

Viùõu.NOM.SG

when EMPH

dha

„vad

vÆùaõam

support.3SG.IPF bull.ACC.SG.M

madacy£taü reeling.with.excitement.ACC.SG.M

On the “syntax of silence” in Proto-Indo-European 199

This point becomes clearer when one uses a generative model of description that I have applied to PIE several times in the last years.13 There are two main sentence structures (Figure 1) and (Figure 2) in the SOV-language PIE. If the verb is fronted to C-position (cf. (7b), (8b)) or if it remains in its original sentence final position in the first part of the sentence (cf. examples (5b), (6b)), then forward gapping can be applied in the second part of the coordinate construction. As pointed out in section 3. of this paper, I attribute this Janus-like behaviour of V in sentence final position to analogy (abduction). If the verb remains overt in sentence final position in the second part of the sentence, then only backward gapping can be triggered in coordinate constructions (examples (5a), (6a), (7a), (8a). In Table 1 and Table 2 the arrows indicate the direction(s) of the possible ellipses. CP (=S’) qp XP C’ TOPIC qp W1-W2 C IP (=S) ru XP I’ VP 6

(I)

...........

V

Figure 1. Structure of the PIE sentence with arrows indicating possible directions of ellipses CP (=S’)

CP (=S’) W1 XP TOPIC1 qp XP C’ qp TOPIC2 W2 C IP (=S) ru XP I’ (I)

VP 6 ...........

V

Figure 2. Structure of the PIE sentence with Chomsky adjunction and arrows indicating possible directions of ellipses

200

Thomas Krisch

4.4. Reconstructing the Proto-Indo-European situation I propose to reconstruct the situation of Latin, Greek and Vedic Sanskrit as the Proto-Indo-European state: a) PIE was an SOV language which allowed V-to-C-movement. b) PIE thus allowed not only for backward gapping (predicted by the direction of government in an SOV-language) but it allowed also for forward gapping. This peculiarity is reflected in Latin, Greek and Vedic Sanskrit. How does Hittite fit into this picture? I do not know of any good case in Old Hittite which would force us to assume V-to-C movement for this language. Perhaps the possibility of a forward ellipsis (5b) in Hittite represents a last piece of evidence in Hittite that V-to-C-movement once existed also in this branch of Indo-European. If this interpretation is correct, it would provide us with an instance where Hittite, the oldest attested AIEL, exhibits syntactic innovation.14

5. Ellipsis and information structure Apart from structural points of view (cf. sections 3. and 4. above), there exist a number of other approaches to deal with ellipsis in modern linguistic theory. I restrict myself here to approaches that interpret ellipsis as an epiphenomenon of the TOPIC-FOCUS distinction.15 This type of functional explanation in my view does not contradict structural views but supplements them. Kuno emphasizes the importance of TOPIC-FOCUS (/theme-rheme) for ellipsis in his famous “pecking order of deletion” (Kuno 1980:132) “Pecking order of deletion Delete order [sic! correct “older” TK] (less important) information first, and newer (more important) information last”. Klein’s formulation is a bit more precise: (10)

Klein (1993: 791) “Genau jene lexikalischen Einheiten, die eine beibehaltene Topik ausdrücken, können p-reduziert werden” (those and only those elements that express maintained topic/thematic information may be preduced).16

Merchant (2001) starts from a different approach but goes in a direction similar to Klein’s (1993) formulation. Merchant's focus condition on ellipsis (Merchant 2001: 38) reads “A constituent α can be deleted only if α is e-GIVEN” whereby e-GIVEN means (in a slightly simplified way) that

On the “syntax of silence” in Proto-Indo-European 201

there is an antecedent for α which entails the deleted constituent by an existentially bound variable and vice versa. If you take the sentence in (11), (11)

Abby sang.

you can abstract an existentially bound variable of the form in (12) (12)

∃x sing (x)

In a sentence like (13) (with VP-ellipsis), (13)

Abby sang because Ben did SING.

one can recover the verb SING because an existentially bound variable like the one in (12) may be abstracted (“∃-type shifting”). This view makes it a bit easier to understand the possibility of “sloppy identity” (as exemplified by the Greek example (7b)). The verbal form édPke “he gave” (3Sg.Aor) appears in the first part of the coordinate construction and thus represents thematic material for the second part of the coordination. An interpretation of édPke analogous to the one in (12) would be something like “there exists an x who gives”. This abstraction of a variable evidently is strong enough to make it possible to interpret the elided form as a plural (perhaps by thinking about the variable x as a collective expression). With an example taken from German, Klein (1993: 774) claims that for forward ellipses only the “lexical content” is important and not morphological marking. For backward ellipsis, Klein (1993: 797) formulates the rule in (14): (14)

“Identisches Endstück in parallelen Konjunkten kann beim ersten Vorkommen p-reduziert werden” (the first instance of an identical piece of the final part in parallel coordinated constructions can be p-reduced).

In other words, Klein claims that there must be complete formal identity of the elided part and the overt part in backward ellipsis (cf. Klein 1993: 773). The ungrammatical German example (15) (Klein 1993: 774) confirms this claim:17 (15)

*weil ich because I trinkst drink.2Sg.Prs

Bier beer

TRINKE DRINK.1Sg.Prs.

und du and you

Wein wine

202

Thomas Krisch

Our Vedic example [(8a), repeated as (16)], too, meets Klein's requirement in (14): the verbal form jayema “we want to win” which appears in the second part of the coordinate construction is exactly the form that is missing in the first part of the coordinate construction. (16)

Vedic Sanskrit RV 6,75,2

dh@nvanà gx dh@nvanBj¡ê jayema “with the bow (we want to win) the cows with the bow we want to win the battle.” dh@nvanà

gx

JAYEMA

dh@nvanà

bow.INS.SG cow.ACC.PL.F WE.WANT.TO.WIN 

Bj¡ê

jayema

battle.ACC.SG

win.1PS.PL.OPT.PRS

bow.INS.SG

But our Hittite example for backward ellipsis [(5a), repeated as (17)] shows sloppy identity in backward gapping: (17)

Hittite StBoT 12 (Neu 1970; Hittite thunderstorm ritual), Rs. III, 21´ G]IŠ DINNANA.GAL.GAL LÚ.MEŠIal-li-ri-e[„ i„-]Ia-mi-an-zi “The big I„tarinstrument [probably a lyra, TK] (resounds), the priest-singers resound” LÚ.MEŠ GIŠ DINNANA GAL.GAL IŠ)AMAI Iallirie„ I„tar-instrument big.big RESOUNDS priest-singer.NOM.PL i„Iamianzi resound.PRS.3PL

This means that Klein’s claim in (14) cannot be a universal one. Languages seem to differ in this respect. This can also be shown by the following Russian examples from Philippa Cook, taken from the description of a project of Lang (2006: 4): (18)

(Russian) a a. Ja nauŠnuju sta'tju !ITAJU I scientific article READ.1SG but Šitaet. read.3SG vodu i b. Ja *PIL / PILA I *DRANK.SG.M / F water and vodku vodka

on he

detektiv detective story

Anna Anna

pila drank.SG.F.

On the “syntax of silence” in Proto-Indo-European 203

Russian tolerates differences in person in backward ellipses (18a) but not differences in gender (18b). (18b) is only grammatical if ja “I” refers to a female person. The exact conditions for the appearance of backward gapping in AIELs and in PIE await further investigation. A thorough analysis of a large corpus of several AIELs will be necessary for such a project.

6. Object deletion as a phenomenon of Proto-Indo-European syntax This section deals with deleted objects in AIELs and with the possibility to reconstruct this type of ellipsis for the proto-language.

6.1. Latin data Consider first the Latin example in (19), taken from “Amphitruo”, a comedy by the greatest Roman comic playwright, Titus Maccius Plautus (254184 B.C.): (19)

Plaut. Amph. 387-392 (God Mercurius, having taken the shape of Sosia, a servant of Amphitruo, beats up Sosia, because Sosiaa (rightly) claims that hea is Sosia. Mercurius wants Sosiaa to deny that hea is Sosia ) Merc. Ego sum Sosia ille quem tu dudum esse aiebas mihi. Sos. Obsecro ut per pacem liceat te alloqui, ut ne vapulem. Merc. Immo indutiae parumper fiant, si quid vis loqui. Sos. Non loquar nisi pace facta, quando pugnis plus vales. Merc. Dic si quid vis, non nocebo. Sos. Tuae fide credo? Merc. Meae. Sos. Quid si falles? Merc. Tum Mercurius Sosiae iratus siet. “Merc. I am this Sosia about whom you just told me that you are him. Sos. I implore you that you allow me to speak to you in a friendly way, that you do not beat me up. Merc. Well, there shall be armistice for a short time if you want to say anything. Sos. I shall not speak unless peace has been made because you are the stronger one with your fists. Merc. Say SOMETHING, if you want TO SAY anything, I shall not harm YOU. Sos. May I believe in your honesty? Merc. Yes [lit. mine] Sos. What if you will deceive ME? Merc. Then Mercurius will be angry with Sosia.”

204

Thomas Krisch

It is evident from this example (cf. the deleted material in the translation) that the immediate context gives us hints to interpret the elliptical elements. Let us first take a closer look at line 391: (20)

Plaut. Amph. 391 si quid vis Dic (ALIQUID) Say.IMP.2SG (SOMETHING) if anything.ACC.SG want.2SG.PRS LOQUI, non TIBI nocebo YOU.DAT.SG harm.1SG.FUT SAY.INF.PRS not

In (20), the imperative dic is construed without an object, an “ellipsis” provided for by the Latin lexicon (actually changing the semantics of dic to “speak”18), and with the sentential adjunct (a conditional clause) si quid vis. In verse 389 the verbal form loqui (which has to be supplemented in verse 391) appears in the phrase si quid vis loqui uttered by the same speaker (Mercurius). The verb loqui in verse 389 thus represents thematic, already known material and therefore neatly meets Klein’s criterion cited in (10). This type of verbal ellipsis resembles forward gapping in a sense but goes beyond it since the distance between the antecedent and the gap is bigger, there is no coordinate construction involved and the elliptic element is not the finite auxiliary vis but the infinite part of the verbphrase (loqui). The Latin verb nocFre “harm” is normally construed with a direct object in the dative case19 “to harm somebody”. An “unspecific” reading (see also note 15) without the object in the dative case is achieved by using the figurae etymologicae noxam nocFre or noxiam nocFre, which mean something like “to harm a harm”.20 Since this construction is not present in our case (20) we have to assume an ellipsis of the dative object TIBI here. This ellipsis is also based on given thematic material. All of the sentences uttered by Mercurius before our example (20) contain a pronoun of the second person singular either explicitly (cf. (19) verse 387) or implicitely [as “pro” in the verbal form vis in verse 389 and in the verbal forms dic and vis in the immediate context before nocebo, cf. also (19)]. In Latin, the dative object with the verb nocēre behaves like a normal direct object. As a good piece of evidence for that one can show that, in passive constructions, this dative may be changed into a nominative case: (21) Vitr. 2,9,14 Larix ... ab suci vehementi amaritate ab carie aut tinea non nocetur ... “The larch ... is not harmed by decay or by a caterpillar because of the enormous bitterness of its sap ...” Larix .... non nocetur not harm.PRS.PASS.3SG larch.NOM.SG

On the “syntax of silence” in Proto-Indo-European 205

In verse 392 of (19), repeated as (22), the object of the verb falles (which normally governs an accusative case) is missing: (22)

Plaut. Amph. 392 Quid si (ME) What if (I.ACC.SG )

falles? deceive.FUT.2SG

Also this object can be easily retrieved by the context of a dialogue situation.21 The omission of direct objects is such a common phenomenon in Latin that quite complex constructions can be found, cf. (23). Here the verbs with missing objects are put in bold letters. The normal valency frames of these verbs are listed in (24). (23)

Tac. Ann. 12,46,18 – 12,47,1 (cf. Luraghi 1997: 250) qua necessitate Mithridates diem locumque foederi accepit castelloque egreditur. (47) Ac primo Radamistus in amplexus eius effusus simulare obsequium, socerum ac parentem appellare; adicit ius iurandum, non ferro, non veneno vim adlaturum; simul in lucum propinquum trahit ... “Under this compulsion, Mithridates accepted a day and a place for the completion of a contract and quitted [lit. “quits”, TK] the fortress. (47) And first, Radamistus threw himself into his embraces, simulated [historical infinitive, TK] respect, called [historical Infinitive, TK] HIM father-in law and father, swore [lit. “swears”, TK] an oath too that he would do no violence TO HIM, neither by sword nor by poison. At the same time he drew [lit. “draws”, TK] HIM into a neighbouring grove ...” appellare... socerum ac parentem EUM father-in-law.ACC.SG and father.ACC.SG HE.ACC.SG.M call.INF vim non ferro non veneno EI not sword.ABL.SG not poison.ABL.SG HE.DAT.SG.M violence.ACC.SG adlaturum ESSE22 bring.ACC.SG.M. PTCP.FUT BE.INF ... in lucum propinquum simul EUM at the same time HE.ACC.SG in grove.ACC.SG nearby.ACC.SG trahit draw.3SG.PRS

206

Thomas Krisch

(24)

a. appellare “call, refer to as” (+ acc (someone) + acc (as something/ someone)) Tac. Ann. 4,52,19 Afer primoribus oratorum additus, divulgato ingenio et secuta asseveratione Caesaris qua suo iure disertum eum appellavit. “Afer was ranked among the foremost orators, through his publicly known ability and through the subsequent assertion of the emperor by which he called him ‘eloquent by his own right’”. suo iure disertum eum appellavit own.ABL.SG right.ABL.SG eloquent.ACC.SG him.ACC.SG he.called b. vim afferre “do violence to” ([lit. vim (violence.ACC) afferre (carry.to)] (with dat.)23 Liv. 39,54,6 (an accusativus cum infinitivo construction) oppidum quoque aedificare coepisse, quod indicium esset nec agro nec urbi ulli vim adlaturos venisse. “(they [scil. the Gauls, TK ] said that) they also had started to build a town, and that would be an indication for the fact that they had not come in order to do violence to the country or to any town.” nec agro nec urbi ulli and.not country.DAT.SG and.not town.DAT.SG any.DAT.SG vim adlaturos venisse carry.to.PTCP.FUT.ACC.PL come.INF.PERF violence.ACC.SG c. trahere “draw” + acc. (someone / something) + prepositional phrase (denoting the GOAL) Tac. 4,21,4 Et spreta potentia Augustae trahere in ius Urgulaniam domoque principis excire ausus erat. “He also dared to hand over Urgulania to the courts neglecting the power of Augusta and to summon (her) from the palace of the emperor.” trahere in ius Urgulaniam Urgularia.ACC.SG.F draw into law.ACC.SG.N

Thus, in the case of appellare in (23) a pronoun (3SG.M) in the accusative is missing, in the case of vim afferre in (23) there is an ellipsis of a pronoun (3SG.M) in the dative and in the case of trahit in (23), a pronoun of the accusative (3SG.M) is missing, all of which refer to the same person addressed by the form eius “his” (lit. “of him”, GEN.SG.M of the anaphoric personal pronoun of the third person singular is, ea, id) in the immediate context before. Eius, again, (carrying thematic material, as an “anaphoric” pronoun) refers back to Mithridates in the sentence before. Thus, we have a neat “chain” of thematic material here, first producing an anaphoric element (eius) and then a series of null elements.

On the “syntax of silence” in Proto-Indo-European 207

6.2. Greek data The same type of object ellipsis is ubiquitous in Homeric Greek. Consider example (25). The verbs with missing objects are put in bold letters. (25)

Hom. Il 5, 22-24 (Dares, a priest of the god Hephaistos, has two sons, Phegeus and Idaios. In the Troian war, they fight against Diomedes. Diomedes kills Phegeus and would have killed Idaios as well, but Hephaistos rescues Idaios)

oudè gàr oudè ken autòs hypékphyge kËra mélainan, / all’ HÇphaistos éryto, sáPse dè nyktì kalýpsas, / hPs dÇ hoi m¼ págchy gérPn akachÇmenos eíē. “Nay, he [Idaios, TK] would himself not have escaped the black goddess of death, but Hephaistos guarded HIM, rescued HIM, enfolding HIM in darkness so that the aged one [= Dares TK] might not be utterly fordone with grief ”24 all' MIN HÇphaistos éryto, sáPse but HE.3SG.ACC Hephaistos guard.3SG.IPF.MID rescue.3SG.AOR nyktì MIN kalýpsas dè MIN night.LOC.SG HIM hide.PTCP.AOR.NOM.SG.M and HIM

Here, again, we are dealing with a thematic “chain”. The name Idaios is the explicit subject of the sentence before (Hom. Il. 5,20). The anaphoric pronoun autós refers back to this person. The following missing direct objects to the verbal forms éryto, saPse, and kalýpsas all refer back to the same person. The normal valency frames of the verbs with elliptic objects in (25) are exemplified in (26): (26)

a. erý(e)sthai “guard” + acc. (someone) Hom. Il. 13,554–555 péri gár ra PoseidáPn enosíchthPn / Néstoros hyiòn éryto “for mightly did Poseidon, the Shaker of Earth, guard Nestor’s son” péri gár ra PoseidáPn all.around for indeed Poseidon.NOM.SG enosíchtPn Néstoros hyiòn Shaker.of.Earth.NOM.SG Nestor.GEN.SG son.ACC.SG éryto guard.3SG.IPF.MID

208

Thomas Krisch b. sᮼzein “rescue”+ acc. (someone)

Hom.Il.17,452

óphra kaì Automédonta saÔseton ek polémoio “that you two [scil. two horses TK] will also rescue Automedon out of the war” óphra kaì Automédonta saÔseton that also Automedon.ACC.SG.M rescue.2DU.FUT ek polémoio out.of war.GEN.SG c. kalýptein “hide” + acc. (someone/something) Hom. Il. 4,461 tòn dè skótos ósse kalypse “and darkness enfolded his eyes” (lit. and darkness enfolded him, the eyes”25) tòn dè skótos ósse kalýpse he.ACC.SG.M and darkness.NOM.SG eye.ACC.DU hide.3SG.AOR

The null object construction is not allowed in the languages continuing Latin and Ancient Greek, viz. the Romance languages and Modern Greek.26 On the other hand, the attestation of the “null object” type of ellipsis in Ancient Greek and in Latin make this construction a candidate for PIE reconstruction.

6.3. Vedic data The data from Vedic Sanskrit fit in with what was just said about Latin (6.1.) and Greek (6.2.), cf. (27). The verbs with missing objects are put in bold letters. (27)

RV 2,35,1 ( cf. van der Wurff 1997: 345)

£pem asçkùi vàjay£r vacasyxê c@no dadhãta nàdy¢ g¡ro me | apxê n@pàd à÷uh‚mà kuv¡t s@ sup‚÷asas karati j¢ùiùad dh¡ || “Eager for reward I have poured out (my) eloquence. The child of the river may accept my songs [gíro] with favour. Will he, the rapidly rushing Apàm Napàt [name of a god in the waters, lit. “offspring of the waters” TK], make THEM [scil. the songs, TK] well adorned? Will he enjoy THEM [scil. the songs, TK]?” 27

apxê 

n@pàd 

à÷uh‚mà

water.GEN.PL offspring.NOM.SG.M rapidly.rushing.NOM.SG.M

T ᘇS

kuv¡t 

s@ 

sup‚÷asas 

INTERR

he.NOM.SG.M

well.adorned.ACC.PL.F THEY.ACC.PL.F28

karati 

T ᘇS 

j¢ùiùad 

dh¡

make 3SG.AOR.SBJV THEY.ACC.PL.F enjoy.3SG.AOR.SBJV EMPH

On the “syntax of silence” in Proto-Indo-European 209

There is thematic continuity of the elided null objects with the word for “songs” (gíro) in the first line of our example (27). The thematic “chain” is not as “dense” as it was in the Latin and the Greek examples (23) and (25). There is no “intermediate” stage with an overt pronoun in (27). The normal valency frames of the verbs with elliptic objects in (27) are exemplified in (28): (28)

a. kar- “make” + acc. (someone/something) + secondary predicate (adjective) RV 10,18,6 ih@ tv@ùñà suj@nimà saj¢ùà dãrgh@m xyuþ karati jãv@se vaþ “May Tv@ùñç [Indian god of creation, TK], giver of good birth, being gracious, make long the life-span for your life here.”

ih@  tv@ùñà 

suj@nimà 

here Tv@ùñç.NOM.SG.M

giving.good.birth.NOM.SG.M

saj¢ùà 

dãrgh@m

xyuþ

gracious.NOM.SG.M

long.ACC.SG.N

life.span.ACC.SG.N

karati

jãv@se

vaþ

life.DAT.SG. you.GEN./DAT.PL.CLIT make 3SG.AOR.SBJV b. joù - “enjoy” + acc. (something)29 RV 1,25,18 etx juùata me g¡raþ “He shall enjoy these songs from me” 

etx 

juùata 

me

this.ACC.PL.F

enjoy.3SG.INJ.MID

I.GEN / DAT. SG (CLT)

g¡raþ song.ACC.PL.F

6.4. Hittite data In Hittite, there are not many examples for null objects. The specialist for Hittite syntax, Silvia Luraghi (2005: 244) remarks “Transitive verbs ... only very infrequently occur with N[ull] O[bject]s”. She brings an example from the relatively young (Neo-Hittite) Annals of Mursilis for this phenomenon (Luraghi 2005: 244) and mentions the Old Hittite law texts (Luraghi 2005: 242), where null objects occur. In addition to that I have found several attestations for this construction in the Old Hittite thunderstorm ritual. Three examples are cited in (29). The verbs with missing objects are marked with bold letters.

210

Thomas Krisch

(29) StBot 12 (Neu 1970, Hittite thunderstorm ritual) RS. III, 44'–46' LÚ SÌLA.ŠU.DU8 NINDAIar-ši-in EM-­A GE6 / LUGAL-i pa-a-i ta p@r-ši-K[a LÚ ] SÌLA.ŠU.DU8 e-ip-zi ta LÚ GIŠBANŠUR pa-a-i .... “The cupbearer gives sour dark bread to the king and he cuts IT. The cupbearer takes IT and he gives IT to the table servant ...” LÚ SÌLA.ŠU.DU8  NINDAIarši-n  EM­A  GE6  30 Cupbearer  bread-ACC.SG.COMM sour.ACC black AN LUGAL-i  pài ta king-DAT.SG give.3SG.PRS and IT-ACC.COMM. (CLT) LÚ SÌLA.ŠU.DU8 AN paršiKa cut.3SG.PRS.MID cupbearer IT.ACC.SG.COMM. (CLT) ēpzi  ta  (AŠ)31 take.3SG.PRS  and  (HE.NOM.SG.COMM. (CLT)) LÚ GIŠBANŠUR pài AN  give.3SG.PRS IT.ACC.SG.COMM. (CLT)  table.servant32

The first sentence of (29) shows the full valency of the verb “to give”, piKe-IIi (appearing in the form pài).33 The last part of the cited text exhibits ellipsis of the accusative object with the same verb. In between these two sentences there are two further examples of null objects (elliptical accusative objects). In all of these cases there is thematic continuity with the word for “bread” in the first line of the cited text. Like the ègvedic example (27), the Hittite example in (29) does not show “intermediate” pronominalization. In (30) one can find examples for the normal valencies of the verbs paršKe-a “cut” and ēpp/ app- “take, seize” in Old Hittite:  (30)  a.  paršKe-a “cut” + acc. (something) StBoT 12, (Neu 1970; Hittite thunderstorm ritual) RS IV 25' NINDA Iar-ši-in-na pár-ši-Ka “and he cuts bread” NINDA Iaršinn-a páršiKa bread-ACC-and cut.3SG.PRS.MID b.  ēpp/ app- “take, seize” StBoT 8 (Otten / SouŠek 1969; Old Hittite ritual for the royal couple) Vs. II 19 ma-a-an MUŠENIa-a-ra-na-an Iu-š[(u-Wa-an-da-an ap-pa-an-zi)] ...“if one seizes [lit. “they seize”, TK] a living eagle ...” màn  MUŠENIàrana-n  IušuWanda-n  app-anzi if eagle-ACC.SG.COMM living-ACC.SG.COMM seize-3PL.PRS

On the “syntax of silence” in Proto-Indo-European 211

6.5. Germanic data (?) Like the modern Romance languages (cf. section 6.2. above), the modern Germanic languages are said to lack clear instances of object ellipsis, whereas this phenomenon is claimed to exist in older Germanic dialects.34 But there are many questions open here. We can cite one instance from Old High German which suggests, in our view erroneously, that this older state of German could have been more permissive in this respect than modern German, namely the Straßburg oaths: (31)

Old High German Straßburg oaths (Steinmeyer 1916=1971: 82) Oba Karl then eid, then er sinemo bruodher Ludhuuige gesuor, gileistit, indi Ludhuuuig, min herro, then er imo gesuor, forbrihchit, ob ih inan es iruuenden nemag ... “If Charles keeps the oath which he swore to his brother Ludwig, and Ludwig, my lord, (on his part) breaks THE OATH / IT which he swore to him, if I cannot prevent him from it, ....”

In Modern German the corresponding sentence with ellipsis would be ungrammatical: (32)

Modern German *... und (wenn) Ludwig, mein Herr, DEN EID/den, den er ihm schwor, bricht,

A grammatical rendering of (32) in Modern German would be .... und (wenn) Ludwig, mein Herr, den (Eid), den er ihm schwor, bricht (with an overt correlative pronoun (den) as head of the relative clause). The Romance text of the oath shows an overt pronoun in the accusative (lo) in the corresponding construction.35 In my opinion, the lack of a (correlative) pronoun as head of a relative construction in the Old High German text (32) [Ludhuuuig, min herro, then (relative pronoun) er imo gesuor] could have been immediately influenced by Latin which showed such constructions.36

6.6. Reconstruction The examples taken from AIELs and discussed in sections 6.1–6.4. showed deletion of the object. This leads us to the conclusion that PIE had a rule that allowed deleting objects which refer back to already mentioned “thematic” material. This “null object” type of ellipsis thus is operating in a

212

Thomas Krisch

forward direction and meets Klein’s criterion cited in (10). Modern IndoEuropean languages like the Romance languages, the Germanic languages (also partly in their old attestations) and Modern Greek do not permit null objects any more (cf. sections 6.2. and 6.5).

7. Conclusions a)

b)

Gapping was possible in the ancient Indo-European languages and can be reconstructed for Proto-Indo-European. It operates backwards and forwards in these languages and in the protolanguage and it is determined by syntactic (basic word order) factors and by factors of functional sentence perspective. Object ellipsis was possible in the ancient Indo-European languages and in Proto-Indo-European. It operates in forward direction in these languages and in the protolanguage and it is determined by factors of functional sentence perspective.

Appendix: Abbreviations AIEL=Ancient Indo-European Language; Amph.=Amphitruo; Ann.=Annales; AOR=Aorist; Ar.=Aristophanes; Bell. Gall.=De Bello Gallico; C=position of the “Complementizers”; Caes.=Caesar; CLT=clitic; COMM=genus commune; CP=Complementizer Phrase; EMPH=emphasizing particle; Hdt.=Herodotos; Hom.=Homeros; Il.=Ilias; INJ=injunctive; INTERR=interrogative particle; IP=Inflection Phrase; IPF=imperfect; Liv.= Livius; MID=middle voice; OPT=optative. Mil.=Miles Gloriosus; PERF=perfect; Ph.=Phormio; PIE=Proto-Indo-European; Plaut.=Plautus; preduced=phonologically reduced; Ra.=Ranae; RV=ègveda; StBoT=Studien zu den Bo÷azköy–Texten; Tac.=Tacitus; Ter.=Terentius; TK=Thomas Krisch; TOPIC=Topic or contrastive focus position; TOPIC1=topic position; TOPIC2=focus-position; Vitr.=Vitruvius; VP=verbal phrase; W=Wackernagel particle (W1= enclitical sentence connectors; W2= other enclitics); XP=any phrase

On the “syntax of silence” in Proto-Indo-European 213

Notes 1.

2.

3.

4.

5.

6. 7.

8.

Interlinear glosses of examples follow the Leipzig glossing rules to be found at the following URL: http://www.eva.mpg.de/lingua/resources/glossingrules.php (April 2, 2008). Additional abbreviations used in this article are listed as an appendix before the notes. Many thanks go to Jürg Fleischer for numerous suggestions, to Thomas Lindner for helping the author with the Straßburg oaths, to Hermann Bieder for confirming the Russian data in example (18), to an anonymous reviewer of this article for some useful hints and to Christina Katsikadeli and Stefan Niederreiter for proofreading the text. All remaining shortcomings remain in the responsibility of the author, of course. Cf. e.g. the following examples for the “right hand” which show nominalization of the word for “right” and do not express the (feminine) word for “hand” (put in parentheses) lat. dext(e)ra f. (scil. manus), germ. die Rechte f. (scil. Hand), gr. dexia/ f. (scil. cheír); ved. d@kùiõa- m. (scil. hásta-). A similar case is the ellipsis of the head noun in the following ancient Greek syntagm és Haidou “to the (houses of) Hades” (e.g. Ar. Ra. 69). Blakemore (2002: 71): “Pragmatics does not simply enter when linguistic decoding fails; on the contrary, the linguistic system is subservient to pragmatic inference in the sense that it functions as an aid to the inferential system.” Of course, I am aware of the fact that the exact semantic interpretation of English but and German aber is more complex, as is clear from the analyses of Abraham (1979: e.g. 92) and Lang (1977: e.g. 168). As is well known, the term “implicature” has been made popular in linguistics by Grice (1975). The concept of conventional implicature is viewed upon in a critical way by Bach (2006; in press), especially p. 14 (cited after http//userwww.sfsu.edu/~kbach/TopTen.pdf (seen April 2, 2008)). Bach wants to get rid of the notion of “conventional implicature” which is “counterintuitive” for him because to his opinion the conjunction but states something rather than implicates something. I cannot follow this way of argumentation, though. By hearing aber “but” the hearer starts worrying and expects additional information. Cf. also Maling (1972: 103). Of course, not every case of V-to-C movement looks like an SVO structure. Consider the following German coordinate sentences where the first position in the sentences is occupied by an adverb: Gestern liebten die Kinder Fisch, aber heute lieben sie Yesterday loved the children fish but today love they Gemüse. vegetables “Yesterday the children loved fish but today they love vegetables”. Cf. e.g. Krisch (1984: 44–47); Krisch (1992: 157) with reference to Andersen (1973).

214 9.

10.

11.

12. 13. 14. 15. 16.

17.

18.

19.

20. 21.

22.

Thomas Krisch The Hittite texts are first given in their cuneiform spelling using the traditional transcription system. Sumerograms are given in capital letters, as determinatives they are written in upper case. Akkadograms appear in capital italicized letters and Hittite words appear in syllabic writing with a hyphen between the syllables (broad transcription). In glossed words the hyphens of syllable writing are removed (“narrow transcription”) and if hyphens are used there, they indicate morpheme boundaries. The importance of these daughter-languages for the syntactic reconstruction of PIE lies in the fact that they are the oldest attested AIELs and that we have a great amount of texts at our disposal from all of them. Since backward gapping is only attested in SOV languages (cf. section 3.) this may in turn count as quite safe diagnostics for the status of these languages as SOV. Cf. also Krisch (2001: 169–170); Krisch (1997: 305); McCone (1979: 224). Krisch (1998, 2002, 2004). Another case of syntactic innovation in Hittite is proposed in Krisch 1990: 72. Cf. e.g. Kuno (1982); Klein (1993); Merchant (2001). By p(honologically)-reduced Klein means that ellipsis takes place in the phonological component of grammar and that the elided material is still present in syntax. Forward ellipsis, on the other hand allows sloppy identity of person in German. The sentence “weil ich Bier trinke und du Wein TRINKST” is grammatical (cf. Klein 1993: 774). This type of ellipsis in the lexicon creates an “unspecific” reading of the verb and this type of ellipsis of an object normally is not counted as “null object” (cf. e.g. Luraghi 2005: 235). Therefore, in example (20), I have put the elided object (SOMETHING) into parentheses. We shall mainly deal with “referential null objects” (Luraghi 2005: 235) here, but, admittedly, it is not always clear in the texts whether one has to do with an “unspecific reading” (a possibility provided for by the lexicon) or with a “referential null object”. This dative object originally was an adjunct with “malefactive” meaning. It is a causative formation to the PIE root *ne,- (“to get lost”, “to vanish”, “to die”) with the original meaning “bring death to someone”. Cf. LIV (2001: 452) with footnote 9. Cf. e.g. tab. XII, 2a (Law of the Twelve Tables) Si servus furtum faxsit noxiamve noxit ... lit. “If a slave has committed theft or has harmed a harm ...” The interpretation of fallere as an optionally intransitive verb with the meaning “afflicting harm” (an “unspecific reading”) already provided for by the lexicon cannot be ruled out completely here. Therefore, in (22), I put the elided object (ME) into parentheses. Cf. also footnote 18. This is an AcI-construction dependent from adicit ius iurandum “he adds an oath”. Adlaturum esse is the periphrastic future infinitive of afferre in Latin.

On the “syntax of silence” in Proto-Indo-European 215 23. Afferre and the form adlaturos (PART. FUT) are suppletive members of a paradigm. Our example (24b), taken from Livy exhibits this suppletive future participle like the Tacitus-example (23). 24. I do not translate hoi (Dat.Sg.M) “him” here in order to avoid confusion. Hoi refers to the old man. 25. The schema kat' holon kai meros (“Construction of the whole and the part”) is a typical stylistically marked construction in Ancient Greek and in other AIELs. cf. e.g. Smyth (1963=1956=1920: 267): “... a verb may take two objects, one denoting the person, the other the part especially affected by the action.” 26. For a more detailed analysis cf. e.g. Luraghi (2005: 246). 27. Geldner (2003=1951: 321) interprets both of the interrogative clauses of our translation as emphatic assertions. Hettrich (1988: 154) translates only one of our interrogative clauses as interrogatrive and the second one as emphatic. 28. The Sanskrit word gír- “song” is a feminine noun. 29. There is also an example with instrumental case (RV 10,6,4) and one example with mixed cases (accusative and instrumental RV 5,39,4). The forms of this verb have to be analysed all as forms originally belonging to a middle paradigm, cf. Joachim 1978: 82. 30. Sumerograms (Logograms) often are not characterized morphologically and can serve several functions. In this case, the Sumerogram functions as a nominative case. 31. As a rule in Hittite “no subjects of transitve verbs are ever cliticized” (Luraghi 1990: 41). Therefore I put the deleted nominative clitic (AŠ) in parentheses. 32. In this case, the Sumerogram functions as a dative case (cf. also note 30). 33. The Hittite verbal forms are cited in the form in which they are cited in Oettinger 1979. 34. Cf. Luraghi (2005: 247–248) for Old English and Old Norse examples. 35. Si Lodhuuigs sagrament, quĊ son fradre Karlo iurat, conservat, et Karlus meos sendra de suo part ñ lostanit, ... . Having consulted the facsimile [in Becker 1972: 28–29 and the facsimile to be found at the URL http://www.fh-augsburg.de/~harsch/germanica/Chronologie/09Jh/Strassburger Eide/eid_text.html (seen 3 April 2008) (Enneccerus 1897: 24–26, 34–36 non vidi)] I agree with the interpretation of the much discussed (cf. e.g. Elcock 1960: 335–336, especially footnote 1 on these pages) ñ lostanit in the cited passage as ñ lo franit [in the late Carolingan minuscle writing ductus the group fr may look very much like st; franit can be a writing variant for fraint “breaks” (Thomas Lindner, Salzburg, personal communication; cf. also Elcock 1960: 339)], whereby ñ (probably an abbreviation for the dative of nomen. “name”) is a fill-in for a proper noun, in this case Lodhuuige (in the dative case). The Romance text can be translated as follows: “If Ludwig keeps the oath that he swore to his brother Charles, and Charles, my lord, breaks (franit) it (lo) to Ludwig (ñ), as far as he is concerned... ”. This text shows an accusative pronoun (lo) that refers back to sagrament “oath”. There is no relative clause in the Romance text.

216

Thomas Krisch

36. The construction without the pronominal head was possible in Latin, cf. Plaut. Mil. 367–368 PHIL(OCOMASIUM) Tun me vidisti? SCE(LEDRUS) Atque his quidem hercle oculis. PHIL Carebis OCULIS, credo, qui plus vident quam ID quod vident “PHIL You saw me? SCE And even, alas, with these eyes PHIL You will have to do without THE EYES which see more than THAT what they see” (cf. also Hoffmann / Szantyr (1965: 555–556). No Latin text is attested for the Straßburg oaths but one may assume that it existed, because the speech of Ludwig to the troops is handed down in Latin only (cf. also de Boor 1971: 48). Elcock (1960: 337–338) even goes so far as to reconstruct the Latin original of the text. If this assumption (not representing the current commnis opinio, cf. Schmidt-Wiegand 1995: 379) is correct, then the use of the ellipsis in the Old High German text (without a correlative demonstrative pronoun) could have been immediately influenced by the Latin original.

References Abraham, Werner 1979 But. Studia Linguistica 23: 89–119. Andersen, Henning 1973 Abductive and Deductive Change. Language 49: 765–793. Bach, Kent 2006 The Top 10 Misconceptions about Implicature. In Drawing the boundaries of Meaning: Neo-Gricean Studies in Pragmatics and Semantics in Honor of Laurence R. Horn, ed. Betty Birner and Gregory Ward. Amsterdam/Philadelphia: John Benjamins. Cited after http://userwww.sfsu.edu/~kbach/TopTen.pdf (seen 2 April 2008). Becker, Siegfried 1972 Untersuchungen zur Redaktion der Straßburger Eide. Frankfurt a.M.: Peter Lang. Blakemore, Diane 2002 Relevance and Linguistic Meaning: The semantics and pragmatics of discourse markers. Cambridge: Cambridge University Press. de Boor, Helmut 1971 Die deutsche Literatur von Karl dem Großen bis zum Beginn der höfischen Literatur. Mit einem bibliographischen Anhang von Dieter Haacke. 8. Auflage. München: C.H. Beck‘sche Verlagsbuchhandlung. Elcock, W. D. 1960 The Romance Languages. London: Faber & Faber Limited. Enneccerus, Magda 1897 Die ältesten deutschen Sprach-Denkmäler in Lichtdrucken. Frankfurt a.M.: F. Ennecerus.

On the “syntax of silence” in Proto-Indo-European 217 Gaeta, Livio and Silvia Luraghi 2001 Gapping in Classical Greek Prose. Studies in Language 25: 89–113. Geldner, Karl Friedrich 2003 Reprint. Der Rig-Veda aus dem Sanskrit ins Deutsche übersetzt und mit einem laufenden Kommentar versehen. Cambridge, Massachusetts and London, England: Harvard University Press. Original edition, Cambridge, Massachusetts: Harvard University Press, 1959. Grice, H. Paul 1975 Logic and Conversation. In Syntax and Semantics 3: Speech Acts, eds. Peter Cole and Jerry L. Morgan, 41–58. New York/San Francisco/London: Academic Press. Hettrich, Heinrich 1988 Untersuchungen zur Hypotaxe im Vedischen. Berlin/New York: Walter de Gruyter. Hoffmann, J. B. and Anton Szantyr 1965 Lateinische Syntax und Stilistik. München: C.H. Beck‘sche Verlagsbuchhandlung. Joachim, Ulrike 1978 Mehrfachpräsentien im ègveda. Frankfurt a.M.: Peter Lang. Klein, Wolfgang 1993 Ellipis. In Syntax: Ein internationales Handbuch zeitgenössischer Forschung, eds. Joachim Jacobs, Armin von Stechow, Wolfgang Sternefeld and Theo Vennemann. 1. Halbband, 763–799. Berlin/New York: Walter de Gruyter. Krisch, Thomas 1984 Konstruktionsmuster und Bedeutungswandel indogermanischer Verben: Anwendungsversuche von Valenztheorie und Kasusgrammatik auf Diachronie und Rekonstruktion. Frankfurt a.M.: Peter Lang. 1990 Das Wackernagelsche Gesetz aus heutiger Sicht. In Sprachwissenschaft und Philologie: Jacob Wackernagel und die Indogermanistik heute, eds. Heiner Eichner and Helmut Rix, 64–81. Wiesbaden: Ludwig Reichert Verlag. 1992 Analogische Prozesse in der lateinischen Sprachgeschichte. In Latein und Indogermanisch: Akten des Kolloquiums der Indogermanischen Gesellschaft, Salzburg, 23.–26. September 1986, eds. Oswald Panagl and Thomas Krisch, 155–181. Innsbruck: Institut für Sprachwissenschaft. 1997 Delbrücks Arbeiten zur Wortstellung aus heutiger Sicht. In Berthold Delbrück y la sintaxis indoeuropea hoy: Actas del Coloquio de la Indogermanische Gesellschaft Madrid, 21–24 de septiembre de 1994, eds. Emilio Crespo and José Luis García Ramón, 283–309. Madrid: Ediciones de la UAM and Wiesbaden: Ludwig Reichert Verlag.

218

Thomas Krisch

1998

Zum Hyperbaton in altindogermanischen Sprachen. In Sprache und Kultur der Indogermanen: Akten der X. Fachtagung der indogermanischen Gesellschaft Innsbruck 22.–28. September 1986, ed. Wolfgang Meid, 351–384. Innsbruck: Institut für Sprachwissenschaft. 2001 „Man kann sich ein Klavier ja auch um den Bauch binden“. In Fremd und Eigen: Untersuchungen zu Grammatik und Wortschatz des Uralischen und Indogermanischen in memoriam Hartmut Katz, eds. Heiner Eichner, Peter-Arnold Mumm, Oswald Panagl, Eberhard Winkler with collaboration by Roland Hemmauer, Susanne Knopp and Velizar Sadovski, 155–174. Wien: Edition Praesens. 2002 Indogermanische Wortstellung. In Indogermanische Syntax: Fragen und Perspektiven, ed. Heinrich Hettrich with collaboration by JeongSoo Kim, 249–261. Wiesbaden: Ludwig Reichert Verlag. 2004 Some aspects of word order and sentence type From Indo-European to New High German. In ANALECTA HOMINI UNIVERSALI DICATA. Arbeiten zur Indogermanistik, Linguistik, Philologie, Politik, Musik und Dichtung: Festschrift für Oswald Panagl zum 65. Geburtstag, eds. Thomas Krisch, Thomas Lindner and Ulrich Müller with editorial collaboration by Michael Crombach, Stefan Niederreiter, Helga Panagl and Ursula PaviŠiŠ. Vol I, 106–129. Stuttgart: Akademischer Verlag Hans-Dieter Heinz. Kuno, Susumo 1980 Functional Syntax. In Syntax and Semantics 13: Current Approaches to Syntax, eds. Edith Moravcsik and Jessica R. Wirth, 117–135. New York/London/Toronto/Sydney/San Francisco: Academic Press. Lang, Ewald 1977 Semantik der koordinativen Verknüpfung. Berlin: Akademie Verlag. 2006 Parallelismus in der Grammatik: Reichweite, Status, Herkunft. Förderungsantrag (Antrag auf Gewährung einer Sachbeihilfe) P 6. Berlin: Humboldt-Universität. Cited after http://www.zas.gwz-berlin.de/ research/projects/p6.pdf (seen April 3, 2008). LIV 2001 Lexikon der indogermanischen Verben: Die Wurzeln und ihre Primärstammbildungen, ed. Helmut Rix with collaboration by Martin Kümmel, Thomas Zehnder, Reiner Lipp and Brigitte Schirmer. 2d. revised and extended ed. by Martin Kümmel and Helmut Rix. Wiesbaden: Ludwig Reichert Verlag. Luraghi, Silvia 1990 Old Hittite Sentence Structure. London/New York: Routledge. 1997 Omission of the Direct Object in Latin. Indogermanische Forschungen 102: 239–257.

On the “syntax of silence” in Proto-Indo-European 219 2005

Null Objects in Latin and Greek and the Relevance of Linguistic Typology for Language Reconstruction. In Proceedings of the Fifteenth Annual UCLA Indo-European Conference, eds. Karlene Jones-Bley, Martin E. Huld, Angela Della Volpe and Miriam Robbins Dexter, 234–256. Wahington D.C.: Institute for the Study of Man. Maling, Joan M. 1972 On gapping and the order of constituents. Linguistic Inquiry 3: 101– 108. McCone, Kim Robert 1979 Aspects of Indo-European Sentence Patterns and Their Role in the Constitution of the Old Irish Verbal System. Oxford Phil. Diss. Merchant, Jason 2001 The syntax of silence: Sluicing, Islands and the Theory of Ellipsis. Oxford: Oxford University Press. Neu, Erich 1970 Ein althethitisches Gewitterritual. Wiesbaden: Otto Harrassowitz. Oettinger, Norbert 1979 Die Stammbildung des hethitischen Verbums. Nürnberg: Verlag Hans Carl. Otten, Heinrich and Vladimir SouŠek 1969 Ein althethitisches Ritual für das Königspaar. Wiesbaden: Otto Harrassowitz. Praust, Karl 2003 A Missing Link of PIE Reconstruction: The Injunctive of *H1es- ‘to be’. In Proceedings of the Fourteenth Annual UCLA Indo-European Conference, eds. Karlene Jones-Bley, Martin E. Huld, Angela Della Volpe and Miriam Robbins Dexter, 112–144. Washington D.C.: Institute for the Study of Man. Richards, Norvin 1998 The Full Pursuit of the Unspeakable. In NELS 28 (2): 153–168. Ross, John Robert 1970 Gapping and the order of constituents. In Progress in Linguistics: A Collection of Papers, eds. Manfred Bierwisch and Karl Erich Heidolph, 249–259. The Hague/Paris: Mouton. Schmidt-Wiegand, Ruth 1995 ‚Straßburger Eide‘. In Die deutsche Literatur des Mittelalters: Verfasserlexikon, ed. Burghart Wachinger, Vol. IX, 377–380. Berlin/New York: Walter de Gruyter. Smyth, Herbert Weir 1963 Reprint. Greek Grammar. 2d. ed., revised by Gordon M. Messing. Cambridge: Harvard University Press. Original edition, Cambridge: Harvard University Press, 1956.

220

Thomas Krisch

Steedman, Mark 2000 The Syntactic Process. Cambridge, Massachusetts: MIT Press. Steinmeyer, Elias von 1971 Reprint. Die kleineren althochdeutschen Sprachdenkmäler. Dublin/Zürich: Weidmann. Original edition, Berlin: Weidmann, 1916. van der Wurff, Wim 1997 Syntactic reconstruction and reconstructability. Proto-Indo-European and the typology of null objects. In Linguistic Reconstruction and Typology, ed. Jacek Fisiak, 337–355. Berlin/New York: Mouton de Gruyter.

Information-structural categories in the main texts of early German inheritance

Word order variation and information structure in Old High German: An analysis of subordinate dhazs-clauses in Isidor Eva Schlachter

Abstract One of the earliest texts written in Old High German exhibits an extraordinarily high percentage of verb-early position in subordinated clauses introduced by dhazs. This paper shows that copula and auxiliary verbs end up more frequently in a higher clause position than main verbs. It demonstrates that these verb types occur with different information-structural values and concludes that they relate to different sentence structures. At the same time it is a contribution to the philological debate about differences between the so-called quotation syntax and the rest of the text.1

1. Introduction Until recently it was considered to be a well-established fact that Old High German already displayed the same verbal syntax as Modern German (Fourquet 1974, Lenerz 1984): in main clauses the finite verb shows up in second position (Verb-second) while in subordinate clauses it occurs in final position (Verb-end). This complementary distribution led to the bynow classic generative analysis of den Besten (1977): in main clauses the finite verb ends up in the Complementizer-position (linke Satzklammer ‘left sentence brace’), while in subordinate clauses it remains in the final INFLposition (rechte Satzklammer) since the Comp-position is blocked by a lexical complementizer. Old English, which shows a similar distribution, received an analogous analysis in Kemenade (1987). However, more recent studies on Old and Middle English (Pintzuk 1993, 1999; Kiparsky 1996; Fuß and Trips 2002) support the view that there are more positions than the ones standardly assumed. There is a high number of embedded clauses showing a non-final verb position, which is identified with an INFL-medial grammar.2 Since

224

Eva Schlachter

there are at the same time many cases of subordinations with the finite verb in final position, it is assumed that INFL-final and INFL-medial grammars coexisted side-by-side in the same speaker (Kroch 1989). The ultimate change, then, is seen as the result of a competition process between these two grammars.3 For Old High German, however, contradictory data have been known for more than 100 years. Diels (1906: 164) considered early verb placement (or Verb-second) in subordinate clauses as a quite common fact, triggered by the intonation of the first “word” (1906: 168). Rannow, who investigated exclusively the syntax of Isidor in comparison to the Latin original, also regarded intonation to be responsible for non-canonical ordering, but considered it only as a relevant factor in explaining the occurrence of the postverbal parts of the clause (1888: 119). Finally, when later work pointed out the existence of early verb placement in subordinate clauses, it was mainly analysed as a type of extraposition in support to maintain the Verb-end analysis (Lenerz 1984: 129, Borter 1982: 198). In contrast, Tomaselli (1995) and Fuß (1998) have argued for an INFL-medial position in Old High German. Schlachter (2004), adopting this analysis, has shown that the postverbal constituents in embedded Verbsecond clauses are to be interpreted as focussed. Robinson (1997) has investigated all embedded clauses of the Old High German Isidor translation on the basis of their semantic and syntactic function in relation to the main clause, the active/passive status, and the kind of selecting verb. His data seems to show that Verb-end is more common in adverbial and relative clauses than in indirect questions and complement dhazs-clauses. Nevertheless, he notes that in adverbial dhazs-clauses in which the postverbal constituent is not a sentential constituent or a predicate the verb occurs in 50% of all cases in a higher position and speculates that such clauses seem to be less subordinated, in “that their meaning is often the main content of the whole sentence in which they appear, and almost always asserted information” (1997: 83). The main problem with all these analyses is that they only focus on the type of clauses that fit the given explanation. For instance, Robinson’s group of adverbial dhazs-clauses is a very small one indeed. But what is the reason for early placement in the other cases? Tomaselli bases her analysis on a few Verb-third main clauses, which are very rare in Old High German, and Schlachter mainly argues with copular constructions. Thus, the results cannot be generalized.

Subordinate dhazs-clauses in Isidor 225

This paper therefore wants to narrow the gap between theoretical explanation and empirical findings. It aims at a classification of the different causes of early verb placement in subordinate clauses, thus creating a secure starting point for further theoretical work. We limit the empirical base of this study to the dhazs-clauses, because the syntactic status of dhazs as formal marker of subordination is not in doubt in the majority of cases. In addition, the dhazs-clauses have played an important role in the traditional view of how subordinate structures developed (cf. Müller and Frings 1959). As a data base we chose the earliest prose text, the so-called Old High German Isidor from the end of the 8th century. The paper is organized as follows. After a short presentation and evaluation of the Old High German text we present a word class distinction of the finite verb in early position and a classification of the postverbal constituents. In the next section it is shown that these constituents have different values of focus. The results and some of the consequences are discussed in the final section. 2. The significance of the Isidorian treatise The texts of the so-called Isidor group vary a great deal with respect to their length, which is certainly one of the reasons why they get different degrees of scholarly attention. The text most known and studied is a translation of a theological treatise written in Latin by Bishop Isidor of Sevilla (560–636) De fide catholica ex veteri et novo testamento contra Iudaeos ‘About the catholic faith from the Old and New Testament against the Jews’. Together with the Latin text it has been handed down in a codex BN, Ms lat. 2326 in Paris named (P). Until folia 22 the Old High German text is written side by side with the Latin text, then the Old High German part remains empty up to folia 34, where there is no more space left for the translation (Krotz 2002: 20). It has been edited by Hench (1893) and more recently by Eggers 4 (1964), the latter of which will be the basis for the quotations in this paper. Since the beginning of the treatise is not recorded in (P), Eggers restores parts of it from a codex from the monastery in Mondsee (M) which is nowadays kept in the Austrian National Library of Vienna. This manuscript consists of five small texts: beside the translation of the Isidorian treatise, which is a fragment of only five pages, it contains parts of the Gospel of St. Matthew, fragments of a further treatise and two more homilies. All the texts were transcribed from non-Bavarian texts into the Bavarian language (Matzel 1970: 45).

226

Eva Schlachter

The Isidorian translation was written in the last decades of the 8th century (Eggers 1964: VIII) in the southwest Franconian dialect and originates from the area of Lorraine (Sonderegger 2003: 129). The basic motivation for its origin is the theological debate about the nature of Christ: is he the real son of God or was he ‘as human being’ adopted by God the Father? The dispute between adherents of the Trinity doctrine and followers of Arianism had undergone a revival at the end of the 8th century, at which time Arianism was known as Adoptianismus, a movement represented by Felix von Urgel. This heresy was condemned by church councils in Regensburg in 792 and in Frankfurt in 794. The text defends the doctrine of the Trinity, arguing with quotations from the Old Testament which show “that Christ, God, and the Holy Spirit are, contrary to the proclamations of the followers of Adoptianism, for example, one equally in essence and in substance” (Nordmeyer 1958: 33). Nordmeyer supposes that the text “was intended to be read aloud, perhaps before a mixed audience of clerics and lay persons assembled in council to judge the Felician heresy” in order to “account with some plausibility for the translator’s rhetorical and exegetical additions and enlargements of his text” (1958: 33). If Nordmeyer is right, the huge number of deviations from the Latin text would find a functional explanation: the translator tried to make the content of the treatise as explicit as possible so that even a theologically uneducated lay public would be able to understand it. Because of its relative independence from the Latin original, the Old High German Isidor version is considered an excellent piece of translation (Lippert 1974: 28). Nevertheless, it‘s exactly because of this outstanding quality that it hasn’t always counted as a good object for syntactic investigation. Gering, as quoted in Robinson (1996: 3) points out that the Isidor “is so free a translation that we cannot be quite sure whether a deviation from the Latin is required by the grammar of German, or simply better style. Such a deviation in Tatian is so infrequent, however, that when one occurs it m u s t be because of German grammar“. We will argue, however, that the Isidor is a well-suited text for the study of information structure for the following reasons. First, stylistic effects are grounded in the grammatical system. Scrambling in Modern German, for instance, which refers to the non-canonical sequencing of constituents in the middle field, is certainly governed by grammatical rules that, at the same time, create stylistic or informationstructural effects.

Subordinate dhazs-clauses in Isidor 227

Second, the frequency of a certain structure in the corpus is indicative of its being a core property of grammar. In anticipation of one of the results, we found 50% of early verb positions in subordinate dhazs-clauses. Since a huge number of these occurs independently of the Latin original and since these clauses are part of a text which has been evaluated as an outstandingly good translation, we have to take them as part of the Old High German syntax. Third, since we conclude that both verb positions (the early and the final one) are part of the Old High German syntax, we can have a closer look at the ordering of the other constituents. Deviations from the Latin text can lead to interesting clues about different structures of German and Latin with respect to information flow. The influence of the Latin original is a factor which has to be controlled for although it is not as important as, for instance, in the Tatian translation (see Petrova and Solf, this volume). On the contrary, it seems that the Latin original of the Isidorian treatise was revised in order to make it more “Germanic”. Ostberg (1979), who investigated all Latin versions of the text, points ot that as “the [Latin] version represented by [Paris-Latin] and [Mondsee-Latin] is consistently closer to the German text than any of the other [manuscripts] examined, it would appear that the translation effort was closely connected with the drafting of the [Latin] ‘Vorlage’ itself” (Ostberg 1979: 205). He furthermore concludes that „most of the alternations apparent in the translation, especially those involving clarification and emphasis, must – in the absence of evidence to the contrary – continue to be regarded as the translator’s own contribution.“ (Ostberg 1979: 205) Furthermore, there is another factor besides the Latin influence which has to be controlled for: the so-called quotation syntax, where the author quotes parts from the Old Testament to strengthen his arguments. Matzel (1970: 357) therefore differentiates between two techniques of translation: the argumental parts should be considered as free paraphrasing which mainly aims to explain the difficult theological reasoning, while the quotation parts are much closer to the Latin original using more archaic means of expression. However, regarding the status of the quotation syntax, a systematic investigation is still lacking about the differences (and similarities) between the two systems, differences which, for example, refer to the attraction of the relative pronoun or the collocation of the definite article (see Matzel 1970: 357, note 855). An example of the different evaluation of syntactic phenomena with regard to quotation syntax is the view on Verbfirst sentences. Robinson (1994) concludes that Verb-first sentences, which

228

Eva Schlachter

mainly occur in the quotation syntax, aren’t an Old High German pattern at all, but a foreign one. However, this interpretation can’t be correct as Verbfirst patterns are well established in all Germanic languages. Hinterhölzl and Petrova (2005) show for the Old High German Tatian that verb-initial placement can be analysed as encoding a coordinating discourse relation with mainly two functions: first it coordinates pieces of discourse at the same level of text structure (typically of narration), and second, it serves to return from a “subordinating”, elaborating discourse part to the main line of discourse structure (Petrova 2006: 162). Although the precise status of the quotation syntax is still unclear, its potential influence should nevertheless be taken into account. In sum, the treatise was intented to be as clear as possible in order to convince the hearers with its theological content. We therefore expect it to use the tools of information structure in a consistent manner. The results, however, should be checked against other Old High German texts to make sure that they are not simply idiosyncratic properties.

3. Syntactic subgroups of dhazs-clauses 3.1. Methodology and first results As already mentioned, we have built up a corpus of subordinate dhazsclauses of the treatise. Dhazs can occur in adverbial clauses, complement clauses and relative clauses. Relative dhazs-constructions have not been considered since their status as main or subordinate clauses is often unclear. Compare (1a), where dhazs is analysed as a demonstrative pronoun evoking a Verb-second configuration,5 and (1b), where it is interpreted as a relative pronoun of a relative clause with an extraposed subject: (1)

Dhazs ni saget apostolus noh forasago that not says apostle nor prophet ni bifant not found-out a. ‘that neither the apostle says nor the prophet found [it] out’ b. ‘neither what the apostle says nor the prophet found-out’ ... nec apostolus dicit nec propheta conperit. (II.3, 103)

Subordinate dhazs-clauses in Isidor 229

Even if the Latin original points to a certain translation, we don’t mechanically transfer it to the Old High German text when a different interpretation is possible. Consider, for instance, the following clause, where the Latin text suggests a verb-final interpretation of the first dhazs-clause. At the same time it could be interpreted as a relative pronoun or as a correlative 6 demonstrative pronoun of a main Verb-second clause. (2)

Dhazs ir chundida dhazs dher ‘that he announced that the got. God’ Ut eundem spiritum ostenderet esse deum.

selbo same

gheist ist spirit is

(III.10, 248) Without considering such unclear cases, we found 56 clauses introduced by dhazs in the whole manuscript. More than half of these, 30, display an early verb position. On the basis of the standard analysis of Old High German as a Verb-end language, this is a spectacular result. Before we start to analyse these clauses we have to make sure that early verb placement is not simply conditioned by the dominance of the Latin syntax or the quotation style, whatever it might be. Table 1. The influence of Latin and quotation syntax

V-early position parallel to Latin 11 deviant from Latin 19 quotation syntax 9 part of argumentation 21

58% 51% 82% 47%

V-end position 8 18 2 24

Total 42% 49% 18% 53%

19 37 11 45

A first look at the two subsystems ‘parallelisms’ and ‘deviations’ leads to the conclusion that verb position is not primarily conditioned by the Latin. 58% of the clauses with Verb-early position are parallel, which is only slightly higher than the 51% which are deviant. Within the deviant clauses Verb-early and Verb-end position are nearly equally represented. These results point to the fact that the translator selected the verb position very deliberately and independently of the Latin original. Therefore, the Latin influence on verb placement is not a significant factor.

230

Eva Schlachter

With respect to quotation syntax we come to a different result: while in the theological argumention parts Verb-early and Verb-final placement are equally represented (48% : 52%), the 11 dhazs-clauses occurring in quotation show a different distribution. With 9:2 cases, Verb-early in quotation syntax is much more common than Verb-end. This extraordinarily high proportion of Verb-early placement points to the fact that quotation syntax indeed behaves differently. We might hypothesize that these Verb-early positions are in fact due to a Latin influence. But a comparison reveals that only 3 cases are totally parallel to the Latin original, while 4 sentences differ with respect to verb placement and in one case only the nominal constituents are changed. Thus, quotation syntax indeed seems to be more than just a copy of the Latin text although it might have been influenced by it. Table 2. Position of the dhazs-clause and clausal function

Dhazs-clause extraposed preposed adverbial clause complement clause

Verb-early position 30 0 8 22

Verb-end position 23 3 5 21

The position of the subordinated clause doesn’t seem to be decisive: nearly all of the sentences are extraposed; the only three instances of preposing out of the group with Verb-end cannot be regarded as significant. Thirteen of the total of 56 clauses are adverbial ones; more than half of these, 8, show an early verb position. However, 22 of the remaining 43 complement clauses also display an early verb position, which is a little bit more than half of all cases. Thus, we conclude that the degree of syntactic integration cannot be the decisive factor. Rather it seems to be the distribution of the informational content between main and subordinated clauses.

3.2. Types of finite verbs At this point we understand ‘early verb position’ as a purely descriptive term for the position of the finite verb. It could thus be a INFL–medial position as proposed by Tomaselli (1995) and Schlachter (2004), and it has recently also been interpreted as a TP selecting a ȣP by Weiß (to appear). It could also be a Verb-end position as suggested by Axel (2007), leading to

Subordinate dhazs-clauses in Isidor 231

the consequence that all postverbal material has to be interpreted as extraposed. Before interpreting the structural position of the finite verb we want to determine its status as an auxiliary, copula or main verb in order to see if a more finely differentiated word class distinction helps to identify preferences for certain positions. Robinson observes that “the verb uuesan ‘be’ is different from all other verbs. In subordinate clauses it frequently ends up before predicate nominals or adjectives, or, if they precede, before the subject” (1997: 149) and he states that “this is clearly not just because the specific form of the verb found is a ‘light element’” (1997: 149).7 Apparently he is refering to the copula status of the verb. However, as table 3 shows, copulas and auxiliaries show the same syntactic behaviour: Table 3. Position and type of the finite verb

auxiliary verb main verb copula verb

Verb-early 12 9 9

Verb-end 8 12 6

Copulas and auxiliaries prefer early placement with 60:40% (if such a small number of examples allows a statistically relevant result at all). Main verbs, on the other hand, seem to prefer the end position with nearly the same percentage. If we look at the clauses with early auxiliary placement with respect to their lexical fillings, we find the following distribution: Table 4. Position and type of the auxiliary

ist sii uuari uuerde uuard uurdi mahti scolda

Verb-early 5 2 1 1 2

Verb-end 1

6 1

1 1

232

Eva Schlachter

The table shows a nearly complementary distribution: the auxiliary forms based on the verb be mainly appear in an early verbal position, whereas the forms with uuerdan tend to appear in final position. Thus Robinson’s statement about the significance of the verb uuesan can be strengthened and modified: uuesan ends up in early position with an extraordinarily high frequency, differently than other verbs. Since its behaviour is the same as a copula and as a auxiliary verb and since the other auxiliary verbs prefer final position, one of the reasons for early placement could be its light phonological and semantic weight. On the other hand, the forms of uuerdhan which prefer the final position behave rather as main verbs, which could be due to the fact that they are still at the very beginning of the grammaticali8 zation path from main verb to auxiliary. 9 The 9 copula constructions are mainly selected by verbs of saying; consider for instance: (3)

dhoh ir in cyres nemin quadhi, dhazs ir he although he in Kyros name saidSUBJ that ioh druhtin ist got is God and Lord ‘Although he used Kyros’ name when he said that he is God and the Lord’ deum et dominum ita esse testatur dicens (III.2, 151)

Nevertheless, the selecting main verb doesn’t seem to be decisive for early verb placement since Verb-end clauses occur with the same type of verb: (4)

Hear quidit umbi dhazs christus got here says about that Christ God ‘Here is said that Christ is God and the Lord’ Quia christus deus et dominus est.

endi and

druhtin ist. Lord is.

(III heading, 130) Fourquet (1939: 139–140) already observed that in clauses with only one nominal element (as he called it) the final verb could occur either in final or in early position. But in clauses with two nominal elements as in (5) the finite verb behaves quite regularily, usually occuring between these two elements. We will return to this type of clause in section 4.

Subordinate dhazs-clauses in Isidor 233

(5)

dhazs ir gote uuas that he GodDAT was ‘that he was equal to God’ …esse se equalem deo

ebenchiliih equal

(V.3, 406)

3.3. Syntactic functions of postverbal elements Conforming to the three verbal classes (main verb, copula and auxiliary) we distinguish between the following kinds of postverbal elements. Main verbs typically select a DP or a PP, copulas select a predicative nominal or adjective and auxiliaries either are followed by a DP and a non-finite verbal form or they introduce the verbal complex followed by a prepositional phrase. Accusative or dative objects never occur after the non-finite verb; they always follow the finite verb. Only prepositional phrases tend to be placed at the end of the clause either after the finite or after the non-finite verb. We found the following distribution: Table 5. Syntactic functions of the postverbal constituents

Vmain + DP/PP

9

Vcop + DP/AP predicative

9

Vaux + XP + Vnon-finite

8

Vaux + Vnon-finite + PP

4

The group [Vmain + DP/PP] is an important one for the claim of an INFLmedial position since examples with postverbal accusative or dative objects or with pronominal forms are usually not analysed as extraposed. We found 10 the following examples: (6)

“dhazs ih fora sinem anthlutte that I in front of his face dheodun,..” imu him peoplePL ‘That I subjugate his peoples in front of him’ ut subiciam ante faciem eius gentes

hneige subjugate

(III.2, 153)

234 (7)

Eva Schlachter

“dhazs dhu firstandes heilac chiruni“ that you understandSUBJ holy secret ‘that you understand the holy secret’ ...et archana secretorum… (III.2, 159)

(8)

“dhazs uuerodheoda druhtin sendida that army Lord sent dhir“ you ‘that the Lord of the armies sent me to you’ quia dominus exercitum misit me ad te

mih me

zi to

(III.9, 236) (9)

dhazs imu arsterbandemu siin fleisc ni flesh NEG that heDAT dieParttPres/DAT his enigan unuuillun. any decomposition ‘That when he died his flesh did not decompose’ quia moriens caro eius non uidit corruptionem ...

chisah saw

(IX.12, 719) (10)

dhazs sie ni eigun eouuihd, that they NEG own anything, dhar uuidar setzan. there against oppose ‘that they have nothing to object’ dum non habeant quod proponant

huuazs sie what they

(V.5, 430) (11)

Endi dhazs mittingart firleizssi diubilo and that earth leaveSUBJ/PRET devilsGEN endi auur aruuegodi zi sines drugidha his idols and again returnSUBJ/PRET to scheffidhes huldin. mercy CreatorGEN ‘and that the world leaves the deceptions of the devils and again returns to its Creator’s mercy’ Omissisque mundus dĊmonum simulacris reconciliaretur gratiĊ conditoris. (V.10, 507)

Subordinate dhazs-clauses in Isidor 235

(12)

dhazs ir dhoh in dheru chihuurfi zi that he after all in thisFEM/DAT turnsSUBJ/PRET to rehtnissa uuerchum. gotes minniu endi zi to justice’s work GodGEN love and ‘that after all he turns by the virtue of this to God’s love and to the work of justice’ ut uel per ipsam reuerteretur ad amorem dei et operationem iustitiĊ (V.10, 501)

(13)

dhinem fordhrom“ “dhazs dhu faris zi that you go to your ancestors ‘that you go to your ancestors’ ut uadas ad patres tuos (IX.2, 622)

The first three examples with postponed accusative object or postponed pronouns are part of quotation syntax. Examples (9) and (10) display accusative objects independently of quotation syntax. Both occur in negated clauses with postverbal indefinite pronouns or adjectives which have to be interpeted as focussed. Only (11) exhibits no specially marked accusative object, neither influenced by Latin nor quotation syntax. It could, however, be interpreted as a main clause. Interestingly (12) and (13) are constructed exactly as the Latin original. Nevertheless the clause final PPs in (12) and (13) have to be regarded as features of a Germanic syntax and are typically analysed as extrapositions. We conclude that the evidence for a structural early verb position within this group is based on just a few examples, which are mainly part of quotation syntax. The group [Vcop + DP/AP predicative], in contrast, doesn’t seem to be influenced either by quotation syntax or by the Latin text. We only find deviations from the Latin original as in (14) or (15), the only example which seems to imitate at least the ordering of the last constituents being (16): (14)

got dhazs ir selbo Christ ist chiuuisso that he himself Christ is certainly God druhtin ioh and Lord ‘that he himself, Christ, is certainly God and the Lord’ Quia idem deus et dominus est (III.1, 135)

236

Eva Schlachter

(15)

“dhazs ir gote uuas that he GodDAT was ‘that he was equal to God’ Esse se equalem deo

ebenchiliih…” equal

(16)

dhazs dher selbo gheist ist got. that the same spirit is God ‘that the same spirit is God’ Ut eundem spiritum ostenderet esse deum.

(V.3, 406)

(III.10, 248) One possible analysis of the types [Vaux + XP + Vnon-finite] is to regard them as cases of verb projection raising with the finite verb occurring in final position and the verbal projection with the non-finite verb extraposed to the right (cf. Axel 2007: 98-104) Between the finite and the non-finite verbal form only little material is included: out of a total of 8 cases the subject appears 5 times, as in (17) and (18), and only once does an object (19), an adverb (20) or PP (21). (17)

dhazs fona dhemu almahtigin fater dhurah inan father through him that from theDAT almighty uuordan ist al is all created ‘that all is created by the almighty father and by means of him’ quando a patre per illum cuncta creata esse noscuntur (II.3, 99)

(18)

dhazs dhar ist christ chizeihnit that there is Christ meant ‘that there is meant Christ’ intellege christum (III.2, 148)

(19)

“dhazs druhtin dhir ist huus zimbrendi“ house building that Lord youDAT is ‘that the Lord is building a house for you’ quod Ċdificaturus sit domum tibi dominus (IX.2, 620)

Subordinate dhazs-clauses in Isidor 237

(20)

dhazs (...) christ iu ist that (...) Christ yet is ‘that Christ came long ago’ (...) christus olim uenisse cognoscitur

lange long ago

quhoman come

(V.6, 455) (21)

dhazs ir selbo gotes sunu uuard in was in that he self GodGEN son chiboran born ‘that God’s son himself was born as a human’ eundem filium dei natum in carne monstremus

lihhe body

(V.1, 381) As Weiß points out (to appear: 7), clauses like (17) with an extraposed pronominal are not grammatical in modern dialects which allow verb projection raising. He therefore assumes that such constructions with a right adjoined VP are also not possible in Old High German and proposes an analysis with leftward movement of the finite verb. However, independently of the proposed analysis all of these clauses have to be regarded as examples of a typical Old High German syntax. The group [Vaux + Vinfinit + DP/PP] only consists of four examples. They all belong to the argumentative part of the treatise.11 (22)

unseres druhtin dhazs dhiz ist chiquedan in that this is said in our LordGEN nemin name ‘that this said in the name of our Lord’ Quod in persona specialiter christi domini nostri accipitur. (III.3, 174)

(23)

dhazs ir sih auur dhurah hreuun that he himself again through remorse chinisti. mahti chigarauuan zi could prepare to salvation ‘that he could prepare himself for salvation by remorse’ ut per penitentiam reparari possit ad ueniam (V.10, 498)

238 (24)

Eva Schlachter

dhazs ir iesus uuardh chinemnit in that he Jesus was named in iesuses bauhnungum dhes chiuuarin JesusGEN naming theGEN trueGEN ‘that he was called Jesus in naming the true Jesus’ ut iesus nominaretur ad significandum illum uerum iesum. (IV.2, 545)

These examples have to be evaluated like the ones in (12) and (13): although they apparently copy the Latin original, they are instances of a genuine Germanic syntax which is still used in present-day German. In the next section we try to find out if all the postverbal positions are motivated by the same information-structural purpose.

4. The Influence of information-structural factors We start our investigation with the hypothesis that the last or the postverbal parts of the clause are reserved for focussed information. This assumption is a very old one (cf. Rannow 1888: 119; Erdmann 1886: 190). Lacking the notion of focus it was understood as a specially stressed position. Since access to pronunciation in historical texts is not available, one method to find the focussed parts of the clause is text interpretation. Admittedly this isn’t a very transparent instrument either since discourse often allows different interpretations. Only in some cases do we find lexical items which we can take to be reliable indicators for focus interpretation. Before interpreting the information-structural values of the dhazs-clauses with early verb position, we have to define the concepts to be used. As a starting point we adopt Molnár’s (1993) approach to information structure, where Focus-Background, Topic-Comment and Theme-Rheme are located on different layers (for a detailed discussion, see Petrova and Solf in this volume). Thus, Topic and Focus, which are often interpreted as complementary categories, may overlap under certain circumstances. The TopicComment layer will not be treated here since topics, which are the better studied concept of this dichotomy, typically occur at the beginning of the clause and not at the end. The notions of Theme and Rheme refer to the status of discourse referents and are classified by the features [new] and [given]. The former means “newly introduced into discourse”, while the latter covers previously mentioned or accessible entities. In the theological argumentative Isidor treatise the given items are, for instance, God, Christ, the Holy Spirit,

Subordinate dhazs-clauses in Isidor 239

the prophet, and so on, which are under discussion throughout. A first approximation of the concept of Focus might use the notions of highlighting or marking of the relevant part of the utterance (cf. Petrova and Solf, this volume) Thus, Focus can be used to highlight old or new information. If it overlaps with new information, it is to be characterized as “new information Focus”, but this overlap is not obligatory. It also may coincide with old information. Other definitions of Focus recognize the potential overlap of Focus with old information and try to save the notion of “newness” by defining it as, for instance, a “new state of information in the addressee’s mind” (Lambrecht 1994: 210, quoted by Petrova and Solf, this volume). Nevertheless, such a definition cannot explain the rhetorical relations of the argumentative theological treatise we are dealing with in a satisfying way. The often repeated or derived claim that Jesus is God can never be considered as new information; it is always a known proposition wether it is accepted or not. We therefore adopt the semantic definition of Focus proposed by Rooth (1985, 1992) and developed by Krifka (2007: 6). This says that “Focus indicates the presence of alternatives that are relevant for the interpretation of linguistic expressions.” In our example the excluded alternative in interpretation is the heretic view that Jesus should be regarded as a human being and not the son of God. This special pragmatic use could also be described as a means to “correct or confirm information” (Krifka 2007: 11), which can be considered as a case of contrastive Focus (Krifka 2007: 21). In any case, these functions of Focus, including new information Focus and contrastive Focus, should also be subsumed under the general definition of Focus as exclusion of alternatives. Using this definition we try to determine typical interpretations of the postverbal material and to find a motivation for its position, be it extraposed or not.

4.1. DPs or PPs after the finite main verb We cannot confirm the traditional view that the constituents after the finite main verb represent focussed information if we understand focus in the sense of exclusion of alternatives. Sometimes the reading of exclusion is only possible if we extend the focus phrase and include the predicate. Restriction to the postverbal NP leads to an interpretation which can better be described as “new information” or as rhematic material. Consider, for example, (7), repeated here for convenience as (25).

240 (25)

Eva Schlachter

“dhazs dhu firstandes heilac chiruni“ that you understandSUBJ holy secret ‘that you understand the holy secret’ ... et archana secretorum... (III.2,159)

The ‘holy secret’ is not contrasted with something else; its reference is vague as it is newly introduced within the quotation part. The focus rather includes the predicate answering the question: “What do you want me to do?” In contrast, the postverbal personal pronouns in (8), repeated here as (26), which typically indicate givenness, seem to contradict the observation that the postverbal material is new. As already mentioned, however, these sentences are quotations from the Old Testament. Being imported into the text of the Isidor treatise the pronouns cannot refer to known material. They count, therefore, as unknown material. (26)

“dhazs uuerodheoda druhtin sendida that army Lord sent dhir“ you ‘that the Lord of the armies sent me to you’ quia dominus exercitum misit me ad te.

mih me

zi to

(III.9, 236) Only the continuation of the text in (27) explicitly suggests an interpretation for the referent of the pronoun mih in (26): (27)

Huuelih ist auur nu dhese druhtin which one is however now this Lord fona uuerodheoda drohtine chisendit, nibu from army Lord sent, if not auur dher selbo druhtin nerrendeo christ again this same Lord Redeemer Christ ‘Which one, however, is this Lord sent by the Lord of the armies, if not just the same Lord Redeemer Christ? (III.9, 238)

The same holds for the postverbal preposition phrase in (13), here (28).

Subordinate dhazs-clauses in Isidor 241

(28)

“dhazs dhu faris zi dhinem fordhrom“ ‘that you go to your ancestors’ ut uadas ad patres tuos. (IX.2, 622)

In fact, we do not know who this person addressed by dhu is, nor who his ancestors are. Thus, if our interpretation of the postverbal parts in quotation syntax is on the right track, we have found one of their characteristics: they typically introduce new material. Nevertheless, in the argumentative part there are some cases where the postverbal constituent is focussed in the sense that other alternatives are excluded. They typically involve negations with a focussed indefinite adjective or pronoun where the excluded alternative consists of the positive counterpart. Thus, the postverbal constituents only provide focussed information if they include a focus-inducing element: (29)

dhazs imu arsterbandemu siin fleisc ni his flesh NEG that heDAT diePartPresDAT enigan unuuillun. any decomposition ‘that when he died his flesh did not decompose’ quia moriens caro eius non uidit corruptionem ...

chisah saw

(IX.12, 719) (30)

dhazs sie ni eigun eouuihd, that they NEG own anything, dhar uuidar setzan. PART oppose ‘that they have nothing to object’ dum non habeant quod proponant

huuazs sie what they

(V.5, 430) 4.2. Constituents within or after the verbal complex Our findings that material after the main verb is only interpreted as new information unless it is accompanied by a focus-inducing pronoun or adjective does not seem to be true if we look at the examples (22) and (24): both consist of a complex DP where the Genitive and the adjective can be interpreted as contrastively focussed. We conclude that in addition syntactic

242

Eva Schlachter

complexity can also be a motivation for postverbal Focus. This seems to be true if we look at the simple DP zi chinisti ‘to salvation’ in (23), which cannot be evaluated as focussed. The picture is different for constituents between the finite and the nonfinite verb. Except for (33), they all have to be interpreted as contrastively focussed. (31) (32) (33) (34) (35)

ist al uuordan (II.3) ist christ chizeihnit (III.2) ist huus zimbrendi (IX.2) ist lange quhoman (V.6) uuard in lihhe chiboran (V.1)

We thus conclude that focussed (contrasted) material is not inserted in relation to finiteness, but in relation to the main verb and that it prefers the preverbal position. Postverbal focussed material apparently needs a special lexical or syntactic marking to be licensed. However, in establishing this generalization a methodological problem may arise: having no access to pronunciation we have no means to decide wether a constituent which is not lexically marked can be interpreted as focussed or not. I am aware of this problem, but the only way to deal with this issue is to rely on the weaker method of textual interpretion.

4.3. Identificational and predicative constructions Typical focussed information occurs in copular constructions that are analysed as identificational clauses,12 where the second NP serves to fix the identity of the first NP: (36)

dhazs ir selbo Christ ist that he self Christ is druhtin Lord quia idem deus et dominus est (...)

chiuuisso certainly

got God

ioh and

(III.1, 135)

Subordinate dhazs-clauses in Isidor 243

(37)

dhazs ir that he (III.2,151)

ist is

(38)

dhazs iesus ist that Jesus is dominum esse iesum

got God

ioh and

druhtin Lord

druhtin Lord (VI.2, 549)

The context is always the same: after a quotation or argumentational part the author confirms that the previously mentioned Christ has to be God and Lord, where we interpret the last NPs as being focussed. None of the constituents delivers new information. Nevertheless, other interpretations are possible: it could be that the first part of the equation is the prominent one (if ir selbo or iesus were stressed). It is even possible that the copula itself is stressed, which leads to an interpretation in which the truth value of the assertion is focussed (cf. Höhle’s Verum-Fokus 1992). Nevertheless, we assume that it is the postverbal constituent which is focussed, the sentence adverb chiuuisso in (36) can be seen as an indication of this interpretation. In this respect copulas are identical to auxiliaries, which also appear before the focussed material. We thus get the following ordering: (39)

Background / Verbaux/cop / Focuscontr / Verbmain/

New information Focus (lex. induced) Focus (syntact. compl.)

5. Results We have argued that only by distinguishing different types of dhazs-clauses can we arrive at insightful explanations of the interaction between sentence structure and information structure. The widely held belief that the preferred position for focussed constituents is the clause-final one must be modified in the sense that not all constituents at the end are focussed in the same manner or are focussed at all. Prepositional phrases often simply provide new material, and have to be interpreted as rhematic. Sometimes they provide given material as in (23). The majority of Verb-early clauses headed by main verbs also lack postverbal contrastive focus, the postverbal material has to be classified as new rather than contrastive. In this case the new information focus com-

244

Eva Schlachter

prises the entire VP. At the same time it has been shown that these clauses mainly occur in the quotation parts of the text. The situation is different for auxiliaries and copulas in early position, which are usually followed by a contrastive constituent. A typical case is the identificational copular clause, where the postverbal nominal element is contrasted with alternative interpretations. The majority of these structures are independent of the Latin or quotation syntax. Furthermore, the observation that the verb uuesan (to be) plays an outstanding role in early verb placement can be strengthened and modified. Not only as a copula verb but also as an auxiliary uuesan prefers an earlier position in the clause. We therefore assume a structural Verb-early position for auxiliaries and copula verbs, which, following Fuß and Trips (2002), we identify with ȣP. This genuine Old High German structure is then used for structures with early main verbs as they typically occur in quotation syntax. They have the common feature of a special information-structural pattern and point to a different stylistic system, which could be described either as archaic or as contact induced. At the same time there are many Verb-end structures with copulas and auxiliaries (and, of course, main verbs). These belong to the subsystem of subordinated clauses which already existed in Old High German and became generally established in the later development; consider (40) and (41):13 (40)

dhazs christ (…) Ċr allem uueraldim fona that Christ before all worlds by fater uuard chiboran. (the) father was born ‘that Christ before the beginning of the worlds was born by the father’ …ante omnia secula filius a patre genitus esse declaratur. (II.3, 97-98)

(41)

Hear

quidit umbi

dhazs christus got

here says about that Christ God ‘here is said that Christ is God and the Lord’ Quia Christus deus et dominus est.

endi

druhtin ist.

and

Lord

is

(III. heading, 130) The later unification of the syntactic position for finite verbs has the disadvantage of an information-structural ambiguity common to subordinated clauses with Verb-end. If there is more than one constituent before the verbal complex, as for example in (40), it is not necessarily clear whether the preverbal constituent is focussed or if the whole subordinate clause is to be

Subordinate dhazs-clauses in Isidor 245

interpreted as backgrounded. The contrastive reading means that ‘Christ is the son of the divine father and not of a human being’. In contrast, the background reading just summarizes the knowledge of Christ’s birth as a starting point for further argumentation. This interpretation gains support by the fact that Verb-end clauses typically occur at the beginning or at the end of a paragraph. The Verb-end analysis also accounts for clauses where main verbs are followed by constituents which are marked by a focus-inducing element or by syntactic complexity. These are analysed as extrapositions triggered by a special prosodic pattern. Finally, we don’t regard these different verb positions as parts of different grammars but as different means of the same grammar with special discourse functions. It is possible that information structure allows different sentence structures as in the example quoted by Fourquet (mentioned in section 3.2 as (5), here (42)). Fourquet obviously did not see any regularity in the position of the predicative adjective with respect to the dative object except for the fact that the verb apparently has to occur in the penultimate position. In (42) the predicative adjective shows up after the copula and in (43) it is preverbal. In both positions they have to be interpreted as focussed as revealed by the context given in brackets [ ]. (42)

“dhazs ir gote uuas ebenchiliih.” that he GodDAT was equal ‘that he was equal to God [and not just in the same form] esse se equalem deo. (V.3, 406)

(43)

dhazs ir chihoric uuari gote that he obedient wasSUBJ GodDAT that he should be obedient to God [and no longer disobedient] ut esset deo subiectus (V.9, 491)

Following our analysis of copular structures, uuas in (42) occurs in ȣ°. The focussed predicative adjective corresponds to the established informationstructural pattern. Nevertheless, the Verb-end analysis is not excluded since the extraposed constituent is lexically marked: the adjective ebenchiliih ‘equal’ is a composition of eban and chiliih, both meaning ‘equal’. Its specific lexical (and prosodic) structure determines the final position in the clause. In contrast, chihoric ‘obedient’ in (43) has no special lexical focus

246

Eva Schlachter

device and makes use of the usual preverbal focus placement of the ‘regular’ Verb-end construction. The extraposition of gote guarantees that there is no information-structural ambiguity in the sense that a background reading is not available. We are aware of the fact that our results are based on a very small number of examples and might be better described as hypotheses. In future research we hope to test these findings against all the subordinated clauses of the Isidor text.

Notes 1. 2.

3. 4.

5. 6. 7. 8. 9. 10.

11.

12. 13.

I want to thank Svetlana Petrova and Roland Hinterhölzl for interesting discussions and comments on an earlier version of this paper. At this point in the discussion, we don’t distinguish between the different proposals for the landing-site of the finite verb be it VIP as in Kiparsky (1996), ȣP as in Fuß and Trips (2002) or TP and ȣP as in Weiß (to appear). But see Hinterhölzl (2004) for a critical evaluation of the double base hypothesis. We opted for Egger’s edition because it is accessible to a greater public. Since we not only quote the line, but also the chapter and the paragraph, it will be possible to find the quoted clauses in Hench’s edition as well. The Verb-second analysis is based on Jäger’s (2005) account of negation, where the negative particle is clitised to the finite verb. See Fleischer (2006) for methodological problems in general. This fact, already noted by Fourquet (1974: 318–319) has not yet received a satisfactory explanation. This point was stressed by Svetlana Petrova in a personal communication. Seven times with the pattern uuesan/sin + NP/AP, two times with the pattern uuerdan + NP/AP. We don’t consider Verb-first constructions. Many of them are parallel to the Latin text as in the following example: Dhazs ni bilibun ano herrun iudaeoliudi fona iudases chunne ‘that the Jews of the family of Judas are without masters’ Non defuisse principes iudeorum populi ex genere iuda (VIII.1, 581). Again, we won’t consider a Verb-first construction from the quotation part: „dhazs chiendot uuerdhe dhiu aboha ubarhlaupnissi…” That finished beSUBJ this bad offence. Ut consummetur praeuaricatio (V.6, 448). For a recent discussion of copular clauses, see Geist (2006). We take examples with a final verbal complex as cases of Verb-end even if the finite verb occurs in the penultimate position.

Subordinate dhazs-clauses in Isidor 247

References Axel, Katrin 2007 Studies on Old High German Syntax. Left sentence periphery, verb placement and verb-second. Amsterdam/Philadelphia: John Benjamins. Besten, Hans den 1977 On the Interaction of Root Transformation and Lexical Deletive Rules. Ms. University of Amsterdam. Borter, Alfred 1982 Syntaktische Klammerbildung in Notkers Psalter. Berlin/New York: Walter de Gruyter. Diels, Paul 1906 Die Stellung des Verbums in der älteren althochdeutschen Prosa. Berlin: Mayer & Müller. Eggers, Hans 1964 Der althochdeutsche Isidor. Nach der Pariser Handschrift und den Monseer Fragmenten. Tübingen: Niemeyer. Erdmann, Oskar 1886 Grundzüge der deutschen Syntax nach ihrer geschichtlichen Entwicklung. Stuttgart: Cotta. Fleischer, Jürg 2006 Zur Methodologie althochdeutscher Syntaxforschung. Beiträge zur Geschichte der deutschen Sprache und Literatur 128 (1): 25−69. Fourquet, Jean 1939 L’ ordre des élément de la phrase en germanique ancien. Paris: Les Belles Lettres. 1974 Genetische Betrachtungen über den deutschen Satzbau. In Festschrift für Hugo Moser zum 65. Geburtstag. Studien zur deutschen Literatur und Sprache des Mittelalters, eds. Werner Besch, Günther Jungbluth, Gerhard Meissburger and Eberhard Nellmann, 314−323. Berlin: Erich Schmidt. Fuß, Eric 1998 Zur Diachronie von Verbzweit. Die Entwicklung von Verbstellungsvarianten im Deutschen und Englischen. M.A. thesis, Universität Frankfurt/Main. Fuß, Eric and Carola Trips 2002 Variation and change in Old and Middle English: On the validity of the Double Base Hypothesis. Journal of Comparative Germanic Linguistics 4: 171−224. Geist, Ludmilla 2006 Die Kopula und ihre Komplemente. Zur Kompositionalität in Kopulasätzen. Tübingen: Max Niemeyer.

248

Eva Schlachter

Hench, George A. 1893 Der althochdeutsche Isidor. Facsimile-Ausgabe des Pariser Codex nebst critischem Texte der Pariser und Monseer Bruchstücke. Straßburg: Karl J. Trübner. Hinterhölzl, Roland 2004 Language change versus grammar change: What diachronic data reveal about the distinction between core grammar and periphery. In Diachronic Clues to Synchronic Grammar, eds. Eric Fuß and Carola Trips, 131−160. Amsterdam/Philadelphia: John Benjamins. Hinterhölzl, Roland and Svetlana Petrova 2005 Rhetorical relations and verb placement in Early Germanic languages: Evidence from the Old High German Tatian translation (9th century). In Salience in Discourse. Multidisciplinary Approaches to Discourse, eds. Manfred Stede, Christian Chiarcos, Michael Grabski and Luuk Lagerwert, 71–79. Münster: Stichting/Nodus. Höhle, Tilman 1992 Über Verum-Fokus im Deutschen. In Informationsstruktur und Grammatik, ed. Joachim Jacobs, 112−141. (Linguistische Berichte, Sonderheft 4.) Opladen: Westdeutscher Verlag. Jäger, Agnes 2005 Negation in Old High German. Zeitschrift für Sprachwissenschaft 24 (2): 227−262. Kemenade, Ans van 1987 Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris. Kiparsky, Paul 1996 The shift to head-initial VP in Germanic. In Studies in comparative Germanic Syntax II, eds. Höskuldur Thráinsson, Samuel D. Epstein and Steve Peter, 140−179. Dordrecht: Kluwer. Krifka, Manfred 2007 Basic notions of information structure. Interdisciplinary Studies on Information Structure (ISIS) 6: 13–56. Kroch, Anthony S. 1989 Reflexes of grammar in patterns of language change. Journal of Language Variation and Change 1: 199−244. Krotz, Elke 2002 Auf den Spuren des althochdeutschen Isidor. Studien zur Pariser Handschrift, den Mondseer Fragmenten und zum Codex Junius 25. Mit einer Neuedition des Glossars Jc. Heidelberg: C. Winter. Lambrecht, Knud 1994 Information Structure and Sentence Form: Topic, Focus and the Mental Representation of Discourse Referents. Cambridge: Cambridge University Press.

Subordinate dhazs-clauses in Isidor 249 Lenerz, Jürgen 1984 Syntaktischer Wandel und Grammatiktheorie. Eine Untersuchung an Beispielen aus der Sprachgeschichte des Deutschen. Tübingen: Niemeyer. Lippert, Jörg 1974 Beiträge zu Technik und Syntax althochdeutscher Übersetzungen unter besonderer Berücksichtigung der Isidorgruppe und des althochdeutschen Tatian. München: Wilhelm Fink. Matzel, Klaus 1970 Untersuchungen zu Verfasserschaft, Sprache und Herkunft der althochdeutschen Isidor-Sippe. (Rheinisches Archiv 75.) Bonn: Ludwig Röhrscheid. Molnár, Valéria 1993 Zu Pragmatik und Grammatik des TOPIK-Begriffes. In Wortstellung und Informationsstruktur, ed. Marga Reis, 155−202. Tübingen: Niemeyer. Müller, Gertraud and Theodor Frings 1959 Die Entstehung der deutschen daß-Sätze. Berlin: Akademie Verlag. Nordmeyer, George 1958 On the OHG Isidor and its significance for Early German prose writings. Publications of the Modern Language Association 73: 23−35. Ostberg, Kurt 1979 The Old High German Isidor in Its Relationship to the Extant Manuscripts (Eighth to Twelfth Century) of Isidorus ‘De Fide Catholica’. Göppingen: Kümmerle. Petrova, Svetlana 2006 A discourse-based spproach to verb placement in Early West-Germanic. Interdisciplinary Studies on Information Structure (ISIS) 5: 153−185. Petrova, Svetlana and Michael Solf this vol. On the methods of the information-structural analysis of historical texts: A case study on Old High German. Pintzuk, Susann 1993 Verb seconding in Old English: Verb movement to Infl. The Linguistic Review 10, 5−35. 1999 Phrase Structures in Competition: Variation and Change in Old English Word Order. New York: Garland. Rannow, Max 1888 Der Satzbau des althochdeutschen Isidor im Verhältnis zur lateinischen Vorlage. Berlin: Weidmannsche Buchhandlung. Robinson, Orrin, W. 1994 Verb first position in the Old High German Isidor translation. Journal of English and Germanic Philology 93: 356−373.

250

Eva Schlachter

1997

Clause subordination and verb placement in the Old High German Isidor translation. Heidelberg: Carl Winter.

Rooth, Mats 1985 Association with focus. Ph.D. dissertation, University of Massachusetts, Amherst. 1992 A theory of focus interpretation. Natural Language Semantics 1: 75−116. Schlachter, Eva 2004 Satzstruktur im Althochdeutschen. Eine Skizze zur Position des Verbs im Isidor-Traktat des 8. Jahrhunderts. In Beiträge zu Sprache und Sprachen 4. Vorträge der Bochumer Linguistik-Tage, ed. Karin Pittner, Robert J. Pittner and Jan C. Schütte, 179−188. München: Lincolm. Sonderegger, Stefan 2003 Althochdeutsche Sprache und Literatur. Berlin/NewYork: Walter de Gruyter. Tomaselli, Alessandra 1995 Cases of verb third in Old High German. In Clause Structure and Language Change, eds. Adrian Battye and Ian Roberts, 345−369. New York/Oxford: Oxford University Press. Weiß, Helmut (to appear) Die rechte Peripherie im Althochdeutschen. Zur Verbstellung in dass-Sätzen. To appear in: Akten der Arbeitstagung der Indogermanischen Gesellschaft, Jena 2006.

Information structure and word order variation in the Old High German Tatian Svetlana Petrova

1. Introduction Word order variation in the right periphery of subordinate clauses is one of the most striking properties of Old High German (OHG) syntax. It belongs to the central and most vividly discussed topics in the historical treatment of sentence structure in German, namely to the description of the rules and principles that determine the position of the finite Verb (Vfin) in the earlier stages of the German language. The same issue has been intensively discussed with respect to the remaining early Germanic languages as well. In recent generative work on Old English (OE), two different accounts have been put forward. The first one launched by Kemenade (1987)1 attributes surface orders with postverbal constituents in clauses with overt complementizers to extraposition from a uniform SOV base. This kind of operation applies in modern SOV langages as well, especially with PPs and CP-complements which are regulary extraposed to the right of the selecting verb. Additionally, some restructuring operations in verb clusters leading to orders with a tensed auxiliary before the untensed main verb in subordinate clauses (verb raising and verb projection raising) were originally analysed as instances of rightward movement of the VP as well (Haegeman and van Riemdijk 1986). In line with the theoretical discussion on properties of asymmetric SOV languages, structural variation in the right periphery of subordinate clauses in OE was explained as the result of rightward movement while Vfin always remains in its basic position in the end of the clause (see also Tomaselli 1995, 350–351). However, the idea that OE has a uniform SOV structure in the base has been challanged by Pintzuk (1991) who discovered evidence for postverbal phrases, e.g. pronouns and light adverbs, which are excluded form extraposition in modern SOV langauges. To explain structural variation in the data, Pintzuk claimed that OE displays variation in the head-complement parameter in both I(nfl)P and VP. In line with this model, non-Vend orders

252

Svetlana Petrova

in OE are explained partly as a result of leftward movement of Vfin to a clause-medial I(nfl)P, and partly as instances of VO in the base. The basic points in the discussion on word order variation in OE have been also applied to the interpretation of the OHG data. Weiß (2006) discusses word order in complement clauses introduced by dass ‘that’ in the so-called ‘Minor texts’ of the OHG tradition. He is able to derive a great part of the non-Vend orders form a basic SOV order, although he is forced to assume, apart from extraposition of PPs and heavy complements, a series of leftward movement operations according to which Vfin targets two different functional projections (TP and ȞP) below CP. A different approach is pursued by Schallert (2006) who discusses evidence for mixed word OV/VO order in OHG claiming that the early Germanic languages were unspecified with respect to the head-complement parameter in the VP. The present study addresses this complex discussion from the perspective proposed by Hinterhölzl (2004) who relates word order variation to properties of the information-structural organization of the utterance. A first and by now unique empirical investigation on variation in the right periphery of subordinate clauses in OHG is provided by Schlachter (2004). Her analysis reveals that the different placements of Vfin in complement clauses in the OHG Isidor correlates with the iconic separation of the domains of focus and background in the clause. These findings are in line with a long tradition in the descriptive literature which related the principles of verb placement in dependent clauses to stylistic effects and properties of theme-rheme (see the summary in Ebert 1978, 39–43). The foregoing observations suggest that pragmatic considerations play an important role in the explanation of word order variation in the earlier stages of German. Therefore, the aim of the present paper will be to explore in more detail the extent to which information-structural principles are responsible for the different placements of Vfin in subordinate clauses in OHG. Special attention will be placed on the correlation between pragmatic properties of constituents like givenness/novelty, contrast, emphasis and the like, and their positional realization with respect to Vfin.

2. Properties of the database This study analyses data from the OHG Tatian translation, which is the largest prose text of the classical period of OHG. In order to base the observations on genuine OHG structures, we will examine only sentences in which the word order differs from that of the corresponding Latin original.2

Information structure in the Old High German Tatian

253

Project B4 of SFB 632 “Information Structure” has provided a data collection of all clauses displaying differences in word order between OHG and Latin in the text parts assigned to three different scribes (Į, ȕ, and ε). For the purpose of the present study, we will analyse the dependent clauses found in this data collection. First, let us look at the quantitative distribution of Vend vs. non-Vend orders in all conjunctional and relative clauses found in the database. The figures provided in Table 1 clearly show that non-Vend orders are highly frequent in subordinate clauses in OHG. Causal clauses are listed separately because they are ambiguous between coordinate root conjuncts with an extra-clausal connective comparable to modern German denn ‘because’ and subordinate clauses with a lexically filled complementizer. This ambiguity may explain the high number of non-Vend orders among the causal clauses. However, the ratio of non-Vend among the unambiguously subordinate clauses is only slightly below 50 per cent. This means that in nearly half of the conjunctional clauses the scribe decided to depart from the structure of the Latin original but ended up in a structure which is not Vend in OHG3: Table 1. Relative frequency of Vend vs. non-Vend orders in subordinate clauses in the OHG Tatian in the database of the study. clause type

total

Vend

non-Vend

conjunctional clauses

364

190

52,2 %

174

47,8 %

relative clauses

196

129

65,8 %

67

34,2 %

causal clauses

79

29

36,7 %

50

63,3 %

There is crucial evidence suggesting that non-Vend is an authentic native pattern in subordinate clauses in OHG. First, we find cases where Vend order is given in the Latin original but suspended in the OHG translation, see (1a). Second, non-Vend is attested in clauses whose Latin equivalent lacks a finite verb, see (1b–c). In such cases, we can assume that the insertion of Vfin is ruled by native OHG grammar. As the examples show, the scribes disregard the oportunity to preserve or create Vend patterns in subordinate clauses:

254

Svetlana Petrova

(1)

a. thaz [...] thie dar gisehent daz sie sin blinte (T 224, 4–6) that […] those PRT see that they are-SUBJ blind-PL ‘[I have come,] so that […] those who see may be made blind’ lat. ut […] qui uideant caeci fiant b. salige sint thiethar sint sibbisame (T 60, 16) blessed are who-PRT are peaceful ‘Blessed are the peacemakers’ lat. Beati pacifici c. In thie burg/ galileĊ thero namo ist nazar&h (T 28, 4–5) in that city Galilee whose name is Nazareth ‘into that city of Galilee whose name is Nazareth’ lat. In ciuitatem/ galileae cui nomen nazar&h

Moreover, the postverbal domain in OHG hosts types of constituents which do not undergo movement to the right in modern SOV languages, e.g. single (unmodified) NPs (2a), predicative nouns (2b) and adjectives (2c):4 (2)

a. Inti thie thár hab&un diuual (T 59, 1) and who PRT had devil ‘and those who were possessed by the devil’ lat. & qui demonia habebant b. thaz sie hiezzin boanerges (T 59, 22) that they were called Boanerges ‘that they be called Boanerges’ lat. boanerges c. oba thin ouga uuirdit luttar (T 69, 22) if you eye becomes bright ‘if your eye becomes bright’ lat. si fuerit occulus tuus simpex

The number of postverbal single elements after Vfin in subordinate clauses is raised by non-finite forms of main verbs in complex predicates, e.g. in constructions with modal verbs (3a–b), or in the combination of sîn ‘to be’ and uuerdhan ‘to become’ with the past participle (3c), a construction which is formally identical with the passive in modern German:

Information structure in the Old High German Tatian

(3)

255

a. thaz sie Inan mohtin ruogen (T 199, 2) that they Him might-SUBJ accuse-INF ‘that they had something of which to accuse Him’ lat. ut possent accusare eum b. Inti thiethár uuolle mit thír uuehslon (T 65, 12) and who-PRT want-SUBJ from you-DAT borrow-INF ‘and whoever wants to borrow from you’ lat. & uolenti mutuare a té c. nibi ir uuerdet giuuentite/ inti gifremite soso theser luzilo (T 151, 12) NEG-if you become converted-PL and formed-PL like this young [boy] ‘if you do not convert and become like this young boy’ lat. nisi conuersi fueritis/ & efficiamini sicut paruuli

Orders in which the tensed auxiliary precedes the untensed main verb in subordinate clauses have been related to phenomena like verb raising and verb projection raising typical for verb clusters in modern SOV languages as well (Fuß and Trips 2002). However, the degree of grammaticalization of periphrastic forms is questioned in OHG, especially with respect to the formal equivalent of the modern passive construction. According to the common view, this expression is still undergoing a process of grammaticalization from a copular construction to a periphrastic form (Valentin 1987 among others). The presence of inflectional endings on the participle, which agrees in number, gender and case with the corresponding subject constituent as in (3c), is a strong indication of the copular status of the construction. Copular construction, however, are not among the clusters discussed in relation with verb raising or verb projection raising in the literature (Wurmbrand 2004). These observations prompt the view that variation in verb placement in subordinate clauses is a genuine syntactic property of OHG which calls for an alternative description. For this reason, a more detailed analysis of the grammatical and pragmatic properties of constituents in different syntactic patterns in subordinate clauses from the OHG period is needed. The present study is based on the analysis of 100 clauses with Vend order and 100 clauses with non-Vend order which were selected from the main corpus. Representatives of all types of subordinate clauses established above, i.e. conjunctional, relative and causal clauses, are included. For reasons of unambiguous classification, sentences containing only one constituent apart from Vfin have been left aside, as a serialisation of the

256

Svetlana Petrova

type Conjunction – XP – Vfin may be viewed both as V2 and Vend. Furthermore, the influence of the translation technique known for the OHG Tatian has been also reflected in the selection of the database: as the requirement not to shift material across the lines of the manuscript is the main translation principle stated for this text (Masser 1997), examples where the transposition of constituents appears to be blocked by the line break also remained unconsidered.

3. Analysis 3.1. Clauses with Vend order 3.1.1. The placement of discourse-anaphoric material Among the total of 100 instances with Vend order, there are 34 cases in which the only syntactic difference in clause structure results form the insertion of the subject pronoun in the translation while the rest of the sentence remains unchanged. See (4a–b) where the subject pronouns her ‘he’ and ír ‘you-2pl.nom’ have no equivalent in the Latin clause but the order of the PP mit imo ‘with him’ and the direct object zeichan inti uuvntar ‘signs and miracles’ with respect to Vfin is identical in both OHG and Latin: (4)

a. thaz her mit imo uuari (T 88, 26) that he with Him was-SUBJ ‘that he was with Him’ lat. ut cum eo ess& b. nibi ír zeichan inti uuvntar giseh& (T 90, 18) NEG-if you signs and miracles see ‘unless you see signs and miracles’ lat. nisi signa & prodigia uideritis

These examples are revealing as to where subject pronouns are usually placed in OHG. In fact, our group of Vend sentences shows a 100 percent consistency with respect to the placement of subject pronouns, as in all cases investigated here the position chosen for the pronoun in the OHG clause is always the one immediately after the subordinating conjunction, i.e. the so called Wackernagel position (see also Tomaselli 1995, 349). In 28 additional cases, the same syntactic positon is targeted by other pronominal elements in the OHG text, see the direct object thiu ‘these

Information structure in the Old High German Tatian

257

things-3pl.acc.neutr’ in (5a), the indirect object thir ‘you-2sg.dat’ in (5b), and the prepositional object in (5c): (5)

a. tho siu thiu gisah (T 28, 12) when she this-PL saw ‘as she saw these things’ lat. quae cum uidiss& b. unzan ih thir quede (T 40, 28) until I you-DAT tell ‘until I tell you’ lat. usquedum dicam tibi c. soso zi In gisprochan uuas (T 37, 5) as to them-DAT said was ‘as has been said to them’ lat. sicut dictum est ad illum

It can be also shown that pronominal arguments appear to the left of adverbials which according to the standard syntactic assumptions mark the left edge of the VP, see thara and thar ‘there’ in (6a–b): (6)

a. Inti thô her thara quam (T 42, 03) and when He there came ‘and when he came there’ lat. & ueniens b. thô sie thar uuarun (T 35, 22) when they there were ‘as they got there’ lat. cum essent ibi

These facts about OHG sentence structure have not passed unnoticed in the literature. According to Behaghel (1932, 4–6, §1426), word order in early Germanic is subject to an intricate interplay of two basic sets of principles, the first one concerning the informational relevance of sentence constituents and the second one concerning their “physical” properties (ibid. 5) in terms of relative length and phonological heaviness. According to the first set of rules, less relevant information tends to precede more relevant one in the clause. Additionally, the principle of growing constituents requires shorter constituents to precede longer ones in the clause. From this perspective, structures like (4)–(6) show a perfect interplay of the mentioned requirements. It is obvious that pronouns representing previously men-

258

Svetlana Petrova

tioned information can be analysed as less relevant than the rest of the utterance conveying the new information in the discourse. Additionally, they represent short items which tend to be de-accented and cliticised to other tonic elements and thus being typical cases of phonologically light elements which tend to be placed before heavy, full lexical material. This evidence raises the question whether anaphoricity is the factor leading to the identical positional distribution of pronominal arguments in OHG. In order to examine this issue, we will turn our attention to the syntactic realization of full lexical arguments with anaphoric properties in OHG. According to Dittmer and Dittmer (1998, 21), full phrases tend to be less often shifted across Vfin in contrast to pronominal elements in the OHG Tatian. However, our data base yields 8 instances in which full constituents with anaphoric properties are shifted from the postverbal position in the Latin sentence to the preverbal domain in OHG. The reverse transposition is not found in the corpus. Consider (7a–b), in which the discourse-given object DP thén buoh is shifted across the verb against the Latin original in two subsequent clauses: (7)

a. so hér thén buoh int&a (T 53, 21) when He this book opened ‘as he opened the book’ lat. & ut reuoluit librum b. inti mit thiu hér thén buoh bit&a (T 53, 32) and when He this book closed ‘and as he closed the book’ lat. & cum plicuiss& librum

The fact that, just like pronouns, full lexical DPs are also regularly shifted to the preverbal domain when they are discourse-given, strongly supports Behaghel’s rule of relevance. But what about the principle of growing constituents and the role of phonological heaviness, given the fact that we deal with full lexical categories bearing an overt determiner and therefore comprising several syllables? In a historical corpus, we have no opportunity to judge about the prosodic realization of sentence constituents. However, we can draw parallels to the situation in some contemporary intonational languages. Lakoff (1976, 288) presents some well-known facts for English. He shows that full DP-expressions regularly give up accent to the verb when they refer back to a previously mentioned antecedent, see (8a), Lakoff’s (91). At the same time, main accent on a DP blocks its interpretation as an anaphor to a pre-established referent, see (8b), Lakoff’s (92)5:

Information structure in the Old High German Tatian

(8)

259

a. When Harryi entered the room, Mary KISsed the presidenti. b. When Harryi entered the room, Mary kissed the PREsidentj/ *PREsidenti.

So anaphoric reading correlates with de-accentuation, and vice versa. From this we can assume that in OHG too, full DPs were de-accented when used in anaphoric relation to an antecedent in the previous context. Consequently, they do not count as heavy constituents but share the prosodic behaviour of pronouns and light adverbs. Again, Behaghel’s principles cooccur: anaphoric, i.e. informationally less relevant material, as well as phonologically de-accented, i.e. light material, appears early in the clause.

3.1.2. Non-anaphoric information The observations made on the positional realization of discourse-given material are confirmed by the fact that Vend order is found in clauses containing familiar information only, e.g. resuming a pre-established fact or conveying an expected, inferable event. In (9), the entire information in the purpose clause, namely that the first-born son shall be presented to God, is inferrable from the common knowledge of the customs of the Jewish people explicitly refered to in the context (after moyseses euuu ‘according to the Law of Moses’): (9)

Inti after thiu gifulta uuarun taga /[…] brahtun sie Inan thô In and after filled were [the] days/ […] brought they Him-ACC then to hierusalem/ thaz sie Inan gote giantuuvrtitin (T 37, 11–14) Jerusalem/ that they Him-ACC God-DAT presented-SUBJ ‘after the days of her purification they brought Him to Jerusalem to present Him to the Lord’ lat. & postquam Impeti sunt dies/ […] tullerunt illum In herusalem/ ut sisterent eum domino

However, unlike the examples discussed so far, there are subordinate clauses with Vend order in which the preverbal domain does not convey anaphoric or inferable information. Instead, the expressions preceding Vfin can be seen to achieve special prominence over the rest of the utterance for different pragmatic or contextual reasons. Mainly two types of preverbal phrases can be distinguished here: parts of idioms and narrowly focused expressions.

260

Svetlana Petrova

In the first group, the preverbal constituent forms a complex semantic unit with Vfin which immediately follows it, i.e. the pre-verbal phrase forms a kind of an idiom with the following verb. Consider thurft sîn ‘to need’ in (10a) as well as heim uuverban ‘to return home’ in (10b) in which the non-finite part provides the semantic core of the complex predicate: (10) a. uueiz íuuar fater/ uues íu thurft ist (T 67, 29) knows your Father/ what-GEN you-DAT need is ‘your Father knows the things you have need of’ lat. scit enim pater uester/ quibus opus sit uobis b. mit thiu sie heim uuvrbun (T 42, 17) when they home returend ‘when they returned home’ lat. cum redirent Non-verbal elements of complex semantic units cannot be pronominalized or referred back to by any means of anaphoric reference. In other words, parts of idioms fail to display properties which according to Karttunen (1976) are distinctive for referential expressions, i.e. they lack referential status and are not subject to the given/new-distinction. In modern German, non-verbal parts of complex predicates, like e.g. the NPs in Ball spielen ‘to play ball’, Schlange stehen ‘to stand in line’etc., are known to stay in a close relation to the verbal head not only with respect to semantics but to syntax as well. So in basic order, the nominal part has to be left adjacent to the verb: it is not subject to scrambling and does not allow insertion of adverbials or negation elements between itself and its verbal head (Pittner 1998), see (11a–b): (11) a. *dass die Kinder Ball oft/ nicht spielen that the children ball often/ NEG play b. dass die Kinder oft/ nicht Ball spielen that the children often/ NEG Ball play By contrast, ordinary arguments of verbs show no restrictions in this respect, see (12a–b): (12) a. dass die Kinder das Geld oft/ nicht ausgeben that the children the money often/ NEG spend b. dass die Kinder oft/ nicht das Geld ausgeben that the children often / NEG the money spend

Information structure in the Old High German Tatian

261

The incorporation of non-verbal parts of idioms to form a unitary whole with the verbal head is best represented in the case of separable verb prefixes which historically go back to directional or locative adverbials closely related with the verb (Di Meola 2000, 129). In the second group, the position before Vfin is occupied by material revealing properties of narrow, e.g. contrastive or operator-bound focus, though the focus operator is phonologically empty in most of the cases. Consider (13) in which the preverbal position in the embedded question hosts the DP thin zesuua ‘your right hand’ as an alternatives to the DP thin uuinistra ‘your left hand’ mentioned in the preceding main clause: (13) niuuizze íz thin uuinistra/ uuaz thin zesuua tuo (T 67, 4–5) NEG-know-SUBJ it your left hand what your right hand does ‘your left hand should not know what your right hand is doing’ lat. nesciat sinistra tua/ quid faciat dextra tua Furthermore, we find examples in which the preverbal phrase bears properties of exhaustiveness similar to those described for preverbal focus in Hungarian (Kiss 1998). The only contextually adequate reading of (14) is that Jesus spoke about no one else than about the Pharisees. As we can see, the constituent conveying exhaustiveness occupies the preverbal position, all remaining background material precedes: (14)

Inti pharisei […]/ furstuontun thaz her Iz fon In quad (T 204, 22) and the Pharisees […]/ understood that He it about them said ‘and the Parisees realized that he spoke of them [and of no one else]’ lat. & phairsei […]/ cognouerunt quod de ipsis dicer&

Similar effects are given in the relative clauses in (15a–b) which are uttered to exclude any alternative to the divine origin of Jesus. The ability to hear or speak the words of God is restricted to a referent with a special property only, namely to the one sent by God. In this interpretation, the pre-verbal phases fon gote ‘from God’ and got ‘God’ act as focus expressions bound by a phonologically empty focus operator triggering an exhaustive effect. In both cases, the focus expressions are shifted from the postverbal domain in the Latin text to the position immediately before the verb in the OHG translation:

262

Svetlana Petrova

(15) a. ther fon gote ist ther horit gotes uúort (T 219, 1) who from God-DAT is DEM hears God-GEN words ‘who came from God can hear God’s words’ lat. qui est ex deo uerba die audit b. then got santa ther sprihhit gotes uúort (T 57, 26) who-ACC God sent DEM speaks God-GEN words ‘who was sent by God may speak God’s words lat. quem enim missit deus uerba dei loquitur There are also examples in which exhaustive interpretation applies to adjuncts or modifiers, see êrist ‘as the first one’ in (16): (16) íogiuuelih gommanbarn/ thaz uuamba êrist Intuot (T 37, 16–17) every male child who womb first opens ‘every male who opens the womb [shall be called holy to the Lord]’ lat. omne masculum/ adaperiens uuluam The constituent êrist reveals exactly that part of the clause which is crucial to the proper understanding of the utterance: the presentation of a newborn child in the temple according to the Law of Moses applies to the firstborn son only, not to the others. Remarkably, êrist has no proper lexical equivalent in the Latin original but is included in the semantics of the present participle adaperiens ‘opening’. In the OHG sentence, the participle construction is transformed into a relative clause, and the semantics of the Latin participle is split into the focused modifier êrist and a finite verb Intuot ‘open-3sg.pres.ind’. The focus phrase is placed immediately before the finite verb while the object DP uuamba ‘the womb’, which is inferable in this context, is shifted to the position immediately after the conjunction. To sum up, the preverbal parts in the two kinds of patterns considered in (10)–(16), i.e. parts of idioms and narrowly focussed phrases share some important common features with respect to Behaghel’s classification. First, they both represent relevant information, the one with respect to the overall semantics of the complex predicate, and the other with respect to the context. According to this, both types of preverbal phrases can be assumed to carry main stress and consequently to meet the condition of prosodic heaviness. Main stress on the preverbal constituents in (11)–(16) is evident from their focal status. As far as non-finite parts of idioms are concerned, their prosodic behaviour in modern German is rather suggestive. As they carry word stress in the complex unit, they also take the functions of the focus exponent in the clause if there is no other argument suitable to carry

Information structure in the Old High German Tatian

263

main stress, e.g. because it is de-accented in anaphoric use. Exactly this condition applies in the sentences quoted above, where all the rest of the information in the sentences is anaphoric, i.e. de-accented. Therefore, informational relevance and accentuation are two features which link together narrow focus and nominal parts of idioms placed before Vfin.

3.1.3. Interim conclusion Taken together, the clauses with Vend order analysed above prompt the assumption that the domain between CP and Vfin in OHG is organized according to information-structural principles. First, two basic informationstructural domains can be identified here, one preserved for background material, and another one fixed for lexically or contextually relevant, i.e. focused material in the clause. Second, it is obvious that preverbal focus is subject to a further specification. From the examples viewed above, it is clear that it involves single phrases with a contrastive or operator-bound reading, and does not include wide focus or cases of focus projection. Furthermore, there is strong evidence that there are firm principles governing the serialisation of these information-structural domains in the clause: elements belonging to the background are associated with the position immediately following the subordinating conjunction or the relative pronoun while focused XPs has to be left-adjacent to Vfin. In the entire sample of clauses with Vend order, this principle is violated only once: (17) trisiuuet íu treso in himile/ thar noh rost noh miliuua íz nifurmelit (T 69, 15–16) deposit you-DAT treasure in Heaven where neither rust nor moth it NEG-destroys ‘But lay up for yourselves treasures in heaven, where neither rust nor moth destroys them’ lat. Thesaurizate autem uobis/ thesaurus in cello/ ubi neque erugo neque tinea demolitur In (17), the constituent placed left-adjacent to Vfin against the Latin original is the object pronoun iz ‘it-3sg.acc.neutr’. It takes up the previously metnioned antecedent treso ‘treasure’ and therefore represents given information, and the relevant context does not provide any indications leading to the contrastive interpretation of this pronoun. However, this remains

264

Svetlana Petrova

a single occurrence which cannot be taken as representative for the OHG situation in general.6 The results concerning the distribution of the information-structural domains in OHG gained from the analysis of clauses with Vend order may be therefore summarized as follows: (18) CP – [XP….]BGR … – [XP]FOC – Vfin Two relevant questions with respect to the anaylsis of clauses with nonVend order arise from these observations. First, do sentences displaying non-Vend order also show the same distribution of information-structural categories in the preverbal domain, and second, which positional distribution may be provided for wide, e.g. VP-focus in the OHG Tatian.

3.2. Clauses with non-Vend order 3.2.1. The placement of anaphoric material Within the group of non-Vend subordinate clauses, the proportion of cases deviating form the Latin only with respect to the transposition or insertion of the subject pronoun is higher than within the group of Vend clauses. Here, it applies to 52 of all 100 cases; a typical example is given in (19) where the subject pronoun is inserted in the OHG clause but the order of the remaining constituents adheres to that in the original: (19) thaz sie gihórtin gotes uuort (T 55, 2) that they heared-SUBJ God-GEN words ‘in order to hear God’s words’ lat. ut audierent uerbum dei The tendency to place pronominal elements in the domain immediately after the conjunction is confirmed by the syntactic behaviour of nonsubjects in further 19 cases, see the indirect object imo ‘him-3sg.dat.masc’ for lat. ei in (20): (20) só imo gibot truhtines engil (T 35, 2) as him-DAT commanded Lord-GEN angle ‘as the angel of the Lord commanded him’ lat. sicut precepit ei angelus domini

Information structure in the Old High German Tatian

265

There are only 5 cases in which a pronominal argument remains in postverbal position, each time in accordance with the Latin structure, see e.g. inan ‘him-3sg.masc.acc’ for lat. eum in (21):7 (21) thaz sie fiengin inan (T 119, 9) that they arrest-SUBJ Him ‘in order to arrest Him’ lat. ut raperent eum This evidence confirms the view that background elements have their usual position immediately below C°. However, the DP truhtines engil ‘God’s angel’ in (20), which is discourse-anaphoric as well, is not preposed into this domain. But in the group of non-Vend clauses, we nevertheless find 6 instances in which a full anaphoric DP is shifted from a postverbal position in the Latin to a preverbal position in OHG; the reverse transposition, i.e. to shift anaphoric material after the verb against the Latin original, does not occur. See (22) where the entity uueralt ‘the world’ established in the governing root clause is resumed in the following embedded clauses and placed both times before Vfin against the Latin order:8 (22) nisanta got sínan sun/ In uueralt thaz her uueralt tuome/ NEG-sent God His son/ to world that He world condemned-SUBJ uzouh thaz uuerolt si giheilit thuruh inan (T 197, 30–32) but that world is-SUBJ healed though Him ‘God didn’t send his son into the world to condemn the world, but that the world might be saved through Him’ lat. non enim missit deus filium suum/ In mundum ut l[sic!]udic& mundum/ sed ut salute&ur mundus per ipsum In the database, we find examples which provide an interesting minimal pair given with respect to the placement of given vs. new material in OHG. In (23), the DP thin elimosina ‘your charity’ which is mentioned for the first time in the discourse is retained in postverbal position in accord with the original. But in (24), where it represents given material, the same phrase is shifted to the preverbal domain: (23) thanne thú tuos elimosinam (T 66, 29) then you do charity ‘when you give charity’ lat. Cum ergo facies elimosinam

266

Svetlana Petrova

(24) thaz thin elimosina sí in tougalnesse (T 67, 6) that you charity is-SUBJ in secret ‘in order that your charity be done in secret’ lat. ut sit elimosina tua in abscondito Thus, we can conclude that anaphoric material, be it light or heavy lexical material, is placed as a rule adjacent to C°, while material that represents new information is placed after Vfin. The question is whether this is an accidental distribution resulting from the shift of anaphoric material before the verb, or part of a regular tendency applying in OHG independently of the Latin. In order to check this, we shall turn to the analysis of examples in which the OHG sentence contains postverbal material in contrast to the Latin original.

3.2.2. Properties of postverbal constituents in OHG The following examples contain postverbal material only in the OHG version of the text. First, we shall look at instances in which preverbal material of the Latin structure is realized postverbally in OHG. In our data base, this occurs 7 times. In 3 of the examples, the shifted material is a predicative adjective or participle in a copular construction; see (25a)9. In the remaining 4 cases, the shifted constituent is the direct object of the finite verb, see (25b)10. In all cases, the postverbal information is new and therefore part of the domain of new-information (i.e. presentational) focus: (25) a. giueh& uúarlihho/ thaz íuuere namon sint giscribane/ in himile (T 103, 26–28) be happy PRT/ that your names are written-PL in Heaven ‘Be happy that your names are written in heaven’ lat. gaud&e autem/ quod nomina uestra scripta sunt/ in caelis b. thaz in mir habet sibba (T 290, 8) that in Me have peace ‘that in Me you may have peace’ lat. ut In me pacem habeatis Furthermore, there is a group of sentences which are formed independently of the original to translate a nominal group or a participial construction of the Latin text. As the placement of the verb relative to the remaining con-

Information structure in the Old High German Tatian

267

stituents is not influenced by the original, we can assume that the linear order attested here obeys the principles of native OHG syntax. Among these rather valuable sentences, we find 7 instances displaying postverbal material only in the OHG text. Two of them are given in (26):11 (26) a. soso thie lihhazara sint gitruobte (T 68, 23) as the hypoctires are sad-PL ‘like the hypocrites are with a sad countenance’ lat. sicut hypocrite tristes b. thes namo uuas giheizzan simeon (T 37, 24) whose name was called Simeon ‘whose name was calles Simeon’ lat. cui nomen simeon All cases involve copular constructions in which the finite copula (sîn ‘be’, heizzan ‘be called’) precedes the nominal part of the predicate. However, exactly the latter carries the new or relevant lexical information in the particular context while the copula is semantically empty and only represents grammatical features like tense and agreement. Quite interestingly, ‘to be’ and ‘to call’ make up the majority of the cases in the group of examples in (25). Following this, we can assume that predicative adjectives and nouns as parts of copular constructions form a stable class of constituents being generally realized in the postverbal domain in OHG in evident contrast to the Latin structure. This is further supported by the fact that the postverbal realization of the predicative part of copular constructions is also typical for Old English (27) as well as for late-OHG texts, see (28) from the Physiologus (mid-11th century). At the same time, postverbal placement of nominal parts of predicates is ungrammatical in modern German subordinate structures of any kind: (27) On hiera dagum Hengest 7 Horsa […] gesohton Bretene on þam in their days Henges and Horsa […] sought Britain on that staþe þe is genemned Ypwinesfleot (ASChr 449) shore that is called Ebbsfleet ‘In their days [in the days of the reign of Mauritius and Valentiunus] Hengest and Horsa arrived in Britain on the shore which is called Ebbsfleet’

268

Svetlana Petrova

(28) dinen schephare, der dir ist ganemmet oriens (Ph 142)12 ‘your creator who is called Oriens’ These observations prompt the assumption that at the right periphery of the sentence in early Germanic, next to the very well known position of extraposed heavy constituents (PPs, heavy NPs and CP-complements), there was also a position occupied by close arguments of the verb providing new information in the discourse. This assumption shall be elaborated further in two more steps. First, constituents maintained in the postverbal position already given in the Latin structure shall be examined with respect to novelty and focus, and then, postverbal focus shall be compared with preverbal focus argued for in section 3.1.2 above.

3.2.3. The nature of postverbal focus We shall look back at the examples in which apart from the placement of background material, no further syntactic differences between the Latin and OHG structure occur. The task will be to find clues for the retention of postverbal material related to novelty or focus on the constituents. Quite certainly, this cannot be assumed for all examples of the kind. See e.g. the PP in (29) which clearly provides given information, as at that particular point in the story, it is known that Zacharias is still in the temple. However, the postverbal PP in (30) conveys a new direction and thus is associated with new-information focus: (29) Inti uuvntorotun thaz her lazz&a in templo (T 27, 23) and marveled that he stayed in [the] temple ‘and marveled that he lingered so long in the temple’ lat. & mirabantur quod tardar& ipse in templo (30) mitthiu her quam ubar thén giozon / in lantscaf gerasenorum (T 86, 31-32) when he came over the river to [the] country Gergesenes ‘when he had come over the river, to the country of the Gergesenes’ lat. Et cum ueniss& trans fr&um/ In regione geraseorum While examples of the type in (29) will not be interpreted further but left aside as equivalents of today’s cases of PP-extraposition, those like in (30) may be associated with focus in OHG. Among the non-Vend sentences,

Information structure in the Old High German Tatian

269

there are 32 cases in which the phrase retained in postverbal position represents new information. As the examples show, there is no restriction as to the grammatical type of the postverbal phrase: it can be a participle (31a), a PP (31b) as well as an object (31c): (31) a. Quamun thô thie firnfollon man/ thaz sie uuvrdin gitoufit (T 46, 24-25) came PRT the sinful men/ that they became-SUBJ baptized ‘The sinful men also came in order to be baptized’ lat. Uenerunt autem & publicani/ ut baptizarentur b. mit thiu her tho arsteig in skef (T 88, 23) when He PRT went into [a] boat ‘when He got into a boat’ lat. Cumque ascender& nauem c. thaz her giuuente herzun fatero In kind (T 27, 23) that He turned-SUBJ [the] hearts of fathers towards [the] children ‘that He turned the hearts of the fathers towards their children’ lat. ut conuertat corda partium In filios Looking at these examples more carefully, we discover that the new information is not only provided by the postverbal constituents alone but rather comprises the entire VP. In (31a), it covers not only the participle but also the finite verb used to translate the Latin synthetic passive. In (31b), the focus domain includes the verb as referring to a new action together with the discourse-new directional phrase in scef ‘into a boat’. The same applies to the VP in (31c) which assigns new information to the discouse-given referent John the Baptist. In a further group of examples, wide VP-focus spreads over the Vfin and an argument that is being re-activated at that particular point in the discourse. In (32a), a previous event, namely the birth of Christ, is now being related to the king, therefore, the new information is that King Herod also heard about the birth of Christ. In (32b), respectively, the previous action, namely the unexpected catch of fish, is now presented from the point to view of one participant of a group of referents established before. What is achieved here is a kind of change of perspective, or a topic shift with respect to the continuation of the narration. So in these cases, the verb provides together with the postverbal constituent the new-information focus domain, while background information, e.g. the preceding action resumed in the anaphor thaz ‘this’ is placed preverbally:

270

Svetlana Petrova

(32) a. thô thaz gihorta herodes ther cuning (T 39, 17) when that heard Herod the kind ‘when Herod the king heard this’ lat. audiens autem herodes rex b. mit thiu thaz tho gisah simon petrus (T 55, 29) when this PRT saw Simon Peter ‘when Simon Peter saw this’ lat. Quod cum uider& simon petrus These examples confirm the view that apart from cases of PP-extraposition (see (29) above), which is common in modern German as well, the postverbal field in OHG can also host constituents carrying new information, or more precisely, being part of a wide new-information focus domain opened by the finite verb. Therefore, the right periphery of such clauses is structured according to the following scheme: (33)

CP – [XP…]BGR … – VP[Vfin…]FOC

But how does this statement relate to the existence of a preverbal focus position claimed above? The intuitive answer to this question relates to the idea that wide, i.e. new-information focus, and narrow, operator-bound focus, are realized in two distinct syntactic positions distinguished by the placement of the finite verb in the clause. The following section shall provide more empirical support in favour of this view. 3.2.4. Multiple foci in OHG Crucial evidence supporting the existence of two distinct positions for focus material in OHG comes from sentences with multiple foci. Consider (34)–(35) which contain both a discourse-given but contrastively focussed constituent and additional new information after the verb: (34) [b&onte nicur& filu sprehan/ sósó thie heidanon mán = ‘And when you pray, do not use vain repetitions as the heathen do’] sie uuanen thaz sie in iro filusprahhi / sín gihórte (T 67, 23–26) they think that they in their many words/ are-SUBJ heard ‘They think that they will be heard for their many words’ lat. orantes autem. nolite multum loqui/ sicut &hnici.’/ putant enim quia in multiloquio/ exaudiantur

Information structure in the Old High German Tatian

271

(35) [thanne thu fastes/ salbo thin houbit/ Inti thin annuzi thuah = ‘when you fast, anoint your head and wash your face’] thaz thu mannon nisís gisehan/ fastenti. úzouh thinemo fater (T 68, 29–32) that you men-DAT NEG-are-SUBJ seen/ fasting but your-DAT father‘ ‘so that you do not appear to men to be fasting but to your Father’ lat. tu autem cum ieiunas/ unge caput tuum/ & faciem tuam laua/ ne uideatis hominibus/ ieiunans. sed patri tuo Let us now look at the postitional distribution of the different focus types in these examples. Narrow focus on constituents left adjacent to Vfin is evident in both cases. In (34), the PP in iro filusprahhi ‘for their many words’ acts as a focus exponent bound by an empty focus operator yielding the interpretation that only many words guarantee the fulfilment of the prayers. In (35), narrow focus on mannon ‘men-dat.pl’ results from the fact that it forms a constrastive pair with the explicitly mentioned constituent thineno fater ‘your father’. In both cases, the structures may be viewed to be chosen deliberately by the scribe. So in (34), the equivalent of the Latin synthetic passive could also be constructed in the order ‘participle – Vfin’. Instead, the scribe opted for the reverse order, namely ‘Vfin – participle’, which allows to retain the narrowly focused material before Vfin and to place the new information after it. Similarly, the deponens lat. uideatis is dissolved into a periphrastic construction involving the order ‘Vfin – participle’, while the narrow focus is shifted across Vfin against its original position in the Latin sentence. The sentence in (35) is notable in one more respect. According to the standard view, full PPs like in iro filusprahhi are a typical candidate for extraposition in modern German. It is striking, however, that exactly when pragmatic conditions apply yielding a narrow focus on the PP, it is put in preverbal position like any other type of phrase acting as operator-bound focus in the utterance. It is interesting what causes this particular distribution of focus material in OHG. In order to arrive at a plausible explanation to this question, we shall consider two more examples from the OHG Tatian:

272

Svetlana Petrova

(36) nimág ther man Iouuiht intphahén/ noba imo íz gigeban uuerde fon himile (T 57, 6–7) NEG-can the man anything receive/ NEG-if him-DAT it given became-SUBJ from Heaven ‘a man can receive nothing unless it has been given to him from Heaven’ lat. Non potest homo quicquam accipere/ nisi ei fuerit datum a caelo (37) [thisu sprahih íu = ‘these things I have spoken to you’] thaz in mir habet sibba/ In theru uueralti habet ir thrucnessi (T 290, 7–9) that in Me-DAT have peace/ in the world have you tribulation ‘that in Me you may have peace; in the world, you will have tribulation.’ lat. Haec locutus sum uobis/ ut in me pacem habeatis/ In mundo presuram habebitis In each of these sentences, two different constituents receive focus interpretation. One of them is set in an explicit contrastive relation to another entity in the utterance. In (36) this is the participle gigeban ‘given’ which refers to the only way to obtain spiritual power, namely by being given it, not by acquiring it oneself. In (37) the PP in me ‘in me’ is set into contrast to the expression In therru uueralti ‘in this world’. Additionally, there is also material supplying new information to the context. In (36) this is the source of the spiritual power, namely Heaven, and in (37) it is the direct objects sibba ‘peace’ which is also contrasted to the expression thrucnessi ‘pressure’ in the following conjunct. In the Latin version, in both cases the different types of foci are placed on the same side of the verb, after the verb in (36) and before the verb in (37). In the OHG text, however, these two types of foci are spaced in such a way that the contrastive or narrowly focused information is immediately before Vfin while the new one follows it. This invites the assumption that Vfin in OHG is used to avoid the stacking of two different types of focus in one and the same structural domain in the sentence.13 This enables the distinction of the focus types and the disambiguation of focus interpretations in the sentence.

Information structure in the Old High German Tatian

273

4. Results: different positional distribution of focus types in OHG The present paper investigated the role of pragmatic factors for the explanation of syntactic variation in the right periphery of subordinate clauses in OHG. It started with the claim that the standard account on word order in subordinate clauses in the early Germanic languages cannot be maintained without any modifications for a number of reasons. If in accord with the previous literature, we assume that OHG displays a basic SOV order maintained in clauses introduced by an overt complementizer, while exceptions to Vend are due to the extraposition of PPs, heavy NPs or CP-complements as well as to verb raising and verb projection raising in verb clusters, we are in need for an explanation of postverbal material like single NP or nominal parts of copular constructions. Therefore, we approached variation in the right periphery from a different perspective which subscribes to the view that verb placement in early Germanic is a grammatical correlate of pragmatic, discourse-based principles. More precisely, our analysis shows that in early Germanic, there is a tight correlation between the informationstructural properties of sentence constituents and their realization with respect to Vfin. This approach was applied to a sample of 100 Vend and 100 non-Vend sentences from the OHG Tatian which deviate from the structure of their Latin counterparts. First, it was shown that background material regularly appears in the domain immediately below C°, while focus material is found in two different structural positions adjacent to Vfin in the clause. Second, it was possible to account for principles governing this kind of distribution of focus material in the clause. On the one hand, narrowly focused information, e.g. contrastive focus on a single XP as well as operator-bound focus, tends to be placed left adjacent to Vfin. On the other hand, the domain of wide, new-information focus is opened by Vfin while the remaining elements of the focus projection follow it. The different positional distribution of focus types gained crucial support by the realization of multiple foci in OHG which occupy distinct syntactic positions with respect to Vfin. The picture derived for OHG evokes clear parallels to the situation in Yiddish as described by Diesing (1997, 390-396). According to her, the different syntactic realization of object NPs with respect to the selecting main verb triggers three different types of semantic interpretation. As Hinterhölzl (2004, 154) observes, these interpretations correspond to different categories of information structure. Leftward movement of an NP outside the VP associates with definitness and specificity, i.e. with background, while postverbal (in-situ) placement yields an existential reading

274

Svetlana Petrova

of discourse-new indefinite NPs as instances of new-information (presentational) focus. Additionally, both definite and indefinite objects in the position left-adjacent to the verb gain a special, marked status only possible when contrastive or corrective emphasis is put on them, i.e. when they are contrastively focussed. From this Hinterhölzl (2004) concludes that OHG, similarly to Yiddish, establishes two different syntactic positions for contrastive vs. non-contrastive, i.e. presentational focus. One significant difference to the situation described for OHG, however, remains: the position of narrow (contrastive and operator-bound) focus is left adjacent not to the main selecting verb but to Vfin in the clause. This, in turn, fits with the observations of Sapp (2006) on verbal clusters in Early New High German. He reports that contrastive interpretation on the immediately preceding XP is among the most influential factors leading to orders in which the Vfin is placed before all non-finite verbs in verb clusters. The question is why these two focus positions were distinguished in the system of OHG. Two hypotheses can be put forward to explaining this issue. The first one relates to aspects of the prosodic realization of focus especially in cases of multiple foci: as focus is prototypically associated with main stress, the placement of the finite verb between two different types of focus was a means of avoiding a clash of two heavily stressed phrases in one and the same structural domain of the clause. This scenario, however, does not account for the regular association of narrow vs. wide focus with a special position in the clause. This feature is reflected in the second hypothesis claiming that the different positional realization of focus types allows the unambiguous interpretation of the pragmatic value of the constituents involved. It guarantees that preverbal focus is interpreted as narrow XP-focus only, excluding the option of focus projection. In this way, OHG avoids a phenomenon known as ‘focus ambiguities’ in modern German (as well as in a number of other non-related languages). It is wellknown that in modern German, main accent on the rightmost XP in basic order yields both VP- or XP-focus while in scrambled order, the rightmost surface constituent receives an unambiguous contrastive interpretation (Abraham 1992). By contrast, in the system reconstructed for OHG, phrases belonging to the domains of new-information focus surface in postverbal position while preverbal focus only triggeres the option of XP-focus with additional effects of contrast, emphasis, and exclusion of alternatives.

Information structure in the Old High German Tatian

275

Notes 1.

2. 3.

4.

5. 6.

Kemenade (1987) proposes a model according to which OE displays properties of an asymmetric SOV language like modern German or Dutch. She assumes a base-generated SOV order maintained in subordinate clauses introduced by an overt complementizer in the head of a functional projection CP. In root clauses, the empty position of the complementizer is filled by Vfin while an optional movement of another constituent to SpecCP yields V2 in the surface. A similar approach has been proposed by Lenerz (1984) for OHG as well. Some basic differences between OE and OHG consist in the obligatoriness of V-to-C movement in main clauses. For Old English (OE), residual V2 in clauses with syntactic operators in SpecCP (wh- and negation words, sentence adverbials like þa/þonne, or a silent imperative-mood operator) is assumed, while in all other contexts Vfin targets another projection below CP (van Kemenade 1987 and 1997, Eythórsson 1996). By contrast, fronting of Vfin to C in cases with non-operators in SpecC is said to apply regularly in OHG (Axel 2007 and in this volume). Basic methodological considerations in favour of this view were put forward by Dittmer and Dittmer (1998). Schlachter’s statistics (in this volume) comprising all dhazs-sentences counting as unambiguous cases of subordination in the OHG Isidor confirms this picture. As Schlachter also shows, Latin influence has to be definitely excluded as a factor leading to this situation in the OHG Isidor. Pronouns are also excluded form extraposition in modern SOV languages. However, all instnaces involving a postverbal pronoun in our database may be explained as imitations of the original, which also involves a postverbal pronominal object, see (i): (i) thaz sie úz uuvrphin sie (T 76, 2) that they PRT threw-SUBJ them ‘that they threw tham away’ lat. ut eicerent eos Additional instances with postverbal pronouns are found in T 50, 21, T 119, 9, T 122, 15 and T 220, 10. But note that Dittmer and Dittmer (1998) provide examples in which pronouns are placed after Vfin against the Latin original, see (ii): (ii) thiedar giotmotigot sih (T 195, 19) who-PRT humiliates ReflPr ‘who humiliates himself’ lat. qui se humiliate (Dittmer and Dittmer 1998, 148). Uhmann (1991, 200 and 217) provides similar facts for modern German, too. In 2 additional cases involving narrow focus on a single phrase, the pattern Background–XP-Focus–Vfin is obviously blocked by the line break. Consider (i) and (ii) where the phrases thin zesuuua ouga ‘your right eye’ and thin ze-

276

7. 8. 9. 10. 11. 12.

13.

Svetlana Petrova suuúa hant ‘your right hand’ represent a pair of alternatives and therefore receive an interpretation as narrow, contrastive focus: (i) oba thin zesuuua ouga / thih bisuihhe (T 63, 24–25) if your right eye/ you-ACC troubles ‘if your right eye causes you to sin’ lat. quodsi oculus tuus dexter/ scandalizet te (ii) Inti oba thin zesuuúa hant / thih bisuihhe (T 63, 31–32) and if your right hand/ you-ACC troubles ‘and if your right hand causes you to sin’ lat. & si dextra manus tua / scandalizat té Here, the pronoun thih ‘you-2sg.acc’ belonging to the background intervenes between the focus phrase and Vfin. However, it is clear that placing the pronoun in the Wackernagel position and above the focus phrase would violate the line principle. But see Note 4 above. The remaining instances are: T 30, 19–20, T 69, 22–24 and T 84, 10–11. The remaining instances are: T 151, 12, T 224, 4–6. The remaining instances are: T 46, 2–4, T 59, 1, T 89, 26–28. The remaining instances are: T 35, 14–16, T 59, 22, T 60, 3, T 60, 12, T 60, 14, T 60, 16. I owe this example to Richard Schrodt (University of Vienna) who discussed evidence for non-Vend orders in late OHG in a talk “An den Rändern des Satzes. Kommunikative Dynamik im Althochdeutschen” on 17th Feb 2006 at Humboldt University Berlin. In a similar way, Speyer (2008) argues that focus spacing in double-focus constructions is responsible for topicalisation in Old and Middle English.

Primary texts [ASChr]

[T]

The Anglo-Saxon Chronicle. A collaborative edition. Volume 3. MS A. ed. Janet M. Bately, 1986. A semi-diplomatic edition with introduction and indices. Cambridge: D.S. Brewer. Die lateinisch-althochdeutsche Tatianbilingue Stiftsbibliothek St. Gallen Cod. 56. hg. Masser, Achim. 1994. Göttingen: Vandenhoeck & Ruprecht.

Information structure in the Old High German Tatian

277

References Abraham, Werner 1992 Clausal focus versus discourse rhema in Geman: a programmatic view. In Language and cognition 2, ed. Dicky Gilbers and Sietze Looyenga, 1–18. Groningen: Universiteitsdrukkerij. Axel, Katrin 2007 Studies in Old High German Syntax: left sentence periphery, verb placement and verb second. Amsterdam/Philadelphia: John Benjamins Publishing Company. this vol. The verb-second property in Old High German: different ways of filling the prefield. In New Approaches to Word Order Variation and Change in the Germanic Languages, eds. Roland Hinterhölzl and Svetlana Petrova. Berlin: Mouton de Gruyter. Behaghel, Otto 1932 Deutsche Syntax. Eine geschichtliche Darstellung. Band IV. Wortstellung. Periodenbau. Heidelberg: Carl Winter. Diesing, Molly 1997 Yiddish VP order and the typology of object movement in Germanic. Natural Language and Linguistic Theory 15.2: 369–427. Di Meola, Claudio 2000 Die Grammatikalisierung der deutschen Präpositionen. Tübingen: Stauffenburg Verlag. Dittmer, Arne and Dittmer, Ernst 1998 Studien zur Wortstellung - Satzgliedstellung in der althochdeutschen Tatianübersetzung. Göttingen: Vandenhoeck & Ruprecht. Ebert, Robert Peter Historische Syntax des Deutschen. Stuttgart: J. B. Metzlersche Verlagsbuchhandlung. Eythórsson, Thórhallur 1996 Functional Categories, Cliticization, and Verb Movement in the Early Germanic Languages. In Studies in Comparative Germanic Syntax, eds. Höskuldur Thráinsson, Samuel David Epstein and Steve Peter, 109–139. Dordrecht/Boston/London: Kluwer Academic Publishers. Fuß, Eric and Trips, Carola 2002 Variation and change in Old and Middle English – on the validity of the Double Base Hypothesis. Journal of Comparative Germanic Linguistics 4: 171–224. Haegeman, Liliane, and Riemsdijk, Henk van 1986 Verb Projection Raising, Scope and the Typology of Rules Affecting Verbs. Linguistic Inquiry 17: 417–466.

278

Svetlana Petrova

Hinterhölzl, Roland 2004 Language Change versus Grammar Change: What diachronic data reveal about the distinction between core grammar and periphery. In Diachronic Clues to Synchronic Grammar, eds. Carola Trips and Eric Fuß, 131–160. Amsterdam/Philadelphia: John Benjamins Publishing Company. Karttunen, Lauri 1976 Discourse Referents. In Syntax and Semantics 7: Notes from the Linguistic Underground, ed. James McCawley, 363–385. New York/ San Francisco/London: Academic Press. Kemenade, Ans van 1987 Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris. 1997 V2 and embedded topicalization in Old and Middle English. In Parameters of morpho-syntactic change, eds. Ans van Kemenade and Nigel Vincent, 326–352. Cambridge: Cambridge University Press. Kiss, Katalin. E. 1998 Identificational focus versus information focus. Language. Journal of the Linguistic Society of America 74: 245–273. Lakoff, George 1976 Pronouns and Reference. In Syntax and Semantics 7: Notes from the Linguistic Underground, ed. James McCawley, 275–335. New York/ San Francisco/London: Academic Press. Lenerz, Jürgen 1984 Syntaktischer Wandel und Grammatiktheorie. Eine Untersuchung an Beispielen aus der Sprachgeschichte des Deutschen. Tübingen: Max Niemeyer. Masser, Achim 1997 Syntaxprobleme im althochdeutschen Tatian. In: Semantik der syntaktischen Beziehungen. Akten des Pariser Kolloquiums zur Erforschung des Althochdeutschen 1994, ed. Yvon Desportes, 123–140. Heidelberg: Carl Winter. Pintzuk, Susan 1991 Phrase structures in competition: variation and change in Old English word order. Doctoral dissertation, University of Pennsylvania. Pittner, Karin 1998 Radfahren vs. mit dem Rad fahren: Trennbare Verben und parallele syntaktische Strukturen In Zwischen Grammatik und Lexik, eds. Irmhild Barz und Günther Öhlschläger, 103–112. Tübingen: Max Niemeyer. Sapp, Christopher D. 2006 Verb Order in Subordinate Clauses. From Early New High German to Modern German. Doctoral Dissertation, Indiana University.

Information structure in the Old High German Tatian

279

Schallert, Oliver 2006 Hybride OV / VO-Systeme und syntaktischer Wandel zu OV und VO in den germanischen Sprachen. Diplomarbeit, Universität Salzburg. Schlachter, Eva 2004 Satzstruktur im Althochdeutschen. Eine Skizze zur Position des Verbs im Isidor-Traktat des 8. Jahrhunderts. In Beiträge zu Sprache & Sprachen 4. Vorträge der Bochumer Linguistik-Tage, eds. Karin Pittner et al. 179–188. München: LINCOM. this vol. Word Order Variation and Information Structure in Old High German. An Analysis of Subordinate dhazs-Clauses in Isidor. In New Approaches to Word Order Variation and Change in the Germanic Languages, eds. Roland Hinterhölzl and Svetlana Petrova. Berlin: Mouton de Gruyter. Speyer, Augustin 2008 Topicalization and Clash Avoidance. On the interaction of Prosody and Syntax in the History of English with a Few Glimpses at German. PhD Thesis, University of Pennsylvania. Tomaselli, Alessandra 1995 Cases of Verb Third in Old High German. In Clause Structure and Language Change, eds. Adrian Battye and Ian Roberts, 345–369. New York/Oxford: Oxford University Press. Uhmann, Susanne 1991 Fokusphonologie: eine Analyse deutscher Intonationskonturen im Rahmen der nicht-linearen Phonologie. Tübingen: Max Niemeyer. Valentin, Paul 1987 Zur Geschichte des deutschen Passivs. In Das Passiv im Deutschen, ed. Centre de Recherche en Linguistique Germanique (Nice), 3–15. Tübingen: Max Niemeyer. Weiß, Helmut 2006 Die rechte Peripherie im Althochdeutschen: Zur Verbstellung in dassSätzen. Ms. Wurmbrand, Susi 2004 West Gemanic Verb Clusters: The Empirical Domain. In Verb Clusters: A Study of Hungarian, German and Dutch, eds. Kathalin É. Kiss and Henk van Riemsdijk, 43–85. Amsterdam/Philadelphia: John Benjamins Publishing Company.

Verb placement and information structure in the OHG Gospel Harmony by Otfrid von Weissenburg Andreas Lötscher

Abstract This paper gives an overview on some central rules and factors that determine word order in the OHG Gospel Harmony by Otfrid von Weissenburg. An outline of the elementary principles for the verb position first shows that Otfrid has the same differentiation between main clauses and subordinate clauses as later stages of German. In addition, the rule of the early position of the finite verb and the late position of infinite verbal elements creates a “frame and field”-structure for sentences as is known for later periods of German. Otfrid, however, has more freedom in placing elements in front and after the respective elements at the left and the right border of the verbal frame structure. It seems useful to consider the V1-pattern as the basic order in main clauses; from this, V2-order and V-late order in main clauses can be explained by the influence of information-structural factors. In detail, Otfrid’s language seems to represent an intermediate stage of language development, showing a combination of older and newer patterns. On the whole, however, Otfrid’s language shows more regularity and consistency with regard to word order than is usually assumed.

1. Preliminaries 1.1. Purpose of this contribution The purpose of this paper is to investigate in more detail rules and principles that possibly govern word order in Otfrid von Weissenburg’s “Evangelienharmonie” (Gospel Harmony) and thereby to contribute to a better knowledge of the situation in Old High German in this respect. Within the limits of this paper, I will not deliver a full-fledged grammatical description in all details (which would presuppose a general grammatical descrip-

282

Andreas Lötscher

tion of the overall system), but restrict myself to collecting the basic observations that seem relevant to the above question.

1.2. The text Otfrid von Weissenburg’s “Evangelienharmonie” seems to be a prime source for the grammatical description of OHG. It is the most voluminous and an outstanding vernacular work in the OHG period before Notker. It was written in a period (a.d. 863–871), in which no other work of comparable importance and richness had been produced. In addition to its value as a corpus, it could be a valuable counterpart to Notker, who produced his work about 200 years later. In spite of such facts, Otfrid’s work is usually not taken into consideration with regard to questions of word order. The main reason is the fact that it is a poetical work, written in end-rhyming verses. It is usually assumed that authors of poetical works tend to stretch or neglect rules of word order, in order to satisfy the rigid rules of versification and rhyming. In Otfrid’s work, word order is apparently more variable than in later Old High German works such as those of Notker of St. Gallen. Such deviations of the normal use are explained as irrelevant because they are supposed to be caused by rhyme.1 In favour of Otfrid, we can assume that even if Otfrid does not apply the usual word order principles of the core grammar of his language his use is not irregular, but simply follows special rules, be it of an older tradition, be it of poetical language as a subsystem of the overall system. Moreover, to dismiss as irrelevant any empirical facts in Otfrid’s language that deviate from the use in other authors or from the rules that are expected amounts to a petitio principii: After all Otfrid has decided to chose the formulations he uses as a competent writer of his language, and it should be proved from other, independent facts that he not only deviates from the usage of other authors, but has violated basic rules of his own language.2 It has been noted, of course, that Otfrid often violates grammatical rules under “Reimzwang” (cf. Nemitz 1962). But these observations relate almost exclusively to morphosyntactic problems such as congruency. Above all, to be able to classify a particular formulation as a violation of a grammatical rule, we first have to know the grammatical rules as a whole. In historical texts, where the grammatical system has to be inferred from the regularity of the use, a particular formulation can be classified as a violation if it is a rare case, compared with the usual construction, if it cannot be

Verb placement and information structure in the Gospel Harmony by Otfrid

283

explained by specific additional grammatical factors, and if the violation can be attributed to external factors such as “Reimzwang” or problems of performance. None of these cautionary considerations have been made with respect to the word order in Otfrid. Thus, before we can classify a particular word order as ungrammatical, we have to show that it is an exceptional use that does not make grammatical sense. Be this as it may, as long as we do not analyse Otfrid’s language in more detail we cannot assess the justification of any judgment about Otfrid’s language. Thus, I will here take the position that an analysis of the language of Otfrid with respect to the word order might be fruitful and that anyway an unprejudiced description might give valuable insights into word order rules of OHG.

1.3. Some theoretical and methodological considerations One of the key questions in investigating word order is the relation among grammatical and information-structural conditions for word order, in the terminology of Hinterhölzl and Donhauser (2003: 174). If it seems reasonable to base an analysis on this distinction, the question has to be asked what the theoretical premises and implications of such a distinction could be. Any systematic distinction between grammatical rules and informationstructural rules, so it seems to me, presupposes rules on essentially different linguistic levels. Grammatical word order rules by definition function on a purely structural level, whereas information-structural ordering rules are based on pragmatic principles, and probably on cognitive principles of a more general nature. In the following, I will proceed on such assumptions. Whereas grammatical word order rules, which are guided by structural properties of a sentence, seem comparatively easy to identify and define, this is not so for pragmatic principles for the sequencing of constituents. In order to be able to formulate a hypothesis on the relation between pragmatic conditions and word order and to avoid a circular argumentation, we have to determine the pragmatic interpretation of a sentence independent of the grammatical structure. Notoriously, interpretations of the pragmatic properties of an utterance are not easy to make evident, as they imply complex procedures of text understanding, based on conjectures about alleged background knowledge and the purpose of an utterance.

284

Andreas Lötscher

Often, a three-layered approach to this problem is proposed, such as described in Petrova/Solf (in this volume). According to this approach, three different levels of information-structure are to be distinguished: – – –

Given – New Topic – Comment Focus – Background

There are many problems connected with these categories. First, there are definitional problems with all of these pairs of terms. There are many differing definitions, some of them rather vague, others overlapping. E.g. there is no agreement whether “given” should mean ‘shared knowledge’, ‘activated knowledge’, ‘recoverability’, ‘textual accessibility’. Moreover it is not clear what the logical or grammatical category of content the property of ‘given’ or ‘new’ has to attributed to: to an individual as a semantic entity, to an expression denoting an individual or to a constituent within a clause. The same is true for the other pairs (for a discussion of the problems see Petrova/Solf, in this volume). Generally, definitions mostly are made on the basis of rather simple patterns; as soon, as we procede to more complicated utterances in a complex text, the application of the definitions becomes difficult. Second, it is not quite clear how these dimensions correlate. Undoubtedly, givenness has a strong correlation to topicality. Typically, one can make an individual entity a topic only if this individual is known to both of the communication partners; a new individual most often is introduced by way of mentioning it as a comment. However, the reverse is not true: one can treat something as a comment even though it is given and has been introduced earlier in the context: In No, I saw YOUR FATHER yesterday, the expression your father almost certainly is given, but a comment at the same time. The same weak correlation exists between “background” and “givenness”: Only entities that are textually accessible can be mentioned as background, but such entities can be focussed under certain contextual conditions, e.g. if they are disputed with respect to an assertion. It depends on the choice of the topic and the comment of a sentence whether something appears to be part of the background of an utterance. Thus, the distinction between “focus” and “background” possibly is not an independent one, but is dependent on the topic-comment-structure of an utterance. Third, when faced whith a task like that of the present contribution, it is difficult to see how these three layers interact with respect to the surface structure of a sentence. The basic question to be asked here is: How is the

Verb placement and information structure in the Gospel Harmony by Otfrid

285

word order in the text of Otfrid to be explained by grammatical and/or information-structural factors. For information-structural factors, the question is one of correlating form with content. Word order is linear, i.e. it is one-dimensional. How is a three-dimensional content translated in a onedimensional form? Or how can we detect in a one-dimensional form a three-dimensional content? One important methodological corrolary of these considerations is that one should give definitions for the respective pairs strictly in terms of content and avoid taking recourse to surface structure features. Otherwise, form and content are not distinguishable; moreover, circular definitions and/or descriptions result. Focus usually is defined with respect to stress, topic as the first element in a clause, i.e. with respect to word order. In my view, this is not an appropriate approach. Word order and stress are the main formal indications of the information-structure of an utterance. But first, there is no guarantee that there exists a one-to-one correlation between formal means of expression and a particular pragmatic content. E.g. both comment and certain types of topics can bear strong accent; comments can be fronted under certain conditions in New High German. Second, languages can differ. Evidently, all languages do not function the same way in the translation of informations-structural content into surface structures. We do not want to know how a language fits our predefined correlations of form and content, but to find out the way content factors govern the variancy of forms in a particular language. Third, there are possibly more factors influencing word order than just information-strutural ones, and we must be prepared to find other factors and rules govering word order than we expected.3 For the purposes of the present paper, I will restrict here myself to introducing a few basic concepts that will be necessary for the following analyses, omitting further questions and details.4 I will base my analyses on the assumption that the topic-comment distinction is the basic one, and that the distinctions of givenness-newness and focus-background are corrolaries or conditions of it. My main assumption will be that the topic-commentstructure has to be explicated in pragmatic terms, more specifically in terms of the function of an utterance within a communicative context.5 The function of an assertion is to introduce new assumptions, change existing common assumptions or confirm existing common assumptions (Sperber and Wilson 1986: 121). An utterance cannot establish completely new knowledge, but has to relate to some piece of existing common knowledge. We proceed from common knowledge (“given” knowledge in the broadest sense) with the intention to change or confirm certain pieces of knowledge

286

Andreas Lötscher

that seem to us controversial or worth being disputed.6 Making something a “topic” means declaring that this particular part within a common assumption is that part about whose properties one perceives a dispute and wants to introduce new assumptions, change assumptions or confirm assumptions. The corresponding elements introducing new assumptions, changing or confirming the relevant part of the assumptions are called “comment”. I take that which is usually called “background” as that portion of common assumptions connected with the utterance and possibly verbalized that the speaker neither wants to topicalize nor to dispute, that is irrelevant to his communicative intentions, either because he agrees with his audience or because a possible disagreement is not important to him in the given situation. Topics are specially marked if new background assumptions are introduced, specified or a change to other topics takes place. In many situations, in the course of a communication, a topic, once established, can become part of the background. Topics and changes of common assumptions within a communicative process are the result of continuous processes of changing, narrowing down and contrasting pieces of “background” and assumptions disputed.7 This has important implications for the semantic and pragmatic nature of topics and focus: When starting a communication, only elements with a referential counterpart in a context may be disputed; in the course of a communicative process, however, it usually happens that only parts of utterances are discussed, among such that have no referential counterpart or have only grammatical status, e.g. verbal elements, adjectives (predicative, attributive), quantifiers.8 In the following example, in the utterance of B, GOETHE is narrowing down the topic of the discourse (assuming that everybody knows in the given context that among the books there is a volume by Goethe), UNTER is a correction of auf in the preceding sentence; the utterance thus is intended to change a common assumption (‘Where are the books’) with regard to a specific book. In the utterance of C, the “background” for which a change of assumptions will be attempted by the speaker C is narrowed down to UNTER dem Tisch as opposed to AUF dem Tisch (C thus indicates, e.g., that she does not want to dispute the books on the table); she changes the common assumptions about what is under the table in a very general way to ‘many other books’ (Bücher bearing no contrastive value to the accepted background assumptions).

Verb placement and information structure in the Gospel Harmony by Otfrid

(1)

287

A: Ich habe die Bücher auf den Tisch gelegt. B: Der GOETHE liegt aber UNTER dem Tisch C: UNTER dem Tisch sind noch VIELE ANDERE Bücher. ‘A: I have put the books on the table. B: But the book of Goethe is under the table. C: Under the table, there are a lot more other books.’

Whereas the overall problem under dispute may well be covered by the term “topic” in the sense of “expression about whose referent the sentence is” (although this might be only a prototypical case), for other elements that make up only such details of a proposition under dispute, such a definition seems rather odd. The methodological consequences are: We cannot dismiss an element as a topic because it has no referential status; and possibly there are several layers of topics of different status present in an utterance, such as topical elements taken over from previous discourse (“continued topic”) and topics newly introduced or topicalized again. How can we detect the informational structure of a historical written text whose grammar we want to find out? As I said earlier, directly arguing from word order is not suitable, because it begs the question. Stress is not accessible directly in historical written texts, either. In the case of Otfrid, though, we can interprete the accent structure of sentences from the given verse structure; moreover Otfrid has noted accents carefully in his manuscript, although he does not differentiate primary from secondary stress. Basically, we have to interpret the information structure by interpreting the communicative function of an utterance within its context. On the basis of the connectivity of the text (recurrence of expressions referring to entities) and morphosyntactic indications (articles etc.), we can make inferences about the givenness of an entity and consequently about its topicality. But in important respects, our interpretation of the function of an utterance relies on implicatures about its communicative relevance within a given context. Such implicatures are based on assumptions about what a reasonable text should be. Thus, in important respects, the attribution of a particular information-structural value to an expression in a given sentence is the result of a text interpretation based both on universal premises and individual expectations.

288

Andreas Lötscher

2. General observations on the position of verbal elements As mentioned above, the position of verbal elements in Otfrid’s work shows great variety. In a first step I will attempt to establish a general grouping of the different types of order on a more observational level in order to have a basis for more detailed analyses of the particular problems.9

2.1. The position of the finite verb It is important to differentiate between the position of verbs in main clauses and in subordinate clauses, although in the particular sentence the differences are not always so evident.

2.1.1. Main clauses In main clauses, the finite verb (V-fin), can be in the first position (V-1), in the second position (V-2), in a position later than V-2 (V-late) or even at the end of the clause (V-last). V-fin=V-1: (2) Spráchun sie tho zimo sár: \ "meistar, zéllen wir thir wár, wir woltun wízan in giwís, \ war thu émmizigen bíruwis." (II, 7, 17–18) ‘They told him at once: „Master, we tell you in truth we wanted to know surely where you live usually “’ V-fin=V-2: (3) In búachon ist nu fúntan: \ thaz wort theist mán wortan, iz ward héra in worolt fúns \ joh nu búit in úns; Wir sáhun sinaz ríchi \ joh sina gúallichi, thaz was scóni al so frám, \ so sélben gotes súne zam. (II, 2, 31–34) ‘In the books it is found: the word has become man, it was here in the world at his own will and lives among us now. We have seen his realm and his goodness, that was so good, as it befits to the son of god.’

Verb placement and information structure in the Gospel Harmony by Otfrid

289

V-fin=V-late: (4) óuh man thara ládota \ thie júngoron thier tho hábeta. (II, 8, 8) ‘They had also invited the disciples he then had’. V-fin=V-last: (5) Er fíngar sinan thénita, \ then júngoron sar tho zélita, joh sár in tho giságeta \ thia sálida in thar gáganta. (II, 7, 9–10) ‘He pointed with his finger, he told his disciples and explained them at once that salvation was approaching them.’

2.1.2. Subordinate clauses In subordinate clauses, the possibilities of the positions of the finite verb are more restricted, compared to that in main clauses: the usual position either is final (V-last) or later than second (“late position”, V-late): V-fin = V-last: (6) Sie lóbont inan hárto \ frónisgero wórto, joh thánkont es mit wórte \ Kriste themo wírte; Want ér unsih fréwita, \ then gúaton win uns spárota, ther fúrdir uns ni wénkit, \ joh géistlicho drénkit; Thaz únsih es gilúste, \ thera fréwida ni bréste, joh wír zi themo gúate \ io wesen fástmuate. (II, 10, 17–22) ‘They praise him with magnificent words and thank it to Christ the host with words, because he delighted us and saved the good wine for us. that in future does not falter and gives us a spiritual potion, so that we rejoice it and do not lack joy and always be eager to do the good.’ V-fin = V-late: (7) Maht lésan wio iz wúrti \ zi theru drúhtines gibúrti, thaz éngil mit giwúrtin \ iz kundta sar then hírtin; Joh theiz ni wás ouh bóralang \ thaz hériskaf mit ímo sang; wio éngilo ménigi \ fúar thar al ingégini. (II, 3, 11–14) ‘You can read how it happened with the birth of the Lord that the angel with joy announced it to the shepherds: and it was within a short time that the hosts sang with him how the crowd of angels approached them.’

290

Andreas Lötscher

In this classification, some types of positions in main clauses and subordinate clauses are grouped to a simple “V-late”-position. We could differentiate in more detail between, say, Verb-third, Verb-fourth for main clauses and V-late (or “Verb-second-to-last” etc.) for subordinate clauses, as does e.g. Wunder (1965: 482–483). In most cases, such a differentiation does not yield much information about the underlying grammatical facts. On the contrary, on a purely observational level, a more differentiated description suggests more of a grammatical description than can be accounted for. In a clause with, say, four elements and the finite verb placed in the middle of this sequence, it is rather arbitrary or it even begs the question to classify this sequence either as a V3-structure or a V-late-structure.

2.2. The relative position of finite and infinite verbal elements Things become more transparent if we consider the position of nonfinite verbal elements relative to that of the finite verb. Infinite elements appear in Otfrid in the perfect tense, (participle + auxiliary), passive or passivelike constructions (participle + sîn/werdan), modal constructions (infinitive + modal verb) and other infinitive constructions, (e.g. infinitive + biginnan).

2.2.1. Main clauses For the relative positions of a finite verb and a nonfinite element, we have to differentiate between main clauses and subordinate clauses. In main clauses the finite verb always precedes the nonfinite element (apart of some special cases to be discussed later). The position of the nonfinite verbal element relative to other nonverbal elements again is variable: it can be in the ultimate, penultimate or antepenultimate position, and there can be one or more constituents between the finite and the nonfinite verbal element. This means, among other things, that e.g. a direct object of a nonfinite element can be placed before or after this nonfinite element. Taken together, this yields quite a number of combinations of different distributions of the two verbal elements in a clause. Nevertheless, the principle that the nonfinite element has to follow the finite one implies that the nonfinite element statistically is very often in final position.

Verb placement and information structure in the Gospel Harmony by Otfrid

291

V-fin = V-1, V-infin = V-last: (8) wólt er sar mit wíllen \ thaz sin gibót irfullen. (II, 9, 42) ‘He eagerly wanted to fulfil his commandment.’ V-fin = V1, V-infin = V-late: (9) Liaz ínan waltan álles \ thes wúnnisamen féldes; (II, 6, 11) ‘He let him cultivate all the beautiful land.’ V-fin = V-2, V-infin = V-last: (10) Ni wolt ér fon níawihti \ (thoh er so dúan mohti, ob ér thes wolti thénken) \ then selbon wín wirken; (II, 10, 1–2) ‘He did not want at all (although he could have done so, if he had intended) to create this wine (i.e. from the water).’ V-fin = V-2, Vinfin = V-late: (11) Wir scúlun uns zi gúate \ nu kéren thaz zi múate, mit wiu ther díufal so frám \ bisueih then ériston man; Wir sculun dráhton bi tház, \ thaz wir giwárten uns thiu báz. (II,5, 1-3) ‘We shall to our benefit pay attention to that, how the devil deceived very much the first human being; we shall strive thereby to beware of that.’

2.2.2. Subordinate clauses In subordinate clauses both the position of the whole group of verbal elements and the position of the finite element with respect to the infinite element differ from that of main clauses. First of all, in the overwhelming majority of the instances (e.g. in book II, in 90 % of all cases, i.e. in 28 of 31 instances), one verbal element or the whole group has clause final position. As a rule, the whole group is clause final: (12) Scúld bilaz uns állen, \ so wír ouh duan wóllen, (II, 20, 35) ‘Release all of us from our debts, as we will do as well.’ In a few cases the verbal group is followed by another element: (13) Ther evangélio thar quit, \ theiz móhti wesan séxta zit; (II, 14, 9) ‘The gospel says that it could have been the sixth hour.’

292

Andreas Lötscher

In most cases, the two verbal elements follow each other directly. It is possible, however, that they are separated by another nonverbal element: (14) Er tháhta odowila tház, \ thaz er ther dúriwart wás, er íngang therera wórolti \ bisperrit sélbo habeti; (II, 4, 8f.) ‘He thought perhaps that he was the doorman, that he himself had locked the entry to this world.’ As an important difference to main clauses, in subordination clauses, the relative positions of finite and nonfinite elements is variable and does not seem to be fixed by a grammatical principle. The sequence V-non-fin – V-fin is used as often as the reverse sequence V-fin – V-non-fin. The order of verbal elements does not seem to be influenced by their grammatical nature; whether the finite verb is an auxiliary or a modal verb does not influence the frequency of the order. (15) a. Ni wolt ér fon níawihti \ (thoh er so dúan mohti, ob ér thes wolti thénken) \ then selbon wín wirken; (II, 10, 1–2) ‘He did not want at all (although he could have done so, if he had intended) to create this wine (i.e. from the water).’ b. Bigínnu ih hiar nu rédinon, \ wio ér bigonda brédigon joh méistera ther uns ónda, \ sámanon bigónda. (II, 7, 1f.) ‘I begin here to tell, how he began to preach and to gather the teachers he gave us.’ An exception from this principle is the well-known case of conditional clauses, in which the finite verb has clause initial position, as in all German varieties up to present day Standard German: (16) Wil thu iz kléinor reken, \ in wíne gisméken: fon Kríste scalt thu iz zéllen, \ gisteist thu tház irwéllen. (II, 9, 69–70) ‘If you want to interpret it more subtly, taste it as wine, you have to relate it to Christ, if you want to do so.’

2.3. Conclusions From this, we may conclude that the distinction between main clauses and subordinate clauses by means of verb placement is a fundamental gram-

Verb placement and information structure in the Gospel Harmony by Otfrid

293

matical principle in Otfrid. In this respect Otfrid follows the same fundamental rules that are characteristic for all later German variants, with some differences in detail: –

– –

In main clauses the finite verb has to precede a nonfinite element obligatorily; the finite verb is in an “early” position, the infinite element in “late” position. This creates a “frame-and-field structure” comparable to that of all later historical variants of German. In subordinate clauses both elements of the verb group are in “late” or clause final position, but without a strict ordering among them. Otfrid differs from later stages of German such as Notker (cf. e.g. Näf 1979) or Middle High German insofar as in main clauses, the finite verb often appears in the first as well as in a late or the last position.

All this implies that the verb position in Otfrid at least partially is governed by grammatical rules. However, it is not possible to give rules for an absolute position of the respective verbal elements, such as “verb second” or “verb final” position. To speak of “verb first”, “verb second”, “late position” is only a superficial observational characterization that cannot be understood as a grammatical description. In many respects this situation is comparable to the situation in Middle High German. The basic approach to describe the greater freedom of word order in Otfrid, as in Middel High German, to describe the variable position of non-verbal elements with regard to verbal elements as the result of the influence of pragmatic factors such as information-structural conditions, among others.

3. Verb-last position in asyndetic subordinate clauses In Otfrid, clauses are not infrequent that seem to be subordinate from their apparent logical or grammatical connection with the context and or from the mood (subjunctive), although they are only asyndetically linked to it. In (17a), this sentence is a complement to haben, in (17b), the clause is an attributive clause to salida, the relevant clauses in (17c) and (17d) are purpose clauses:

294

Andreas Lötscher

(17) a. Thaz kínd thaz druag thaz wítu mit, \ joh er iz hábeta furi niwíht, er fon thes fáter henti \ tho thar dót wurti. (II, 9, 43–44) ‘The child carried the wood with him, and he considered it as nothing, that he should die from the hands of his father.’ b. Tho uns ward thiu sálida so frám, \ er sélbo in thesa wórolt quam, thaz thiu sin géginwerti \ zi sálidon uns wúrti: (II, 10, 7–8) ‘When the bliss was given to us, that he himself came into this world, so that his presence became our salvation …’ c. Iz ist giscríban fona thír, \ thaz faren éngila mit thír, sie thih biscírmen állan \ joh thíh ni lazen fállan; sie thín giwaro wárten \ jóh thih harto hálten, thaz thin fúaz iowánne \ in stéine ni firspúrne. (II, 4, 57–60) ‘It is written of you that angels escort you that they protect you entirely and do not let you fall. So that they keep you in safety and hold you fast that your feet never hit at a stone.’ d. Bi thiu ílemes io gigáhon \ zi then drúhtines ginádon, er unse wéga irwente \ fon themo fíante; Er únsih ni bisoufe \ áfter themo dóufe. (II, 3, 63–65) ‘Because of this we hurry to come to the grace of the Lord that he turn our ways away from the foe, so that he does not drown us after the baptism.’ In our context, the main relevance of such constructions lays in the fact that the subordination is marked almost exclusively by the late or final position of the verb; this can be interpreted as one more indication that the verb-last position in fact is a grammatical feature of subordinate clauses.

4. Wackernagel position and field structure As a consequence of the variability of the verb positions, in many cases it is not decidable from the superficial word order what the appropriate grammatical description of the verb position or the position of non-verbal elements relative to a verbal element should be. In the following examples, there are at least as many constituents following the verb as are preceding it; sometimes the verb even occupies the position just after the conjunction,

Verb placement and information structure in the Gospel Harmony by Otfrid

295

although, within the given grammatical context, we would classify the clause as a subordinate clause and thus predict a final or at least a late position of the verb: (18) a. Ni mag thaz mán duan nihéin, \ thaz thiono héreren zuein, thaz er irfúlle io follon \ bédero willon. (II, 22, 1–2) ‘Nobody is able to achieve it that he serves two masters, that he fulfils the will of both of them.’ b. Firnim in álawari, \ thaz got ther fáter wari, joh thaz kínd eino \ Krístan bizéino; Then er zi tóde salta \ bi únsih, sos er wólta, noh themo éinigen ni leip, \ io so Páulus giscréip; Wio er sélbo druag thaz krúzi, \ tho er thúlta thaz wízi, joh irstarp tháre \ in thes cruces áltare; (II, 9, 75–80) ‘Learn in truth that god was the father and the only child means Christ, whom he gave away for death for us, as he wanted to, and did not spare it, as Paul has written, how he himself carried the cross, as he suffered the punishment, and there died on the altar of the cross.’ Such examples demonstrate that a purely numerical classification of the verb positions is not a good grammatical description; such a description has to be of a more abstract nature. A passage such as thaz thiono héreren zuein (II, 22, 1), showing a null-subject construction, can superficially be either characterized as a V-1 or as a V-late pattern, or, assuming a late deletion of a subject pronoun, even as an underlying V-2 construction. The appropriate description can only be given by correlating the surface order to a corresponding grammatical derivation, accounting for the whole range of variations. One indication that there is really a field structure is the position of enclitic weakly stressed elements, mostly pronouns. In New High German their regular position is at the left side of the middle field or, in other words, directly at the left bracket of the middle field, the so-called „Wackernagel position”. In main clauses, this left bracket is constituted by the finite verb in V-2-postion, in subordinate clauses by the conjunction or the relative pronoun. The differences between main clauses and subordinate clauses with regard to the position of the finite verb create differences in the position of enclitic elements: in main clauses they follow the finite verb, in subordi-

296

Andreas Lötscher

nate clauses they precede it. A pronoun with the function of an accusative object thus precedes the subject: (19) a. Vielleicht ADV

hat Vfin

dich dein Bruder angerufen. OBJakk SUBJ Vinf ↑ Wackernagel position ‘Perhaps your brother has called you up’. b. dass dich vielleicht dein Bruder angerufen hat. CONJ OBJakk ADV SUBJ Vinf Vfin ↑ Wackernagel position ‘that perhaps your brother has called you up’

These regularities apply for Otfrid’s text as well. Weakly stressed pronominal and other weak elements regularly have their enclitic place in the Wackernagel position. This is clearly borne out in the above examples, cf. e.g. (15a), where pronominal elements follow or precede verbal elements, depending from the type of clause: (15) a. Ni wolt ér fon níawihti \ (thoh er so dúan mohti, ob ér thes wolti thénken) \ then selbon wín wirken; (II, 10, 1–2) ‘He did not want at all (although he could have done so, if he had intended) to create this wine (i.e. from the water).’ The principle of the Wackernagel position for weak elements, incidentally, applies also to V-1-structures: (20) a. Gab er mo ántwurti \ mit súazeru giwúrti (II, 7, 57) ‘He answered him with sweet joy.’ b. Fúart er sar tho thárasun \ then sélbon sinan drútsun; wólt er sar mit wíllen \ thaz sin gibót irfullen. (II, 9, 41–42) ‘He led there his dear son; he wanted wilfully fulfil his order. ’ These observations have several implications. First, the position of enclitic elements indicates that in subordinate clauses in fact verb late position applies. Second, this can be taken as one more indication that both for V-1structures and V-2-structures Otfrid uses a field structure comparable to the field model as we know it from later German periods. A sensible generalization will be that this is the result of a general grammatical rule applicable

Verb placement and information structure in the Gospel Harmony by Otfrid

297

to all clauses; in other words, we have to interprete infinite verbal element as being in verb late position even when there no other element between the finite and the infinite verbal element and even when there is possibly another element following the infinite element, such as in (18a) (thaz thiono héreren zuein). On the basis of these generalizations, we can decide for particular clauses whether it is a main clause or a subordinate clause. When a weakly stressed pronominal element precedes a finite verb as in (18a), we have an instance of a V-late position, even though the finite verb may have the second position in a clause and is followed by other nominal elements. This may be especially relevant for clauses introduced by wanta, which are notoriously ambiguous between main clauses and subordinate clauses; e.g., in the following passage the wanta-clause has to be interpreted as a subordinate clause: (20) c. Sálig sint zi gúate \ thie rózegemo múate, wanta in firtílot thaz sér \ dróst filu mánager; (II, 16, 9–10) ‘Blessed are those with a sorrowful mood, because much consolation removes their sorrow.’

5. Verb-late and verb-final position As a result of the above considerations, the verb in late or final position of any type of clause has to be interpreted as the right bracket of the middle field. Usually, in situations as this, the final position is taken to be the basic position of the verb (finite or infinite), and the position of other elements right of the verb is attributed to extraposition. Apparently, this rule of extraposition works differently in the different stages of German. As to Otfrid, postposition relative to a “late” verb is only possible for certain types of constituents: Complements to a verb only occur in postverbal position if they are “heavy” constituents in a double sense: they are composed of fully lexical material and they carry strong accent, being comment. Otherwise these constituents have to be in the middle field (or Wackernagel position): (21) a. Ther evangélio thar quit, \ theiz móhti wesan séxta zit; (II, 14, 9) ‘The Gospel says that it might have been the sixth hour.’

298

Andreas Lötscher

b. Sie kúndtun thar then líutin, \ thóh si es tho ni rúahtin, thaz ín was queman hérasun \ ther gotes éinigo sun. (II, 3, 25–26) ‘They announced to the people, although they did not realize it at that time, that the only son of God had arrived.’ c. Nu léru ih íuih hárto \ kúrzero wórto, wio ír giduet fóllon \ then drúhtines wíllon. (II, 23, 1–2) ‘Now I teach you in very short words, how you fulfil completely the will of God.’ This restriction does not apply to adverbial constituents: They can follow a final verb even when they are weakly stressed short constituents (cf. also the position of hérasun in example (21b) above): (22) a. Ih mag iz wóla midan, \ mag hiar nídarstigan; ziu scal ih íowanne \ gótes koron thánne. (II, 4, 77–78) ‘I can easily avoid this and stay down here, why should I tempt God then ?’ b. Iz ist giscríban fona thír, \ thaz faren éngila mit thír. (II, 4, 57) ‘It has been written of you that your are escorted by angels’ A particularly interesting case is (22b) in our context. Two phrases are following the finite verb faren, which has to be assumed in “late position” from our former generalizations. That éngila is following the verb, can be accounted for by the extraposition possibility just mentioned: it can safely be assumed that éngila has the IS-function of a comment, being the relevant new information within the given context. Furthermore, it can be assumed with some certainty that mit thír is a weakly accentuated verbal phrase. Thus, an weakly accentuated adverbial element follows an heavily accentuated complement, being a comment, when both are extraposed. Whether this is a general rule or just a possibility remains to be investigated more closely and I leave it open as a question. We can describe this situation in the following grammatical terms: –

–

We can assume that verb final position is the basic position for the respective cases (nonfinite verbal elements in main clauses, both finite and nonfinite verbal elements in subordinate clauses). This verbal final rule can be overruled by extraposition of complements and adverbial elements, with clear differences among the different types of elements:

Verb placement and information structure in the Gospel Harmony by Otfrid

−

−

299

Extraposition of complements can apply under restricted pragmatic conditions. Only a complement with heavy accent being a comment can be extraposed. The general semiotic principle underlying this rule is the tendency for comments to be placed at the end of a clause. Other constituents such as adverbs are free in their position with regard to the clause final verb. Without going further in details we can hypothesize that pragmatic factors rule the position of such elements, too. As far as I can see, extraposed adverbial elements tend to mention contextual parameters already fixed within the context such as time, place or addressee etc.; in other words, they are “background elements”. Of course, other major factors in Otfrid’s case are rhyming and rhythm. But we may assume that the extraposition of adverbial elements is of another nature as that of complements. Semiotically, it has more the character of “Nachtrag” (postscript), grammatically, the extraposition of adverbials places them in a position outside the clause.

6. Verb-first and verb-second position As mentioned earlier, Otfrid displays a much wider variability for the position of the finite verb in the main clause than later German periods. A central question is how this variability is to be accounted for. I will first try to give a description of the use of V-1 and V-2 positions.

6.1. Indications for a grammatical V-2-rule There are some indications that under certain grammatical conditions, a verb-second rule is valid: Some grammatical facts seem only to be explicable in a straightforward way if we assume a verb-second rule as a more general grammatical principle. The usual verb-second rule implies a defined prefield that can only be filled with one constituent. There are certain well-defined cases in Otfrid’s work that allow exactly one constituent in the prefield. If there were no such verb-second rule it should be possible that also in these cases other constituents accompany such constituents, which does not seem to be the case.

300

Andreas Lötscher

6.1.1. Question words One such case are question words in direct questions, which always take the place of the prefield and are not followed by any other constituent: (23) a. wio mág thaz sin firlóugnit, \ thaz hímil theru wórolti ougit? (II, 3, 20) ‘How can this be denied that heaven shows itself to this world?’ b. Waz mag ih zéllen thir ouh mér? \ ther púzz ist filu díofer; (II, 14, 29) ‘What can I tell you more? The well is very deep.’ Without a verb second rule it should be possible that other constituents follow such constituents, a situation that as far as I can see never arises.

6.1.2. Resumptive elements Another case of apparent V-2-structures are resumptive constructions of different types, either with noun phrases or with subordinate clauses, where a left dislocated structure is resumed by an anaphoric expression, which appears in the prefield position. The left dislocated structure must be classified as being outside the field structure (for New High German in the “linkes Außenfeld”, cf. Zifonun et al. 1999: 1577), grammatically outside the clause structure. The nominal type is usually known as “left dislocation” (“Linksversetzung”): (24) a. Unsere áltfordoron \ thie bétotun hiar in bérgon; (II, 14, 57) ‘Our ancestors prayed here in the mountains.’ b. Ther géist ther ist drúhtin \ mit fílu hohen máhtin; (II, 14, 71) ‘The holy ghost is the lord with very high power.’ Another type of such resumptive constructions are subordinate clauses resumed by an adverb or a resumptive pronoun: (25) a. Ther fon ther érdu hinana íst, \ ther scal spréchan thanana er íst. (II, 13, 19) ‘Whoever is from this earth, he shall speak from where he is.’ b. Wio íh iu hiar gibíete, \ thaz hóret ío zi gúate; (II, 19, 13) ‘How I give you commands, listen that to your benefit.’

Verb placement and information structure in the Gospel Harmony by Otfrid

301

Resumptive elements, like question words, are always alone in filling the position in front of the verb, which can again be interpreted as an argument in favour of a V-2-rule in Otfrid. Note, however, that a left dislocated element is not necessarily resumed by a resumptive element. If no such element occurs, the following clause can be organized differently, according to the general rules for sentences without precedent subordinate clause, either with the usual V-2-structure or even with a V-1-structure: (26) a. So sie tho thára quamun, \ thaz héimingi gisáhun, sie núzzun thera héimwisti \ then dág tho mit gilústi. (II, 7, 21–22) ‘When they came there and saw his home, they enjoyed the house that day with delight.’ b. Ob ér sih thoh biknáti, \ jáhi sos er dáti, zaltiz állaz ufan síh: \ ni wúrtiz alles so égislih; (II, 6, 43–44) ‘If he had repented it and admitted it when he had done it, and had taken it upon himself, it would not have become so terrible.’ c. Thie ínan thoh irkántun \ joh múates sih biwántun, giéret er se in thén sind, \ tház sie warin gótes kind. (II, 2, 27–28 ‘Those who recognized him and changed their mind, he honoured them so that they became the children of god.’ Thus, the argument in favour of a V2-rule holds only in case resumptive elements are used. Note by the way that the direct connection of a main clause with a verb after a preceding subordinate clause such as in (26b) and (26c) has to be interpreted as an instance of a V-1-structure and not as a V-2-structure with a subordinate clause filling the prefield slot (“integrated front position”). In later medieval German, subordinate clauses of any type are not allowed to fill the prefield position, as in New High German, but have to be preposed and resumed by a resumptive particle (cf. Lötscher 2005).

6.1.3. Obligatory verb-first constructions There are constructions that apparently demand verb first position obligatorily, namely direct questions and conditional clauses. Verb first position in these cases seems to be a grammatical rule. Such a rule can only have

302

Andreas Lötscher

structural sense if these constructions are opposed to grammatically defined verb second constructions. (27) Lási thu io thia rédina, wio drúhtin threwit thánana? (V, 19, 31) ‘Have you ever read the passage how the Lord appears threatening from there?’ Taken together, such observations seem to indicate that Otfrid obeys a grammatical rule of V-2 comparable to later variants of German.

6.2. Information-structural factors for the choice between verb-first and verb-second constructions Besides V-2-structures, we find V-1-sentences in declarative clauses as well, and they are fairly frequent, in fact. Thus, we cannot assume a V-2rule as an obligatory general principle. This leads to the question what factors could decide for a choice between V-1- and V-2-structures, especially for the cases of V-1- and V-2-structures that are not grammatically specifiable as those mentioned in the previous section.

6.2.1. Topics Hinterhölzl and Donhauser (2003) propose an information-structural account for the alternatives of V-1-sentences and V-2-sentences. They state for Old High German in general that V-2-sentence are used when a new topic is established or a topic is changed, whereas V-1-sentences are used for “all-focus” utterances (utterances without an overt topic) or sentences where an established topic is taken over from the context. Basically, this hypothesis is born out for Otfrid as well. This can be seen in a text sequence such as the following one: (28) 15 “Heil mágad zieri, \ thíarna so scóni, állero wíbo \ gote zéizosto! Ni brútti thih múates, \ noh thines ánluzzes fárawa ni wenti; \ fol bistu gótes ensti! Fórosagon súngun \ fon thir sáligun, 20 wárun se allo wórolti \ zi thir zéigonti. Gímma thiu wíza, \ magad scínenta,

Verb placement and information structure in the Gospel Harmony by Otfrid

25

30

15

20

25

30

303

múater thiu díura \ scált thu wesan éina. Thú scalt beran éinan \ alawáltendan érdun joh hímiles \ int alles líphaftes, Scépheri wórolti \ (theist min árunti), fátere gibóranan \ ebanéwigan. Got gíbit imo wíha \ joh éra filu hóha (drof ni zuívolo thu thés), \ Davídes sez thes kúninges. Er ríchisot githíuto \ kúning therero líuto; (thaz steit in gótes henti) \ ána theheinig énti. Állera wórolti \ ist er líb gebenti, tház er ouh inspérre \ hímilrichi mánne.” Thiu thíarna filu scóno \ sprah zi bóten frono, gab si imo ántwurti \ mit súazera giwurti: … (I, 5, 15–34) ‘ “Hail, beautiful maiden, young lady so beautiful, the dearest of god of all women! Do not be frightened and do not loose the colour of your face: you are full of god’s grace. The prophets have sung of you, the blessed one and they have prophesied you to all the centuries. Brilliant gem, shining girl, you shall be the illustrious mother, you alone. You shall bear the one who alone rules heaven and earth and all living beings. The creator of world, (this is my message), the son who is as eternal as the father. God gives him glory and high honour, (do not doubt about this), and the seat of David the king. He will rule gloriously as a king of the people, (this is in the hand of God), without ending. He will give life to all the world. so that he will unlock heaven for mankind. The beautiful girl spoke to the messenger of the Lord, answered him with sweet joy: …” ’

Passages that confirm the observations of Hinterhölzl and Donhauser (2003) are e.g. the verses 19 and 20, 28, 29 and 33–34: Forasagun in v. 19 is a newly introduced topic, that is resumed and continued by se in v. 20; in v. 19, the expression is placed in the prefield, in v. 20 enclitically after the verb (in the Wackernagel position), the prefield is left empty. A similar sequencing of topic introduction and continuation can be found in verse 33,

304

Andreas Lötscher

where thiu thiarna is introduced as a new topic, (implicitly opposed to the angel speaking, mentioned before in v. 15), which is again continued in pronominal form si in enclitic position in verse 34. Similarly in V. 28 Got is introduced as a new topic and posited in the prefield. A slightly different case is represented in v. 29, however, where the seemingly anaphoric pronoun er (as a continued topic, connected to the complement imu in the preceding sentence) appears in the prefield. On a closer look, er in v. 29 is not a continued topic, but it’s mentioning brings about a topic change from got to er (= the son of God) as well. As hypothesized by Hinterhölzl and Donhauser (2003) for Old High German in general, V-1-sentences occur in Otfrid’s work for statements without a constituent identifiable as a new topic in a textually relevant sense. (29a) and (29b) are characteristic in that a V-1-sentence is opening a text sequence with a new thought or a new textual topic, not having a indivual referent as a single topic, however, about which the sentence is a statement. II, 7, 1 in (29a) is a typical opening of a chapter; ih as the speaker of the sentence cannot be considered as a topic in a precise sense, rather, it is a background element. Similarly, in II, 7, 5, the sentence as a whole is a holistic desription of a situation where Johánnes is not treated as the topic about which the utterence is. II, 6 47 in (29b) is a kind of a gnomic statement; a generic ‚everybody’ without precise referent cannot be a real topic in such a statement either. (29) a. Bigínnu ih hiar nu rédinon, \ wio ér bigonda brédigon joh méistera ther uns ónda, \ sámanon bigónda; Mit zúhtin sier mo húldta \ joh wísduames irfúlta, sant er thíe tho in allahánt, \ so himil thékit thaz lant. Stuant Johánnes gomono éin \ mit sinen júngoron zuein, gisáh er gangan thárasun \ then selben drúhtines sun; (II, 7, 1–6) ‘I begin here to tell how he began to preach and to gather the teachers that he gave us. He won them for himself by teaching them and he filled them with wisdom. Johannes, one of the men, stood there with his disciples, he saw going there God’s son.’ b. Ságe mir nu, friunt mín, \ wio dati só bi then win, thih sus es nu inthábetos, \ so lángo nan gispáratos? Gíbit giwelih mánno, \ ther fríunta frewit gérno, (ih weiz, thu es ínnana bist) \ then fúriston io sar zi érist. (II, 6, 45–48)

Verb placement and information structure in the Gospel Harmony by Otfrid

305

‘Tell me, dear friend, why did you proceed with the wine in this way, that you have hold it back, that you have saved it for such a long time? Everybody who wants to please his friends (I know that this is well known to you) serves the best wine first.’ In other cases, referents are mentioned pronominally that have been introduced earlier in the context. They could perhaps be interpreted as “continued topic”. As a matter of fact, however, such elements cannot considered as textually relevant topics; they rather give background information: (30) a. Thiu muater hórta thaz tho thár; \ si wéssa thoh in álawar, thaz íru thiu sin gúati \ nirzígi thes siu báti. Gibót si then sar gáhun \ then thes lídes sahun, so wás so er in giquáti, \ iz íagiliher dati. (II, 6, 23–26) ‘His mother heared this, but she knew for sure. that his goodness would not refuse what she begged for. She then asked those who were responsable for the beverages to do whatever he would order.’ b. er kérta sih sar widar zín, \ quad: “gúate man, waz skel iz sín?” Spráchun sie tho zimo sár: \ “meistar, zéllen wir thir wár, wir woltun wízan in giwís, \ war thu émmizigen bíruwis.” (II, 7, 16–18) ‘He turned to them and said: “Good men, what do you want?” They anwered him: “Master, we tell you honestly, we wanted to know for sure, where you live everyday.” ’ These principles apparently are not in any case followed consistently. In some cases, pronominal elements resuming a topic referent of a previous clause (“established topics”) sometimes do not appear in the middlefield in the Wackernagel position, but in the prefield; it seems that this is restricted to subjects: (31) a. Thie ínan thoh irkántun \ joh múates sih biwántun – giéreta er se in thén sind, \ tház sie warin gótes kind. Ni quámun sie fon blúate \ noh fon fléislichemo múate; sie wárun er firlórane, \ nu sint fon góte erbórane In búachon ist nu fúntan: \ thaz wort theist mán wortan, iz ward héra in worolt fúns \ joh nu búit in úns; Wir sáhun sinaz ríchi \ joh sina gúallichi, (II, 2, 27–33)

306

Andreas Lötscher

‘Those who recognized him and changed their mind, he led them to the path to become children of God. There origin was not the blood nor carnal will. Earlier, they were lost, now they are reborn from God. In the books it is found that the word has become a man, it came in this world willingly, and now it lives among us. We have seen his kingdom and his goodness.’ b. So er thaz suért thenita, \ ther éngil imo háreta, er híaz inan irwíntan; (II, 9, 51–52) ‘When he lifted the word, the angel hared to him, he commanded him to hold on.’ Especially the pronouns ih ‘I’ and wir ‘we’, which are normally weakly accentuated, can be found before or after the verb, without apparent difference in the informational function. Rather, the reasons are usually rhythmical. Within a theory according to which in sentences with an anaphoric background element, V-1 has to be chosen, such patterns should not be possible, as such an element should be located in the Wackernagel position. One factor that can be made responsible for the fronting of such elements is the fact that subjects in a neutral ordering have to precede all other constituents within the same information-structural level and that this could be a particular reason overriding the Wackernagel position. Apart from this, it has to be taken into account that writers in many cases have some freedom as to how treat given elements such as subject pronouns within the informational structure of an utterance. The context of an utterance does not force automatically the functional interpretation to be given a pronoun with respect to its informational status.

6.2.2. “Frame topics” There are cases of constituent fronting in Otfrid that are not predicted by Hinterhölzl and Donhauser (2003), however. In particular the fronting applies not only for “aboutness topics”, i.e. topics that have as referent the object of a statement and denotate the item an utterance is about, but other elements such as temporal, local or causal adverbs with a deictic function, that function as text connectors to the preceding text or as delimitations of the contextual frame of a statement, i.e. elements that could be classified as cases of “frame-setting elements” (Jacobs 2001):

Verb placement and information structure in the Gospel Harmony by Otfrid

307

(32) a. Fúar er thuruh Samáriam, \ zi einera burg er thar tho quám, in themo ágileize \ zi éinemo gisáze. Tho gisaz er múader, \ so wir gizáltun hiar nu ér, (II, 14, 5–7) ‘He wandered through Samaria. There he came to a borough in this toil, to a resting place. Then he sat down tired, as we told earlier.’ b. Úaptun thar thie líuti \ eino brútloufti themo wírte joh theru brúti \ in sáligeru zíti. […] Thar was Kríst guater \ joh sélba ouh thiu sin múater, (II, 8, 3–4, 7) ‘There was a wedding a happy time for the host and his bride. […] The good Christ was there as well as his mother.’ c. Bi thiu ílemes io gigáhon \ zi then drúhtines ginádon, (II, 3, 63) ‘Therefore we strive to go to the graces of the Lord.’ d. Sie íltun tho bi mánne \ fon theru búrg alle, íltun al bi gáhin, \ tház sie nan gisáhin Innan thés batun thár \ thie júngoron then méistar, tház er thar gisázi \ zi dágamuase inti ázi. (II, 14, 93–96) ‘All people of the village hurried then they hurried to see him. Meanwhile the disciples asked the master

to sit down and to eat the meal.’ 6.2.3. Focussed comments Moreover, not only topics, but also comments may be placed in the prefield, as long as there is no marked topic (new topic or topic change with heavy stress) in the clause simultaneously: (33) a. Druhtin Kríst sar zi imo sprah, \ so er nan érist gisah: “Symon bistu, muates línd, \ joh bistu ouh dúbun kind;” (II, 7, 35f.) ‘Christ, the Lord, told him, when he first saw him: “You are Simon, of peace-loving mind, and you are the son of a dove.” ’ b. Wil thu iz kléinor reken, \ in wíne gisméken: fon Kríste scalt thu iz zéllen, \ gisteist thu tház irwéllen. (II,9,69–70) ‘If you want to interpret it more subtly, taste it as wine, you have to relate it to Christ, if you want to do so.’

308

Andreas Lötscher

c. Tharana maht thu irthénken, \ mit brúnnen thih gidrénken, gifréwen ouh thie thíne \ mit géistlichemo wíne Ih zéllu thir in alawár: \ luzil dránk ih es thar. (II, 9, 23–25) ‘You might think about this and quench your thirst with wather and enjoy your people with spiritual wine. I tell you honestly: I drank little of it here.’ d. Sie sint úbil thrato \ wérko joh thero dáto; míhil ist ir úbili \ thuruh thaz hérza frávili. (II, 12, 89–90) ‘They are very evil in works and in deeds; their evilness is great through their sinful heart.’ e. Zéllu ih thir ouh hiar tház \ bi thiu stéininun fáz: hérza iz sint gidígano \ thero gotes drútthegano. (II, 9, 11–12) ‘I also give you an explanation of the stoneware vessels: they signify the pure harts of the disciples of god.’ Within this group, there are some special cases. One type concerns modal adverbs, often to verba dicendi: (34) a. Tho drúhtin themo mán luag, \ thes ih hiar óbana giwúag, óbaz theih hiar fóra quad, \ thaz er mo hárto firspráh: Hárto sageta er imo tház, \ thaz er mo bórgeti thiu baz; (II, 6, 3–5) ‘When the Lord prohibited to the man, as I told earlier, the fruit I mentioned before and he banned, he told him severely that he should refrain from it.’ b. Scono zált er imo tház \ (so drúhtin io giwón was). (II, 12, 51) ‘He explained it to him in good words, (as the Lord used to do).’ Semantically, the adverbs belong to the verb as a modal adverbial. As to the information structure, undoubtedly, the adverbial and the finite verb together make up part of the comment of the main clause. A second group concerns sentence-initial, prefield position of nonfinite verbal elements. At first sight, such cases seem to contradict the rules stated earlier that nonfinite verbal elements in main clauses follow the finite verbal element and are placed be in V-late position. The examples show that such nonfinite verbal elements can also be fronted like other constituents:

Verb placement and information structure in the Gospel Harmony by Otfrid

309

(35) a. Thó sprah Kríst zi imo sár: \ “giscríban ist in álawar, thaz mánnilih giwéreti, \ selb drúhtines ni kóroti.” (II, 4, 75) ‘Christ said to im then: „It is written truly, That everybody shall fight not to tempt the Lord.” ’ b. Giléitit ward tho druhtin Kríst \ thar ein éinoti ist, (II, 4, 1) ‘Christ was lead into a desert.’ c. Spúan er io zi nóti \ jénan zi úbarmuati, zi gíri ioh zi rúame, \ zi suaremo ríchiduame. Níazan sah er inan tház, \ thaz ímo ju gisuás was; (II, 5, 7–9) ‘He lead him by force to arrogance, to greed and He saw him make use of that was agreeable to him.’ Such verbs undoubtedly are part of comment, too, often together with an end-placed subordinate clause: The verb is not a concept introduced earlier in the text, it is not in contrast to another predicative expression that possibly could have the function of a topic, and they are nevertheless in a position with heavy accent. The motivations for preposing of a comment element instead of leaving it in the usual comment position at the end of the middlefield are often difficult to determine. One factor seems to be that preposing a comment element helps solving other ordering problems in a sentence. In the verses cited in (34), in the usual word order, the adverb as comment would have to be located at the end of the clause, which could make it appear as the sole comment element of the clause. In other cases, preposing an comment element makes it possible to keep another element at the end of the main clause that otherwise would land in a position considered as infelicitous. Often a pronoun belonging to the background part of the utterance would have to be placed in the first – topic – positioning front of the finite verb. If in (35c) the infinitive would be in its grammatically normal final position there could arise rhythmical problems for the demonstrative pronoun “thaz”. Furthermore, the placement of a heavily accented infinite verb usually facilitates an appropriate versification of a sentence; these verbs as a rule occupy the first accent position of the verse. Another possible factor is special emphasis. Such special emphasis can be seen in II, 9, 26 (ex. (33c)): (33) c. Tharana maht thu irthénken, \ mit brúnnen thih gidrénken, gifréwen ouh thie thíne \ mit géistlichemo wíne Ih zéllu thir in alawár: \ luzil dránk ih es thar, (II, 9, 23–25)

310

Andreas Lötscher

‘You might think about this and quench your thirst with wather and enjoy your people with spiritual wine. I tell you honestly: I drank little of it here.’ An element preposed that according to the usual rules is expected in the the end of a sentence is much more marked when it is fronted.10 Both the case of preposed modal adverbs and the case of nonfinite verbal elements raise the additional question why preposing or fronting is necessary or possible for these elements when they form a common comment together with the finite verb in the V-2-positon. Structurally, they should form a single constituent whose internal ordering should not be affected by information-structural factors. The explanation lies in the assumption that in fact, the finite verb as a surface constituent forms a constituent of its own, structurally separated from the other parts of the other elements of the verbal group. Whether the finite verb is placed the V1position or in the V2-position, this position is grammatically fixed and cannot be changed by information-structural factors. In contrast to this, the other elements of the verbal group can be placed independently according to information-structural needs, i.e. they can be placed in front of the clause-initial finite verb. Independently of the grammatical explanation of these facts, they in itself are one more indication that the usual German field-and-frame structure is valid for Otfrid as well.

6.3. Discourse markers Another apparently problematic type of ordering is found in adverbials that represent illocutive specifications of a speech act. They can appear in the prefield position as well, without having a particular contrastive value either as a topic or as a focus: (36) a. giwisso zéllu ih thir nú: \ finfi hábotost thu jú. (II, 14, 52) ‘I tell you for certain, that you had five (husbands).’ b. Thísu selba rédina \ theih zálta nu hiar óbana, bréitit siu sih hárto \ géistlichero wórto; Thoh wíll ih es mit wíllen \ hiar étheswaz irzéllen. (II, 9, 1–3) ‘Although this events that I telled earlier about are open to many different spiritual interpretations nevertheless I will explain some of them.’

Verb placement and information structure in the Gospel Harmony by Otfrid

311

From a grammatical point of view, elements as giwisso in (36a) are adverbials to the verb. Thoh in (36b) on the other hand is a concessive particle. In the usual word order, in the verses cited in (36), these elements would normally have to be located at the end of the clause. This could make them appear as the sole comment element of the clause. This could motivate preposing the element the point of view of the information structure of such sentences, the adverbial and the finite verb form the comment of the main clause. Semantically and pragmatically, however, these elements are not adverbials in the strict sense, but are discourse markers, expressing a perlocutive attitude of the speaker or the perlocutive connection of an utterance to the context. As such, these expressions are not part of the propositional content of the utterance and cannot be integrated into its informational structure. It is quite usual that such elements are placed in front of a sentence. Thus, such sentence patterns do not contradict the earlier findings; rather, they show that besides the information-structural factors, there are more factors influencing the order of elements in a sentence

6.4. Conclusions In section 6.2., it was attempted to explain the choice between V-1- and V2-structures along the pragmatic lines proposed by Hinterhölzl and Donhauser (2003). It was observed that their theory can be corroborated partially; but there are more constituents that may be in prefield position than predicted by their hypothesis, namely constituents that are not topics in a strict sense, but rather frame delimitations or textual connectors on the one side and comment elements on the other hand. This seems to be a rather heterogeneous collection of elements, which makes it difficult to derive a simple information-structural rule for explaining the choice between V-1and V-2-structure. Looking from a different angle, however, a rather simple principle guiding the distribution can be formulated. Obviously, there are different factors that may induce the fronting of an element: Elements may be fronted if they are topics, if they serve as textual connectors, if they are discourse markers, if they are given special emphasis as comment, or in some special cases, to circumvent ordering problems that otherwise might impair a correct pragmatic interpretation. The particular factor inducing the fronting or the particular pragmatic function of a constituent has no particular relevance for the position in the prefield, however; any element that takes the first position among the non-verbal constituents in a particular utterance for particular pragmatic or textual reasons is fronted left to the

312

Andreas Lötscher

finite verb, except when it is a weakly accentuated element to be located in the Wackernagel position. The place in front of the verb thus does not have a specific quality marking the element taking that place as having a particular pragmatic function in the sentence. This principle has some resemblance to the V-2-rule in present day Standard German. The main difference is that unlike in Standard Present Day German, the prefield is not to be filled obligatorily, but only when there exists an element that is marked pragmatically or stylistically. Looking at the ordering rules from the point of view of the verb, we may conclude that in the case where there is no particular reason for any element to be fronted, the position of the finite verb is the leftmost position in the sentence and that this is the basic, i.e. the unmarked case. V-1 thus appears to be the basic order. These considerations seem to contradict the findings in the first subsection, namely that there must be a V-2-rule for grammatical reasons. In the light of the previous analyses, these findings can be integrated in the following way: Grammatically definable elements such as question words or grammaticalised elements such as resumptive pronouns and particles form a separate group of preposing, i.e. preposing for strictly grammatical reasons, as opposed to preposing for information-structural reasons. A common restriction in this system seems to be that exactly one constituent can be fronted for grammatical reasons. For the fronting rule for pragmatical reasons, the system is more variable, as will be shown in section 7. One problem remains, namely the rule that yes-no-questions are marked by V-1-position. If this is a grammatical rule, the question arises how this grammatical rule is to match with a basic V-1-structure for the “unmarked” case. Superficially, simple V-1-structures in assertive sentences and in questions cannot be distinguished. There are two possible answers to this question: Either the V-1-order represents the same basic order as that for assertions; questions just do not have a topic usually. Or there may be a structural difference between assertions and yes-no-questions: For questions there is no place open in front of the verb to be filled optionally. I must leave this question open to further discussion.

7. Verb-late/verb-last position of the finite verb in main clauses Besides Verb-first and Verb-second structures, Otfrid uses Verb-late or Verb-last position for the finite verb in main clauses as well. The pattern is by no means rare and not exceptional in Otfrid’s work.

Verb placement and information structure in the Gospel Harmony by Otfrid

313

(37) a. Sin wórt iz al gimeinta, \ sus mánagfalto déilta al io in thésa wisun \ thuruh sinan éinegan sun (II, 1, 33f.) ‘His word decided it all, divided it up in diversity, all in this manner through his only son.’ b. So er nan zi ímo brahta, \ Kríst inan irknáta, (II, 7, 53) ‘When he brougth him to him, Chris recognized him.’ c. Druhtin Kríst sar zi imo sprah, \ so er nan érist gisah: “Symon bistu, muates línd, \ joh bistu ouh dúbun kind”(II, 7, 35f.) ‘Lord Christ said to him, when he saw him the first time: “You are Simon, of a gentle mind, and you are the son of a dove.” ’ d. “Séhet”, quad er, “hérasun, \ war geit ther drúhtines sun;[…]” Thiu wórt sie sar intfíangun \ joh after ímo giangun (II, 7, 11, 15) ‘ “Behold”, he said, “where God’s son is going […]” They accepted his words at once and followed him.’ e. María thaz bihúgita, \ joh Kríste si iz giságeta. (II, 8, 12) ‘Maria noticed it and told it Christ.’ f. Ein scaf er stántan gisah, \ thaz was zem ópphere gimah(II, 9, 59) ‘He saw a sheep standing there that was suitable for a sacrifice’ g. Zít ward tho giréisot, \ thaz er gíangi furi gót; ópphoron er scólta \ bi thie síno súnta, (I, 4, 11–12) ‘The time had arrived that he should appear with got He should make a sacrifice for his sins.’ The question is how this pattern is to be interpreted within Otfrid’s system so far described. One can think of the following two possibilities: –

–

“Verb-last-pattern”: We may hypothesize that Otfrid uses the grammatical verb-last pattern similar to that of asyndetic subordinate clauses. This would amount to the hypothesis that Otfrid stretches or even violates the grammatical system at disposition using a pattern reserved for subordinate clauses in main clauses, simply in order to fulfil the necessities of verse structure. “Expanded fronting”: A second possible description consists in the assumption that the Verb-late pattern in main clauses arises from a more general application of the fronting rule of the system described in the previous chapter, together with a rule placing weak elements in a Wackernagel position after the first element. The hypothesis would implicate that in fact such V-late-patterns are the result of the same or at least similar rules as that derived from V-1/V2-structures.

314

Andreas Lötscher

Certain aspects speak in favour of the second approach. The overwhelming number of main clause V-late structures show a specific well-defined distribution of constituents: The first constituent is an element with heavy accent; this element is followed by one or more weak constituents, such as anaphoric pronominal constituents (“continued topic” or background elements). The elements with a heavy accent are fronted under the same conditions as that discussed with V-2-patterns: They are heavily stressed topics (contrastive topics or new topics), as in (38a) – (38e), or comments, as in (38f) and (38g): (38) a. Sin wórt iz al gimeinta, \ sus mánagfalto déilta al io in thésa wisun \ thuruh sinan éinegan sun (II, 1, 33f.) ‘His word decided it all, divided it up in diversity, all in this manner through his only son.’ b. So er nan zi ímo brahta, \ Kríst inan irknáta, (II, 7, 53.) ‘When he brougth him to him, Chris recognized him.’ c. Druhtin Kríst sar zi imo sprah, \ so er nan érist gisah: “Symon bistu, muates línd, \ joh bistu ouh dúbun kind;” (II, 7, 35f.) ‘Lord Christ said to him, when he saw him the first time: “You are Simon, of a gentle mind, and you are the son of a dove.” ’ d. “Séhet”, quad er, “hérasun, \ war geit ther drúhtines sun […]” Thiu wórt sie sar intfíangun \ joh after ímo giangun;(II, 7, 11,15) ‘ “Behold”, he said, “where God’s son is going […]” They accepted his word at once and followed him.‘ e. María thaz bihúgita, \ joh Kríste si iz giságeta. (II, 8, 12) ‘Maria noticed it and told it Christ.’ f. Ein scaf er stántan gisah, \ thaz was zem ópphere gimah (II,9,59) ‘He saw a sheep standing there that was suitable for a sacrifice’ g. Zít ward tho giréisot, \ thaz er gíangi furi gót; ópphoron er scólta \ bi thie síno súnta, (I, 4, 11–12.) ‘The time had arrived that he should appear with got He should make a sacrifice for his sins.’ The weak constituents, on the other hand, are placed in enclitic attachment to this heavy constituent in front position. If the underlying structure followed the verb-final clause structure the ordering among weak and heavy elements should be the reverse one. This position of weak elements enclitically after the first heavy element of the sentence can be seen as the typical Wackernagel position in the original Indo-European version.

Verb placement and information structure in the Gospel Harmony by Otfrid

315

Such a sequence of constituents thus can be seen as a combination of fronting heavily stressed topic or comment constituents under the same conditions as described earlier and the enclitic attachment of weakly accentuated constituents directly to this first constituent, instead of being attached to the right of the verb. The difference between V-2-structures, as described above, and V-late-structures consists in the fact that for V-late-structures, we have a different version of the Wackernagel position and the absence of the restriction of the fronting to a single constituent.11 As this concerns only fronting for pragmatical reasons and not fronting for grammatical reasons, this could be an argument in favor of the hypothesis that fronting for grammatical reasons is a rule different from that of fronting for pragmatical reasons. The question remains, why the V-2-structures are so more frequent than V-late-structures. For one thing, the information-structural value of the verb might have an influence. One important question is, however, how in Otfrid’s system the Wackernagel position is to be defined. In the present account, we have to assume two different versions. Depending on which version of the Wackernagel rule is applied, the result is either a V-2structure or a V-late-structure. The choice of the particular version of the Wackernagel rule may depend on extragrammatical factors, such as rhythm, rhyming necessities etc. But eventually, we can speculate that we are confronted with two versions of the constituent defining the Wackernagel position belonging to two different diachronic periods. The older one, of course, is the definition of the Wackernagel position by the first constituent with heavy accent. In the newer one, the Wackernagel position is defined by the verb in V-2-positon. In this view, Otfrid is using two patterns from two diachronically different systems, a phenomenon that is typical for intermediate stages in language change. We can speculate with some plaubility that the older stage has been carried on in poetic works as the new type was developed. Thus, the V-late pattern in main clauses can be interpreted as typical for poetic language. Apart from the specific properties of V-late-patterns mentioned, one more argument in favour of the second approach of a description can be found in the fact that V-late ordering appears also in other early sources of OHG, such as the Isidor-translation Contra Iudaeos.12 In sentences as the following ones the sequence of elements and their informational characteristics is similar to the patterns described above – a heavy constituent as a first element is followed by a weak pronominal constituent: (39) a. erino portun ih firchnussu. (Isidor 157) ‘I shatter iron doors.’

316

Andreas Lötscher

b. Dhes martyrunga endi dodh uuir findemes mit urchundin dhes heilegin chiscribes. (Isidor 516) ‘We demonstrate his martyrdom and his death with evidence from the Holy Writings.’ Sentences as these have been the object of discussions on several occasions, such as in Lenerz (1984), Tomaselli (1995) and Ramers (2005), with varying grammatical interpretation, which take directions comparable to the proposals put forth above, although the relatively small number of examples makes it even more difficult to draw general conclusions from the facts. Independently from the grammatical interpretation, such sentences seem to show that the possibility of V-late structures existed in early Old High German outside of Otfrid’s work. Furthermore, Tomaselli (1995) points to the fact that in Old English similar structures can be found even more frequently. Thus it is not unfounded to assume that Otfrid did not invent such a word order rule by forcing grammar but that he used a traditionally given pattern. It has to be added, though, that V-late structures can be found that do not fit into this explanation and apparently can only be explained as strict verb-final patterns. (40) a. So was er io mit ímo sar, \ mit imo wóraht er iz thar; so wás ses io gidátun, \ sie iz allaz sáman rietun. (II, 1, 15f.) ‘He thus was with him all the time, he did it together with him; whatever they did, they planned it together.’ b. Stuant Johánnes gomono éin \ mit sinen júngoron zuein, gisáh er gangan thárasun \ then selben drúhtines sun; […] Er fíngar sinan thénita, \ then júngoron sar tho zélita, joh sár in tho giságeta \ thia sálida in thar gáganta. (II, 7, 5–6,9–10) ‘John, one of the men, stood there with two of his disciples, he saw passing there god’s own son. […] He (= John!) pointed with his finger, he told his disciples and explained them at once that salvation was approaching them.’ In sentences as these, the weak pronoun is in the first position and a heavy constituent is following it.13 This cannot be explained by the factors developed above. These sentences show the characteristics of strict verb-final structures, used e.g. in asyndetic subordinate clauses. Such patterns are rare exceptions, though, and they cannot put to question the explication given for the more frequent cases.

Verb placement and information structure in the Gospel Harmony by Otfrid

317

8. Conclusions We may summarize the results of the above observations and considerations as follows. Although the order of elements in Otfrid’s work, especially with regard to verb position, is more variable than e.g. in Notker, his use is by no means irregular; on the contrary, it is possible to find relatively clear patterns and principles: –

–

–

The position of verbal elements clearly follows grammatical rules: There are distinct grammatical differences between main clauses and subordinate clauses, especially with respect to the relative position of finite and nonfinite verbal elements. These rules indicate that Otfrid applies a system of “frame-and-field-structure” comparable to that of later variants of German. Moreover, for yes-noquestions and conditional clauses, the verb position is clearly defined as V-1. This implies that Otfrid has a system with grammatically defined verb positions. Several factors seem to blur the rules of “Rahmenbildung” and verb position: There are no absolute rules for the position of the verbal elements at the left side and at the right side of the clause. This variability can be attributed to pragmatic influences, especially information-structural factors. Pragmatic factors take a more important role in the ordering of elements than in later stages of German; in other words, grammatical rules are more restricted in their range, and ordering is subject to pragmatic factors more frequently. One major difference of Otfrid to later systems is the variability of the position of finite verb in the main clause. It can be placed in the first, the second and even in a late position. The choice can be attributed to specific pragmatic factors: − V2-structures and V-late arise from preposing elements under specific conditions. Most often, these factors are of information-structural nature, such as topicalization and focussing comment. Other elements that tend to be preposed are textual connectors and discourse markers. As a consequence, it is not possible to attribute a specific pragmatic value to the position in front of the finite verb, restricted only to informationstructure. Rather, there are several different factors that play a role in preposing, and this place is open to any element that is to be preposed for any reason whatsoever.

318

Andreas Lötscher

−

–

–

The V-1-structure in this view appears as a structure that is used in absence of such specific factors; in a certain sense, V1 thus appears to be the basic order. − V-late-structures may be captured by the same principles, applied in a more general way, if we assume a diachronically older definition of the Wackernagel position. This implies that Otfrid is applying ordering rules originating in different diachronic stages of Germanic languages. Looking at the Old High German Isidor translation, we notice that Otfrid is not an isolated case. In one way or other, V-1 and V-late patterns can be found in the Isidor translation as well, although the material is not abundant, which makes an interpretation for Isidor more difficult. Moreover, Robinson (1996) correlates some of the more marginal patterns to special conditions; e.g. he characterizes the V-1 pattern as a “distinctively foreign pattern, which [...] led the Isidor translator to use it as a flag to mark Biblical quotations [...]“ (Robinson 1996: 25f.) In the light of Otfrid, this statement could be revised to some extent Otfrid shows some preferences to V-2-structures in main clauses when fronting specific elements. This might show a tendency to restrict the number of elements in front of the verb to a single constituent, although this cannot be an absolute rule as in later stages of German. It may well be that with his combination of freedom of word order left to the finite verb and his preference for and partial grammaticalization of V-2-structures, Otfrid combines newer developments with the use of older pattern that are vanishing in his own period. This would be no unusual trait for a poetical text such as his Gospel Harmony.

There certainly remain many problems and apparent exceptions to be considered, and at a closer look, we can certainly find more descriptional problems. Nevertheless, I hope to have made plausible that, what concerns word order, Otfrid’s language is not as irregular as is often assumed.

Notes 1. 2.

“Die Späterstellungen bei Otfrid sind reimbedingt und damit keine aussagekräftigen Belege”. (Schrodt 2004: 196). As P. Piper notes in a commentary to O II, 4, 97: “Der Reim kann wohl den Dichter zu weniger gebräuchlichen Wendungen und Constructionen, zur Ver-

Verb placement and information structure in the Gospel Harmony by Otfrid

3.

4.

5.

6.

7. 8. 9. 10.

11.

12. 13.

319

wendung von Flickwörtern u. dgl. verleiten, nie aber dazu, sprachlichen Unsinn zu schreiben.” (Otfrid 1882, 1: 161) Cf. Jacobs (2001). Jacobs lists varying semantic attributes of different kinds of “topics”. His list seems to be incomplete, though; moreover it remains open to questioning whether any expression that shows the linear and accentual properties of a topic on the surface structure level must be classified as a topic semantically. It is impossible, of course, to give here an overall outline of the theoretical questions in this domain. E.g., I leave out the discussion of different kinds of topics such as « aboutness-topics », « familiarity topic », « contrast topic » etc. (cf. Frascarelli/Hinterhölzl 2007), as these distinctions play no role for the following discussion. Other terms are explained in its context. In Lötscher 2006, this dimension is called “informational relevance”.We do not only discuss knowledge, but also hypotheses and other concepts in nonreal worlds, sometimes within ad hoc modalities. Cf. the notion ‘disputable’ in Büring (1997: 700). I assume that it is part of a communicative act of a speaker to put something specific to dispute in a communicative act (granted that it must be disputable for some reason). But things are not simply disputed because they are disputable; most assumptions are disputable, but we do not discuss them anyway. This dynamic aspect of topic treatment wthin a text gives place to the different types of topics mentioned in fn. 4. The topic-comment interpretation of adjectives and quantifiers is discussed in more detail in Büring (1997). The examples are usually taken from book II. The fronting of comment elements to the prefield has been a common possibility in all periods of the history of German, although it apparently has a not the same stylistic and rhetorical functions at all times. Tomaselli (1995) treats similar patterns in the Old High German translation of Isidor as V-3-patterns. Whether this is correct or not, could only be decided on a broader set of examples than are attested in Isidor. In any case, the situation in Otfrid shows that a purely numerical definition (“V-2”, “V-3” etc.) on the basis of a few examples is not sufficient for a general grammatical description. More examples are given in Schrodt (2004: 202f.) Actually, the pragmatic function of sie in II, 1, 16 in example (40a) is not altogether clear; we could assume as well that the pronoun has a special topic function within the context, which would make this example an instance in favour of the “fronting”-hypothesis.

320

Andreas Lötscher

References Text sources Isidor 1964

Der althochdeutsche Isidor. Edited by Hans Eggers. Tübingen: Niemeyer.

Otfrid 1882/87 Otfrids Evangelienbuch. Edited by Paul Piper. Vol. I: Einleitung und Text. Vol. II: Glossar und Abriss der Grammatik. Freiburg: J.C.B. Mohr. Otfrid von Weissenburg 1973 Otfrids Evangelienbuch. Edited by Oskar Erdmann. 6th edition by Ludwig Wolff. (Altdeutsche Textbibliothek Nr. 49, Ed. 6). Tübingen: Niemeyer.

Linguistic work Büring, Daniel 1997 The Meaning of Topic and Focus. The 59th Street Bridge Accent. London and New York: Routledge. Donhauser, Karin and Roland Hinterhölzl 2003 Die Rolle der Informationsstruktur bei der Herausbildung von Wortstellungsregularitäten im Germanischen. In Informationsstruktur. Die sprachlichen Mittel der Gliederung von Äußerung, Satz und Text. Finanzierungsantrag 2003-2007, ed. Caroline Féry, Potsdam und Berlin: Universität Potsdam. Frascarelli, Mara, and Roland Hinterhölzl 2007 Types of Topics in German and Italian. In Information Structure and the Architecture of Grammar: A Typological Perspective, eds. Susanne Winkler and Kerstin Schwabe, 87–116. Amsterdam/Philadelphia: John Benjamins. Hinterhölzl, Roland, Svetlana Petrova and Michael Solf 2005 Diskurspragmatische Faktoren für Topikalität und Verbstellung in der althochdeutschen Tatianübersetzung (9. Jh.). In Approaches and Findings in Oral, Written and Gestural Language, eds. Shinichiro Ishihara, Michaela Schmitz and Anne Schwarz, 143–182. (Interdisciplinary Studies on Information Structure 3.) Potsdam: Universitätsverlag Potsdam. Jacobs, Joachim 2001 The dimensions of topic-comment. Linguistics 39: 641–681. Lenerz, Jürgen 1984 Syntaktischer Wandel und Grammatiktheorie. Tübingen: Niemeyer.

Verb placement and information structure in the Gospel Harmony by Otfrid

321

Lötscher, Andreas 2005 Linksperiphere Adverbialsätze in der Geschichte des Deutschen. Beiträge zur Geschichte der deutschen Sprache und Literatur 127: 347–376. 2006 Die Formen der Sprache und die Prozesses des Verstehens. In Text – Verstehen. Grammatik und darüber hinaus, eds. Hardarik Blühdorn, Eva Breindl and Ulrich H. Waßner, 19–45, Berlin: de Gruyter. Näf, Anton 1979 Die Wortstellung in Notkers Consolatio. Berlin/New York: Mouton de Gruyter. Nemitz, Werner 1962 Zur Erklärung der sprachlichen Vertöße Otfrids von Weißenburg. Beiträge zur Geschichte der deutschen Sprache und Literatur (T) 84: 358–432. Petrova, Sevtlana and Michael Solf this vol. On the methods of the information-structural analysis of historical texts. A case study on Old High German. In New Approaches to Word Order Variation and Change in the Germanic Languages, eds. Roland Hinterhölzl and Svetlana Petrova. Berlin: Mouton de Gruyer. Ramers, Karl Heinz 2005 Verbstellung im Althochdeutschen. Zeitschrift für Germanistische Linguistik 33: 78–91. Robinson, Orrin W 1997 Clause subordination and verb placement in the old high German “Isidor” translation. Heidelberg: Universitätsverlag C. Winter, cop. Schrodt, Richard 2004 Althochdeutsche Grammatik. Tübingen: Max Niemeyer. Sperber, Dan and Deirdre Wilson 1986 Relevance: communication and cognition. Oxford: Blackwell. Tomaselli, Alessandra 1995 Cases of Verb Third in Old High German. In Clause Structure and Language Change, eds. Adrian Battye and Ian Roberts, 345–369. Oxford: Oxford University Press. Wunder, Dieter 1965 Der Nebensatz bei Otfrid. Heidelberg: Carl Winter. Zifonun, Gisela (et al.) 1997 Grammatik der deutschen Sprache. Vol. 3. Berlin/New York: de Gruyter.

Translating information structure: A study of Notker’s translation of Boethius’s Latin De Consolatione Philosophiae into Old High German Rosemarie Lühr

Abstract In this paper we are studying information structure on the basis of Notker’s Old High German translation of Boethius’s Latin De Consolatione Philosophiae. Our questions are: How and to which extent is the information structure of Latin converted into Old High German? What insights into the information structure of Old High German do we gain from this? To answer these questions we have to describe specific semantic characteristics of the Latin and of the Old High German language and to compare Classical Latin with Late Classical Latin. Furthermore we must discuss whether there are any specific characteristics of the information structure in Notker’s translation of rhetorically marked word structures which are particularly prominent in his poetic source. It will be shown that Notker did not convert the hyperbaton into Old High German, but, where appropriate, he used other figures of speech or rhetorical word order, especially in exclamative sentences and causal sentences introduced with wanda, which are mostly in explanatory parts of the text. Concerning focus and topic, all the linguistic means which mark these information structural entities are investigated: focus particles, emphatic pronouns, word order, contrastive structures, which resemble I-topicalisation in modern German vs. initial position, anaphoric and demonstrative pronouns, the dissolution of the relative connection of sentences, a construction which does not exist in Old High German. In sum, it will be shown that Notker’s representation of the information structure of the Latin original is first of all contingent on his didactic purposes. In addition Notker’s handling of the information structure demonstrates a fundamental difference between the two languages compared here with regard to the positioning of the kinds of foci: Old High German: structural focus – verb – emphatic focus vs. Latin: verb – emphatic focus – structural focus.

324 Rosemarie Lühr 1. Introduction Translations are a possible means of coping with the methodological problems arising from the study of the syntax of historical linguistic corpora. For the lack of (native) speaker competence can be compensated in parts by comparing the source language and the target language, cf. Petrova and Solf (in this volume). In the historical stages of German the relevant contrast is the one between Old High German and Latin as it is reflected in the language of the translators. One Old High German author who explicitly comments on this is Notker Labeo (Notker III of St Gall; Notker the German). In her groundbreaking study of Notker’s translation of Martianus Capella Glauch therefore rightly quotes Notker’s letter to the Bishop of Sitten: “Da die Klosterschüler ohne das Vorstudium gewisser Disziplinen die kirchlichen Bücher nicht vollständig verstehen könnten und er wünsche, dass sie Zugang zu diesen Büchern hätten, wage er es, lateinische Texte in unsere Sprache zu übertragen und das syllogistisch, figürlich und rhetorisch Ausgedrückte mittels Aristoteles, Cicero oder eines anderen artes-Schriftstellers zu erhellen.“ [As the pupils of the convent school could not fully understand the ecclesiastical books without having previously studied certain subjects and as he wished them to have access to these books, he was taking the liberty to transfer Latin texts into our language and to use Aristotle, Cicero or other writers of the artes poeticae to clarify what has been expressed by syllogisms, figures of speech and by rhetoric.] (2000: 29). And indeed, “Notker is unique in his time with his pedagogically motivated work as a translator and with his attempt to use German prose for educational and academic purposes”. Strictly speaking, Notker’s individual achievement lies in the “Verbindung von Texterklärung und Text überhaupt mit Übersetzung” [in combining explanations of texts or texts in general with translation] (Schröbler 1953: 153). It is generally assumed that in his translations from the Latin, Notker “always takes the original as his starting point and (...) never really departs from its contents or structure” (Glauch 2000: 60). The question is, however, whether this holds true for all levels of language. Of particular interest in this context is the information structure as one “organizational level”, because we can assume a priori that a good translator always tries to imitate the informational structure of the original. To answer this question, we need to contrast the focusbackground and the topic-comment structure of the Latin and of the Old High German text. For although we can assume that the categories of information structure are universal concepts, this, however, does not apply to

Translating information structure 325

the ways different languages realize them. Therefore the specific lexical, morphosyntactic and topological strategies for marking categories of information structure of specific languages are particularly interesting for us. This problem becomes particularly pertinent if there is no fixed syntactic position, for example for marking focus: Because if we have no focus in situ (Hetland and Molnár 2001: 621; cf. also Jacobs 1986: 123), focussing one of the constituents means that we have to adapt syntax structures accordingly. Therefore the first object of our research is: How and to which extent is the information structure of Latin converted into Old High German? From this the second question arises: What insights into the information structure of Old High German do we gain from this? In order to learn more about this, we will begin with describing specific semantic characteristics of the Latin and of the Old High German language. In this context we will also compare Classical Latin with Late Classical Latin. We will also have to discuss whether there are any specific characteristics of the information structure in Notker’s translation of rhetorically marked word structures which are particularly prominent in his poetic source1. In Latin, the hyperbaton is a particularly important linguistic device expressing information structure. As the hyperbaton has a potential for distinct effects of the information structure, which may have inspired imitation even in nonclassical languages, Notker may also have used these kinds of structures for representing the information structure of Latin in Old High German. Moreover, we have to examine both languages as to whether there are syntactic differences of position with regard to realizing various kinds of focus. Linguistic research in general distinguishes between a structural focus in the innermost embedded phrase of a sentence on the one hand, and a different form of narrow or emphatic focus on the other hand (Abraham 1986 and 1992; Doherty 2002: 30–43). We have evidence that in Latin both the structural focus and the emphatic focus are usually serialized on the right hand side of the verb, whereas in Old High German one focus is placed on the left hand side and the other one on the right hand side. In special cases, in addition, focus in Latin seems to be positioned in the left periphery. We also have to keep in mind the structuring of given and inferable information2 – aspects such as continuous, shifting3, familiar4, aboutness5, discourse6 and contrastive topics7 – as Notker may have used these topics in various forms and degrees. In general, however, we can say that due to Notker’s linguistic creativity each act of translation as such is a unique act. For example, for pedagogical reasons Notker frequently changed the syn-

326 Rosemarie Lühr tactic structure of his source when translating it into Old High German8. To make sentences more comprehensible, he often added short subordinate clauses – especially causal clauses – which can also be interpreted as part of the information structure. But we need to look at the translations in their entirety in order to get a more general picture about Notker’s syntax. We choose Boethius’s De consolatione Philosophiae as our basic text, for as a platonic dialogue and a disputation, this text uses the devices of rhetoric and dialectics in almost every part of the text9. The text was written in Late Classical Roman Prose, but it also contains poems, so it is a prosimetrum. Boethius wrote the original manuscript shortly before his execution in AD 524 or 525. In the Early Middle Ages, the text was hidden for three centuries; from the second half of the 9th century onwards it was used as a textbook in schools. Today there are still 500 manuscript copies (Gruber 1978: 13 ff.). It has been shown that Notker edited the manuscripts available in the Abbey of St. Gall in connection with the commentaries prevalent at the time (Tax 1986: XIXff.) and that he used the Alemannic variety of Old High German10. As a rule, the basis for Notker’s translations was the specific, slightly modified version of the Latin text which he prefixed to his translations, whereas commentary manuscripts were only used occasionally (Glauch 2000: 145). Therefore we can compare the word order of the individual Latin text with that of the immediately following Old High German translation.

2. Particularities of the Latin and the Old High German syntax pertinent to the information structure 2.1. Classical Latin vs. Late Classical Latin 2.1.1. Classical Latin We will compare those parts of the syntax of Classical Latin which are relevant for the information structural arrangement to contemporary German in order to highlight the differences. These will be grammatical vs. rhetorical word order and the hyperbaton11. First, let us look at word order in Classical Latin. As is generally known, the usual or grammatical or traditional word order is as follows: Das Subjekt eröffnet den Satz, das Prädikat beendet ihn, die Objektsund Adverbialbestimmungen werden zwischen diese beiden Satzteile

Translating information structure 327

eingeschlossen, und zwar so, dass sie um so näher am Subjekt bzw. Prädikat stehen, je enger sie zu einem von beidem gehören (sog. SOP-Stellung). (...) Diese gewöhnliche Wortstellung ist nur selten streng eingehalten. Häufig richtet sich die Stellung der einzelnen Satzteile nach der Betonung, der Menge an neuer Information, der Deutlichkeit, dem Wohlklang, der Abwechslung oder nach anderen stilistischen Gesichtspunkten. So entsteht die sog. rhetorische (okkasionelle oder invertierte) Wortstellung, die von den Absichten des Sprechers bestimmt ist. Die für den Gedanken wichtigsten Satzteile nehmen häufig die am stärksten betonten Stellen im Satz ein, meistens Satzanfang und Satzende. [The sentence begins with the subject and ends with the predicate; object and adverbial clauses are inserted between these two parts in a way which moves them closer to the subject or to the predicate respectively, depending on which of these two they are more closely referring to (so-called SOP-position) (...) This common rule for word order is rarely followed. Often the position of the individual parts of the sentence is determined by emphasis, the amount of new information, clarity, melodiousness, variation or by other stylistic aspects. Thus we arrive at the so-called rhetorical (inverted or okkasionell [occasional]) word order which is determined by the intentions of the reader. The parts of the sentence which are most relevant to the ideas frequently take the most emphatic positions in the sentence, which is mostly at the beginning or the end.] (Menge 2000: 575) If you look at these word order patterns from the point of view of information structure, you have to examine whether Classical Latin can be described as “discourse configurational”12. In this case, the word order is not determined by the grammatical relation between the respective constituents of the sentence but serves special discourse needs. The topic position, for example, can be occupied by subjects, objects, indirect objects, etc., if these constituents function as the topic on the information structural level. The syntactic structure may be the following: (1)

P1 Topic

–

P0 Focus

–

V Verb

–

X pragmatically unmarked

With respect to topics, modern German, for example, is said to be “discourse configurational”, for it is assumed that modern German has a special topic position in the middle field (Frey 2000). Second, a hyperbaton is a figure of speech in which words that syntactically belong together, such as noun and attribute, verb and adverb, etc., are separated from each other for emphasis. This kind of unnatural or

328 Rosemarie Lühr rhetorical separation is possible to a much greater degree in highly inflected languages, where sentence meaning does not depend closely on word order. In Latin and Ancient Greek, the effect of hyperbaton is usually to emphasize the first word. It has been called “perhaps the most distinctively alien feature of Latin word order”13. In Classical Latin, the hyperbaton is considered as grammatical, if that kind of separation is possible in an elevated prose style: “Man kann sogar von einer Tendenz in der gehobenen Prosa (und erst recht in der Poesie) sprechen, syntaktisch zusammengehörende Wörter zu trennen, falls dies zwanglos möglich ist, ohne den Eindruck der Künstlichkeit hervorzurufen” [One can even speak of a propensity of elevated prose (and even more so of poetry) to separate words if this is informally possible without creating an impression of artificiality] (Menge 2000: 581). The insertion of enclitics such as pronouns is considered to be a point in case, but also the use of conjunctions such as autem, enim, igitur, quoque, -ne in the second position in the sentence, all of which are actually instances of Wackernagel particles. But in principle, any kind of word can be placed between words which syntactically belong together: (2)

Cicero, fam. 3,9,3 Tuis incredibiliter studiis your-Abl incredibly-Adv studies-Abl ‘I’m made incredibly happy by your studies’

delector I’m made happy

If that construction is translated directly into English we would expect sentences such as (3a,b) which violate Ross’s Left Branch Condition. In focus constructions in English, the right branch of a noun phrase cannot remain in situ. Note that (3c), which is grammatical, is not a split construction in which raw is a modifying adjective of the noun oysters but a case of secondary predication: (3)

a. *Which has he invited friend to dinner? b. *The RED he bought car last week, the BLUE he has had car for years c. Raw he used to eat oysters. (Devine and Stephens 2000: 4f.)

Apart from the SVO-order, (3a) and (3b) are also ungrammatical in contemporary German. It follows that the hyperbaton in form of the Latin example (2) is not a possible syntactic structure of New High German14.

Translating information structure 329

2.1.2. Late Classical Latin So far, there is no comprehensive description of Late Classical Latin which constitutes the beginning of Middle Latin (from 500). “Antike Syntax ist [in der Regel] eine moderne Ableitung aus antiken Autoren.” [So we can still say as a rule, ancient syntax is a modern derivation from ancient authors.] (Kindermann 1998: 42)15 The question now is, whether “discourse configurationality” actually applies in Late Classical Latin. If you go through what the handbooks tell us, you will find: In order to achieve a special effect, the end of the sentence is frequently – especially in the writings of poets – reserved for the main term which dissolves the tension or suspense which was built up before. Zwischenstellung [sandwiched position] of the verb, positioning the verb somewhere between other constituents, especially between subject and object, is said to be natural if there is a syntactic connection between the final word and the following sentence, as for example in Petronius. Apuleius is also said to frequently put the verb in the third position from the end of the sentence if it is followed by a noun with an adjective or a preposition (Hofmann and Szantyr 1965: 404). Therefore there seems to be no fixed position for topic and focus in Late Classical Latin. Regarding the hyperbaton the handbooks will tell you: Rhetorical schooling, which took parts of its rules from the poetry of the classical period, and the Klauseltechnik, the periodic style with various forms of ending periods, resulted in a frequently rather unnatural way of handling this kind of word order in post-classical prose. Petronius, for example, frequently uses this word order in the vulgar passages of his novel Satyrica. This is interpreted as a sign that the use of this order had become a general tendency16.

2.2. Old High German Syntax The position of the structural focus depends on the position of the verb. Näf (1979: 114) describes the following rule for the “personal form” of the verb, i.e. the finite verb, in declarative sentences in the translation of Boethius’s Consolatio philosophiae: (4)

Rule 1: In declaratives the personal form of the verb is in second position

This rule also applies in sentences opened by a conjunction or a connective such as unde, uuanda, noch, aber, nube, ioh, alde, sunder (Näf 1979: 125

330 Rosemarie Lühr ff). This is illustrated in (4)–(5) which display no possible pattern in modern German. (5)

I 6,18f. Et rigant and they moisten .i. i.e. Únde and

non not fúllent fill

ora elegi. i. miseri. faces-Acc mourner-Gen i.e. poor-Gen fictis false-Abl sie they

ueris true-Abl

fletibus tears-Abl

mîniv óugen. mit my-Acc eyes-Acc with

érnestlichên drânen honest-Dat tears-Dat

Lat. ‘And they moisten the face of the mourner, i.e. the poor, with true, i.e. not false tears’ OHG ‘And they fill my eyes with honest tears.’ (6)

I 9,9ff. Et abstulerant and they had dragged away quisque each of them Únde and stúc- /chen. pieces-Dat

uuâren were

particu-/ las pieces-Acc

quas Rel-Acc

poterat. could sie they

ána-uuert away

dîe îogelicher Rel-Acc each of them

mít with

íro their-Gen

besuérben dragg off

máhta. could

Lat. ‘And they had dragged away as many pieces as each of them could.’ OHG ‘And they went away with the (their) pieces which each of them could drag.’

Nonetheless, we can rightly claim for Notker’s language that, apart from a few relics, we have verb-second position in declaratives17. Näf (1979: 187f.) also puts up a rule for the verbal Satzklammer [sentence bracket]18: a structure where the finite and the infinite parts of a compound verb are separated from each other and placed at some distance within a sentence. Thus they form a kind of bracket for the other parts of the phrase: (7)

Rule 2: “Im Hauptsatz steht die Infinitform auf der dritten oder einer späteren Stelle, und zwar (...) vor oder nach substantivischem Subjekt und substantivischen kasuellen Ergänzungen (...) vor oder nach (pronominalen oder substantivischen) Präpositionalgruppen (...) vor oder nach Satzadjektiven und Adjektivadverbien.” [In the main clause the infinite form is in the third

Translating information structure 331 or in a later position, either before or after subject noun and case-determined noun complements (...) before or after a prepositional group (pronominal or nominal) (...) before or after sentence adjectives and adjective-adverbs.]

For the complete Satzklammer, however, we have to add examples of sentences with AcI-construction: For instance, in contrast to the unmarked word order of Classical Latin, in the following sentence from the Latin original the agent of an AcI-construction is in final position. The result is a focus on the right margin of the sentence. In (7) Notker does not adopt this word order but converts it into a different Old High German construction with unmarked word order, in this case with a verbal bracket: (8)

I 6,28f. Et and

dolor pain

Únde and

léid hábet sorrow has

iussit let

inesse suam approach his-Acc míh/ me

aĊtatem19./… time/age-Acc

álten getân. old-Acc made

Lat. ‘And the pain made his (old) age approach.’ OHG ‘And the sorrow has made me old.’

We also have to note that sometimes exbraciation, i.e. the positioning of words outside the sentence bracket, may also be due to rhythmical reasons: (9)

II 100,14ff. Et relicta prestantia and left aside-Prt.Abl preference-Abl uirtutisque. virtue-Gen Únde and

postulatis you demand

conscienitĊ/ sermunculis. conscience-Gen idle talk-Abl

premia de alienis rewards-Acc of strange-Abl

nehéina/ uuára tûondo déro stíuri no-Acc attention paying-Prt-Adv Art-Dat leading-Dat

déro Art-Gen

geuuízzedo. álde conscience-Gen or

uuéllent ir will/want you

déro Art/Gen

háben. have-Inf

tíu Rel

fóne of

déro túgede./ Art-Dat virtue-Dat

uuórto words-Gen

dáng thanks-Acc

ánderên other-Dat

châmen. came20

Lat. ‘And, after the preference of conscience and of virtue has been left disregarded, you demand rewards for strange idle talk.’ OHG ‘And although you did not pay attention to the precedence of conscience and of virtue, you want to be thanked for the words which came from others.’

332 Rosemarie Lühr Presumably for rhythmical reasons, the voluminous genitive object déro stíuri déro geuuízzedo. álde déro túgede as part of the focus here is positioned after the adverb of the present participle, whereas in nehéina uuára tûondo [not paying attention] the other part of the focus nehéina uuára again is placed on the left side, before the present participle. In contrast to the position of the one-word verb, Näf’s “either ... or”- rules do not point to a distinct change of position of complex predicates in comparison to early Old High German. But the following sentences clearly demonstrate the obvious differences to New High German: (10) a. Er hat seinen Großvater besucht He has his grand-father visited b. *Er hat besucht seinen Großvater He has visited his grand-father c. Er hat ihn besucht He has him visited d. *Er hat besucht ihn He has visited him Whereas in present-day German only a. and c. are acceptable, for Notker a., b. and c. are possible options. Näf does not give any further rules for the word order which could be relevant for the information structural arrangement of Latin and of Old High German. Therefore further particularities can only be demonstrated by working with a text, in our case a continuous passage from the consolatio, mainly Book I, Chapter 1.

3. Information structure in the Latin original and in the Old High German translation In analyzing the topic-comment and the focus-background structure we implicitly adopt the scalar representation of features attested to the informational status of discourse referents, cf. Petrova and Solf (in this volume). The following features are constitutive for topicality: givenness/accessibility, referentiality, definiteness21, early position in the sentence. We also agree on the differentiation of a new information focus and a contrastive focus. One kind of focus is also the I-Topicalisation (see below). Let us say in advance that we can take it for granted that Notker did not only command the rules for word order in Old High German, but of course also those for Latin. His knowledge of Latin was brilliant, as we can see from his Latin insertions, see (10). In his amendment, Notker retains the

Translating information structure 333

aboutness topic22 at the beginning of the sentence. This correlates to the linguistic usage in Latin (Menge 2000: 576). (11)

I 10,15f. Hunc uero innutritum eleaticis studiis. atque this-Acc scilicet brought up-Acc eleatic-Abl studies-Abl and achademicis academic-Abl

s. non pa-/tior mihi not I tolerate me

subtrahi. taken away-Inf.Pass

Lat. ‘to wit/scilicet, I do not tolerate that he who was brought up with eleatic and academic studies will be taken away from me.’

3.1. Hyperbaton In the hyperbaton of the Classical languages, the modifier can either be topicalized as focus and other words can be positioned between modifier and head or the topicalized head as topic is separated from the focal modifier which follows later. A further option is splitting a wide focus into a primary focus containing the head and, succeeding other words, a second focus – the modifier23. One example of a focal modifier is: (12)

a. b. I 9,22 Ac toruis and sinister-Abl Ióh and

<X

inflammata inflamed

Y2head>

luminibus. looks-Abl

trôlicho séhendíu. threatening-Adv looking

Lat. ‘and with sinister looks inflamed’

In Latin the past participle (passive) inflammata refers to philosophy which appears in the shape of a woman. This participle separates a preceding adjective from the word it refers to. Notker uses a syntactic structure with a familiar vernacular word order for this, an adverb preceding a present participle (active) “and threateningly looking”. In this case the adverb takes the position reserved for the structural focus.

334 Rosemarie Lühr The difference between these two structures shows the map of information structure onto syntactic structure: (12)

c. AP A

A’

toruisi A

DP

inflammata N

Ai

luminibus

d. AP AdvP

A

trôlicho

sehendiu

Consider also the following case: (13)

I 10,4ff. HĊ sunt these are

enim quĊ namely who

necant / infructuosis kill infertile-Abl

affectuum . affections-Gen

uberem fertile-Acc

segetem seed-Acc

Tíz/ these

tîe Rel

sínt are

ézisg seed-Acc mít with

tîen Art-Dat

tero Art-Gen

den uuûocher Art-Acc harvest-Acc rationis reason-Gen dórnen thorn-Dat

spinis thorns-Abl

fructibus rationis. fruit-Abl reason-Gen únde and

dén Art-Acc

ertémfent. / stifle

uuíllônnes. desire-Gen

The semantically corresponding translation of the Latin sentence is: Because it is them who stifle the fertile seed of reason with the infertile briars

Translating information structure 335

and thorns of the affections24.The literal translation, however, is: ‘To wit, they are those who kill with infertile thorns of affections the rich seed by fruits of reason.’ So “the rich seed by fruits of the reason” takes the place of “the seed of reason made fertile / rich by fruits”: The modifying adjective uberem (fertile, rich) is separated from the word it refers to – segetem (seed) – by the adjective complement “by fruits of reason” and receives special emphasis by appearing in front. The regular order would have been putting the whole adjective phrase at the end (Rubenbauer, Hofmann and Heine 1977: 327): segetem fructibus rationis uberem (Menge 1961: 536). At this point of his rendition Notker doesn’t go into such fine details and translates: “These are who stifle the harvest and the seed of reason with the thorns of desire.” Now, in Latin the phrase uberem segetem fructibus rationis, which contains the structural focus, stands on the right-hand side of the finite verb after the ablative phrase infructuosis spinis affectuum: In a broader sense, this is an adverb of manner and constitutes a set of alternatives25 which is why it has special emphasis. So we can say that in Latin we have emphatic focus here. In contrast, the corresponding phrase in Old High German mít tîen dórnen uuíllônnes appears immediately right of the finite verb – also an emphatic focus – while the phrase with structural focus uuûocher únde dén ézisg tero rationis is placed on the left. With regard to the position of foci in the languages compared here, this means that the order in Latin may be: verb – emphatic focus – structural focus whereas that in Old High German would be: structural focus – verb – emphatic focus. This distribution of the foci also applies to the subordinate clause: (14)

I 7,20ff. ... dum … Et when and

lacrimabilem signarem I would chronicle tearful-Acc

officio/ help-Abl

stili pen-Gen

…Únz … Únde íh …when and I temo screíb mít wrote with Art-Dat

sús âmerlicha so wailful-Acc grífele. pen-Dat

querimoniam . complaint-Acc

chlága complaint-Acc

Lat. ‘...when I ....and the tearful complaint chronicled with the help of a pen’ OHG ‘...when I ... so/such wailful complaint wrote with the pen’

A further example of a hyperbaton, this time with a possessive pronoun as modifier, is the following:

336 Rosemarie Lühr (15)

I 9,19ff. Et di-/ctantes and dictating Únde and

mír me

meis uerba my-Dat words-Acc

trâne récchende . tears-Acc producing

mít / with

fletibus. weeping-Dat íro uuórten. her-Gen words-Dat

Lat. ‘and dictating the words to my flood of tears’

First of all, we have to note again that in Latin the structural focus uerba again appears after the predicate while in OHG the accusative trâne is placed before the predicate. But as far as the hyperbaton is concerned, in Latin meis – the possessive pronoun in dative – which refers to the dative word fletibus (flood of tears) is positioned between the participle dictantes and its object uerba. Notker’s translation is different: “únde mír trâne récchende mít íro uuórten” [and me tears producing/bringing forth with her words“] with “mír” [me] as a pertinent or possessive dative before the object trâne [tears]. Admittedly, the separation of possessive pronoun and reference word has again been converted into a familiar vernacular structure of Old High German. However, Notker imitated the emphasis put on the Latin possessive pronoun in the hyperbaton by deciding on a pertinent dative rather than a possessive pronoun for its representation. But the “mít íro uuórten” outside the Satzklammer here is an emphatic focus which parallels the structural focus of the Latin uerba. So what we actually have here is a re-structuring of the different kinds of focus: The Latin structural focus uerba is turned into the emphatic focus mít íro uuórten which is put outside the Satzklammer, whereas the phrase meis ... fletibus, which had been emphatically focussed by the hyperbaton, now appears as structural focus trâne with pertinent dative. To resume: The construction of the hyperbaton cannot be converted directly into Old High German26. In Latin, it does not indicate a fixed focus position. But it shows that this language is not “discourse configurational”, as the hyperbaton is not only optional but also appears in various types. With regard to Old High German, the examples discussed in (11) to (14) clearly show the extent to which and the techniques by which Notker tries to convert information structural distinctions of Latin into Old High German. The examples show that it is highly probable that there are distinct differences between Latin and Old High German with regard to the order of the various kinds of focus: in Old High German: structural focus – verb – emphatic focus; in Latin: verb – emphatic focus – structural focus.

Translating information structure 337

3.2. Rhetorical word order When converting complex sentences, participle constructions, imperative sentences and interrogative sentences in poetic texts, Notker shows no particularities of information structure in comparison to that of prose texts. (By the way, here Middle Latin doesn’t show any differences to Classical Latin either). Therefore, we can disregard the difference between poetry and prose hereafter. But like causal clauses, exclamative sentences are part of the rhetorical word order patterns, which is why they are of particular interest with regard to the distribution of the information structural entities.

3.2.1. Exclamative sentences For a start, we find exclamative sentences in which Notker uses verb-final position in Old High German instead of the Latin verb-second position. Verb-final position also occurs in that kind of exclamative sentences in New High German which are consistent with the order in subordinate clauses. (16)

Poetry I 7,6f. Eheu . quam oh how Áh oh

ze_ alas

uuênegen poor-Acc

surda deaf-Abl sêre .

aure auertitur ear-Abl he scorns uuîo how

úbelo bad

ér / he

miseros. poor-Acc

die Art-Acc

gehôret. listens to

Lat. ‘Oh, how he with deaf ear the poor scorns’ OHG ‘Oh alas, how bad(ly) he to the poor listens.’

While in Latin the structural focus, accusative miseros, is again placed on the right side of the verb, in Old High German die uuênegen is placed on the left27, like in a subordinate clause. On the other hand, we can assume that like in New High German the main accent of the focus is on the adjective adverb. But sometimes Notker also follows the word order of his source; in the following example this is the postposition of the subject as emphatic focus after the verb, i.e. VS-position : (17)

Poetry I 11,10f. Heu quam hebet mens . oh how is faint soul

mersa fallen

precipiti into the depth

profundo. bottomless

338 Rosemarie Lühr Áh oh

uuîo how

hárto very

síh / itself

mûot . soul

káhes kestúrtez sudden fallen-Prt

mísse-hábet lets (be)disconcert(ed)

mánnes man-Gen

ín in

grûoba. abyss-Acc

dia Art-Acc

Lat. ‘Oh how faint is the soul fallen in the bottomless depth’ OHG ‘Oh how much lets itself be disconcerted man’s soul suddenly fallen into the abyss’

Compared to (16), Old High German in (17) uses exbraciation, positioning the subject after the verb group. There are cases with evidence for postposition of the subject (2.3.1.). But sometimes Notker complete re-structures exclamations: In (18), for example in Latin the focus of the sentence consists of the exclamation “felix mors hominum” [blessed death of man], while Notker uses a categorical sentence (Steube and Späth 2002: 238). He explicitly introduces a discourse referent which functions as topic of the statement in the following sentence: (18)

Poetry I 7,3ff. Felix mors blessed death dul-/cibus sweet-Dat Táz that

íst is

hominum . men-Gen

annis . years-Dat

et and

quĊ Rel saepe often

nec not uocata called-Prt

se itself uenit comes

inserit inserts mestis sad-Dat

sâlig tôd . blessed death

tér Rel

in in

lústsa- / mên pleasant-Dat

zîten times-Dat

nechúmet . not comes

únde and

in in

léit-sámên / painful-Dat

geuuúnstêr wished-Prt

netuélet. not hesitates

Lat. ‘Blessed death of men which does not push itself into the sweet years and, if he is often called, comes to the sad ones.’ OHG ‘That is a blessed death which does not come in pleasant times and on painful times if you wish for it does not hesitate.’

Considering its form, the Old High German sentence could be an answer to the question “How do you define a blessed death?” In this, the copulative construction in Táz íst sâlig tôd displayst he information structural distribution: The finite verb marks the beginning of the focus domain in the sentence.

Translating information structure 339

Exclamatory sentences are more emphatic and easier to memorize than declarative sentences and so for pedagogical reasons, Notker sometimes also converts Latin declarative sentences into Old High German exclamatory sentences: sentence (19), for example, is an exclamation with postposition of the verb which represents a Latin declarative statement with VO-position: (19)

Poetry I 7,7f. Et sæua . and cruel

claudere close-Inf

negat flentes refuses crying-Prt.Acc

oculos. eyes-Acc

Únde / uuîo and how

úngerno unwillingly

ér he

betûot closes

íro (t)he(i)r-Gen

uuéinonten crying-Prt.Acc

óugen. eyes-Acc

chéligo cruel

Lat. ‘And cruel [death] refuses to close the crying eyes.’ OHG ‘And how unwillingly he cruel(ly) closes her/their crying eyes.’ 28

Notker’s rephrasings also underline his didactically motivated endeavour to make the Latin text readily accessible and as comprehensible as a schoolbook. Due to their expressive character exclamatory sentences are particularly suited for that purpose.

3.2.2. Causal clauses Notker often added causal clauses as explanations. So, for example, in (20) and (21) he changes a simple Latin sentence into a complex one. Here the causal clause – wanda-sentences in the function of subordinate clauses29 – follows the matrix sentence. They substantiate statements made in the main clause, that is to say weil connects two propositions (Lühr 2007). From the point of view of information structure, that kind of causal clauses add focal elements to the rest of the background of the matrix sentence. They add an overall commentary to this sentence, but they have a topic-comment structure. In (20) and (21) this is a continuous-topic. In the following examples there is no correlate to deswegen in the matrix sentence: (20)

I 8,4ff. Nam nunc for now

quidem certainly

cohibebat contracted

communem common-Acc

mensuram measure-Acc

hominum./ men-Gen

sese ad herself to

340 Rosemarie Lühr Uuánda éina uuîla . for Art-Acc time-Acc

kezúhta contracted

si she

hára hither

únsermo our-Dat

méze . measure-Dat

uuánda / si as she

humana human-Acc

áhtôt. regards

zu to

uuîlon sometimes

síh herself

Lat. ‘For now certainly she contracted herself to the common measure of men/mankind’. OHG ‘For a short time she contracted herself hither/down to our measure as she sometimes regards human things’. (21)

I 8,6ff. Nunc uero now but

uidebatur she seems

pulsare touch-Inf

cacumine / top-Abl

summi highest-Gen

uerticis. crown (of the head)-Gen

Ándera other-Acc

uuîla tûohta time-Acc seemed

den hímel rûoren. Art-Acc sky-Acc touch-Inf uuánda si as she

cĊlum . sky

si she

mít/ with

mír me

óbenahtigemo highest-Dat

hóubete head-Dat

astronomiam uuéiz. astronomy-Acc knows

Lat. ‘But now she seems the sky with the top of the crown of her head to touch’ OHG ‘The other time seemed she (to) me the sky with the crown of her head to touch as she astronomy knows’.

We find evidence of a different structure in the following sentence: (22)

I 10,25ff. At ego and I

cuius acies Rel-Gen eye

lacrimis . tears-Abl

nec not

quĊnam esset who was/would be auctoritatis . dignity-Gen

caliga- / rat . was dark

dinoscere recognize hĊc this

obstipui. I became silent

mersa immersed

possim . I could-Subj.Pres

muli-/er tam imperiosĊ woman so great-Gen

Translating information structure 341 Áber ut

íh I

erchám feared

míh tô myself there

dáz Art

uuîb uâre woman as/would be

uárentíu. acting-Prt

íh I

bechénnen . recognize-Inf

uuánda mír as me

fóllez full

sô so

nemáhta not could daz Art

dés . / uuér of-Gen who

geuuáltîgo powerful-Adv sia / she/her-Acc óuga eye

tímbereta . became dim

trâno tears

Lat. ‘And I, whose eyes dark of tears were dark with tears, so that I could not recognize who this woman of such great dignity was, became silent’. OHG ‘But I feared myself for the following reason: Who this woman was who acted so powerful(ly) couldn’t I recognize because my eye became dim full of tears.’

In Old High German the indirect interrogative clause uuér dáz uuîb uuâre sô geuuáltîgo uárentíu is placed before the matrix verb, whereas in Latin it appears after the matrix verb. And in place of the Latin relative clause cuius acies caligarat ‘whose eye was dark, immersed in tears’ after the ego at the head of the sentence, Notker uses a causal clause uuánda mír daz óuga tímbereta . fóllez trâno ‘because my eye became dim, full of tears’ at the end of the compound sentence. But this uuánda-sentence states the reason why Boethius didn’t recognize the woman. That is to say, the causal clause again refers to the propositional level of the matrix sentence and consists mainly of focal elements (only mír reverts to íh from the matrix sentence). The best New High German translation for the conjunction is “deswegen weil” because Notker uses the word dés in the cataphoric construction Áber íh erchám míh tô dés ‘But I became silent because of that’ which precedes the compound sentence. So Notker broke down this passage into its components and represented the train of thoughts step by step. Thus he arrives at a much more precise presentation of the logical structure, but at the same time at a different distribution of information structural units concerning the contents of the wanda-sentence: In Latin, the relative clause cuius acies caligarat ‘whose eye was dark’ serves to set the frame while the causal clause uuánda mír daz óuga tímbereta . fóllez trâno ‘because my eye became dim, full of tears’ in Old High German has focal function. From this pragmatic analysis it follows

342 Rosemarie Lühr that what we have here is a wide focus. This confirms data from language typology which suggest that in many languages that kind of focus is positioned on the right periphery of the sentence. Yet another case is the following: (23)

I 35, 20ff. Qui tum who then

credidit has trusted

larga large-Acc

negantibus refusing-Dat

sulcis . furrows-Dat

cum when

sydus sign of the zodiac

cancri cancer-Gen

elusus fide mock(s) faith-Abl

arbores. trees-Acc

Tér sô . he/who so

dô / when

hízza heat-Acc

pergat shall proceed

ad to

súnna sun

in cancro méistûn in cancer most-Acc

téta fílo caused much

sâta sowed

in in

ácher . field-Acc

uuánda as

únzît uuás . tér untime(ly) was he

bedíu therefore

chórnlôsêr without corn

únde and

déro of them-Gen

diu the

graue heavy

inĊstuat radiis phĊ- /bi rages rays-Abl Phoebus-Gen

cereris. Ceres-Gen

quernas oak-Acc

semina / ... seeds-Acc

iz it

ze to the

nére may nourish

hólz wood

únuuílligen unwilling-Acc gánge may go

éi-/chelôn . acorns collect-Inf

síh . himself

Lat. ‘Who, then trusted many seed to the reluctant furrows when heavy from the rays of the sun the sign of the cancer rages (with heat), he will go disappointed in his faith to Ceres to the oak trees.’ OHG ‘He who, when the sun in the sign of the cancer caused most heat, sowed much in the unwilling field. As it was untimely he may go therefore without corn to the wood to collect acorns and with these may nourish himself.’

The wanda-sentence here is also to be found in a commenting passage, but in this example it provides background information: uuánda iz únzît uuás ‘as it was untime(ly)’ names the heat described before as the cause for the failure of the seed to grow. And this sentence is placed before the matrix sentence.

Translating information structure 343

So, as far as information structure is concerned, Notker, on the one hand, uses newly added or rephrased causal clauses as focus material: The wanda-sentences connect propositions, they appear without or with a correlate and they constitute a wide focus on the right periphery. On the other hand, such wanda-sentences contain background information; in that case they precede the rest of the sentence.

3.3. Marked foci In the topic-comment and in the focus-background structure, focus and topic are the central terms. Therefore we have to look for linguistic devices which distinguish these two information structural entities in the two languages we are comparing here. The pertinent elements for the marking of foci are focus particles or emphasizing pronouns, the word order and instances of contrast.

3.3.1. Focus particles, emphatic pronouns Like the Latin quoque, the word auch [also/too] functions as a focus particle in Old High German. But while in Latin quoque follows the word it refers to, in Old High German auch precedes its word of reference30: (24)

I 40,25ff. Tu / quoque you too

si if

uis want

claro clear-Abl Úbe óuh if also

lumine. light-Abl tû uuél-/lêst you want

chîesen see

diu uuârhéit the-Acc truth-Acc

cernere to see

mít with

uerum truth

cláten clear

óugôn eyes

Lat./OHG ‘If you also want to see the truth with clear eyes’

In some instances, however, there is no equivalent of Old High German óuh in the Latin original. Admittedly, in the following example it is not quite clear whether we have a modal particle or a focus particle meaning “...self”31:

344 Rosemarie Lühr (25)

Poetry I 11,22ff. Et ui- / ctor and winner quĊcumque whatever

comprehensam condensed

habebat had

stella star

exercet traverses

numeris . numbers-Abl

ua- / gos cursus . wandering-Acc course-Acc

flexa per uarios orbes . curved through various circles-Acc Únde and

uuíssa er knew he

zálo . number-Dat tûot . does

óuh too tîe Art-Acc

feruuállotíu departing

tîe / uérte be Art-Acc courses after dehéin some-Nom

in in

planeta star

ánder-/ ro planetarum other-Gen planets-Gen

uérte . course-Acc

Lat. ‘And it [the spirit] had as winner condensed into numbers which star traverses the wandering courses which is curved through various circles’ OHG ‘And he knew also the courses by numbers which some star makes by departing into other planets’ courses’

In contrast, examples of evidence with a preceding sélb ‘self’ in combination with the definite article as a means of emphasis are undisputed: here, sélb comes close to the meaning of the focus particle NHG sogar ‘even’. In contrast to NHG, however, it is still flectional. So, for example, in the commentary passage below, the focussing sélb is the equivalent of Latin ipse ‘self’: (26)

II 131,17ff. Nam quid because what regum . kings-Gen

ego I cum as

demonstrem . may show

plena full-Acc

Uuáz what

íh I

tárf shall

disseram shall speak ipsa selves-Acc

tantĊ imbecillitatis? so much-Gen weakness-Gen ságen / fóne say of

dero chúningo . uuîo uuéih Art-Gen kings-Gen how weak sélben die selves Art-Acc

de familiaribus about household-Abl regna / kingdoms-Acc

tîe they

chúninga kings-Acc

dîen gesuâsôn Art-Dat relatives sîn . are geóuget shown

sîd as

íh I hábeo have

Translating information structure 345 sô so

uuéiche? weak-Acc

Lat. ‘Because what shall I speak about the household of the King when I show that the kingdoms [the reign of kings] (them)selves are full of weakness?’ OHG ‘What shall I say of the relatives of the kings, how weak they are, as I the kings themselves have shown so weak?/ ... as I have shown even the kings so weak?’

But sélb also occurs independent of an immediate Latin source. In a free translation of the Latin original, for example, we find: (27)

III 185,17ff. Uel re- / currat or may run back

circulum circle

micans twinkling

nox night

pingitur. is decorated

Álde or

hóhor higher

gestíge-/ nez . climbed

óuh also

hímel sky-Acc

astri . quocumque star-Gen wherever

sélben even

dén Art-Acc

erréiche. reach

Lat. ‘or it [the spirit] wanders through the circle of stars, wherever the twinkling night is decorated’ OHG ‘and also when it has climbed higher, even reached the sky ‘

It is mainly its use in the exclamatory sentence, which shows that Notker employed the focussing function of sélb quite deliberately – he is about to explain the usage of the Latin word vallum which has just been used: (28)

I 37,26f. Tér Art-Nom

zûn fence

die Art-Nom

bóuma hîezen trees-Nom were called

hîez was called

uallum . wall/rampart

sélben even

ualli ! entrenchment piles-Nom

‘The fence was called wall. Even the trees were called “ualli”, i.e. entrenchment piles!’

346 Rosemarie Lühr 3.3.2. Word order A special feature of OHG syntax is that subjects are placed late, e.g. at the very end of the sentence, if they introduce a new discourse referent which is taken up as topic in the following sentence. When these new discourse referents are introduced, they are in the focus domain of the sentence in which they are established. Ex. (29)–(30) illustrate that it is not only the position of the pre-verbal accusative aboutness topic which is maintained, but also that of the subject at the end of the sentence in the Old High German translation: (29)

I 9,7ff. Eandem this-Acc

tamen32 ue- / stem . yet garment

quorundam some-Gen

uiolentorum violent-Gen

manus. hands

Tîa sélbûn the same-Acc

uuât garment-Acc

hábe- / ton had

súmeliche some

nôt-núnftara. thug

sciderant had torn to pieces

ferbróchen torn

Lat. ‘This garment, however, had [been] torn to pieces [by] some violent hands.’ OHG ‘The self-same garment [by] some thugs had [by] torn.’ (30)

I 6,26f. inopina Uenit enim came namely unexpectedly

senectus old age

malis. / misfortune-Abl

uuánda mír for me

íst is

árbéiten / trouble

zûo geslúngen . happened

spûotîg fast

properata hastened úngeuuândo . unexpected

fóne of

álti. (old) age

Lat. ‘There came namely unexpectedly old age, hastened by misfortune.’ OHG ‘For to me unexpectedly because of pain came fast old age .’ [Old age, which comes fast, came to me unexpectedly because of trouble.]

The same word order pattern involving postverbal subjects can be also found in interrogative sentences:

Translating information structure 347 (31)

II 50,13ff. Nos alli- / gabit ad us binds to alienam . foreign-Acc Sól shall

míh me

getûon . make-Inf

constantiam . constancy-Acc

inexpleta insatiate

cu- / piditas greed

uuíder mînemo against my-Dat tero Art-Gen

nostris moribus our-Dat customs-Dat hominum? mankind-Gen

síte customs-Dat ménniskôn mankind-Gen

stâta / constant

úneruúlta unfulfilled

gîrhéit? (Näf 1979: 197) greed Lat. ‘Us shall to a constancy, which is foreign to our customs/character, bind the insatiate greed of mankind?’ [Shall the insatiate greed of mankind bind us to constancy which is foreign to our character?] OHG ‘Shall me against my customs constant make mankind’s unfulfilled greed?’ [Shall mankind’s unfulfilled greed make me constant/resistant against my usual customs?]

3.3.3. Cases of contrast

Contrast is an important means of structuring information for Notker and he uses it in various forms. In Latin, it is not only focussed accusative objects which are positioned right of the verb, but also focussed dative objects. Notker adopts this information structure but not the syntax: In place of the dative object he uses a directional compound construction: (32)

I 10,8ff. Hominumque men-Gen-and

mentes assuefaci- / unt senses they accustom

Únde ménniskôn and men-Gen sîe they

nelôsent not-release

mûot sense

morbo . non liberant. illness-Dat not they liberate

stôzent sie ín_ dia / súht. drive them in Art-Acc illness-Acc siæ them-Acc

nîeht. not

Lat. ‘And men’s senses accustom them to the illness, not liberate (them).’ OHG ‘And men’s senses drives them into the illness, they release them not.’

348 Rosemarie Lühr What we have here is the rhetorical figure disiunctio, a form of the isocolon: Coordinated sentences show a difference of meaning which is based on the negation of a positive term33. Thus we arrive at a contrast between the wide foci stôzent sie ín_dia súht and nelôsent siæ nîeht. Similarly, conjunctions can be used for describing contrasts: (33)

II 50,10ff. Ius est right is

mari . ocean-Dat

Ċquore . surface-Abl

nunc now

nunc now

blandiri stra- /to flatter-Inf smooth-Abl

inhorrescere shudder-Inf

procellis storms-Abl

ac and

óuh also

stílle calm

sîn . be

mít with

strûben síh soar-Inf itself

fóne from

uuínde . únde wind-Dat and

fluctibus. floods-Abl Ter Art

mé-/ re mûoz ocean must

sléhtero smooth-Dat

ébene ! surface-Dat

uuîlon sometimes

óuh also

fóne from

uuéllôn . waves-Dat

Lat. ‘The right of the ocean is it, now to flatter with (a) smooth surface, now to shudder with storms and floods.’ OHG ‘The ocean also has to be calm with smooth surface, sometime soar with the wind and the waves.’

In Old High German óuh – óuh are set parallel to each other to connect structures with antithetical meaning. Compare, for example: (34)

I 10,2ff. QuĊ non who not

modo only

dolores pains-Acc

eius. uerum his-Acc but

dulcibus sweet-Abl

uenenis. poison-Abl

Tîe who

sîn sêr nîeht his-Acc pain-Acc not

ímo him

nullis remediis no-Abl remedy-Abl insuper moreover

éin only

fouerent . would cure ale- / rent would nourish

nehéillent . not-cure

Translating information structure 349 núbe / but

ióh also

íro their-Gen

mêront . mít increase with

sûozemo sweet-Dat

éitere poison-Dat

uuórto. words-Gen

Lat. ‘Who not only with no remedy his pain can cure, but moreover want to would nourish him with sweet poison.’ OHG ‘Who his pain not only not cure, but even increase it with the sweet poison of their words.’

With the double conjunction non modo (...) verum insuper, OHG nîeht éin (...) núbe ióh ‘not only (...) but even/also’ after a negative statement, another – a contrasting – statement is emphasized. Through the negation of the first phrase, the contrastive parallel structures Lat. remediis foverent dolores (...) alerent dulcibus venenis, OHG sêr (...) héillent (...) mêront mít sûozemo éitere ío uuórto gain particular weight. Notker retained this structure in OHG because the combination of negation, contrast and parallelism makes it easier for his pupils to memorize the gist of his statements. In other cases, expressions of contrast in parallel structures in Old High German are placed in initial positions; positions, which in New High German are typical of a contrastive accent, i.e. I-topicalisation34. It is possible that these expressions already had this particular prosodic quality in Old High German. So, for example, in (35) the introductory phrase án dero zéseuuûn – án dero uuínsterûn describes frames which function as topics of contrast; therefore their pragmatic effect could have been a contrastive accent: (35)

I 9,13ff. Et and uero but

gestabat carried sceptrum. sceptre-Acc

Án/ in

dero zéseuuûn Art-Dat right-Dat

dero uuínsterûn Art-Dat left-Dat

quidem dextra libellos . now the right books-Acc

trûog si carried she

sinistra the left

bûoh..../ án book-Acc in

sceptrum ! sceptre-Acc

Lat. ‘And it carried now the right [hand] books, the left, however, a sceptre.’ OHG ‘In the right [hand] she carried books ... in the left a sceptre.’

350 Rosemarie Lühr Compare also to ze_níderost and ze_óberôst as frames in (36): (36)

a. I 8,27ff. ... Harum these-Gen legebatur was read Ze_ At

extremo / extreme-Abl ʌ ʌ

intextvm weaved into-Prt

níderost the bottom

kescríben written-Prt

in in

taz Art

án on

margine . margin-Abl grecum. Greek

dero uuâte . / stûont Art-Dat garment-Dat stood

chrîecheska Greek

p. ... / p

Lat. ‘At its extreme margin a woven-in Greek ʌ was [to be] read.’ OHG ‘At the bottom of the garment was written the Greek ʌ .’ b. In in Ze_ at

superiore upper-Abl

uero but

legebatur was read

ș. ș

óberôst stûont / theta. the top stood theta

Lat. ‘on the upper one, however, was [to be] read ș.’ OHG ‘at the top stood theta.’

In a similar way, the subjects taz chrîecheska p und theta at the end of the sentences express contrasts, that is to say alternatives from a set of comparable entities. Both, in Latin and in Old High German we are dealing a focus – as can also be seen from Latin uero ‘but’. However, it is not only in cases as these, but also in other contexts that the end of the sentence in Old High German is a position for a focus which forms a contrast with another expression of focus. What is added in (37), for example, are the contrastive frames êr and nû which Notker uses as parallels (unlike the use of quondam in the Latin passage): (37)

I 6,13ff. Qui peregi who I completed

quondam carmina formerly songs-Acc

florente flowering-Abl

studio . zeal-Abl ...flebilis ...lamentable

cogor I am forced

inire begin-Inf

mestos modos. sad-Acc tunes-Acc

Translating information structure 351 Íh-tir I-who

êr téta formerly did

máchôn nû make now

frôlichív / joyous-Acc

nôte perforce-Adv

sáng . songs-Acc

íh I

chára-sáng. dirges-Acc

Lat. ‘[me] who I completed formerly songs in flowering lamentabl(y) I am forced to begin sad tunes.’ OHG ‘I, who formerly made joyous songs I make now perforce dirges.’

A further example of the reinforcement of the meaning of contrastive expressions of Latin by parallel syntactic structures in Old High German is the following one: (38)

I 6,23ff. Gloria fe- / licis glory happy-Gen solantur they comfort

olim once

uiridisque blooming-Gen

iuuentĊ . youth-Gen

nunc35 mea fata . mesti now my-Acc fate-Acc sad-Gen

Êr uuâ- / ren formerly were

sie they

nû now

míh me-Acc

trôstent sie comfort they

gûollichi mînero ornament my-Gen

senis. old man-Gen iúgende youth-Gen

.

álten . mî- / nero old-Acc my-Gen

mísseskíhte. misfortune-Gen Lat. ‘The glory of formerly happy and blooming youth now comforts the fate of the sad old man.’ OHG ‘Formerly they were the ornament of my youth. Now they comfort me old (man) in my misfortune.’

In this example, Notker uses contrasts by building two main clauses, in which now the frames êr and nu (Lat. olim and nunc) function as contrastive topics at the head of the respective sentences. A further contrast consists in the contrast foci iúgende and álten. So Notker employs contrastive structures, which underline and emphasize the existing contrasts of the Latin original, in order to clarify the respective passage for didactic reasons.

352 Rosemarie Lühr 3.3.4. Left peripheral foci in Latin

In the Latin emphasized pronoun vos and in the left peripheral focus recte facere in (39) we could have a structure comparable to I-topicalisation: (39)

II 100,12ff. Uos autem you but

nisi recte facere nescitis . rightly act-Inf you don’t understand except

ad in front of

populares auras . / et inanes the people’s-Acc ears-Acc and inane-Acc

rumores. gossip-Acc Ír you

neuuéllent don’t want

úmbe / líuto for people’s-Gen úppigen empty-Acc

áber but

nîeht not

lób praise-Acc

réhto fáren . rightly act únde and

âne except

úmbe for

líument fame-Acc

Lat. ‘But you don’t understand (how) to act rightly, except in front of the ears of the people and inane gossip.’ OHG ‘But you don’t want to act rightly, except for the praise of the people and the empty fame.’

In contrast to Latin, in Old High German the finite verb appears before the infinitive (Bolli 1975: 167). But similar to Latin, the prepositional phrase âne úmbe líuto lób . únde úmbe úppigen líument has been extraposed.

3.4. Constructions with marked topic positions

In this context we have to look at word order and pronouns.

3.4.1. Word order The Old High German sentence in the following example deviates from the Latin original: (40)

I 10,22f. His ille this-Abl that

chorus multitude

increpitvs . / scolded

deiecit lowered

Translating information structure 353 humi to the ground Tô there

mestior rather sad

snífta níder lowered down

dáz Art

uultum. look-Acc sús so

erstóuta ge- / zuâhte. scolded flock

Lat. ‘That by this scolded multitude lowered rather sad the look to the ground [The multitude scolded by that, looked sadly to the ground].’ OHG ‘There lowered down (the look) the thus scolded flock.’

The Old High German sentence contains a given discourse referent after the adverbial dô and the finite verb. The adverbial dô in initial position can also be found in other Old High German texts, for example in the OHG Tatian translation. This pattern typically appears in contexts in which previously mentioned material does not function as the aboutness topic of the sentence36. The topic function of Latin ille chorus incretitvs or OHG dáz sús erstóuta ge is indicated by the deixis pronoun ille which Notker translates with a definite article.

3.4.2. Pronouns Even apart from examples as (40) above, we generally have to consider the various forms of pronouns in Latin when looking at their translations into Old High German. As a rule, subject pronouns are dropped in Latin when they represent continuous topics. The accusative object also frequently remains unexpressed due to object drop. The oblique cases are represented by the forms of the pronouns is, ea, id. While these anaphorically allow the implication of nouns in sentences further away, they can also have a demonstrative function and refer back to immediately preceding contents words or facts, just as the demonstrative pronoun hic, haec, hoc does. This means that the Latin pronouns is, ea, id are ambiguous. Notker, however, can disambiguate the reference by choosing a proper equivalent in OHG. After the disastrous influence of the sirens has been described, the discourse continues with the sentences quoted in (41): (41)

I 10,13ff. Nihil quippe not certainly Án at/of

in eo lede- / rentur shall be belittled in the

démo neinfûore the-Dat not may be lost

mír me

operĊ nostrĊ . troubles ours

nîeht mînero nothing (of) my-Gen

354 Rosemarie Lühr ár- / béito. troubles-Gen Lat. ‘Not, certainly, shall by this our troubles be belittled.’ OHG ‘Of this may not be lost any of my troubles.’

In Latin, the anaphoric in eo is placed on the right of the verb, Notker uses án démo (instead of the less emphatic dârána) and places at the beginning of the sentence. While Notker retains the position of the focal subject at the end of the sentence, he decided on a topic marking which deviates from the Latin one. In keeping with the position of the demonstrative pronoun dieser, the position is at the head of sentence. (42)

Poetry I 11,19ff. Hic quondam liber . assuetus aperto this one once free used open-Abl ire go-Inf

in in

Tíser uuás this one was

Ċtherios ethereal-Acc

meatus . ... paths-Acc

ke- / uuón used

dénchen think-Inf

cĊlo space-Abl

án of

die Art-Acc

hímel-férte … celestial paths-Acc Lat. ‘This one was once free(ly) used to go through the open space in ethereal paths....’ OHG ‘This one was used to think of the celestial paths ...’37

Of particular interest, however, is how Notker translates Latin relative pronouns used in a continuative function (relativischer Anschluss), for this structure is unfamiliar in German. In case a subordinate and a matrix clause share a common subject, it is expressed by a relative pronoun put before the conjunction in Latin. Notker does not retain this structure but rather substitutes the relative pronoun by a personal pronoun as the referent is taken up again due to topic promotion, see (43): (43)

I 8,8ff. QuĊ cum she when ipsum cĊlum. (it)self sky-Acc Sô when

si she

altivs / higher

extulisset caput . had raised head-Acc

etiam even

penetrabat. penetrated dáz hóubet Art-Acc head-Acc

hô / high

ûf erbúreta. raised up high

Translating information structure 355 sô so

úber slûog surmounted

iz it

ten Art-Acc

HÍmel . sky-Acc

Lat. ‘When she higher had raised the head, she even penetrated the sky.’ OHG ‘When she the head raised up high, it surmounted the sky.’

Sometimes he chooses a possessive pronoun for a substitute: (44)

I 8,23ff. Quarum speciem their-Gen appearance-Acc uetustatis age-Gen

obduxerat . covered

quĊdam neglectĊ / certain neglected-Gen

caligo mist

Íro bílde uuâ- / ren its-Gen appearance were

fóre of

álti uersáleuuet ... age-Dat darkened

Lat. ‘Their outer appearance had covered a certain, caused by neglected age, mist [a certain mist caused by neglected age]’ OHG ‘Its outer appearance by age was darkened ...’

More frequently we find a demonstrative pronoun as a substitute: (45)

I 9,1ff. … gradus in steps after …

léi- / ter-sprózen ... rungs of a ladder

Quibus esset on these would be superius higher-Acc Áfter on demo Art-Dat

modum / Art-Acc

scalarum. ladder-Gen álde or

ascensus . / ascent

stégon stûofa. steps of a ladder ab from

inferiori ad lower-Abl to

elementum. part-Acc

dîen these-Dat níderen lower-Dat

man stîgen one climb-Inf pûohstabe zu demo letter-Dat to Art-Dat

máhti . fó- / ne could from óberen. higher-Dat

Lat. ‘... steps in the manner of ladders. On these would an ascent from the lower to the higher part (be possible).’ OHG ‘... rungs or steps of a ladder. On these could one from the lower letter to the higher climb.’

356 Rosemarie Lühr But there are even more alternatives: In the following example, Notker seems to take the relative sentence connection with Lat. quas for a relative pronoun and places the noun phrase at the head of the sentence: (46)

I 8,20ff. Quas which-Acc Tîa Art-Acc

ipsa self uuât garment

texuerat she had woven sî she

íro her-Dat

manibus hands-Abl sélbiu self

suis . her-Abl uuórhta ! made

Lat. ‘Which (the garment) she herself with her own hand had woven.’ OHG ‘The garment she herself made!’

But as we can see from the exclamation mark, Notker changes this sentence into an exclamative sentence with verb-final position. As Lat. quas serves both as a relative and an interrogative pronoun, the sentence could be transformed into a type of sentence which contains an interrogative pronoun. As we expected, the translation of pronouns with topic function, especially the relative connection of sentences again demonstrates the sophistication of Notker’s translations.

4. Summary The first objective of our research was to answer the question of how Notker converted the information structure of Latin into Old High German. The result is a complex picture: Notker did not convert the hyperbaton, a typical linguistic phenomenon of Latin, into Old High German ((12) to (15)), but, where appropriate, he used other techniques to imitate the emphasis implied in the separation of elements which belong together (15). However, we do find evidence of rhetorical word order in Notker: We discussed exclamative sentences and causal sentences introduced with wanda. There is an increased use of exclamative sentences in the poetic parts of the Latin original. In these sentences, Notker sometimes changed the position of the verb (16), sometimes retained the word order of the original (17) and sometimes completely restructured the sentence, as in the instance of the pre-position of an all-focus sentence (18). Notker also transforms Latin declarative sentences into exclamative sentences ((19), (46)).

Translating information structure 357

As far as causal sentences are concerned, he frequently added them as an explanation of the text ((20), (21), (23)) or to make a passage more stringent (22). If they appear after the matrix sentence, they contain focus material and form a wide focus ((20), (21), (22)). If they are put in front, they provide background information (23). As topic and focus are the marked elements of the opposition in the topic-comment and the focus-background structure, all the linguistic means which mark these information structural entities are of particular interest for us. Concerning the focus, these are focus particles and emphasizing pronouns: OHG óuh, Lat. quoque ((24), (25)) or sélb, which Notker uses not only as an equivalent to Lat. ipse (26), but also independent of any Latin source ((27)–(28)). Word order also serves as a marker of focus as long as we have a subject as a new discourse referent. In this function it appears as an emphatic focus at the end of the sentence ((17), (29), (30), (31)). A special characteristic of Notker’s style as a translator is his incorporation of foci into contrastive structures38. We find the imitation of the rhetorical figure of disiunctio (14), the use of parallel conjunctions ((33)–(34)), but also examples which resemble the New High German Itopicalisation with two contrast topics at the beginning of the sentence and two contrast foci at the end of the sentence ((35)–(38)). In contrast to this, in Latin you also find foci on the left periphery (39). For marked topic constructions (ausgezeichnete Topik-Konstruktionen), word order is also relevant, as is suggested by sentences with covert initial position with adverbial dô (40).We also have to take into account the use of pronouns in the languages we are comparing here: the representation of Lat. ille by the definite article (40), the clarification of the ambiguous Lat. is by a demonstrative der at the head of the sentence (41), a position which is occupied by demonstrative pronouns in general (42), and the dissolution of the relative connection of sentences. For this construction, which does not exist in Old High German, Notker chooses various forms of translation, the personal pronoun (43), the possessive pronoun (44), and, most frequently, the demonstrative pronoun (45). Finally, since in Latin the relative pronoun and the interrogative pronoun can be identical, there is the option to transform a Latin declarative sentence with relative connection into an exclamative sentence (46). All in all, what we can definitely say about Notker’s representation of the information structure of the Latin original is, that he certainly had a firm grasp of its structures. First and foremost, this becomes obvious in cases where the allocation of the informational entity focus coincides in the two languages, but where Notker structures the foci independent of the

358 Rosemarie Lühr Latin source. If he doesn’t imitate the information structure, he has good reasons for doing so: it is either because of the basic differences between the two languages or because of his didactic purposes. For Notker’s primary aim always was to make the text comprehensibly to his pupils. The second objective of the research points beyond Notker; for from Notker’s handling of the information structure we can draw a conclusion for Old High German; i.e. that it is more than likely that there is a fundamental difference between the two languages compared here with regard to the positioning of the kinds of foci: Old High German: structural focus – verb – emphatic focus vs. Latin: verb – emphatic focus – structural focus ((13), (15)). In addition, our discussion shows that neither Latin nor Old High German is a ‘discourse configurational’ language.

Notes 1.

So far there is no general agreement as to whether Notker actually translated the metrical passages in a metrical form. In some passages, however, one can certainly detect a particular rhythm (Glauch 2000: 170). 2. Prince (1981); Frascarelli & Hinterhölzl (2007). 3. In the case of a “continuous topic” the topic of the preceding sentence is retained, whereas in the case of a “shifting topic” there is a change of topic (Speyer 2007). 4. Gundel (1988). The dialogue partners are already familiar with these topics. 5. Cf. Frey (2000: 138) following Reinhart (1981): “Topiks sind die Ausdrücke, über deren Referenten durch die Sätze Aussagen gemacht werden – Topik ist eine Kategorie des pragmatischen ‘Worüber’” [Topics are expressions about whose referents the sentences make statements – topic is a category of pragmatic ‘aboutness’]. 6. The discourse topic is about a new topic (Frey 2000; Späth 2005). 7. Contrastive topics can be found, for example, in the so-called I-topicalization. Cf. Lang and Umbach (2002). 8. There is verifiable evidence that Notker based this on a tract on grammatical and syntactical problems (in four parts) from St. Gall (Tax 1986: XXII). 9. Glauch (2000: 172). For Christian influences cf. Mohrmann ([1976] 1984: 302ff.). 10. The Old High German text was written around 1025; even though most of it was written by a scribe he was probably working under the aegis of Notker (Tax 1986: XXVI; XLIII). 11. A further particularity occurs in the compound or periphrastic tenses, the socalled conjugatio periphrastica of the type scripturus fui, eram, etc., and pas-

Translating information structure 359

12. 13. 14. 15. 16. 17.

18. 19. 20. 21. 22. 23. 24. 25.

sive verbal paraphrases of the type amatus fui [I have been loved], amatus fueram [I had been loved] (as opposed to the synthetically constructed active amavi, amaveram). These forms increase in the course of the development of Middle Latin and of Romance languages: From Vitruvius onwards, the est factus-type more and more prevails until in Vulgar Latin and in the Romance languages it completely replaces its rival factus est (Hofmann and Szantyr 1965: 405). Moreover, the victurus sum - and tradendus sum-types begin to rival the simple future; in addition, paraphrases with auxiliaries, modal verbs and aspect verbs with infinitive begin to spread; e.g. habeo, possum, volo, debeo, incipio (Hofmann and Szantyr 1965: 313ff.; Stotz 2004: 323ff.; 445f.). Furthermore, in Middle Latin the infinitive is also used after censere, putare, credere, iudicare, consentire, petere, rogare, permittere, timere, as well as in a final function in abiit manducare, or after facere, curare in the sense of “have something done; make someone do something” (Kindermann 1998: 42. For the use in Boethius cf. Dienelt 1942: 114ff. and 132). This construction is also known in German. For the following discussion of the information structure, however, the compound tenses are irrelevant. Dik 1995; for Ancient Greek see Matiü (2003: 578ff.) following Kiss (1995: 2001). Devine and Stephens (2006: 524). For potentially comparable structures in German cf. Krisch (1998: 373ff.). For particular aspects of syntax cf. now Stotz (2004). Hofmann and Szantyr (1965: 690). But as Krisch (1998: 353ff.) shows, the hyperbaton is a vernacular phenomenon even in Old Latin. While we find verb second, verb first and verb third position of the finite verb in the Old High German Isidor and in the Old High German Tatian, “scheinen ca. 200 Jahre nach der Isidor-Übersetzung und ca. 170 Jahre nach der TatianÜbersetzung alle Stellungsmöglichkeiten außer der Zweitstellung verloren gegangen zu sein” [except for verb second position, all the other potential positions seem to have been lost about 200 years after the Isidore translation and about 170 years after the Tatian translation] (Näf 1979: 146). Cf. also Borter (1982: 51ff.). Ms. Ċtatem. Näf (1979: 137 and 196). Cf. Umbach (2001). Cf. also Büring (1997); Chierchia (1995); Asher and Lascarides (1998) on the concept of aboutness. Cf. Devine and Stephens (2000) for the Greek language. “Sind sie es doch, die mit dem unfruchtbaren Dorngestrüpp der Leidenschaften die fruchtreiche Saat der Vernunft ersticken” (Gothein 1932: 9). Speyer (2003: 15).

360 Rosemarie Lühr 26. The reproduction of the position of Wackernagel-particles would result in other “non-German” structures. So Notker places the “light” pronoun tíz after the conjunction and the subject pronoun íh (Näf 1979:338ff.): I 7,20f. ipse. Haec dum mecum tacitus reputarem this-Acc when with me silent-Prt I would think self Únz íh tíz suîgendo in mîne- / mo when I this-Acc silently in my-Dat

mûote áhtota. mind-Dat thought over

OHG “Als ich dies schweigend in meiner Vorstellung überdachte” [When I this silently in my mind thought over] A word order such as: “Dies als ich schweigend in meiner Vorstellung überdachte” [This when I silently in my mind thought over] would certainly have been ungrammatical in Old High German. 27. But in Latin you can have pronouns between the verb and the structural focus: I 6,15f. scribenda. Ecce lacerê camenê dictant / mihi see sad muses dictate me (what) to write Síh see

no . now

léidege musê . lêrent sad muses teach

míh me

scrîben. write-Inf

Lat. ‘See, sad muses dictate me what to write’ OHG ‘See now, sad muses teach me to write’ Notker in contrast, chooses an AcI in unmarked word position. 28. Cf. also: Poetry I 11,11ff. Et / relicta propria luce ... tendit ire and left behind-Prt.Abl own-Abl light-Abl strives go-Inf externas / external-Acc Únde and demo Art-Dat

in in

tenebras darkness-Acc

uuîo gnôto / how urgent-Adv lîehte in light-Dat in

iz tánne it then dia Art-Acc

îlet . ûzer hurries out uínstri. darkness-Acc

Lat. ‚and (after) the own light been left behind, strives [the mind] to go out into the darkness’ OHG ‘And how urgently it then from the light into the darkness hurries.’ 29. Because of the verb final or verb late position, we can be certain here that we have a subordinate clause; cf. Lötscher (in this volume). 30. Cf. also I 14,15f.

Translating information structure 361 An part

ut that

tu you

quoque also

mecum with me

rea . accusation-Acc

agiteris may be moved

falsis criminationibus? false-Abl accusations-Abl Ínno . part

dáz that

léidúngôn. accusations-Dat

óuh too kemûot tortured

tû you

gescúldigotív . fóne accusations-Acc from

lúkkên false-Dat

/ uuérdêst . may be

Lat., OHG ‘Or that you, too, from false accusations may be tortured’. 31. For the difference between óuh/auch ‘also’ as a focus particle and as a modal particle in declarative sentences cf. Thurmair (1989: 155). 32. Here a pronoun has been separated from its word of reference by the interposition of tamen [however] in the Wackernagel position (Hofmann and Szantyr 1965: 398). Strictly speaking, we therefore do not really have a hyperbaton here. 33. Lausberg (1990: 114). 34. For the intonation pattern cf. ¥Otto ist ins \Kino gegangen (und ¥MarIA in die \Oper [¥Otto went to the \cinema (and ¥MarIA to the \opera] (Frey 2000: 149 with further reading). 35. In Latin, an adverb can be placed between the verb and the structural focus on the right hand side. Cf. the position of the adverb nunc [now] before mea fata. 36. Hinterhölzl and Petrova (2005); Donhauser, Solf and Zeige (2006: 9f.). 37. Cf. also: I 6,30f. uertice intempestiui cani. Funduntur pour crown of the head-Abl untimely hairs Fóne from

dîen / these-Dat

díngen things-Dat

grâuuên turn grey

íh I

ze_ únzite. at “untime“-Dat Lat. ‘There pour from the crown of my head the untimely hairs.’ OHG ‘From these things I turn grey before time.’ For word order in Latin cf. Menge (2000: 577). 38. For contrast in general cf. Molnár (2002).

362 Rosemarie Lühr References Abraham, Werner 1986 Word order in the middle field of the German sentence. In Topics, focus, and configuarionality, eds. Werner Abraham and Sjaak de Meij, 15–38. Amsterdam: John Benjamins. 1992 Clausal focus vs. discourse rhema in German: A programmatic view. Language and Kognition 2: 1–19. Asher, Nicholas and Alex Lascarides 1998 Bridging. Journal of Semantics 15: 83–113. Bolli, Ernst 1975 Die verbale Klammer bei Notker. Untersuchungen zur Wortstellung in der Boethius-Übersetzung. Berlin/New York: Mouton de Gruyter. Borter, Alfred 1982 Syntaktische Klammerbildung in Notkers Psalter. Berlin/New York: Mouton de Gruyter. Büring, Daniel 1997 The 59th Street Bridge accent. On the Meaning of Topic and Focus. Phil. Diss. Tübingen. Chierchia, Gennaro 1995 Dynamics of Meaning. Anaphora, Presupposition and the Theory of Grammar. Chicago/London: The University of Chicago Press. Devine, Andrew M. and Laurence D. Stephens 2000 Discontinuous Syntax. Hyperbaton in Greek. New York/Oxford: Oxford University Press. 2006 Latin Word Order: Structured Meaning and Information. Oxford: Oxford University Press. Dienelt, Karl 1942 Sprachliche Untersuchungen zu Boethius’ Consolatio philosophiae. Glotta 29: 98–128; 129–138. Dik, Helma 1995 Ancient Greek word order. Amsterdam: Gieben. Doherty, Monika 2002 Language Processing in Discours. A key to felicitous translation. London: Routledge. Donhauser, Karin, Michael Solf and Lars Erik Zeige 2006 Informationsstruktur und Diskursrelationen im Vergleich: Althochdeutsch – Altisländisch. In Grenzgänger. Festschrift zum 65. Geburtstag von Jurij Kusmenko, eds. Antje Hornscheidt, Kristina Kotcheva, Michael Rießler and Tomas Milosch, 73–90. Berlin: Nordeuropa-Institut.

Translating information structure 363 Frascarelli, Mara and Roland Hinterhölzl 2007 Types of Topics in German and Italian. In On Information Structure. Meaning and Form, eds. Susanne Winkler and Kerstin Schwabe, 87– 116. Amsterdam/Philadelphia: John Benjamins. Frey, Werner 2000 Über die syntaktische Position der Satztopiks im Deutschen. ZAS Papers in Linguistics 20: 137–172. Glauch, Sonja 2000 Die Martianus-Capella-Bearbeitung Notkers des Deutschen. Bd. I: Untersuchungen. Tübingen: Max Niemeyer. Gothein, Eberhard 1932 (transl.): Anicius Manlius Severinus Boethius: Trost der Philosophie. Berlin: Die Runde. Gruber, Joachim 1978 Kommentar zu Boethius ‚De consolatione philosophiae‘. (Texte und Kommentare 9). Berlin/New York: Mouton de Gruyter. Gundel, Jeanette K. 1988 Universals of topic-comment structure. In Studies in Syntactic Typology, eds. Michael Hammond, Edith A. Moravcsik and Jessica R. Wirth, 209–239. Amsterdam: John Benjamins. Hetland, Jorunn and Valéria Molnár 2001 Informationsstruktur und Reliefgebung. In Language typology and language universals: An international handbook, eds. Martin Haspelmath, Ekkehard König and Wulf Oesterreicher, 617–633. Berlin/New York: Mouton de Gruyter. Hinterhölzl, Roland and Svetlana Petrova 2005 Rhetorical Relations and Verb Placement in Early Germanic Languages. Evidence from the Old High German Tatian Translation (9th century). In Salience in Discourse. Multidisciplinary Approaches to Discourse, eds. Manfred Stede, Christian Chiarcos, Michael Grabski and Lunk Lagerwerf, 71–78. Amsterdam: Stichting, Münster: Nodus. Hofmann, Johann Baptist and Anton Szantyr 1965 Lateinische Syntax und Stilistik. München: Beck’sche Verlagsbuchhandlung (Handbuch der Altertumswissenschaft II, 2,2). Jacobs, Joachim 1986 The Syntax of Focus and Adverbials in German. In Topic, Focus, and Configurationality, eds. Werner Abraham and Sjaak de Meij, 103–128. Amsterdam: John Benjamins. Kindermann, Udo 1998 Einführung in die lateinische Literatur des mittelalterlichen Europa. Turnhout: Brepols-Verlag.

364 Rosemarie Lühr Kiss, Katalin É. 1995 Discourse configurational languages. Oxford: Oxford University Press. 2001 Discourse configurationality. In Language typology and language universals: An international handbook, Vol. 2, eds. Martin Haspelmath, Ekkehard König and Wulf Oesterreicher, 1442–1455. Berlin: Mouton de Gruyter. Krisch, Thomas 1998 Zum Hyperbaton in altindogermanischen Sprachen. In Sprache und Kultur der Indogermanen. Akten der X. Fachtagung der Indogermanischen Gesellschaft, Innsbruck, 22.–28.September 1996, ed. Wolfgang Meid, 351–384. Innsbruck: Institut für Sprachwissenschaft. Lang, Ewald and Carla Umbach 2002 Kontrast in der Grammatik: Spezifische Realisierungen und übergreifender Konnex. In Sprachtheoretische Grundlagen der Kognitionswissenschaft: Sprachliches und nichtsprachliches Wissen, ed. Anita Steube, special issue of Linguistische Arbeitsberichte 79: 145–186. Lausberg, Heinrich 1990 Elemente der literarischen Rhetorik. Eine Einführung für Studierende der klassischen, romanischen, englischen und deutschen Philologie. 10München: Max Hueber Verlag Lötscher, Andreas this vol. Verb placement and information structure in the OHG Gospel Harmony by Otfrid von Weissenburg. In New Approaches to Word Order Variation and Word Order Change, eds. Roland Hinterhölzl and Svetlana Petrova. Berlin: Mouton de Gruyter. Lühr, Rosemarie 2007 Information Structure in Ancient Greek. In The discourse potential of unspecified structures, ed. Anita Steube, 487–512. Berlin: Mouton de Gruyter. Matiü, Dejan 2003 Topic, focus, and discourse structure. Ancient Greek Word Order. In Studies in Language 27 (3): 573–633. Menge, Hermann 1961 Repetitorium der lateinischen Syntax und Stilistik. München: Max Hueber. 2000 Lehrbuch der lateinischen Syntax und Semantik. Völlig neu bearb. von Thorsten Burkard und Markus Schauer. Darmstadt: Wissenschaftliche Buchgemeinschaft.

Translating information structure 365 Mohrmann, Christiane 1984 Some remarks on the language of Boethius ‘consolatio philosophiae’. In Boethius, eds. Manfred Fuhrmann and Joachim Gruber, 302–310. Darmstadt: Wissenschaftliche Buchgesellschaft (Wege der Forschung CDLXXXIII). Molnár, Valéria 2002 Contrast from a contrastive perspective. In Information Structure in a cross-linguistic perspective, ed. Hilde Hasselgård, 147–162. Amsterdam/New York: Rodopi. Näf, Anton 1979 Die Wortstellung in Notkers Consolatio. Untersuchungen zur Syntax und Übersetzungstechnik. Berlin/New York: Walter de Gruyter. Notker der Deutsche 1986 Boethius: De consolatione Philosophiae. Buch I/II. Hg. von Petrus W. Tax. Neue Ausgabe. Tübingen: Niemeyer. Petrova, Svetlana and Michael Solf this vol. On the Methods of Information-Structural Analysis of Texts from Historical Corpora. A Case Study on the OHG Tatian. In New Approaches to Word Order Variation and Word Order Change, eds. Roland Hinterhölzl and Svetlana Petrova. Berlin: Mouton de Gruyter. Prince, Ellen F. 1981 Toward a Taxonomy of Given-New Information. In Radical Pragmatics, ed. Peter Cole, 223–255. New York: Academic Press. Reinhart, Tanya 1981 Pragmatics and Linguistics: An Analysis of Sentence Topics. In Philosophica 27: 53–94. Rubenbauer, Hans, Johann B. Hofmann and Rolf Heine 1977 Lateinische Grammatik. Bamberg: Buchners Verlag. Schröbler, Ingeborg 1953 Notker III. von St. Gallen als Übersetzer und Kommentator von Boethius’ De Consolatione Philosophiae. Tübingen (Heraema N.F. 2). Späth, Andreas 2005 Die linke Satzperipherie und ihr semantischer Beitrag zur Diskurseinbettung. Ms Leipzig. Speyer, Augustin 2003 A Prosodic Factor for the Decline in Topicalisation in English. Ms. Philadelphia. 2007 Die Bedeutung der Centering Theory für Fragen der Vorfeldbesetzung im Deutschen. Zeitschrift für Sprachwissenschaft 26: 83–115.

366 Rosemarie Lühr Steube, Anita and Andreas Späth 2002 Semantik, Informationsstruktur und grammatische Modularität. In Sprachtheoretische Grundlagen der Kognitionswissenschaft: Sprachliches und nichtsprachliches Wissen, ed. Anita Steube, special issue of Linguistische Arbeitsberichte 79: 235–254. Stotz, Peter 2004 Handbuch zur lateinischen Sprache des Mittelalters. Band 4: Formenlehre, Syntax und Stilistik: München: Beck’sche Verlagsbuchhandlung. Tax, Petrus 1986 cf. Notker der Deutsche. Thurmair, Maria 1989 Modalpartikeln und ihre Kombinationen. Tübingen: Max Niemeyer (Linguistische Arbeiten 223). Umbach, Carla 2001 (De)accenting definite descriptions. Theoretical Linguistics 27: 251– 280. Wackernagel, Jacob 1892 Über ein Gesetz der indogermanischen Wortstellung. Indogermanische Forschungen 1: 333–436.

Aspects of word order and information structure in Old Saxon Sonja Linde

1. Introduction Current research on the development of word order in the Germanic languages suffers from the lack of a consistent description of the situation in Old Saxon (henceforth OS) which offers one of the earliest writing traditions of the Germanic group at all and includes various types of word order patterns. Standard descriptions of OS contain very little or no information about word order. While e.g. the grammar book by Gallée (ed. Tiefenbach 1993) only refers to issues of phonology and inflectional morphology, Behaghel’s (1897) comprehensive monograph mainly reflects the syntactic behaviour of different word classes with respect to their property to organize word groups and phrases but lacks any observations on the most intriguing question with respect to our investigation, the description of the linear order of constituents in the clause. In the preface to his monograph, Behaghel himself addresses the incompleteness of his book with respect to word order, cf. Behaghel (1897: vi). A notable exception in this respect is the survey of OS provided by Rauch (1992) who pays attention to the sentence structure in OS records. Rauch (1992: 24–31) accounts for a high degree of variation in word order and establishes a distinction between marked and unmarked patterns on the basis of their relative frequency in the records. In the case of independent declarative sentences, she argues that the (X)VSO order with all arguments following the verb and an optional single non-argument preceding it displays the most common pattern which therefore is defined as unmarked, while other patterns like (X)SVO or (X)SOV are said to be less frequent and therefore marked (1992: 24). In accounting for the functional differences among marked and unmarked patterns, Rauch addresses the issue of their textual role and pragmatic value in the discourse. E.g., she explains the pattern in which the verb precedes all arguments as a device of “concatenation or continuation in a narrative sequence” (Rauch 1992: 30). As shall become clear later, this is a major consideration in the approach pur-

368

Sonja Linde

sued in the present study as well. However, the analysis provided by Rauch raises a couple of questions. First, she discusses early verb positioning in the pattern (X)VSO accounted by her as unmarked on the basis of the example gaf it is iungarum forð ‘[he] gave it [=the bread] further to his disciples’ (Rauch 1992: 28–30). Admittedly, this is an elliptical coordinate conjunct which has no overt subject itself but shares the subject of the previous conjunct expressed in the phrase manno drohtin ‘the Lord of men’. Hence, this example is no suitable representative of the (X)VSO pattern, which really abounds in OS. Second, Rauch explains the initial placement of the verb gaf ‘[he] gave’ in relation to Behaghel’s Rule 2 according to which old information precedes new one in a sentence. In Rauch’s analysis of the example quoted, the verb denotes one of a series of expected actions and therefore carries of old information. At the same time, she seems to overlook that the old information is actually conveyed in the subject manno drohtin ‘the Lord of men’ dropped in this sentence and that the verb merely introduces the new information assigned to it. All in all, Rauch’s discussion on the matter is very brief. It neither results in an attempt to provide a general model of the underlying syntactic structure of the OS sentence nor does it systematically explore all factors leading to variation in the linear order of sentence constituents. Moreover, the classification of marked vs. unmarked order in OS given by Rauch (1992) contradicts the observations made approximately one hundred years earlier by Ries (1880: 5–11). In his view, the basic or regular word order pattern in OS involves the preverbal position of the subject in (S…V…) while the pattern in which the subject (including all arguments) follows the verb (…VS...) is regarded to be secondary, and derived from the basic order for special rhetorical and textual purposes. On closer look, it turns out that the pattern viewed as ‘unmarked’ in Rauch’s terms corresponds to what Ries counts as a secondary, or occasional pattern, and vice versa. Ries builds his proposal on language-internal as well as genealogical considerations. First, he accounts for the fact that the reverse pattern is typical for interrogatives, requests, exclamations etc., i.e. for sentences that bear a special, occasional usage with respect to simple declaratives. Therefore, it is not likely that the syntactic patterns occurring in those functionally ‘marked’ cases should present the basic order in the language. He further points at the fact that the preverbal position of the subject is the basic one in the Indo-European proto-language and a common property of all ancestors except of Celtic which was the only one to generalize verb-first as its basic order. Above all, Ries (1880: 11) explicitly refers to the high frequency of sentences with the so-called reverse order

Aspects of word order and information structure in Old Saxon

369

but argues that quantity of occurrences alone is not decisive for the identification of the basic word order in the system of a language. Apart from these contradictory views, the elaborations in Ries (1880) and Rauch (1992) share one important property: they both promote the idea that word order in OS – although exposing a variety of competing word order patterns – is by no means random but obeys specific rules and principles influencing the positioning of sentence constituents, and especially the placement of the finite verb. They further relate these principles to the broad field of pragmatics and discourse organization, though a more precise characterization of the functions and categories taking primary influence on the syntactic realization of sentence constituents is still missing. In more recent times, a small investigation on the issues of word order in OS appeared aiming to evince that the principles of word order in OS are explainable on purely syntactic grounds, cf. Erickson (1997). This account is nested within the ‘Government and Binding’ model of generative theory and shows that methods and concepts provided for the explanation of the syntactic regularities in other early Germanic languages like Lenerz (1984) for Old High German or van Kemenade (1987) for Old English are adaptable to the situation in OS. Following this approach, OS exhibits a base-generated SOV order retained in embedded clauses introduced by an overt complementizer whereas independent clauses display a variety of features related to a generalized V2-pattern like in modern German or Dutch. The aim of the present study is to account for both views on OS syntax in more detail and to trace back the influences of both syntactically as well as pragmatically based factors for the explanation of the word order patterns attested in OS records. The study is based on the observation of a large amount of empirical data that exceed the examples discussed in the previous literature. The results of the investigation give proof that both types of governing principles – discourse-configurational as well as purely syntactic ones – can be identified as motivating word order in OS. These observations allow for the assumption that OS is a language in which two competing systems govern the linear ordering of constituents.

2. The Source OS is attested in two major texts, Heliand and Genesis, which are both alliterative rhyme poems dated back to the 9th century. Besides, there are some prose texts, e.g. some liturgical texts or tax lists (Heberegister). Un-

370

Sonja Linde

fortunately, the prose texts are of a rather small size, and research on OS is based on the two major texts of the Heliand and the Genesis. Of these, the Heliand is the more extensive one by far including about 6000 long verses. Therefore, it was taken as the basis of this investigation. The OS Heliand was handed down in two major manuscripts, C (Ms. Cotton. Calig. A. VII, London, British Library) and M (Cgm 25, München, Bayerische Staatsbibliothek) as well as in four fragments. The text itself is subdivided into 72 chapters called ‘fits’ which in the original manuscripts are indicated by initial capital letters. In manuscript C, fits are additionally enumerated by Roman numbers.1 Quite often, the beginnings of smaller episodes or periods show initial capitalization as well. It is clear that the Heliand manuscripts expose a sophisticated system of capitalization according to which initial capitals function as indicators of the text organization provided by the writer. Unfortunately, none of the current editions of Heliand reflect these potentially valuable text structuring devices but represent the graphical layout of the manuscripts normalized according to modern German usage. Since no complete facsimile edition and no diplomatic edition exist, evidence on the formal means of text organization provided in the manuscripts is not available to us in full extent. The OS data are taken from the old but rather well-accepted Heliand edition by Sievers (1878). It reflects both major manuscripts in parallel columns and presents the original writing in glosses in case the edition differs from the manuscript. In terms of punctuation and capitalization, Sievers applies the rules of modern German usage. All examples cited are taken from manuscript M according to Sievers (1878) and from manuscript C only if there is no equivalent in M.

3. Properties of V2 in OS? Well-known generative work (e.g. cf. Lenerz (1984) for Old High German, van Kemenade (1987) for Old English) assumes that the early Germanic languages expose a base-generated verb final order. In independent clauses, the inflected verb is raised regulary to a vacant C0-position, and additionally, the SpecC-position is usually filled thus yielding a generalized V2-pattern like in modern German and Dutch. A similar view has been proposed by Erickson (1997) for OS as well. In fact, the data presented in his investigation strongly support this view. In this section, we shall investigate in more detail the assumed V2-properties of OS and discuss some further points concerning this question.

Aspects of word order and information structure in Old Saxon

371

A great part of the observable data indeed prompts the view that OS displays some typical features of a SVO- language. In dependent clauses like complement clauses (1) or relative clauses (2) the finite verb appears at the end of the sentence following all its arguments: (1)

[so uuas thero liudio thau = ‘so it was the custom with the people’] that that erlo gihuulic obean scolde COMP that-ACC noblemen-GEN everyone celebrate should (Hel 2732) ‘(…) that everyone of the noblemen should celebrate that (i.e. the birthday of the monarch)’

(2)

[buide imu be theru brudi = ‘He was living with the woman’] thiu er sines broder uuas the-FEM before his-GEN brother-GEN was (Hel 2706) ‘(…) who had been his brother’s bride before.’

As an exception, PPs and CP-complements are often extraposed, cf. the adjunct phrase an Galilæo land in (3): (3)

tho he im mid is gesidon giuuet eft an COMP he he-DAT with his companions went back to Galilæo land [for im (…)] (Hel 2290) Galilee land went he-DAT ‘When he came back to Galilee with his company, he went …’

In root clauses, on the other hand, the finite verb occurs in the second position like in modern V2-languages, where it is typical that the specifier position of the C-domain is occupied by any constituent regardless of its grammatical relation to the governing verb or to its information-structural value in the context. This property of V2-languages may be traced back in OS data as well. In (4), the constituent in front of the verb is the pronominal subject siu ‘she’. In case that an object or adjunct is preposed, the subject remains post-verbally. This is shown in (5) and (6) accordingly where the pronominal subject follows the verb:

372

Sonja Linde

(4)

siu uuelde tho ira geba egan she wanted then her gift have ‘She wanted to have her reward’

(Hel 2769)

(5)

that scoldun sea fiori thuo fingran that-ACC should they four then fingers-DAT

scriban write (Hel 32)

‘These four were to write it down with their fingers’ (6)

so helde he thea haltun man so healed he the lame men ‘so healed the lame people’

(Hel 2357)

Furthermore, the preverbal position may host a syntactic operator as the wh-phrase in a direct question (7), the aboutness-topic of an utterance (8) as well as newly added information, i.e. focal material (9): (7)

huui uueldes thu thinera modar manno liobosto gisidon how wanted you your mother man dearest do sulica sorga (Hel 821) such worries ‘How could you cause your mother such worries, dearest sun?’

(8)

thea liudi stodun umbi that helaga hus (Hel 101) the-PL people-PL stood-PL around the holy house ‘The people stood around the temple’

(9)

Gabriel bium ic hetan Gabriel am I called ‘I am called Gabriel’

(Hel 120)

However, the straightforward analysis of OS as a V2-language faces a number of methodological and empirical problems which shall be discussed briefly below. One of them concerns the proper differentiation of root vs. embedded clauses in the data from OS. Bearing in mind that the system of complementizers is not fully grammaticalized at that period in the sense that subordinating conjunctions are not distinguishable from adverbials and pronouns in each case, a high number of sentences can be interpreted both as root and dependent clauses. This mainly concerns some adverbial clauses (10) as well as relative clauses (11):

Aspects of word order and information structure in Old Saxon

373

(10)

uuarun im an Nazarethburg thar the neriondio were they-DAT in Nazareth there the saving Krist uuohs undar them uuerode (Hel 782) Christ grew among the people ‘They were in Nazareth, where the saving Christ grew up among the people.’ ‘They were in Nazareth. The saving Christ grew up there among the people.’

(11)

Sie ni uueldun it thoh farlaten ac hetun thar they NEG wanted-PL it yet leave-off but ordered-PL there ledien ford en uuif for themu uuerode guide away a woman in front of the people thiu habde uuam gefrumid (Hel 3840) the-FEM sin sin comitted ‘They did not want to give up but instead they ordered to bring a woman in front of the crowd, who had comitted sin.’ ‘They did not want to give up but instead they ordered to bring a woman in front of the crowd. She had comitted sin.’

Synchronically, the position of the finite verb cannot be used to distinguish root from subordinated clauses. There are clauses which are unambigliously subordinated ones but exhibit post-verbal arguments like subjects (12) or objects (13): (12)

[thit sculun gi uuitan alle iungaron mine = ‘You shall know it all, my disciples’] huand iu fargeben habad uualdand thesaro uueroldes because you forgiven had ruler this-GENworld-GEN (Hel 2434) ‘(…) because the ruler of the world had forgiven you’

(13)

[habda them heriscipie herta gisterkid = ‘He strengthened the heart of their army’] that sie habdon bithuungana thiedo gihuilica (Hel 55) COMP they had defeated folk every ‘(…) so that they had conquered every nation.’

374

Sonja Linde

Moreover, matrix clauses show much more variation with respect to verb placement than is expected in a pure V2-language. In OS, the verb is often found in initial position (14) as well as in later positions (15)2–(16): (14)

habda im the engil godes al giuuisid had them the angel god-GEN everything shown torhtun tecnun (Hel 427) bright-DAT.PL signs-DAT.PL ‘The angel of God had shown them everything with bright signs’

(15)

thuo hie sia an is era antfeng thuru hluttran then he her in his protection took through clear hugi (Hel 5619) mind ‘Then he took her under his protection due to his clear mind.’

(16)

nu ik theses thinges getruon now I this-GENthing-GEN trust ‘Now I trust in this matter’

(Hel 285)

Patterns like (14)–(16) are frequent in the Heliand. Therefore, it seems unjustified to exclude them from the analysis of OS syntax as isolated exceptions picturing archaisms3. The attested data induce most notably two questions: First, how can we explain this confusing coexistence of different word order patterns syntactically, and second, how are the constituents in patterns organized which do not display V2 in the surface. According to Kiparskys (1995) well-known analysis, Old Germanic syntax just started to establish the category of CP. As a consequence, the shift from adjunction to embedding by means of subordinating comple0 mentizers in C took place and so did V-to-C-movement in main clauses as well. Since this syntactic change is supposed to proceed in steps, early attested data may still show variations between CP and S main clauses as reflexes of this development. In fact, the differences in Old Saxon word order shown above can be best explained by assuming this variation, this means that the C-projection is not fully established yet; and therefore V-toC-movement did not occur in all cases. Bearing in mind an analysis like this, still the question arises how to explain the order of elements in non-V2 main clauses since it seems to be dissatisfying just to notice that the finite verb could not move to the C0 slot due to the fact that these clauses have no CP at all. Furthermore, still an

Aspects of word order and information structure in Old Saxon

375

asymmetry between V1 and V-late main clauses occurs. Although in particular V1 clauses can be explained as cases of V-to-C-movement with an empty focus-operator in SpecC4, I want to investigate alternatively whether V1 and V-late in independent declaratives may be attributed to information-structural conditions instead of confining on syntactical factors only. Taking as a starting point some influential ideas in early philological work on the function of different word patterns (Ries 1880, Delbrück 1911) and some more recent approaches on this matter (cf. Donhauser and Hinterhölzl 2003), I will pursue an approach to discerning pragmatic or information-structural factors which account for variation in OS verb placement.

4. A discourse-pragmatic analysis 4.1.

Some basic assumptions

Recent work on the role of information structure and discourse organization for the explanation of different word order patterns in Old High German (Hinterhölzl and Petrova 2005) reveals some regular correlations between pragmatic factors and verb placement in early Germanic languages. In this paper I will show that OS word order also depends to a large extent on pragmatic rules which can be consistently described in terms of information structure. Information structure itself is a complex phenomenon which comprises at least two aspects concerning the organization of an utterance. On the one hand, the utterance can be divided into a topic and a comment about this topic (topic-comment structure), on the other hand, the most relevant information in the utterance makes up the focus as opposed to the background (focus-background structure) (Lambrecht 1994: 117ff and 206ff).5 The topic-comment structure affects the predicational seperations of an utterance and is deeply connected with the concept of aboutness in the sense of Hockett (1958) and Reinhart (1981) or with the notion of familiarity in the sense of Gundel (1988). Usually, the topic expression refers to an entity which is given or presupposed in the particular context. By contrast the level of focus-background is defined in terms of the speaker’s attitude towards the informational relevance of sentence parts. Novelty often correlates with the referent of the focus expression, but also presupposed elements can be emphasized under certain conditions. The pragmatic categories of topic and focus should not be understood as complementary ones. However, there are conditions which prevent the

376

Sonja Linde

establishment of a topic-comment structure in an utterance as is the case in so-called thetic sentences (Drubig 1992, Sasse 1987). On the level of focus-background structure these sentences are analysed as all-focus structures, i.e. the focus domain comprises the entire utterance (cf. Lambrecht 1994: 137ff and 233ff). For the identification of the pragmatic categories of topic and focus in a historical text, features that are grammatical correlates of these categories are most important. So e.g. topics are assumed to be expressed by anaphoric means while focus correlates with prosodic prominence which in the data analysed here may be detected by virtue of rhyme, rhythm and stress. Furthermore, factors of discourse organization play an important role in word order regularities. I refer to the basic assumptions in two of the most accepted models of discourse analysis, the Rhetorical Structure Theory (RST) by Mann and Thompson (1988) and the Segmented Discourse Representation Theory (SDRT) by Asher and Lascarides (2003)6. As a basic principle both share the assumption that utterances in discourse hold certain logical relations among each other yielding a hierarchical structure in the discourse. Although the catalogue of discourse relations is subject to extended discussion, two main types of relation can be broadly distinguished: (i) (ii)

An utterance can hold a dependency relation to a previous one in supplying more information on it; Two utterances can display no dependency relation among each other, i.e. they belong to the same level of discourse organization.

According to Asher and Lascarides the first type of relation is called subordination prototypically instanciated by ‘elaboration’, while the second one is called coordination with ‘narration’ as its prototypical representative. Each structural unit within a discourse is associated either to coordination or to subordination. The types of discourse relations are assumed to be linked to formal correlates such as prosody, tense usage, anaphoricity, linear word order etc. which are governed by parametrical variation. Comparing the discourse function and the information structure of an utterance, one can realize that these correlate in some important points. Based on the definition of the subordinating type of linking, it should be evident that elaboration involves a topic-comment structure. By contrast to that, sentences of the coordinative type help to develop the main line of the narration. Following these considerations, one can assume that sentences of the coordinative type open a new sequence or signal a change in the

Aspects of word order and information structure in Old Saxon

377

narrative setting but do not necessarily display a topic-comment structure. In other words, they share properties of thetic judgements. In a recent application of this model to data from OHG Hinterhölzl and Petrova (t.a.) and Petrova (2006) point at a correlation between the word order and the discourse function of a sentence. Their suggestion especially concerns the placement of the finite verb. To summarize, verb first patterns are said to appear in sentences opening a new text sequence or denoting a change in the narrative setting, thus correlating with coordinating discourse relations. On the other hand, the verb second pattern occurs in contexts elaborating on an already established discourse referent and is therefore prototypical for subordinating discourse relations. In the following section, this model will be applied to OS in order to look whether properties of discourse organization correlate with verb placement in OS as well.

4.2.

Subordinating Discourse Relations in the OS Heliand

Here, we investigate sentences which provide more information on a discourse-given or presupposed referent in supplying additional information on it. Due to their pragmatic nature, we expect these sentences to expose a topic comment division, i.e. to qualify as categorical judgements. In sentences establishing such kind of relations, the V2 pattern dominates. The preverbal constituent in this case is a referential expression conveyed in different forms such as full NPs (17), pronouns (18) and elliptical material (19). These correlate with the topic referent: (17)

[The three Magi enter the place of the birth of Christ] uurekion] fellun te them kinde an kneobeda TOP[thea the-PL strangers fell-PL to the child on knees (Hel 671) ‘The strangers fell on their knees in front of the child.’

(18)

[An old woman named Anna came into the temple] [siu] habde ira drohtine uuel githionod te thanca TOP she had her Lord well served to gratitude (Hel 505) ‘She had served her Lord well in gratitude.’

378 (19)

Sonja Linde

[‘There was an old man’] hugi [that uuas fruod gomoi] TOP[proi] habda ferehtan this was wise man had experienced mind [proi] uuas fan them liudeon Leuias cunnes (Hel 73) TOP was from the people Levi-GEN dynasty-GEN ‘He was a wise man and had much experience and was descended from the dynasty of Levy.’

In most of these cases, the topic is the grammatical subject, however, an object which functions as the aboutness topic is also found in the preverbal position (20): (20)

[‘His name was Simon’] habda giuuisid uualdandas craft langa TOP[Im] led rulers power long-ACC him-DAT had huila (…) (Hel 469) time-ACC ‘The power of the ruler had led him for a long time (…)’

The pattern in (17)–(20) is a prototypical topic comment structure with the topic in the left periphery and the comment following it. The structure Top+V2 is therefore firmly associated with the type of subordinating linking. Syntactically expressed these OS data show instances of V-to-C-movement and SpecC filled by a topic expression for discourse internal reasons. However, besides V2-clauses with a topic expression in the preverbal position, we find a number of main clauses establishing a subordinating discourse relation to the previous context which exhibit a word order which is different from the one described above. More precisely, these sentences involve more than one constituent to the left of the finite verb, see (21)7 and (22).8 (21)

endi siu an iro breostun forstod and she in her heart understood ‘In her heart she realized’

(Hel 292)

(22)

thar ina thiu modar fand sittean under them there him-ACC the mother found sit-INF among the gisidea (Hel 818) crowd ‘The mother found him there sitting in their company.’

Aspects of word order and information structure in Old Saxon

379

In (21) and (22) both arguments and adjuncts are placed in front of the finite verb. Looking at the informational status of the preverbal elements, we notice that they are clearly arranged according to their pragmatic status9. In both sentences, one of the preverbal constituents is the aboutness topic of the utterance, cf. siu in (21) and ina in (22). The remaining constituents left to the verb are also contextually given elements (e.g. thar, thiu modar in (22)) or elements related to already established entities (an iro breoston in (21)). In other words, all preverbal elements share the property of being presupposed. Therefore they constitute the background of the utterance. This is in clear contrast to the information represented by the verb, which is new. This leads to the assumption that in all verb-later cases above, the finite verb seperates background and focus material in the utterance.10 Further evidence for a postverbal position of focus is provided by the syntactic placement of new referents (23) or given entities which are highlighted for reasons of contrast or emphasis (24): (23)

fan adrom Thar imu tegegnes quam new[en idis there him-DAT towards came a woman from other thiodun] (Hel 2984) people ‘There a woman from another region came towards him.’

(24)

endi thar an them uuiha afstod emphasis[mahtig barn and there at the temple stayed powerful child godes] (Hel 797) god- GEN ‘And the mighty son of God stayed in the temple.’

It is important to note that in most of the verb-late main clauses discussed above an adverbial is placed in the initial position. These adverbials bear referential properties as indications for time, place and manner. Therefore they function as frame elements belonging to the background of the utterance. To sum up, syntactic structures occurring in sentences of the subordinating type and showing a verb-late pattern can be represented in the following scheme given in (25). (25)

Background – Vfin – Focus

380

Sonja Linde

The background domain is structured in the way that the frame elements precede the aboutness topic and this is followed by the remaining presupposed elements if any are given. From the perspective of pragmatic principles motivating the syntactic placement of sentence constituents in OS, we can also provide an explanation for structures being ambiguous between root and embedded clauses, cf. (10)–(11) above. Both sentences serve to provide more information on a previously established referent, i.e. they are prototypical cases of subordination in discourse. Quite consequently, they expose the pattern in (25). In this sense, it is not appropiate to speak about parataxis and hypotaxis in our modern understanding of these terms; instead, utterances in OS often obey the rule of organizing the material according to pragmatic principles. Since patterns with the finite verb in a later position than the second one are too frequent to be exceptions and some rules concerning their contextual appearance as well as their linear word order can be identified, one can assume that they reflect an older stage in which CP was not developed yet so that V-to-C-movement cannot take place; the finite verb is expected to remain in its base position. Above we showed that the elements in these verb-late patterns are ordered with regard to their information structural status rather than exclusive of their syntactical status, although word order is surely based on it. Focus elements are assumed to be extraposed. Even though we stated that these structures are frequent, one cannot ignore the fact that the V2-pattern is the more common one in sentences of the subordinated type of discourse relations. Here the preverbal constituent represents the aboutness topic regularly. This is consistent to the fact that subordinate discourse relations add more information about an already established discourse referent, which functions then as the aboutness topic of the utterance. Therefore these patterns depict prototypical topiccomment strutures. All in all, these observations lead to following shemes representing patterns occurring in sentences of the subordinating type: (26)

a. TOP=BG[NP] COM=FOC[Vfin …] b. BG[(Frame)[Top](XP ...)] FOC[Vfin(XP)]

However, if two utterly different patterns, namely a prototypically V2structure involving V-to-C-movement (26a) and a rather pragmatic organized structure analysed as some kind of relict without CP (26b), constitute subordinating discourse relations, the question arises how their discourse internal counterpart – the coordinated discourse relations – is structured.

Aspects of word order and information structure in Old Saxon

4.3.

381

Coordinating Discourse Relations in the OS Heliand

Also in instances belonging to the coordinating type of linking two particular patterns occur. Here, structures in which an adverbial particle precedes the verb in second position (27) and V1-structures (28) prevail. Note that all sentences have an overt subject in postverbal position: (27)

Tho uuard that heuencuninges bodon then became the-ACC heaven.king-GEN messenger-ACC harm an is mode (Hel 159) sorrow-NOM in his heart ‘Then the messenger of the king of heaven felt sorrow in his heart.’

(28)

Bidun allan dag that uuerod for them wait-PRT.PL all day the crowd in front of the uuiha (Hel 174) temple ‘The people waited all day long in front of the temple.’

Both sentences establish a new situation which is settled on the main line of narration. For all that I can tell they seem to be functional equivalents. The sentence in (27) opens a new fit. In an analogous way, (28) opens a new sequence involving a change of place and participant: whereas the preceding context informed us on the events concerning Zacharias inside the temple, the narration switches to a characterization of the people waiting outside. As to the quantitative distribution of these patterns, cases of tho+V2 occur more than twice as often as V1 in this respect. Next to the adverbial particle tho, we find V2 clauses associated to coordinating discourse relations with other adverbial elements in the initial position such as so (29): (29)

So uurogdun ina mid uuordun uuerod Iudeono so accused him-ACC with words crowd Jews-GENc thurh hotean hugi (Hel 5245–46) through full of hatred mind ‘So the crowd of Jews accused him with words full of hatred thoughts.’

Furthermore, there is a group of verbs which inherently have an eventreporting status and focus the entire proposition. As instances of such

382

Sonja Linde

predicates we interpret verba dicendi11, motion verbs and verbs of sensual perception, the latter especially in inchoative meaning. Hinterhölzl and Petrova (t.a.) observe that in Old High German these verbs regularly occur in V1 or tho+V2-patterns respectively. It is argued that these lexemes signal a change in the overall deictic orientation of the situation with respect to place, time and participant. Therefore, instances with these verbs are analysed as typical cases of all-focus sentences inducing a change in the narrative situation. In OS, the situation is quite similar in sentences with verba dicendi (30), motion verbs (31) or perception verbs (32): (30)

tho het he sie an thana sid faran then told he them to that way go ‘He then instructed them to go on their journey.’

(31)

quamun managa Iudeon an came-PL many Jews into ‘Many Jews came to the great hall.’

(32)

uuissun that thoh managa liudi aftar them landa knew-PL that yet many people in the land (Hel 855) ‘But many people all over this country knew that.’

thene the

(Hel 637)

gastseli (Hel 2736) hall

The placement of the verb in the initial position to highlight the entire content of the sentence is assumed to be common to all Germanic languages (Fourquet 1974: 316; cf. also Ries 1880). In other words that the verb opens a maximal focus domain constitutes a very old pattern. Surface-V1-declaratives allow for two types of explanations. On the one hand, one may assume that they derive from general V-to-C-movement with a SpecC slot remaining empty or hosting a silent focus operator, and synchronically nothing contradicts this point of view. On the other hand, some functional contexts were analysed where verb movement to the left periphery seems to have taken place much earlier than overall V-to-Cmovement. In general, imperatives, negated sentences and wh-interrogatives are closely associated to verb fronting in Old Germanic (see e.g. Eyþórsson (1996: 111) and Axel (2007: 52f)). In this paper we do not want to discuss if CP itself or the movement of the finite verb to C0 is optional in Old Germanic. However, it is crucial for our aim to perceive verb fronting under certain circumstances as mentioned above since all of them are related to the purpose of focusing the proposition of the utterance.

Aspects of word order and information structure in Old Saxon

383

V1-patterns in OS Heliand always focus the entire utterance. For all we know structures with a fronted verb associated to focus operations are much older than general V-to-C-movement with additional XP-fronting. In accordance with V-late structures discussed above V1 declaratives seem to represent some kind of reflex of a prehistorical stage of language. On the other hand, tho+V2 corresponds to the need for a lexically filled SpecC slot in declarative main clauses like in modern V2- languages. According to this, all-focus structures in OS are most often found with the adverbial tho, which gradually looses its temporal semantics and develops to a text-structuring discourse-particle. It signals the continuation of the narration and guarantees progress in discourse. The structure tho+V2 is the most wide-spread verb second pattern in the Heliand by far. Besides tho as a focus particle we also find tho functioning as a referential temporal adverbial element in the middlefield which can be topicalised optionally. Therefore, the patterns (i) [[thoParticle] FOC[Vfin ...]] and (ii) BG[thoFrame] FOC[Vfin ...] cannot always be distinguished with certainty. Less frequently, other adverbs as so, nu etc. also function as discourse particles in initial position of an all-focus V2-sentence, cf. (29) above. However, it is interesting to note, that besides adverbs, also the subject can occupy the initial position of V2 sentences associated with coordinating discourse relations. In this case, the subject represents the focus exponent of the sentence, as is illustrated in (33). There are several criteria supporting this analysis: 1) the expression is stressed by virtue of its position in the metrical structure of the verse line; 2) the expression bears the alliteration rhyme (alliterative syllables are given in bold type for clarity); 3) the expression conveys new (or relevant) information. All these criteria apply for the initial element Iohannes in (33); the sentence is analysed as representing a coordinating relation because it introduces a new episode within on the main line of narration: (33)

[A chain of five coordinate conjuncts with V1, which promote the progress of narration as an intermediate sequence] geng thes geres gital Iohannes quam went that-GEN year-GEN quickly John came an liudeo lioht lik uuas im sconi to mankind-GEN light body was him beautiful (Hel 198) ‘The year passed by quickly. John came to the light of mankind. He was beautiful.’

384

Sonja Linde

However, while V2-patterns with an initial discourse particle, especially tho, are extremely frequent, patterns like the one in (33) occur very rarely. Summarizing our findings concerning sentences attributed to the coordinating type of linking we can conclude that the following patterns appear in the OS Heliand in this function: (34)

a. b. b.’

[Vfin ...] [Particle] FOC[Vfin ...] [ [NP] [Vfin ...]] FOC Fexp FOC

As SpecC is surely sensitive to focus elements, it does not surprise that these can be found in this position. The quantitative distribution of these patterns indicates that the verb second pattern with an initial focus particle (34b) exceeds by far. Bearing in mind the overall development of word order in continental WestGermanic, one can assume that the pattern in (34b) replaces the one in (34a).

5. Conclusion It is assumed that the confusing variation of linear word order patterns in Old Saxon Heliand is due to the mixture of two different grammars; however, our text reflexes “the elimination of the variation of CP and S main clauses” (Kiparsky 1995: 162) whereas a good part of main clauses construed as simple S are synchronically reanalysed as CP with a vacant SpecC slot, i.e. V1-declaratives. This point of view is supported by the well established assumption that language change in general takes place gradually and therefore reflexes of an older stage are still found later in changed environments with the possibility of reanalysing “old” properties with regard to the new conditions. The scenario supposed carries out as follows: In a prehistoric language period, say to simplify matters in Protogermanic, only simple S clauses occur. Here the finite verb appears mostly in its base position following all arguments. Moreover, subordinate clauses are adjoined, and linear word order is not only related to the syntactic properties of constituents but also to information structural factors. Background elements are organized preverbally due to their salience and focus expressions are extraposed (see as

Aspects of word order and information structure in Old Saxon

385

a reflex of this protogermanic property the postverbal focus position in OS, e.g. (12), most likely (13), (23) and (24)). Furthermore, in order to focus the whole utterance V1 appears connected to an overt or covert focusoperator triggering verb raising and bearing one of the following features [+wh], [+imperative], eventually [+negation] and, for opening new contents in a discourse, [+coordinative]. This leads to the assumption of a leftperipheral operator-position. However, once CP is established in Germanic, this operator position is reanalysed as SpecC which then opens successively for other elements than focus-operators only; first and foremost SpecC allures the aboutness topic which is predestinated for this position due to the fact that it occurred always preverbally representing the prototypical background element. And in fact, the presumption that SpecC becomes sensitive for any XP gradually is confirmed by the observation that other elements occupying the SpecC slot except for the so-called discourse particle tho and the aboutness topic are extremely rarly found in Old Saxon. All in all, the OS Heliand mirrors a period of syntactic change; this is the establishment of CP in general associated to operations like overall Vto-C-movement, embedding of subordinate clauses in alliance with the development of the category of complementizers and – chronologically later – the generalisation of SpecC hosting any constituent.

6. Acknowledgements The present paper relates to the information-structural analysis provided for my 2006 master thesis ‘Verbstellung und Informationsstruktur im Altsächsischen’, Humboldt Universtität Berlin. The thesis was supervised by the research project SFB 632/B4 ‘Die Rolle der Informationsstruktur bei der Herausbildung von Wortstellungsregularitäten im Germanischen’. I am thankful to the principle investigators of this project Karin Donhauser and Roland Hinterhölzl for giving me the chance to complete my research on OS. I also thank Michael Solf for helpful comments concerning this paper. I am grateful to Svetlana Petrova for help and fruitful discussions during writing this paper.

386

Sonja Linde

Notes 1.

Cf. Bästlein (1991: 213ff). As Bästlein points out, the distribution of initial capitals differs in the single manuscripts. Bästlein challenges the traditional editorial usage to subdivide the Heliand into fits according C only, because only this youngest manuscript structures the text in this way. 2. (15) cannot be analysed as a temporal subclause as it is clearly divided from the previous sentence and is continued as follows: (…) so im is herro gibod ‘as his master told him’. 3. Eyþórsson (1996: 116) describes OHG Vlate declaratives as “isolated (…) examples reflect[ing] a more archaic stage of Germanic” and therefore excludes them from his analysis. OS he states to be regularly V2. 4. E.g. Eyþórsson (1996: 123) considers this solution. 5. For a more detailed discussion see also Petrova and Solf (in this vol.). 6. I follow the model presented in Hinterhölzl and Petrova (t.a.) 7. (21) is definitely not conjoined to a subordinate clause. 8. See also (15) an (16) above. 9. One can be inclined to explain the order of the preverbal elements e.g. in (22) as a realisation of Behaghel’s law of placing longer constituents after shorter ones (Behaghel 1932: 6). However, we think that principles of linear order stated by Behaghel are a correlate of pragmatic distinctions: in the cited OS data the preverbal elements are clearly ordered with respect to their salience. 10. Again, so-called long constituents like PP-complements may be extraposed independently of their information structural status. 11. See also Önnerfors (1997: 120ff).

Primary Text [Hel]

Heliand. Titelauflage vermehrt um das Prager Fragment des Heliand und die Vaticanischen Fragmente von Heliand und Genesis. Ed. by E. Sievers, Paderborn 1878.

References Asher, Nicholas and Alex Lascarides 2003 Logics of conservation. Cambridge: Cambridge University Press. Axel, Katrin 2007 Studies on Old High German Syntax: Left Sentence Periphery, Verb Placement and Verb-Second. Amsterdam/Philadelphia: John Benjamins.

Aspects of word order and information structure in Old Saxon

387

Bästlein, Ulf Christian 1991 Gliederungsinitialen in frühmittelalterlichen Epenhandschriften: Studie zur Problematik ihres Auftretens, ihrer Entwicklung und Funktion in lateinischen und volkssprachlichen Texten der Karolinger- und Ottonenzeit. Frankfurt a.M.: Peter Lang. Behaghel, Otto 1897 Die Syntax des Heliand. Wien: Tempsky. 1932 Deutsche Syntax. Band IV. Heidelberg: Winter. Delbrück, Berthold 1911 Germanische Syntax II: Zur Stellung des Verbums. Leipzig: Karl Teubner. Donhauser, Karin and Roland Hinterhölzl 2003 Die Rolle der Informationsstruktur bei der Herausstellung von Wortstellungsregularitäten im Germanischen. In Informationsstruktur: Die sprachlichen Mittel der Gliederung von Äußerung, Satz und Text. Finanzierungsantrag 2003–2007, ed. Caroline Féry, 173–193. Potsdam: Universitätsverlag. Drubig, Bernhard 1992 Zur Frage der grammatischen Repräsentationen thetischer und kategorischer Sätze. In Informationsstruktur und Grammatik. Linguistische Berichte, Sonderheft 4, ed. Joachim Jacobs, 142–195. Opladen: Westdeutscher Verlag. Erickson, Jon 1997 Some observations on word order in Old Saxon. In Sprache im Fokus, eds. Christa Dürscheid, Karl Heinz Ramers and Monika Schwarz, 95–105. Tübingen: Max Niemeyer. Eyþórsson, Þórhallur 1996 Functional Categories, Cliticization, and Verb Movement in the Early Germanic Languages. In Studies in Comparative Germanic Syntax. Vol. II, eds. Höskuldur Thráinsson, Samuel D. Epstein and Steve Peter, 109–139. Dordrecht/Boston/London: Kluwer Academic Publishers. Fourquet, Jean 1974 Genetische Betrachtungen über den deutschen Satzbau. In Studien zur deutschen Sprache und Literatur des Mittelalters, eds. Werner Besch, Günther Jungbluth, Gerhard Maissburger and Eberhard Nellmann, 314–323. Berlin: Erich Schmidt. Gallée, Johan Hendrik 1993 Altsächsische Grammatik. Tübingen: Max Niemeyer. Gundel, Jeanette 1988 Universals of topic-comment structure. In Studies in Syntactic Typology, eds. Michael Hammond, Edith A. Moravcsik and Jessica R. Wirth, 209–239. Amsterdam: John Benjamins.

388

Sonja Linde

Hinterhölzl, Roland and Svetlana Petrova to app. Rhetorical Relations and Verb Placement in Old High German Tatian. In Salience: Multi-disciplinary perspectives on its function in discourse, eds. Christian Chiarcos, Berry Claus and Michael Grabski. Berlin: de Gruyter. Hockett, Charles F. 1958 A Course in Modern Linguistics. New York: Mac Millan. Karttunen, Lauri 1976 Discourse Referents. In Syntax and Semantics 7: Notes from the Linguistic Underground, ed. James McCawley, 363–385. New York: Academic Press. Kemenade, Ans van 1987 Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris. Kiparsky, Paul 1995 Indo-European Origins of Germanic Syntax. In Clause Structure and Language Change, eds. Adrian Battye and Ian Roberts, 140–169. New York/Oxford: Oxford University Press. Lakoff, George 1976 Pronouns and Reference. In Syntax and Semantics 7: Notes from the Linguistic Underground, ed. James McCawley, 275–335. New York: Academic Press. Lambrecht, Knud 1994 Information Structure and Sentence Form: Topic, Focus and the Mental Representation of Discourse Referents. Cambridge: Cambridge University Press. Lenerz, Jürgen 1984 Syntaktischer Wandel und Grammatiktheorie. Tübingen: Max Niemeyer. Mann, William and Sandra Thompson 1988 Rhetorical Structure Theory: Toward a functional theory of text organization. Text 8 (3): 243–281. Önnerfors, Olaf 1997 Verb-erst-Deklarativsätze: Grammatik und Pragmatik. Stockholm: Alqvist & Wiksell. Petrova, Svetlana and Michael Solf this vol. On the Methods of Information-Structural Analysis in Historical Texts: A Case Study on the OHG Tatian. Rauch, Irmengard 1992 The Old Saxon Language. New York: Peter Lang. Reinhart, Tanya 1981 Pragmatics and Linguistics: An Analysis of Sentence Topics. Philosophica 27: 53–94.

Aspects of word order and information structure in Old Saxon Ries, John 1880

389

Die Stellung von Subject und Prädicatsverbum im Hêliand. Strassburg/London: Karl J. Trübner. Sasse, Hans-Jürgen 1987 The thetic/categorical distinction revisited. Linguistics 25: 511–580.

Subject index

alliteration, 383 analogy, 136, 146, 194, 199 anaphoric use, 60, 263 Behaghel’s Law (Gesetz der wachsenden Glieder), 51–53, 58 clause type, 7, 194, 253 adverbial clause, 153, 228, 230, 327, 372 complement clause, 26, 228, 230, 252, 371 wanda-clause, 323, 339, 341–343, 356 copular construction, 173, 224, 242, 255, 266–267, 273 demonstrative pronoun, 60, 62, 73, 91, 97, 99–101, 104–107, 111– 113, 126, 145, 216, 228–229, 309, 324, 353–355, 357 directionality, 194 discourse adverb, 8, 91, 94, 96–98, 112–113 discourse marker, 109, 112, 310, 311, 317 fronting of, 7, 20, 22, 28, 306, 311 discourse particle, 105, 112–113, 383–385 discourse relation, 228, 376–378, 380–381, 383 discourse status given (discourse-old), 2, 8–10, 21–22, 48, 51, 59–61, 73, 83, 91, 96–98, 100–103, 108, 133–141, 145–148, 152, 200, 204, 238, 240, 243, 252, 258–260, 263, 265, 268–270, 284–287, 298, 306, 311, 325, 332, 353, 375, 377, 379–380

new (information), 8, 10, 21, 48, 51–52, 59, 62, 67, 72–75, 76, 78– 79, 83, 96, 100–101, 133–136, 139, 141–142, 145–147, 149–153, 238–241, 243, 258, 260, 265–274, 284, 298, 327, 332, 368, 379, 383 double base hypothesis (DBH), 4, 46, 49 ellipsis, 8, 96, 191–197, 200–208, 210–212 expletive es, 19, 30, 33–34, 36 focus ambiguity, 274 focus background structure, 332, 343, 357, 375–376 focus type, 271–274 contrastive focus, 52, 141, 149, 150, 153, 239, 243, 273, 276, 332 emphatic focus, 323, 325, 335– 337, 357–358 narrow focus, 73–74, 86, 263, 271 new information focus (presentational focus), 52, 149– 150, 266, 268–270, 273–274 sentence focus (thetic judgement), 24, 152, 302, 342, 356, 367, 377, 382–383 structural focus, 52, 323, 325, 329, 333, 335–337, 358 VP-focus, 149, 264, 269, 274 gapping backward gapping, 8, 191, 193– 194, 198–200, 202–203 forward gapping, 8, 191, 193– 194, 196, 198–200, 204 grammar competition, 45

392

Subject index

head parameter, 46 hyperbaton, 11, 323, 325–329, 333, 335–336, 356 information-structural category (IScategory), 6–9, 11, 45, 49, 51, 58, 64, 132–133, 138, 142, 144–145, 154, 264 background, 3, 9, 51–52, 122, 133, 140, 144, 148, 238, 243, 245–246, 252, 261, 263, 265, 268–269, 273, 283–286, 299, 304–306, 309, 314, 324, 332, 339, 342, 343, 357, 375–376, 379–380, 384–385 contrast, 10, 21–22, 52, 73, 141, 149, 252, 258, 272, 274, 309, 343, 347–351, 357, 375, 379 focus, 3–7, 20, 23, 48, 51, 58, 67, 73–74, 86, 97–98, 132–133, 140– 142, 144, 148–153, 200, 224–225, 235, 238–246, 252, 259, 261–264, 268–274, 284–286, 310, 323–325, 327–329, 331–333, 336–368, 342–347, 350, 352, 357, 361, 375–376, 379–385 topic, 3–7, 22, 30, 63, 73, 111– 112, 137–140, 199–200, 269, 284–287, 302–307, 309–312, 314–315, 317, 323–325, 327, 329, 332–333, 338–339, 343, 346, 349, 352–354, 356–357, 375–380, 383 implicature, 192–193, 197, 287 language change, 1, 63, 67, 74, 78– 81, 87, 122, 315, 384 change in the external language, 86 change in the internal language, 79 language (stage)s, Early Middle English (EME), 4, 45, 62–63, 94, 113

Early New High German (ENHG), 61, 72, 274 Greek, 8, 121, 165, 191, 196–198, 200–201, 207–209, 212, 328 Hittite, 8, 191, 194–195, 197–198, 200, 202, 209–210 Latin, 8–11, 17–19, 48, 121–131, 161, 163–164, 191, 195, 198, 200, 203–205, 211, 227, 229, 253, 325–329, 331–339 Middle English (ME), 4, 8, 62, 67, 83, 91, 94, 96–97, 99, 112–113, 223, 276 Modern German, 10, 285, 295, 300–301, 332, 337, 341, 344, 349, 357 Modern Icelandic, 24, 68–71, 82– 83, 86–87 Old English (OE), 3–5, 7–8, 46– 47, 51, 54, 62–63, 67, 71–73, 91– 92, 94, 96–99, 101, 109–113, 121, 143, 165, 223, 251–252, 267, 316, 369–370 Old(er) Icelandic (OI), 2, 4–5, 7, 46, 51, 62, 67–72, 74–75, 78, 82– 84, 86–87 Old High German (OHG), 2–5, 17–18, 46, 51–52, 58, 122–127, 142–144, 148, 154, 161, 164–178, 180–181, 211, 223–229, 237, 244, 251–268, 270–274, 281–283, 302– 304, 315–316, 323–326, 332, 335–343, 348–353, 356–358, 369–370, 375–377, 382 Old Saxon (OS), 2, 63, 122, 143, 181, 367–372, 374–378, 380–385 Sanskrit, 8, 191, 197–198, 200, 202, 208 Yiddish, 52, 72, 273–274 Latin influence, 5, 123, 227, 229, 230 deviation from Latin, 125, 226– 227, 229, 235 syntactic loan, 124, 131

Subject index left dislocation, 137, 147, 300 left periphery, 5, 6, 7, 11, 82, 106, 111, 113, 325, 357, 378, 382 multiple foci, 270, 273–274 Notker’s Anlautgesetz, 181 null object, 203, 208–212 null subject (pro), 26, 28–29, 32, 295 object deletion (see null object) object shift, 63, 82–83, 83, 86–87 operator movement, 19, 22, 36 palaeographic clue, 8 accent diacritics, 8, 143, 165–176, 179, 181 capitalization, 141, 143, 152, 180, 181, 370 (text) initials, 180–181 punctuation, 161, 180–181, 370 word separation, 8, 161–162, 177–179, 181 phase impenetrability condition (PIC), 57 phase, 53, 57–58, 61, 63 phonological phrase, 50, 56 prosodic constraint (prosodic condition), 52–53, 57–58, 61–62 prosody, 1, 6, 8, 63, 143–144, 168, 170, 180–181, 376 quotation syntax, 223, 227–230, 235, 241, 244 Rahmenbildung (see sentence bracket) reconstruction, 198, 208, 211 remnant movement, 71 rhythmic activation principle, 50 right periphery, 7, 10, 251–252, 268, 270, 273, 342–343

393

Satzklammer (see sentence bracket) scrambling, 8, 59, 70, 72, 73, 107, 108, 109, 110, 226, 260 sentence bracket, 5, 7, 10, 31, 62, 223, 317, 330–331, 336 exbraciation (see extraposition) extraposition, 5, 10, 32, 45, 48– 49, 224, 235, 245–246, 251–252, 268, 270–271, 273, 297–299, 331, 338 sloppy identity, 197, 201–202 specificity, 59, 72–73, 106–108, 273 stylistic fronting, 7, 19, 22–25, 29– 30, 35–37, 68, 82 subject position, 32, 103–104, 106– 107, 111 texts Heliand, 11, 143, 181, 369–370 374, 377, 381, 383–385 Isidor, 7, 10, 17–19, 22, 123–125, 162, 224–227, 238, 240, 246, 252, 315–316, 318 Notker’s Consolatio, 7, 9–11, 17, 21, 27, 60, 62, 123, 143, 164 – 168, 170–171, 175, 181, 282, 293, 317, 323–326, 330–333, 335–339, 341, 343, 345, 347, 349, 350–351, 353–354, 356–358 Otfrid’s Gospel Harmony (Evangelienbuch), 10, 60, 123, 125, 143, 164, 166, 168–172, 175–176, 180–181, 281–283, 285, 287–288, 290, 293, 296–297, 299, 301–302, 304, 306, 310, 312–313, 315–318 Tatian, 9–10, 17–18, 20–22, 27, 33, 35, 48, 51, 123–127, 131, 143–144, 150, 154, 166, 171–173, 175– 176, 178–181, 226–228, 252–253, 256, 258, 264, 271, 273, 353

394

Subject index

topicalization, 57–58, 68, 102, 139, 317 topic comment structure, 133, 138, 147, 284, 324, 339, 375, 376, 377 topic type aboutness topic, 21, 140, 151, 306, 333, 346, 353, 372, 378–380, 385 continuous topic, 325, 339, 353 contrastive topic, 314, 325, 351 discourse topic, 34, 325 empty topic (topic drop), 33–34 familiar topic, 138–140, 146, 325, 375 frame setting topic, 140, 341 hanging topic, 137, 147 shifting topic, 325 typological field, Mittelfeld (middle field), 8, 10, 22–26, 29, 50, 54, 56, 58, 63, 226, 295, 297, 327 Nachfeld (see exbraciation, extraposition) Vorfeld (prefield), 17–36, 138, 147, 299–301, 303–305, 307–308, 310–312, 370, 375, 378, 382–385 universal base hypothesis (UBH), 50, 52 verb movement, 67, 70, 82, 111, 382 verb projection raising (VPR), 236, 237, 251, 255, 273 verb raising (VR), 251, 255, 273, 385 V-to-C-movement, 8, 19, 191, 200, 374–375, 378, 380, 382–383

verb placement residual verb second, 3 verb early, 223, 229, 230, 231, 243, 244 verb first (V1), 19, 26–36, 227, 281, 291, 293, 299, 301–302, 310, 312, 368, 375, 377, 381–385 verb last (Vend), 253, 255–256, 259, 263–264, 273, 293–294, 312, 313 verb late, 296–297, 312–313, 360, 379–380 verb second, (V2), 17–18, 21, 24, 26–30, 32–37, 45, 63, 101, 223– 224, 228–229, 256, 281, 293, 299–302, 310, 312, 317, 330, 337, 369–372, 374, 377–378, 380–384 verb third (V3), 17, 224, 290 VP-intraposition, 54 VP-topicalization, 57 Wackernagel-position, 256, 294–297, 303, 305–306, 312–315, 318 word order, basic, 2, 5, 45, 79, 212, 369 change, 2, 6–7, 45, 50–51, 58, 62– 64, 91 mixed, 8, 46, 49, 52, 71–72 OV order, 2, 8, 45, 68, 71–72, 74, 80–83, 252, 273, 369 unmarked, 7–8, 50–51, 58, 60, 63, 331 variation, 1, 2, 6–10, 45–46, 49, 51, 58, 121–122, 153, 224, 251– 252 VO order, 2, 4–5, 8, 45–46, 49, 68, 71–72, 252, 328