Grammaticalization Second Edition
Thi s is a general introduction to grammatical ization, the change whereby lexical t...
1030 downloads
8408 Views
6MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
Grammaticalization Second Edition
Thi s is a general introduction to grammatical ization, the change whereby lexical terms and constructions come in certain l i nguistic contexts to serve grammatical functions, and, once grammatical i zed, continue to develop new gra m matical fu nc tions. Thus nouns and verbs may change over time i nto gra mmatical elements such as case markers, sentence connectives, and auxi l iaries. The authors synthe size work from several areas of l i nguistics, i ncluding h istorical l i nguistics. dis course analysis, and pragmatics. Data are drawn from many languages i ncluding Ewe, Finnish, French, Hindi, H ittite, Japanese, Malay, and especially Engl ish. This second edi tion has been thoroughly revised with substantial u pdates on theoretical and methodological i ssues that have arisen in the decade si nce the fi rst edition, and includes a signi ficantly expanded pibliography. Particular at tention is paid to recent debates over directionality in change and the role of gra mmatical ization in creolization. Grammaticalization will be a valuable and stimu lati ng textbook for all l inguists i nterested i n the development of grammatical forms. Readers in anthropology and psychology will also appreciate the insights it offers i nto the i n teraction of language and structure and use. PAUL J. HOPPER
is Pau l M ellon Di stingui shed Professor of H umanities at Carnegie Mellon University. His publications incl ude Gralllmaticalization (co authored with Eli zabeth Closs Traugott, Cambridge, 1 993), A Short Course in Grammar ( 1 999), The Limits of Gra/llmaticalization (co-edi ted with Anna Giacalone-Ramat, 1998), and Freqllency alld the Emergence of Linguistic Structure (co-edited with Joan Bybee, 200 1 ). ELIZABETH CLOSS TRAUGOTT
is Professor of Linguistics and Engl ish at Stanford University. Her publications i ncl udeA History of English SYlltax ( 1 972), Linguistics for Students of Literature (co-authored with Mary L. Pratt, 1 980), Grammaticalization (co-authored with Paul J. Hopper, Cambridge, 1 993), and Regularity in Semantic Change (co-authored with Richard B. Dasher, Cambridge, 2001).
CA M B R I DG E T EXT B O OKS I N L I NGU I S T I CS Gel/eral editors:
p. AUSTIN, J. BRESNAN, B. COMRIE.
W.DRESSLER.C.J. EWIlN. R. LASS, D.LIGHTFOOT. I. ROBERTS. S. ROMAINE. N. V. SMITH
II/ this series 1'.1'1. MATTHEWS MO/l"IIJIIIgy B.C O MRIII
Second edition
Aspect
R. M. KIlMI'SON
Selllal/tic Tlreory
HistoriClJI Lil/Kuistics
T IIYNON
and O. DAHL
J.ALLWOOD.I..-G.ANDERSON
D. B. FR Y l1,e Plry. R.A.HUDSON
Sociolinguistics
A. J. ELLIOTT
Child Lal/guage
p. II. MATTII ·
Second ed it ion
Tra11.ttil1'11/atiol/al SYI/tax
A.RADFORD L. I I AU ER
Logic iI/ Lil/guistics
:s IIfSpeeclr
EI/Klislr Word-Forlllatiol/
s.c. LEVINSON PI'CIglllatics G. DROWN
and
G. YULE
R.IIUDDLESTON R. LASS
Discourse AI/alysis
IlItmductiol/ til tire Gl'Clllllllar 1!f'EIIglislr
Plrlll/oIOKY
B.COMRIE
Tel/se
w. KLEIN Secol/d Lallguage Acquisitiol/ A. J. WOODS. p.FLETCHER
and
A.H UG�I ES
D. A. CR USE Lexical Sellla/llics
A.RADFOR 0
M. G 1\ RM 1\ N
Statistics iI/ Lallguage Studies
Tral/.ttil1'1l1atiol/al Grullllllar PsydllJlil/Kuisties
G.G.CORBETT
Gellder
II. J.GI EGERICII EIIglisllPlrol/ology
R. C 1\ NN
Fllrlllal Sellla/llies
J. LA V ER
Pril/ciples 1!f'Plul/letie.l·
F.R.I'A LM ER
Grallllllatical Roles and Relatiol/s
M. A. JON ES Fllul/datiolls 1!f'Frel/d, Sy/ll/lX A.RADFORD
SYl//tJctic Tlreory alld tire St1'llclllre 1!f'EIIKlislr: a Millilllalist Appmaclr
R. D. VAN VALIN, JR. A. DURANTI
and
R. J.LI\ pOLLA
Sy/llax: Struclllrl'. Meallillg alld FUllctiol/
LillgllislicAI/tlrmpoloKY
I/IIOllaliol/ Sec on d ed ition and I'.TRlIDGll.L DialeclllloKY Second edit i o n
A.CRlITTIlNDIlN J. K.n l AM BERS C .LYONS
Det . illilel/e.f.I'
R. KAGIlR
Oplimalily Tlreory
J. 1\ . �I 0 L M All I/Ilmductioll 10 Phlgills alld Creole.f
C. G. CORDETT NUlllber c. J. IlWEN
and
F.R.I'A LM ER
H.VAN DER HlI LST
Mood alld Modalil)'
D . J . D LA KE Case Second E.GUSSMAN
TirePlrol/ological St1'llclllre I!f' Words
Second ed ition
ed i t i on
Plrollology
M. YII'1;lIIe
w. CROFT T.11",llIgy alld Ulliversals
F .COULMAS L. W HITE
Second edition
W rilillg Systellls: WI illtmduClion111 Ilreir IiIlKUislic allaly.
Secolld umKuaKeAcquisitioll a"d Ulliversal Grullllllar
I'. J. HOI'PER
and
E. C. TRAUGOTT
Grallllllaticalimtioll
Second ed it i on
Grammaticalization Second Edition
PAUL J. HOPPER Paul Melloll Distillgllis/led P mfessor of Humallities, Carnegie MellcJII Ulliversity
ELIZABETH CLOSS TRAUGOTT Professor of Lillguistics alld Ellglish, Stallford Ulliversity
",:':"", CAMBRIDGE :::
UNIVERSITY PRESS
CAMBRI DGE U N I V E RS I TY PRE S S Cambridge. New York. Melbourne. Madrid. Cape Town. S i ngapore. Sao Paulo. Delhi
Cambridge Uni versity Press The Edi nburgh B u i lding. Cambridge. C B 2 8RU. UK Published in the Un i ted States of A merica by Cambridge Un i versity Press. New York www.can lbridge.org I n formation on this title: www .cambridge.org/97 805 2 1 804 2 1 9
© Paul 1. Hopper. Elizabeth Closs Traugoll 1 993. 2003 This publ ication is in copyright. Subject to stat utory except ion and to the prov isions of relevant col lective l icensing agreements. no reprod uction of any part may take place without the writlen permission of Cambridge Un iversity Press. First publi shed 1 993 Second edition 2003 Fourth prillling 2008 Printed in the Un i ted K i ngdom at t he Un iversity Press. Cambri dge A cata/olille record./ill" 'his /Jllblicatioll is tlI'ai/ab/e fhnn the British LibrCII)'
Librw:v
(!f" COI/IIIl!,u Caw/olluillg ill Publicatioll daw al'l'lied./ill"
ISBN 978-0-5 2 1 -00948-5 hardback I S B N 978-0-5 2 1 -804 2 1 -9 paperback
Cambridge Uni versity Press has no responsibi lity for t he persistence or accu racy of URLs for external or t hi rd - party internet websi tes referred to i n this publ ication. and does not guarantee that any cOlllent on such websites is. or w i l l remain. accurllle or appropriate.
In memory of Dwight Bolinger 1907-1992
Contents
List offigures List of tables Preface and acknowledgments List of abbreviations 1
Some preliminaries 1.1 Introduction 1 .2 What is a grammatical i zed form? 1 .2 . 1 A prel i m i nary classi fi cation of grammatical forms 1 .2.2 Clines 1 .2.3 Periphrasis versus affixation 1 .3 Some further examples of gram maticalization 1 .3 . 1 Lets 1 .3 .2 1 .3 . 3
1 .4 1 .5 1 .6
2
The 2. 1 2.2 2.3 2.4
3
A West A frican complementizer Agreement markers Grammaticalization and language structure Grammatical ization and the directionality of language change Concl usion history of grammati.calization
Introduction Earl ier research on grammatical i zation Research on grammaticalization from the I 960s to the 1 990s Recent trends in research on grammatical i zation
Mechanisms: reanalysis and analogy 3. 1 Introduction 3.2 Some background assumptions about change 3.2. 1 3 .2.2
Induction, deduction , abduction Who i s the language learner? 3.2.3 The question of genetic endowment 3 .2.4 Innovation versus spread
page x i i i xiv xv xvIII
1 I 4 4
6 7
9 10 13 15
16 16 17
19 19 19 25
30
39 39 40 41 43 45
46
x
Contents 3.3
3.4 3.5 3.6 3.7
4
Pragmatic factors
4. 1 4.2
4.3
4.4 4.5 4.6
5
Reanalysis :'1.3 . 1 The French i n flectional future 3.3.2 The Engl ish modal au x i l i aries The i ndependence of reanalysis and gram matica l i zation 3.4. 1 Word-order change Analogy/rule genera l i zation The di fferential effects of reanalysis and analogy Conclusion
Introduction Inferenci ng and meaning change 4.2. 1 Semantics and pragmatics 4.2.2 Relati onshi ps between senses of a form: homonymy and polysemy 4.2.3 Conversational and conventional i n ferencing The role of pragmatic i n ferenci ng in gra mmaticalization 4. 3. 1 Metaphorical processes 4.3.2 Metonymic processes Metaphor and metony my as problem sol ving Pragmatic enrichment versus "bleachi ng" Conclusion
The hypothesis of unidirectionality
5. 1 5.2
5.3
5.4
5.5 5.6
5.7
5.8 5.9
Introduction Generalization 5.2. 1 Genera l i zation of meaning 5.2.2 Genera l i zation of gra mmatical fu nction Decategori al ization 5 . 3 . 1 A noun-to-affix c l i ne 5.3.2 A verb-to-affi x cline 5.3.3 Multiple paths Some processes participating i n unidirecti onal i ty 5.4. 1 Special i zation 5.4.2 Divergence 5.4.3 Renewal A synchron ic resu l t of unidirectional i ty : layeri ng Frequency 5.6. 1 Frequency effects 5.6.2 Synchronic studies of frequency 5.6.3 Diachron ic studies of frequency Cou nterexamples to unidirecti onal ity The uses of unidirecti onal i ty i n reconstruction Conclusion
50 52 55 58 59 63 68 69
71 71 74 76 77 78 81 84 87 92 94 98
99 99 1 00 101 1 04 1 06 1 10 III
1 14 1 15 1 16 1 18 1 22 1 24 1 26 1 27 1 28 1 29 1 30 1 38 1 39
Contellts 6
Clause-internal morphological changes 6. 1 6.2
6.3 6.4
6.5 6.6
7
Grammaticalization across clauses 7. 1 7.2
140 1 40 1 40 1 42 1 43 151 1 54 1 59 1 65 1 65 1 68 1 72 1 74
175 1 75 1 76 1 79 181 1 83 1 84
I ntroduction A cl i ne of clause-comhining consuuctions 7.2. 1 Paratax i s 7.2.2 Hypotaxis 7.2.3 Subordi nation The gra mmatical i zation of cl ause l i nkers Examples of the development of complex sentence constructions 7.4. 1 ThaI-complementation i n Engl ish 7.4.2 Quotative say-constluctions i n Akkadian 7 .4.3 Rel ative clauses i n Engl ish and H i ttite From complex to si mple clauses 7.5. 1 From cl ause chai n i ng to verb i ntlecti on in Lhasa 7.5.2 Two conjoi ned cl�uses reanalyzed as a si ngle clause 7.5.3 From main cl ause construction to sententi a l adverb i n contemporary English Some counterexamples to unidirectionality in clause combi ning Conclusion
209 21 1
Grammaticalization in situations of extreme language contact
212
7.3 7.4
7.5
7.6 7.7
8
I n troduction Morphologization 6.2 . 1 Some characteri stics of clitics 6.2.2 Positions of cli tics 6.2.3 Semantic "relevance" as a factor i n fusi on and morpheme order 6.2.4 Phonological concom i tants of morphologi zation The development of paradigms Argu ment-structure marking: fu nctional-seman t i c hiera rchies and morphological generalization 6.4. 1 Object mark i ng in Persian 6.4.2 Ergative case marki ng: a statistical perspective Loss Conclusion
XI
8. 1 8.2
I ntroduction B a s i c characteri stics o f pidgins and creoles 8.2. 1 Some characteristics of pidgins 8.2.2 Some characteristics of creoles
1 90 1 90 1 94 1 96 204 204 206 207
212 213 214 216
XII
Contents 8.3
Implications of pidgins and creoles for language change Child versus adul t language acqu isition Simpl i fi cation and elaboration Specific implications of pidgi ns and creoles for grammatical ization 8.3. 1 8.3.2
8.4
9
219 219 222 224
Summary and suggestions for further work
231
Notes References Index of names Index of languages General index
234 237 265 270 272
Figures
3. 1 A model of language change 3.2 Schema o f the development of auxil iary be going to 4. 1 Revised schema of the development of auxiliary be going to 7 . 1 Properties relevant to the c l i ne of clause combining
page 4 1 69 93 1 79
Tables
::1.1
Grammatical ization of VO word order in English between AD 1 000 and AD 1 500 Approx imate proportion of compound verbs i n Indo-Aryan 5. 1 languages Ratio of compound verbs i n Marathi and H i ndi-Urdu 5.2 accord i ng to semantic class of main verb Buryat Mongolian pronouns and verb endi ngs 6. 1 Pol ish tonic and cl i tic forms of the copula 6.2 Bonding of cl itic copu la to verb stem in Pol ish, AD 1500 6.3 to the present 6.4 Ditl"eren tial un iverbation of preterit verb and person-number suffix in Pol ish dialects Affixal aspect-tense-mood forms 6.5 Old Icelandic presen t indicative retlexive verb forms 6.6 Pre-Sanskrit noun intlection 6.7 Sanskrit and pre-Pal i forms of the copu la 6.8 Pre-Pali and Pal i forms of the copu la 6.9 6. 1 0 Differential intlection of the aorist in B u lgarian d ialects 6. 1 1 Differential i n l'lection of past participles i n Bulgarian dialects 6. 1 2 Old English strong adjective singu lar i n fleclion Occurrence of that with think and guess versus al l other 7. 1 verbs
page
67
113 114 141 146 146 147 1 54 160 1 62 1 63 1 63 1 64 164 1 72 209
Preface to the second edition and acknowledgments When we wrote the first ed ition of this book i n the early 1 990s, our aim was to present an overview of gram matical i zation for the benefi t of those students o f li ngu istics t o whom this was a new o r o n l y vaguely fam i l iar framework for u n derstanding l i nguistic phenomena. We defi ned gram matical i zation a s the process whereby lexical i tems and constructions come in certain l i nguistic contexts to serve grammatical fu nctions, and, once grammatical ized, con t i nue to develop new grammatical fu nctions. We also characterized it as the process w hereby the proper ties that distinguish sentences from vocabulary come into bei n g diac h ronica l l y or are organ i zed synchronically. In the last ten years gram matica l i zation has become a maj or field of study, and work on the topic has flourished in both "fu nctional" and "formal" frameworks, that is, on the one hand frameworks that privilege the interplay of language structure and use and consider language as both a cog n i ti ve and a commun icative force, and on the other frameworks that privilege language struc ture and consider language pri marily from the perspective of i ntern a l i zed systems. As work has progressed , it has become clear that the defi n i tion of gralll malical i zation as a "process" has been m isleading. To some i t has suggested that gram matical i zation i s conceived as a force w i th an impetus of i ts own independent of language learners and language tlsers . Th is was never i ntended . Only people can change language. The chal lenge presen ted by the kinds of d ata of central con cern to those who work on grammatical ization is that morphosyntactic changes are repl icable across languages and across times; furthermore, they h ave a very slrong tendency 10 go in the same direction, e.g. lexical verb to auxi l i ary, nom inal adpositions to case marker, not vice versa. To avoid further term inolog ical confu sion, w e now define gram matical ization a s t h e change whereby lexical i tems and constructions come i n certai n l i nguistic contexts to serve grammatical functi ons and, once grammatica l i zed, continue to develop new gram m atical fu nctions. In Ihis second ed ition we h ave mai n tai ned the basic structure and content o f the original book, but because grammatical ization has come to be s o widely studi ed i n Ihe last decade we have added substantial discussion of recen t i ssues where they seemed germane. Lingu istic theory has also undergone some rad ical changes s i nce Ihe book was original l y writte n ; for example, within generative theory ru le-based
xvi
Preface and acknowledgments
grammars have been challenged by constraints-based grammars (see developments in Optimal i ty Theory sparked by A . Prince and Smolensky 1 997); and issues of the relationship between structure and use have come to be of central concern to most theoretical perspectives on l i nguistics. We have not attempted to address the variety of issues that form the current theoretical dialog. That would require not only revision, but reconceptual ization, of the original book. Instead, we have main tained the mai n outlines of the original work, deleted some obviously outdated analyses and suggested poi n ters where relevan t to newer o nes. Many terms in l i nguistics serve two functions, one to describe properties of language (e.g., syntax, morphology, phonology), the other to name the study and theory of that property. The term grammatical ization is no exception. It refers not only to changes observable across time, but also to an approach to language study, one that highlights the interaction of use with structure, and the non-discreteness of many properties of language. In this book we do not advance any particular theory of language, although the framework can be characterized as "integrative functional ist" (Croft 1 995) i n that we consider l ingu istic phenomena to be system atic and partly arbitrary, but so closely tied to cogn itive and social factors as not to be self-contained. We focus on observations that lead to claims about what sorts of concepts an adequate theory must accou nt for. We discuss the kinds of i nterpreta tion of data that flow from the approach of gram maticali zation, as wel l as the ki nds of data that one would look for in studying some aspect of a language from this perspective. The study of grammaticalization touches on many of the topics that have been central to work in l i nguistics, whether synchronic or diachronic, most particularly the domains of morphosyn tax and morphology. No attempt is m ade in this book to cover every topic in an encyclopedic way, n or could it be. Some basic knowledge of l inguistics is needed. We assume that readers have at least worked through one of the standard i n troductory textbooks, and have ei ther had a course i n h istorical l inguistics or have carefully read a recent textbook i n this field. Our deepest thanks to the numerous people who contributed directly or i ndi rectly to either the first or the second edi tion ofthis book, or both . Among them Luc Baron ian, Joan Bybee, Brady Clark, Will iam Croft, Andrew Garrett, A. G i anto, T. G ivon, the late Joseph Greenberg, Claude Hagege, Bernd Heine, Suzanne Kemmer, Pau l Kiparsky, Christian Lehman n , Douglas Lightfoot, Therese Lindstrom, John McWhorter, Fri tz Newmeyer, John Rickford, S arah Roberts, Devyani Sharma, Scott Schwenter, Eve Sweetser, Sandra Thompson, and Max Wheeler deserve special mention . Olga Fischer and her students, especially Marije Bogers, Anthony Glass, and Brad Philpot, drew our attention to many poi nts in the first edition that deserved clarification, as did Claire Cowie. We have been especially i ndebted to the l ate Suzanne Fleischman and to Nigel Vincent, who commented on drafts of the first edition, and to Roger Lass and Arnold Zwicky,
Preface a nd acknowledgments
XVI I
who made detail ed comments on the publ i shed version . Thei r carefu l and insightfu l comments were of i nestimable value. Than ks too to t h e many students and fac u l ty who attended our sem i nar at the Linguistic Society of America's Li ngu istic I n stitute at Stanford University i n 1 987, where many of the ideas for the writing or this book were developed, and in subsequent courses that we have taught i n our respective i nstitu tions and at summer sessions around the worl d . Julia Hard i n g and John McWhorter provided i nvaluable h e l p with editorial matters for the fi rst edition, Kay McKechnie, Rob Podesva, and Elyse Nakaj i ma for the second . We would also l i ke to thank Penny Carter and Judith Ayl i ng for their help in getting the first edi tion to press, and Andrew Winnard for his support of the second ed i tion . We acknowledge the opportunities provided by G uggenheim Fel lowships (Hopper i n 1 985 , Traugott i n 1 983) for the study of many of the aspects of gram m at icalization d iscussed here. Elizabeth Traugott further acknowledges a Fel lowsh i p a t the Center for Advanced S tudy of the Behavioral Sciences at Stanford ( 1 983). In writing this book we debated many points with respect for each other's different views. No two authors covering as wide a range of issues currently open to theoretical and methodological debate as are touched on in this book cou l d expect to agree w i th every concept p u t forward, and w e are n o exception . We have welcomed the inspiration that our differing perspectives have provided , and we hope that this book w i l l encourage further debate i n the l i nguistic community.
Abbreviations
Linguistic terms ABL ABS ACC A OV AGR AGT AOR ASP AUX BEN CAUS CL COMP COMPL CON] CONTIN COP OAT OEF OEM OET OUR EM PH ERG FEM FUT GEN GER
ablative absolutive accusative adverb agreement marker agent aorist aspect auxil iary benefactive causative noun classifier complementizer com pletive conj unction continuative copula dative defi n ite marker demonstrative determ i ner durative emphatic ergative fem inine future gen itive gerundive
Abbreviatiolls
ILL IM P I M PF INCL INDEF INF INSTR lOB] LOC M MASC M ED NEG NEUT NF NOM NP NPrel OB] OM OV P PART PA RTIT PERF PL PM PN POSS POST PREP PRES PRO PROG PURP QUOT R ECIP REFL R EL SG
i l lative imperativt' imperfect i nclusive i nde fi n i te i n fi n i tive i nstru mental i ndirect object locative modal mascu l i ne medial negative neu ter non-final marker nom i native noun phrase anaphoric NP i n a rel ative clause object object marker o�iect-verb word order particle participle parti tive perfect pl ural pred icate marker personal name possessive postposition preposi t i on presen t pronoun progress ive purposive quotative marker reciprocal reflex ive relative si ngular
XIX
xx
Abbreviations
Srel SUB SUBJ SUBJUNCT SW TNS TOP TRANS V VO VP 1 2 3
relative clause subordi nator subject subjunctive switch-reference marker tense topic transitive verb verb-object word order verb phrase firs t person second person third person
Sym bols >
grammatical word> clitic> inflectional affix
Each item to the right is more clearly grammatical and less lexical than its partner to the left. Presented with such a c l i ne, l i nguists wou ld tend to agree that, i n so far as they schematical ly reflect cross- l i ngu istic general i zations, the poi n ts (labels) on the cl ine could not be arranged i n a differen t order, although i ndividual i tems m ay violate the order language-spec i fical ly (Andersen 200 1 ). A n u m ber of such cl i nes have been proposed, based on the many difl"erent d i mensions of form and mean i n g that are found i n language. General ly, they i nvol ve a u n i d i rectional progress i o n in bonded ness. that i s , i n t h e degree of cohesion of adjacen t forms that goes from loosest ("periphrasis") to tightest ("morphology"). It is often difficu l t to establish firm bou ndaries betwee n the categories repre sen ted on c l i nes. and i ndeed the study of grammatical i zati on has emerged in part out of a recognition of the general fl u i d i ty of so-cal led categories. It has also emerged out of recognition that a given form typically moves from a poi n t o n the left of the c l i ne t o a poi nt further o n the right, i n other words. that there i s a strong tendency toward unidirectionality i n the h i s tory o f i ndividual forms. We will discuss u n idirectional i ty and ways of conceptual izing the c l i ne i n some deta i l in Chapter 5 . 1.2.3
Periphrasis versus affixation
O ften the same categories can be expressed by forms at d i fferent places i n the c l i nes. Thus i n English w e have expressions that are "phrasal" o r "peri phrastic" ( l iterally "occurri ng in a rou ndabout fash i on") such as (6): (6)
a. have waited (perfect tense-aspect) b. the household of the queen (possessive) c. more interesting (comparative)
It is also possi ble to ex press tense-aspect, possession, and the comparative through aflixes or changes i n ternal to the stem word. In this case the c a tegories are bound to a host and are said to be expressed "morphological ly" or "affixally"as in (7): ( 7)
a. waited (past tense affixed contrast
-ed); sang (past tense signaled by intemal change:
sillg)
b. the receptionist's smile (possessive affix c. longer (comparative
-s)
-er)
The d i sti nction between the peri phrastic and morpholog i c a l expression of a category is i m portant for the study of gramm atica l i z at ion bec a u se of two diachronic
8
I Some p relimin aries
tendencies. One is for periphrastic constructions to coalesce over time and become morphological ones. While this and other tendencies are discussed i n more detail later, especially in Chapter 6, a couple of examples follow: (a) Defin i te nouns are marked i n many European and other languages with an article that i s separate from the noun, for example, Engl ish the newsp aper, French la rue ' the street,' German die Stadt ' the city, ' etc. In such languages defi n i teness is marked periphrastically (cf. English the five yellow newsp apers, where the article is at some d istance from the noun). But i n some languages this sign of definiteness is an affix, which can usual ly be shown to derive from an earl ier definite article or demonstrative. Thus in Istro-Romanian 3 the Latin demonstrative ille ' that' now appears as a suffix on nouns marking both defi niteness and case, as i n : (8)
gospodar-i-Ior boss-PL-DEF:GEN 'of the bosses'
Here -i marks plural and -lor is the definite geni tive plural suffix deriving from Latin illorum, the masculine genitive plural of ille. S i m i larly i n Danish, -en i n dreng-en ' the boy' and -et in hus-e t ' the house' are defi nite s i ngular markers for common gender and neuter nouns respectively, and have their origin in earl ier postposed demonstratives (cf. Old Norse ulfr-inn ' wolf- the' from ·ulfr hinn 'wolf-that' ). In the modern l anguages they cannot be separated from the preceding stem . (b) Various tenses and aspects of verbs are formed either w i th aux i l iary verbs (i.e., periphrastic tense-aspect) or with verbal suffixes (Le., morphologi cal tense-aspect). Thus i n H i ndi the present tense is formed periphrastically by a verb stem plus the verb to be: (9)
miii kursii par baithaa hiiii. chair on sit: MASCSG be: I SG 'I sit on a chair. '
1
In Swahil i , o n the other hand, basic tenses such a s the future are formed morpho logically, w i th prefi xes on the verb: ( 1 0)
Wa-ta-ni-uliza. they-FUT-me-ask 'They will ask me. '
Morphological tense-aspect formations can often be shown to have developed out of earl ier periphrastic ones. The Romance l anguages supply n umerous ex amples of this, such as the Italian future c antaremo ' we w i l l sing' or the French future (1I0US) ch an terons from Latin c an t are h abemus, l i terally ' we have to sing.' We discuss this kind of development in the Romance l anguages in Section 3.3. 1 .
J.3 Some furthe r examples of g rammaticalizatiol1
9
The second diachronic tendency that makes the periphrasis/bondedncss distinc tion important is an example of what is known as "renewal" - the tendency for periphrastic forms to replace morphological ones over time. Where a long histor ical record is available, the process of renewal can be seen to occur repeatedly. The French future form just mentioned, for example, i s the i nflectional form (1I0IlS) chaflterolls ' we w i l l sing . ' B u t its Latin source, cafltare habemus, was a periphrastic future that eventually replaced an older morphological future, cantabimus, after competing with it for several centuries. Th is form in turn evidently contai ns the verb *b" IIInos ' we are, ' inherited from Indo-European, and can be reconstructed as an earlier periphrastic construction 'kama bh umos. French IlOUS chanterons is itsel f being replaced by nous allons chanter, l i teral ly ' we are going to sing.' S omething like the following sequence of changes can therefore be esta b l ished: (I I)
Pre-Latin
Latin
French
.? ' kanta bh umos > cantabi mus cantare habemus > chanterons allons chanter> ?
At each attested stage two (or more) constructions compete (typically separated from one another by some nuance of mean ing such as ' we w i l l ' versus 'we are about to' ), and eventual l y the periphrastic one wins out, undergoes coalescence of the two elements that comprise it, and may in turn be replaced by a new periphrastic form (Hodge 1 970 provides examples of the renewal by periphrasis from several language fami l ies) . The terms "renewal" and "replacement" are somewhat problematic because they may suggest functional identity over time, and even gaps to be filled. In fact, however, i t is not on l y the forms cantabimus and cantare habemus that d ille r; their exact semantic fu nctions and syntactic distributions ditler too, in so far as the overall set of tense options is necessari l y' different once the two forms coexist (other changes were also occurri ng elsewhere in the system, further reducing any potential identity). Unfortunately our available l i nguistic vocabulary or "metalanguage" for expressing the relationship between earlier and l ater l i nguistic phenomena is poor. We will not attempt to change it here, but w i l l fol l ow custom and use terms such as "replacemen t" and "renewal," on the understand ing that there is no exact identity over time (and, as w i l l be discussed in Section 5.4.3, there are no gaps to be filled). 1.3
S o m e fu rth e r exa m p l es of gra m m atical ization
We turn now to some rel atively detailed examples of g rammaticali zation to i l lustrate several of its characteristics, and some of the problems of defining instances of it u niquely.
10
J Some preliminaries
1 . 3. 1
Lets
An i nitial example w i l l be chosen from contemporary standard Engl ish also known as Present-Day Engl ish (or PDE for short). We beg i n with this ex ample because it i l l ustrates vividly that grammatica l i zation i s an everyday fact of language. It results in not only the very fam i l iar constructions of language such as be going to, but also many of the highly structured, sem i-autonomous "formal idioms" of a language that m ake it un ique, but are often regarded as peri pheral (Fi ll more, Kay, and O' Connor 1 988). In PDE there i s a construction i nvolving a second-person imperative with the verb let: ( 1 2)
a. Let us go. (i.e .. release us) b. Let yourself down on the rope. c. Let B i l l go. (i.e release B i l l ) .•
The understood subject of let is you. The objects of let i n (a), (b), and (c) are all ditl"erent: us, yourself, Bill, and may be passivized, e.g . : ( 1 2)
d. We were let go.
Alongside the ordinary imperative construction with let in ( 1 2a-c) there is a con struction sometimes called an "adhortative" ( i nvol v i ng urg i n g or encouragi ng), as i n : ( 1 3)
Let's go to the circus ton ight.
Quirk, Greenbaum, Leech , and Svartvik ( 1 985 : 829) refer to this construction as a "first-person imperative." Here the subject of let is u nderstood as T as in somethi ng l i ke 'I suggest that you and I . . ' Us is also the subject of the dependent verb rather than the object of let, and can therefore not be passivized : ( 1 2d) i s the passive of ( 1 2a), not of the first part of ( 1 3). Qu irk et al. note the spread of let 's in very col loq u i al English to the s i ngular of the first person : .
( 1 4)
Lets give you a hand. (i.e .. let me give you a hand)
(We will represent the form as lets when the subject i s other than the fi rst-person plura\.) Quirk et al. describe the lets here as "no more than an i n trod uctory particle" ( 1 985 : 830). In some varieties of Engl ish, the first-person-plural i nclusive subject itS of lets has been rei n forced by you and J as i n : ( 1 5)
Let's you and I take 'em on for a set. ( 1 929. Faul kner. Sartor;s 1 1 1 . 1 86: OED let 1 4.a)
It has even been extended beyond first-person subjects of the dependent verb. The fol lowing examples are ti'om Midwestern American speakers :
1. 3 Some Jttrther examples of g rammaticalizatioll
( 1 6)
I I
a. Lets you and him fight.
h. Lets you go first , then if we have any money le ft I' l l go.
W h i le ( 1 6a) was perhaps jocular (a third party eggi n g on two others), the contex t of ( 1 6b) was quite neutral . In other i nstances there is no second- or third-person subject pronoun, and lets simply conveys the speaker's condescendi ng encourage ment, e.g., in address ing a c h i l d or a trucu lent person : ( 1 7)
(Co le 1 97 5 :
a. Lets wash your hands.
268 )
h. Lets eat our liver now, Betty.
The devel opment of the lets construction i l l ustrates a n u m ber of characteristics o f grammatical ization. Among these are: (a) ( 1 2) shows that a fu l l verb let 'al l ow, permi t ' has al tered i ts semantic range in some way. We w i l l suggest that grammatical i zation i n its early stages often. perhaps always, i nvol ves a shift i n mean ing (Chapter 4 ; see also Traugott 1 98 9 : Heine, Claud i , a n d Hlin nemeyer 1 99 1 a ) . Furthermore, as mentioned in con nection with b e goillg to , th i s k i nd o f s h i ft occurs only i n a highly spec i fic context, i n th is case o f the imperative Let /IS A first approx i mation would be to say that the earl ier idea o f perm ission or al lowing has become extended in one part of i ts parad igm to incl ude a further one of suggesting or encourag i ng someone to do someth i ng. The sense o f let has become less spec i fi c and more genera l ; at the same time it has become more centered in the speaker's atti tude to the s i tuat i o n . Th i s new construction h a s been available si nce t h e fourteenth century (Traugott 1 (95). (b) ( 1 6) shows that the range of possible subjects of the verb dependent on lets is bei ng extended from first-person plural to other person s . Th i s was presumabl y made possible b y the fact that we/us i n Engl ish may b e i n terpreted a s i nclusive o f the addressee ( , I and you ' ) o r exclusive o f the addressee ( ' I a n d another o r others ' ) . So long a s the di stribution of let 's i s consistent with first-person-plural su�jects i n the dependent verb (e.g . , ' let's indulge oursel ves' ), i t may s t i l l b e usefu l t o analyze it as let + / I S . But th i s d i stri bution has now spread to other persons, as suggested by example ( 1 4 ), Lets give you a hand (said by one i nd i v i d ual to another), where lets is s i ngular. As men ti oned in connection with be going to, earl ier mean i ngs and fu nctions typical l y persist. Thus ( 1 3- 1 7) coexist with ( 1 2). Furthermore, the semantic changes proceed by small steps (perm ission to suggestion, fi rst to second to third person). 4 (c) A Ilrst-person-pl ural pronoun us became c l i ticized (let 's) , and from the word-pl us-c l i t i c complex a s i ngle word was formed, lets. A s s uggested above, so long as the di stri bution of this form is consistent w i t h the fi rst-person-pl ural subjects of the dependent verb, i t may sti l l be usefu l to a n a lyze i t as a c l i ticized . . •
12
J Some preliminaries
form of us. But when this distribution spreads to non-first-person-plural subjects, we are not synchronicall y j ustified in conti nuing to do so. The final s of lets, then, is losing i ts status as a separate morpheme, and is i n the process of becom i ng a simple phonemic constituent of a (monomorphem ic) word . The historical trajectory: (let) us > (Iet)'s > (Iet)s
illustrates a more general shift of word > affix > phoneme
(cf. Givan 1 979: 208-9; Hopper 1 994)
(d) Once the monomorphemic stage has been reached, then the form becomes subject to further reduction. S ince [ts] is often reduced in rapid speech to the sibilant, it is not surprising that lets [lets] often becomes lets [ \cs] . It even goes further and in very colloquial speech is c1iticized and attached to the fol lowing verb: sgo, sfight. (e) Like other emergent constructions, lets in some sense fi xes, or routi nizes, a meaning or discourse function which was formerly freer (see Hopper 1 987). It singles out one comb ination (in this case, let + us) from what was once a more extensive paradigm of equivalent forms, as i n ( 1 8), and specializes it i n a newly emerging function, the adhortative: ( 1 8)
Let h i m speak now or forever hold his peace.
This new function is provisional and relative rather than permanent and absolute; lets may not survive. However, for now a disti nctive new grammatical resource has entered the language and is available to speakers for the building of interactive discourse. (f) A fi nal comment about the development of lets is that, although the stages are clearly very local and appear somewhat marginal, nevertheless they are part of a typological change affecting Engl ish. This is a shift which has been in progress for over two thousand years from an essential ly "object-verb" system (as in her saw) with case and verb i nflections, in other words, affixal constructions, to an essentially "verb-object" system (as in saw her) with prepositions and phrasal verb constructions, i n other words, periphrastic constructions. We will discuss word order shifts i n more detail in Section 3.4. 1 . Here it must suffice to mention that in Old Engl ish, as in some other older Indo-European languages, the adhortative was expressed by the subj unctive, as shown in ( 1 9) (though a phrasal form with utan also existed). ( 1 9)
C i l d binnan O ritegum nihta sie gefu lwad. child within thirty nights be: SUBJ UNCT baptized 'Let a child be baptised within thirty nights.' (c. 690, Law Ine 1 . 1 )5
/.3 Some further examples of grammaticalizatioll
13
The developmen t of lets, then, is to be seen as among the class of innovations that are leading to a phrasal expression of the modalities of the verb, replacing an earl ier inflectional expression. It is part of the very general change from a morphological way of expressing a function to periphrasis discussed in Section 1 .2.3. The ri se of the numerous aux i l i ary and auxil iary-like verbs and expressions of Modern Spoken English (such as may, be going to, keep V-ing, and others) is symptomatic of the same trend, which has been ongoing in English for many centuries (see Krug 2(0 1 ). 1 .3.2
A West African complementizer
Our examples so far have for the most part i l l ustrated the development of verbs i nto grammatical markers of the kind usual ly associated with verbs, specif. ically tense, aspect, and mood . We turn now to a wel l-known example of a verb being grammaticali zed i nto a connective, in this case a complementizer that i ntro duces a fi n i te complement c l ause. A fi n i te complement clause is equivalent to an English that-clause in such constructions as : (20)
I know that her h usband is in jai l .
The verb which h a s the position o f know in such sentences i s cal led the "matri x verb," and the clause introduced by the complementizer that is the "complement clause." Lord presents data from a number of African and Asian languages in which a locutionary verb meaning 'say ' has come to function as a complemen tizer. Exotic as it may seem, such a construction is by no means unknown in English, cf. : (2 1 )
If/Say the deal falls through, what alternative do you have?
We will c i te examples from Lord's work on languages of Wes t Afi'ica, all of them related members of the Kwa group of Niger-Congo spoken i n Togo and Ghana, especially from Ewe (the examples that fol l ow are from Lord 1 976: 1 79-82). The process leading to the grammatical ization of a 'say ' verb i nto a complemen tizer evidently begins when a general verb meaning 'to say ' i s u sed to rei nforce a variety of verbs of saying i n the matri x clause. In Ewe, for example, if the matri x verb is the general verb be 'say, ' no further complementizer is n eeded : (22)
Me-be me-w:>-e. I-say I-do-i t ' I said, " I did it."!1 s a i d that I did i t . '
However, if some verb of saying other than b e is the matrix verb, as a complementizer:
be
must be used
14
I Some preLimina ries
(23)
Me-gbb be I say
me- w o -e.
say I-do-it
'I said that I did it.
(where gbb is a different verb meaning 'to say ' ). The next stage is one in which be comes to be used as a complemen tizer after a whole range of matrix verbs, i ncluding, for example: gbb 'say ' tt b ' write' 15 eje edzi 'agree' ( l it. 'accept reach top' ) x;)se 'bel ieve' nya ' know' bU ' think' v 5 ' lear, be afraid ' kp 5 'see' ttb ' forget' se ' hear, perceive' na ' make sure' The verbs included are verbs o f speaking, cognition, and perception. S i nce these are verbs which in most languages can have objects that are propos itions ( i .e., clauses), there is an obvious sy ntactic and semantic relationship between them and 'say. ' Even so, the mean i ng and morphology of the 'say ' verb is essenti ally lost in the process of grammatical ization as a complementizer. For example, in (24) we see that be may no longer take verbal affixes such as person markers (compare lIIe-df 'I-want' ), nor may it productively take tense-aspect markers. ( 24)
Me-df
be
map-Ie
awua Qew6.
I-want say I - S U B J UNCT-buy dress some ' I want to buy some dresses. '
Furthermore the original mean ing of 'say ' i n s u c h sentences is n o t easy t o recover. Al though some of i ts origi nal context is maintai ned ( i t remains a form that in tro duces a noun clause), it has become avai lable to many more contexts. From being a verb that introduces something said, it has become generalized to in troducing other kinds of clauses, such as reports of things seen or though t. As with Engl ish be going t o and lets, the Ewe exam ple shows not only a se mantic but also a structural adjustment. Not only does the verb 'say' extend and perhaps even lose its original mean ing of say i ng, but a construction originally consisting of two i ndependent clauses is reanalyzed as a matrix verb plus a com plement c lause in troduced b y a complementizer. For example, (25 ) is reanalyzed as (26):
1. 3
(25)
Megbb be
I -say
(26)
Some further examples of g rammaticalizatioll
15
[mew:>e] .
say I-do-it
'1 said I did it . ' Megbl::l [be mew:>e ] .
I-say
[say
I-do-it]
'I said that I did it . '
We w i l l return later to fu l ler d i scussion o f reanalysis i n Chapter 3 . For the present, it is important to recognize that both semantic and structural reanalysis are m Entstehen tier gram mati kal ischen Formen und i h ren EinftuB aufdie Ideencntwic k l u ng") given i n 1 822 he suggested that the grammatical structure of human l anguages was pre ceded by an evol utionary stage of language in which only concrete ideas cou ld t t)
20
2 The history of grammaticalization
be expressed . Grammar, he suggested, evol ved through distinct stages out of the col location of concrete ideas (Humboldt 1 825 ). At the first stage, only thi ngs were denoted, concrete objects whose rel ationships were not made expl icit in utterances but had to be inferred by the l i s tener. In modern terms, we might designate this stage as a "pragmatic" or "discourse-based" stage (Giv6n 1 979: 223). Eventually certain of the orders in wh ich the objects were presented became hab i tual, and this fixing of word order i n troduced a second stage (we might nowadays cal l i t "syntactic"). At this stage, some words began to waver between "material" ( i .e., concrete) and "formal" (i.e., structural or grammatical) mean ings, and some of them would become specialized for functioning i n more relational ways in u tterances. In the third stage, these functional words became loosely affi xed to the material words; in modern terminology this might perhaps be called a stage of "cl i ticization ." In this way "agglutinative" pairs arose, dyads consisting of a material word and a relational word. In the fourth stage these aggluti native pairs became fused into synthetic, single-word complexes. There were now stem and (inflectional) affixes that contai ned s i m u l taneously material and grammatical meani ngs; we might think of this as a "morphological" stage. At this fourth stage, too, some of the function words would continue their l ives as purely formal indicators of grammatical relationships. The functional l i fe of words was reflected in their forms and meanings; during long usage mean i ngs became lost and sounds were worn down. It is no coincidence that Humboldt's four stages correspond q u i te closely to a ty pology of languages that was in the air during the first decades of the n ineteenth cen tury. According to this typology, there were three basic types of l anguage: Isolating (Humboldt's stage II), Agglutinative (stage III), and Inflectional or Synthetic (stage IV). Humboldt's proposal can be thought of as an account of these types in evolutionary terms, supplemented by an assumed pre-stage (Humboldt's stage I). He eventual ly developed this idea i n to a series of further speculations about lan guage typology and the relationship between language and cultural evolution. (A usefu l account of Humboldt's later ideas on language can be found i n Humboldt 1 988 [ 1 836] , and R. Harris and Taylor 1 997 [ 1 989] : 1 7 1 -84.) By the end of the ni neteenth century a clear tradition in the study of gram matical ization had been establ ished, lacking only the name i tself. A picturesque account of the origi ns of grammatical forms and their evolution is to be found in the survey of l i nguistics by the German neogrammarian Georg von der Gabelentz ( 1 89 1 ). Gabelentz ( \ 89 1 : 24 1 ) invites his readers to visual ize l i nguistic forms as employees of the state, who are hired, promoted, put on hal f-pay, and fi nally retired, while outside new appl icants queue up for jobs ! Forms "fade, or grow pale" ("verblassen"); their colors "bleach" ("verbleichen"), and must be covered over w i th fresh paint. More gri mly, forms may die and become
2.2 Earlier research on grammaticalization
21
"mummified" ("mummifiziert"), l i ngering on without l i fe as preserved corpses ( 1 89 1 : 242). Gabelentz articulated many of the insights basic to work on grammaticali zation. He suggests that grammaticalization i s a result of two competing tendencies, one tendency toward ease of articulation, the other toward disti nctness. As relaxed pronunciations bring about sound changes that "wear down" words, disti nctions become blurred. So new forms must step in and take over the approximate function of the old ones. For example, the Lati n first-person-singular future tense of a verb such as video 'I see, ' videbo, is formed with a suffix -bo which was once * b" wo, a first-person-singular form of the verb ' to be' used as an auxi l i ary. An old periphrastic construction, that is, a complex of a mai n verb and an auxiliary verb (vide + b" wo), was col lapsed i n to a single inflectional form . But l ater this form too "wears down" and is replaced by new periphrastic forms such as videre "abeo 'I have to see.' Somewhat later, this idea was to be articulated again by Mei l let under the rubric of "renewal" ("renouvellement"). A second i nsight developed by Gabelentz is that this is not a l inear process, but rather a cycl ical one. Whereas for Humboldt's generation synthetic (inflectional) languages l i ke the classical Indo-European languages represented an evolutionary endpoint, Gabelentz noted that the process of recreation of grammatical forms i s recurrent, and that the conditions for t h e cycle are always presen t i n l anguages. Moreover, even the idea of a cycle is an oversimpl ification . Gabelentz speaks instead of a spiral, in which changes do not exactly replicate themsel ves but parallel earlier changes i n an approximate manner. Gabelentz's work, unlike Humboldt's, is informed by the awareness of geolog ical timespans, which made it psychologically possible to thi n k of m u ltiple cycles of l inguistic change. It also reflects an expanded knowledge of the variety of human languages and of h istorical texts, especially i n the Indo-European l anguages that the neogrammarians and their predecessors had studied so e nergetically, now for two or three ful l generations. Yet Gabelentz's discussion of the origins of gram matical forms and thei r transformations covers only a couple of pages i n his entire book. Although the germs of l ater work on grammatical ization are contained here, it was Antoine Meil let who first recognized the importance of grammaticali zation as a central area of the theory of language change. Meil let was also the first to use the word "grammaticali zation," and the first linguist to d evote a special work to it. Meillet's use of the term "gram matical i zation" to designate the development o f grammatical morphemes out of earl ier lexical formatives is clearly descended from Humboldt's and Gabelentz's insights. It was also anchored i n a m ore positivistic view of language, which stressed regularity in l i ngu istic change and systematic ity in synchronic description. As Meillet hi mself noted, the first generation of
22
2 The history of grammaticalizatioll
lndo-European ists had speculated i n tensely about the origi ns of gram matical forms. But their results had been random and u nrel iable. Moreover, they had i nsisted on placing these results in a "glottogonic" context, that is, the context of a su pposed evolutionary l i ne that would lead back to the actual origins o f language. But this l i ne of i nvestigation had now fal len i n to disrepute. Mei llet showed that what was at issue was not the origins of gram matical forms but their transforma tions. He was thus able to present the notion of the creation of gram matical forms as a legiti mate, indeed a cen tral, object of study for l i nguistics. In his article "L' evolution des formes gram maticales" ( 1 9 1 2), Meil let descri bes how new grammatical forms emerge through two processes. One is the wel l known fact of analogy, whereby new paradigms come i n to being through formal resemblance to already established parad igms. (An example of analogy i n recent English would be the replacement of the plural shoen by shoes through analogy to such established plurals as stones. ) The second way in which new grammatical forms come into being, Meil let suggested, is through grammatical ization, "the passage of an autonomous word to the role of grammatical element" ( 1 9 1 2 : 1 3 1 ). Meil let il lustrates the synchronic result of this process with the French verb erre 'to be, ' which ranges in mean ing from a fu l l exi stential ontological sense, as in je suis celui qui suis 'I am the one who is [ l i t. am] , ' to a somewhat less fu l l locative sense in je suis chez moi 'I a m a t home,' t o a n almost redundant sense i n je suis malade 'I a m i l l , ' je suis maudit 'I am cursed,' and t o a purely grammatical function as a tense-aspect auxil iary in je suis parti 'I l eft, ' je me suis prol1lelle 'I went for a walk . ' The most significant, a n d remarkable, part of t h i s fu ndamental article is Mei l lel's confident assertion : "These two processes, analogical innovation and the attri bu tion of grammatical character to a previously autonomous word. are the only ones by which new grammatical forms are constituted . The details may be complex in any individual case; but the pri nciples are always the same" ( 1 9 1 2 : 1 3 1 ) . Later in the same article, Meil let goes even further. Analogy can only operate when a nucleus of forms has already emerged to which new forms can be ass i m i lated . So anal ogy is ru led out as a pri mary source of new gram matical forms. Therefore, "the only process left is the progressive attribution of a grammatical role to autonomous words or to ways of grouping words" ( 1 9 1 2: 1 32). In every case where certainty is possible, Mei llet contends, this is the origin of gram matical forms. Noth ing stands i n the way of assum i ng that when allowance has been made for analogical extension the same kind of source can ultimately be attributed to forms o f unknown or uncertain origin also. Considering that during the neogrammarian period all i nvestigations of gram matical morphology had been essentially investigations of analogy, Mei l let's state ment was sweeping and rad ical . Wri ting of the transformation of autonomous
2 . 2 Earlier research
011
grammatiealizatioll
23
words i n to grammatical roles, he says: "The importance [of this] is in fact deci sive. Whereas analogy may renew forms i n detai l , usual l y leav i ng the overal l plan o f the system untouched, the 'grammatical i zation' of certai n words creates new forms and i ntroduces categories which had no l i nguistic expression. It changes the system as a whole" ( 1 9 1 2 : 1 33). "Grammatical ization," then , is seen as a change which affects individual words. B ut it is evidently also meant to be extended to phrases. Indeed , the combi n i ng o f words into set phrases and their eventual amal gamation is presented i n the first part of the article as a defi n i ng feature o f the evt!nl. In the French fu ture represen ted by je vais jaire ' I w i l l do,' l i terally 'I am going to do,' vais no longer contains any perceptible sense o f 'going. ' In je jerai 'I wi l l do,' the fu sion has gone even further, with no analytic trace remai n i n g of tht! original Latin phrase jaeere habeo 'I have to do. ' It is a l oss, Mei l let suggests, o f express ivi ty. A novel way of putti ng words together becomes commonplace ("banal"). In the t!xtrt!me case, tht! phrase even ceases to be analyzable as con tai n i ng mort! than ont! word , but i ts members are fused together ("soude") as one. Th i s phrasal col l ocation is itse l f usually a replacement for an a l ready ex i stent form which has become commonplace. Consequently, gram maticalization tends to be a proct!ss of replac i ng older grammatical categories w i th n ewer ones having tht! same approxi mate val ue: i n flected futures (ama-bo ' I shal l love ' ) are rt!placed by pt!ri phrastic futures (amare habeo ' I have to l ove' > 'I shal l love ' ) , which in turn are fust!d (Fr. aimerai 'I shal l l ove' ), and so on. At the end of the article he opens up the possibi l i ty that the domain of gram matica l i zation m ight be extended to the word order of sentences ( 1 9 1 2 : 1 47-8). In Lati n , he notes, the role o f word order was "expressive," not grammatical . (By "expressive," Mei l let means someth ing l i ke "semantic" or "pragmatic.") The sen tence 'Peter slays Pau l ' cou ld be rendered Petrus PaulullZ caedif, Paulum Petrus caedit, eaedit Paulum Petrus, and so on. In modern French and English, which lack case morphemes, word order has primarily a grammatical val ue. The change has two of the hal l marks of gram matical ization: (i) it involves change from ex pressive to gram matical mean i ng; ( i i ) it creates new grammatical tools for the language, rather than merely mod i fy i ng already ex istent ones. The grammatical fi xing of word order, then, is a phenomenon "of the same order" as the gram matica l i zation of individual words: "The expressive value of word order which we see i n Lat i n was replaced by a grammatical value. The phenomenon i s of the same order as the 'grammatical ization ' of th is or that word ; i nstead of a single word, used with others in a group and taking on the charac ter of a 'morpheme' by the effect of usage, we have rather a way of group i ng words" ( 1 9 1 2: 1 48 ) . We st!t!, then, that i n this i n itial study o f grammaticalization, Meil let already points to appl ications o f the term that go far beyond the simple change from lexical to gram matical mt!an ing of si ngle words. Indeed, if we pursue his argument to its
24
2 The history of grammaticalization
logical conclusion, it is difficu l t to see where the boundaries of gram matical iza tion could convinci ngly be drawn. If the fixing of word-order types is an example of grammaticalization through constant usage, could not al l constructions which have been called "grammatical" constructions be said to have their ulti mate ori gins in such habitual collocations? Evidently, how far we shal l be prepared to extend the notion of "grammatical ization" will be determi ned by the l i m i ts of our understanding of what it means for a construction to be "grammatical" or have a grammatical fu nction. We will suggest i n Chapter 3 that, at least at this stage in our understanding of grammatical ization, word-order changes are not to be included, although they are deeply i nterconnected with it. Meillet also anticipated other themes i n the study of grammatical ization which are still at i ssue. One of these is how grammatical ization comes about. He attributes grammaticalization to a l oss of expressivi ty in frequently used collocations, whose functions may then be rejuvenated through new collocations fi l l i ng more or less the same role. Yet often a "loss of expressivity" seems i n su fficient to capture what happens i n grammatical ization. Some of his own i l lustrations challenge such a motivation. For example ( 1 9 1 2: 1 38-9), the Modern German word heute ' today ' can be traced back to a presumed Old High German phrase hiu tagu, the i nstru mental of two words meani ng ' this day ' (compare Gothic himma daga and Old High German hill jaru ' this year [ instr.] " Modern German heuer). It is, first of all , a l i ttle startling to find a change of this kind discussed under the rubric of grammaticalization, since heute might be more appropriately thought of as i l lus trating the emergence of a new lexical i tem rather than of a grammatical formative. The change in Old High German from *hiu tagu to hiutu is a change only from an adverbial phrase to an adverb, and i t is questionable whether the later form is less meaningful ("expressive") than the earl ier. Yet there i s surely a difference in Modern German between heute and an diesem Tage 'on this day ' that needs to be characterized in some way. Evidently some different way of talking about meani n g change is needed . We return to these issues and a better understanding of the relationship between grammatical ization and lexicalization i n Section 5 .6. Accompanyi ng this loss of expressivity is a supposed weakening ("affaiblisse ment") of phonological form and of concrete mean ing ( 1 9 1 2 : 1 39). Meil let's ex ample is the development of the Modern Greek future tense morpheme tha, whose origi n i s in an older construction thelO ilia 'I wish that' (with ina from a still earlier Mila). The change included the fol lowing stages ( 1 9 1 2: 1 45): thell> ina > thell> na > then a > tha
and the semantic development is from ' wish, desire' to ' future tense. ' It is not difficult to see "weakening" in the phonological process, since there is undeniably a shortening and hence a loss of phonological substance. B u t it is not so obvious
2.3 Research /rom the 1 960s to the 1 990s
25
that the concomitant semantic change should also be seen in the same way. Like all the writing on grammaticalization at his time, and much si nce then, Meil let's account of grammatical ization i n general i s couched in terms which stress deficits of various kinds : l oss, weakening, attrition. Such metaphors suggest that for al l his l i nguistic sophi stication there is still a slight residue of the "classical" atti tude toward language in Meillet's thought, the attitude that equates change with deterioration. Still, this first full-length paper on grammaticalization, i n which the term itself is proposed, is astonishingly rich in its insights and the range of phenomena which are analyzed. Subsequent work on grammaticalization has modi fied, sometimes quite rad ically, Meillet's views, and many more substantive examples have been described, but time and agai n the germs of modern ideas on grammatical ization are to be found, implicitly and often expl icitly, in this initial paper.
2.3
Resea rch o n gra m matica l ization fro m the 19605 to the 19905
After the work of Meil let in the fi rst two decades of the century, the topic of grammatical ization was taken up mainly by Indo-Europeanists. Many other scholars who saw themselves as historical linguists, but not necessarily Indo Europeanists, did not concern themselves with grammaticalization as a subdisci pline or even as a topic i n its own right. The term is consistently overlooked i n the textbooks of synchron ic and historical linguistics of the period. Indeed the tradi tion of what C . Lehmann has called "amnesia" about grammatical ization extends up to the presen t, for the word does not appear in the i ndex of Hock's Principles of Historical Linguistics ( 1 99 1 [ 1 986]), even though some of i ts principles do, nor does it figure in recent textbooks' of l i nguistics such as Fi negan and Besnier ( 1 989). Only very recently do we find the leading h istorical textbooks, such as McMahon ( 1 994), Trask ( 1 996) and Campbell ( 1 999) devoting significant space to grammaticalization. In the mid twentieth century "mai nstream" l inguistics was strongly synchronic in its approaches and assumptions, which meant that historical factors, i ncluding grammatical ization, were of secondary interest. Language change came to be seen as sets of rule adjustments, beginning with one stage and ending with another, but there was l i ttle i nterest in the gradual steps that must have been i nvolved in between : "the treatment of change as the change i n rules between synchronic stages isolates the description of change from the change itself' (Ebert 1 976: viii-ix). The only significant studies of grammaticalization during this period were done by Indo-Europeanists such as Kurytowicz (especially 1 964, 1 976 [ 1 965])
26
2 The histOlY of grammatica lizat ion
and Watki n s ( 1 964) who worked outside the domi nant theoretical parad igm. B u t their work, unfortunately, was read al most exclusively b y other Indo-Europeanists. Significantly, Mei l l et's student B enven iste, i n an article "Mutations of l i nguistic categories" written i n 1 968, found i t necessary to repeat m uch of what Mei l let had said in 1 9 1 2 concern ing the grammatical ization of au x i l i ary verbs out of lex ical verbs such as 'have, hold. ' B enveniste coi ned a new word, "au x i l i ation," to re fer to this process. Even though he used several of the very same examples which had been proposed by Meil let (e.g., the Modern Greek tha future fr.o m an earl ier tile/a ina), at no point in the paper did he expl icitly refer to Mei l let's work or use the term "grammatic(al)ization" or its equ i valent. That such an infl uential l i nguist as Benveniste could appear to be starting afresh in the study of the origins of grammatical categories indicates the extent to which Mei l Iet's i nsights had become submerged by twentieth-century structuralism. We have seen that grammaticali zation presents a chal lenge to approaches to language which assume di screte categories embedded in fi xed, stable systems. It is therefore not surpris i ng that grammatical ization again appears as a major theme of general (as opposed to specifically Indo-European) l i nguistics in the contex t of the ques tion ing of autonomous syntactic theory which occurred i n the I 970s. Duri ng this decade the growing interest i n pragmatics and typology focused attention on the predictable changes i n language types. Li ngu ists thereby (largely unconsciously) revived the same l i ne of i nvestigation that had been dropped earlier i n the century, a l ine which went back at least to Humboldt. An early paper by G i van perhaps be gan th is revival (Hopper 1 996: 220-2). Entitled "Historical syntax and synchronic morphology : an archeologist's field tri p," it announced the slogan "Today's mor phology is yesterday's syntax" (G ivan 1 97 1 : 4 1 3), and showed with evidence from a number of African languages how verb forms that are now stems with affixes could be traced back to earl ier col locations of pronouns and i ndependent verbs. If one of the main tenets of twentieth-century structuralism, especially as de veloped in the United States, was homogenei ty, another was the arbi trari ness of language, that is, its alleged i ndependence fi·om external factors such as the nature of th ings i n the world (the referents of language). Saussure had drawn attention to the arbi trari ness of the sign, for example, to the total independence of a word such as dog of the an imal it names. B u t he also stressed the fact that arbitrari ness is l i m ited by associations and "relative motivations." These include word compou nd ing as i n twenty-five, derivational affixation as in French pommier 'apple- tree' (pol1ll1le 'apple' + -ier), cerisier 'cherry-tree' (cerise 'cherry ' + -ier), and i n flectional para digms such as Latin dominus, domini, domino 'master-NOM, master-GEN, master DAT. ' I ndeed, he regarded grammar, the set of structural rules, as setting l i m i ts on the arbi trari ness and the chaotic nature of language ( 1 986 [ 1 922] : 1 30).
2.3
Research /rom the J 960s to the J 990s
27
One name given to the pri nciple that ensures non-arbi trari ness is "iconicity." Iconicity is the property of simi larity between one item and another. The philoso pher Peirce made a useful disti nction between imagic and diagrammatic icon icity. 2 Imagic icon icity is a systematic resem blance between an item and its referent with respect to some characteristic (a photograph or a scu lpture of a person are i magic icons). Diagrammatic icons are systematic arrangements of signs. None of the signs necessari ly resembles its referen t in any way, bu t, cruc ial ly, the relationsh ip among the signs m irrors the relationship among the icon 's referents: "those [ icons] which represent the relations . . . of the parts of one th i ng by analogous relations to their own parts are diagrams" ( 1 93 1 : Vol . 2, Part 277). For example, the model of language change in Chapter 3 is an iconic diagram of the relationsh ip between grammars of di fferent generations of speakers . It is diagrammatic iconicity which is of chief i mportance in l i nguistics, and which has suggested signi ficant i nsights in to the organization of language and i n to grammatical ization in particular. A very wel l-known example of diagram matic icon ici ty i n language is the tendency for narrative order to match the order of events described ; if the order is not matched, then some special marker or "diacritic" (usual ly a grammatical form) must be used . Thus Caesar's famous Velli, vidi, vici 'I came, 1 saw, 1 conquered' is a much ci ted example of the way in which order of mention mirrors order of action de scribed (see, e.g., lakobson 1 964 [ 1 960) ); any other order would require complex structures such as ' Before 1 conquered, I came and 1 saw. ' Another wel l-known example of diagrammatic iconicity in language is the way i n which pol iteness (soc ial d istance) is typically reflected in language by complex morphology and formal vocabu lary (often itself complex in structure), as exempl i fied by Good //laming (versus Hi!), Would you please pass the butter (versus Can J have the blltter?IPass the butter!) . Al though icon icity was a major topic i n much European l i nguistics, especially in the approach known as "semiotics'? or "semiology," i t was largely ignored as a principle i n American l i nguistics in the first three quarters of the twentieth century, when i n terest was focused on the arbi trari ness of language. Attention to icon icity was, however, renewed by several l i nguists working with issues germane to gram matica l i zat ion, most notably lakobson ( 1 966), Haiman ( 1 980, 1 983, I 985a), and Giv6n ( 1 985), who laid the foundations for much recent thinking on the subject. The value of the pri nciple of iconicity is most apparent in the context of cross linguistic work, and it is not coincidental that the period when iconicity came to be recognized aga i n was also a period of i nterest in typology of languages. Th is was a period, too, of i n tense i n terest in language universals, and some l i n guists began applying the idea of grammatical i zation to general problems of syn chronic descri ption that had arisen i n the course of the search for these universals. The work o f Li and Thompson was especially infl uential among those working
28
2 The history of grammaticalizatiofl
on h istorical issues. In their studies of serial-verb constructions in Chinese and other languages they showed that verbs could be reanalyzed as prepositions and case markers, and thus revived i n terest in the question of how categories come into being. For example, according to Li and Thompson ( 1 976a) in seventh-to ninth-century Chinese a "verb" ba occupied an ambiguous status between verb and prepositional case marker. In the example that fol l ows, these two possibi l i ties are suggested by two different translations of the same sentence: (I)
Zul bii zhO-gen-zr xl k!ln. drunk ba dogwood-tree careful look 'While drunk, (i) I took the dogwood tree and carefully looked at it.' (ba ' take') ( i i ) I carefully looked at the dogwood tree. ' (ba accusative case) (8th century AD, Du-fu poem; ci ted in Li and Thompson I 976a: 485) =
=
Such contexts, Li and Thompson argued, provide the staging for a reanalysis of the former "verb" ba ' take' as a m arker of the direct object of the verb and the collapsing of what had once been a sequence of two clauses (interpretation (i» into a s i ngle clause (interpretation (ii» . 3 Li and Thompson's work on word order and especially on topicalization showed, as Meil let's had done, that syntactic as wel l as morphological development was governed by constraints analogous to or even identical to morphological changes identified as i nstances of grammatical ization . Consider, for example, the distinction between a "topic" such as the i n i tial noun phrase in (2), as opposed to the in itial noun phrase i n (3): (2) (3)
That new yacht of his, he has spent a fortune on it. That new yacht of his has cost h i m a fortune.
In the first sentence, that flew yacht ofhis is said to be a "topic," wh ile i n the second sentence the same noun phrase i s a "subject." At this time there was considerable interest among l i n guists in arriving at a defi nition of the notion of "subject" that would be cross-l i nguisticall y valid. One of the ch ief obstacles was that in many languages the "subject" of a sentence was l i ttle more than a noun phrase in a very loose rel ationship to the verb (for i nstance, in some languages the sentence 'fire engines l ast night not have s leep' would be the normal way of say i ng 'I couldn't sleep last n ight because of the fi re engines ' ) . A n u m ber of properties, depending on the language i nvolved, m ight distin guish topics fro m subjects. For example, there might be verb agreement between subject and verb but not between topic and verb; sometimes subjects but not top ics could be referred to by a reflexive pronoun in the same clause. Moreover, i n some l anguages there appeared t o b e no o r very few topics, b u t a strongly de veloped notion of subject, while in other languages the topic appeared to be the
2.3 Research /rom the 1 960s to the 1990s
29
usual role of a primary noun phrase. These facts had often been noted , but Li and Thompson's work, by plac i n g them i n the context of grammatical ization, revealed the diachronic relationshi p between the two categories. The d i tference between a "topic" and a "subject" was h ypothesized to be one of degree of grammaticaliza tion only: "Subjects are essentially grammatical ized topics . . . " (Li and Thompson 1 976b: 484). This work suggested to many li nguists at the time that a diachronic perspective might offer m ore than merely an interesting historical comment on synchrony ; synchronic "facts" were i nd istinguishable from the diachronic and discourse pragmatic process they were caught up in (see, for example, Bol inger 1 975). Greenberg's empirical cross- l ingui stic study of word order (Greenberg 1 966a) was foundational for the l anguage universals movement. In this work a number of absolute and dependent ("impl icational") generalizations about syntax were claimed, such as a statistical correlations between verb-fi nal word order, post positions, and genitive precedi ng possessed noun in the possessive construction. Clearly, a diachronic perspective on these universals was possible; not only could changes in word order be understood typologically, but synchronic syntax and morphology could be seen as the temporary - and not necessarily stable - reflexes of ongoing shifts. Other cross- l i nguistic work by Greenberg that was sem inal for work i n grammaticali zatio n i ncluded his study of the development of demonstra tives i n to articles and ultimately gender markers (noun classifiers) via agreement markers (Greenberg 1 97 8a), and of numeral systems and their structure in terms of the order of elements i n the numeral phrases, and their syntactic relation to the noun head (Greenberg 1 97 8b). The first of these was primari ly historical in focus, the second primarily synchron ic, but both highl ighted the importance of a dynamic approach to language structure. To return to our brief history of work i n grammaticalization, Giv6n's book On Understanding Gramma r ( 1 979) was a highly i n fl uential , if sl ightly idiosyn cratic, summing up of the decade's thought on these matters . It firmly placed all l inguistic phenomena i n the framework of "syntactici zation" and "morphologiza tion" (terms which G iv6n preferred to "grammatical ization"), and emphasized the essential functional dependency of l i nguistic rules and categories. The forms of speech, G iv6n proposed, were to be viewed as bei ng located on cli nes and as shifting between such poles as child/adult, creole/standard, unplanned/planned, pragmatic/syntactic. In each pair of these, the first is labile or "loose," the second fixed or "tight," and movement - i .e . , change - is general l y in the direction of the "tighter" pole. He proposed ( 1 97 9 : 209) a path of grammatical ization of the type: discourse > syntax > morphology > morphophonemics > zero
30
2
The histOlY of grammaticalizatiol/.
Givo n ' s work distil led for the growing com m u n i ty of workers in gram mat ical ization some of the highly relevant research on morpholog i zation i n pidgins and creoles such as that represented by B ickerton ( 1 975), Sankoff ( 1 980), and, in child l anguage acqu isition, by Slobi n ( 1 977). Unfortu nately, a fu l ler integration of soc io l i nguistic and developmental research with research on gram matical ization sti l l remains t o b e worked out. The l ast decades of the twentieth century were a period i n which cross- l i nguistic projects flourished . The l arge computeri zed data banks that were developed owed much to the original i n spiration of the Stanford Project on Language Universals, headed by Ferguson and Greenberg, which c u l m i n ated i n Green berg, Ferguson, and Moravcs i k 's Universals of Human Language ( 1 97 8), a four-vol ume set of papers. many o f them on topics cen tral to grammatical i zation. Other more recen t proj ects incl ude two in Col ogne, one on Language Universal s and Typology, headed by Seiler and disseminated through the publ ication A rbeiten des KiNner Ulliversalien Plvjekts, the other on Gram matical ization i n African Languages headed by He ine and d i ssemi nated through Afrikallistische A rbeitspapiere. Especially i n lluential has been a survey of the morphology associated with the verb headed by Bybee, i n i ti ated in the early 1 980s at the State Univers i ty of New York at B uffalo. At fi rst a data base of fifty languages was used (exempl i fied by Bybee, Morphology: a Study of the Relation between Meaning and Form, 1 985); l ater i t was expanded to over seventy l anguages (exempli fied by Bybee, Perkins, and Pag l i uca, The Evolution of Grammar: Tense, A spect, and Modality ill the Languages of the World, 1 994). In the early 1 990s a typological study of the European languages (Eurotyp) was headed by Konig of the Free University of Berl i n , the results of which are being publ ished i n a series of n i ne vol umes . Among other cross l i nguistic projects at the time of wri ting are two at Max Planck Inst itutes, one headed by Lev i nson at the Max Planck Insti tu te for Psychol inguistics in Nijmegen, the other by Comrie at the Max Planck Insti tute for Evolutionary Anthropology in Lei pzig.
2.4
Recent trends i n resea rch o n g ra m m atica l ization
The 1 980s saw grammatical i zation (also frequen tly cal led "grammati cization") assu m i ng a significant p l ace as a topic i n its own right i n the re search of a n u m ber of l i nguists. A very i m portant monograph from the earl ier part o f the decade was C. Lehmann's Thoughts 01/. Grammaticalizatioll : a Pm grammatic Sketch (publ i shed as a work i ng paper in 1 982 and then rev ised as a monograph in 1 995 ). It was the first modern work to emphasize the conti nu ity of research from the earl iest period (roughly, Humboldt) to the present, and to
2.4 Recellf trends in research
31
provide a survey of the signi ficant work in grammatical ization up to that t i me. with emphasis on h istorical l i nguistics. B y i ts comprehensiveness and its historical perspective this book provided a usefu l antidote to the apparent assumption of many li ngu ists of the time that the concept of gram matical i zation had been newly invented . Furthermore, Lehmann laid out a set of "parameters" accord i ng to which degrees of grammatical i ty cou ld be measured synchron ical ly and diachron ically. Th ree parameters are associated with each of Saussure's two axes : that of selec tion in one position or slot (the "parad igmatic axis") and that of combi nation in sequence (the "sy ntagmatic axis"). O f relevance on the paradigmatic axis are: I . the "weight" or size of an element (Lehmann refers to "signs") ; weight may be phonological (Lat. ille 'that' has more phonological weight than the French article Ie that derives from i t) or semantic (the motion verb go is thought to he semantically weightier than the future marker go) - "Gram matical i zation rips off the lex ical features until only the gram matical features are left" ( 1 995 : 1 29); 2. the degree to which an element enters into a cohes ive set or paradigm: e.g., Latin tense is paradi gmatical ly cohesive whereas English tense is not (contrast the Lat in w i th its translation i n alllo 'I love,' amabo 'I w i l l love, ' Cllllavi ' I have loved ' ) ; 3 . the freedom with which a n element may be selected ; in Swah i l i i f a clause is transitive, an object marker must be obl igatorily expressed in the verb (given certain semantic constraints), whereas none is requi red i n English. Of relevance on the syntagmatic axis are: 4. the scope or structural size of a construction; periphrasis, as i n Lal. scribere habeo ' write: INF have: I stSg' , is structurally l onger and weightier and larger than in nection, as in Ital . scrivero, 'I shal l write' ; 5 . the degree of bond i ng between elements in a construction (there is a scale from clause to word to mor p heme to affix boundary, 1 995 : 1 54); the degree o f bond ing is greater i n the case of i n llcction than in that of peri phrasis; 6. the degree to which elements of a construction may be moved around ; i n earl ier Latin scribere habeo and habeo scibere cou ld occur in ei ther order, but in later Lat in this order became fixed , which allowed the word boundaries to be erased . As these few examples have shown, the parameters are correlated . Histori cal ly, one can measure the degree of correlation among the parameters for some individual construction (e.g. scribere habeo). Synchron ical ly, one can arrange a variety of historical l y u nrel ated elements with simi lar function on these parameters (e.g., expressions o f possibility such as possible that, maybe, may). Many stud ies of gram maticalization have been wri tten assu m i ng the correctness of these parameters, though as we shall see, some have been questioned, espec ially I in
32
2 The history of grammaticalizatioll
so far as it pertains to semantic weaken i n g or "bleachi ng", 3 in so far as it sug gests that "obligatorification" is a necessary, and 6 in so far as it suggests that gramm aticalizatiol1 necessari ly i nvol ves i ncreased fi x i ng on the syntagmatic axis. Some scholars appear to assume (contra Lehmann) that i f a form has not become seman tically bleached, "obl igatorified," and bonded, then it has not u ndergone grammaticalization . B u t the term "grammatical i zation" i s appl icable to many sit uations i n which these changes are less than categorical . To take an extreme case, French pas "negative particle" has ousted all of i ts competitors except one, point. We would not want to deny that pas (once a lexical noun 'step ' ) is now grammat ical i zed as a negative marker si mply on the grounds that speakers can stil l choose point (once 'dot, poi nt' and now an emphatic negative particle) in i ts place, as i n elle lie tn 'a point aidee 'she didn ' t help m e a bit' versus elle n e m 'a pas aidee 'she didn ' t help me. ' S i nce historical processes are always ongoing, and furthermore are not determ i n istic (i.e., they can a lways be abandoned) we will often not see a completed i nstance of grammaticalization, only one i n progress, and therefore such claims are suspect. Hei ne and Reh 's book Grammaticalization and Reanalysis in African Languages ( 1 984) was perhaps the first to address an entire l inguistic area (Africa) synchroni call y from the poi nt of v iew of grammatical ization. It was a convincing demonstra tion of the power of grammatical ization theory as a tool of descriptive l i nguistics and especiall y of the wide range of l i nguistic facts that grammaticalization could characterize. It provided not only a classification of the phenomena to be addressed, with copious examples, but also an i n d ex of the typical pathways of grammatical ization d iscovered by the authors, particularly with respect to morphosyntax and morphophonology. It goes wi thout say ing that most of these phenomena are not restricted to Africa, but have counterparts elsewhere in the world. Heine has spear headed exhaustive extensions of the i ndex first to additional African languages (see Heine et at. 1 993) and subsequently to the languages of the world (Heine and Kuteva 2(02). B oth Lehmann's and Hei ne and Reh's books focused on morphosyntax, with relatively l i ttle attention to meaning. In the I 980s and especially the 1 990s i ncreas ing attention was paid to semantics and pragm atics as researchers bui l t on Givon's ( 1 979) hypothesis about discourse > syntax > morphology > morphophonemics > zero or explored the cogni tive underpinnings of semantic change. Debate con cerned the role of metaphor and metonymic association i n the flow of speech (see Chapter 4). In her sem i nal book, The Future in Thought and Language ( 1 982), on the development of modals of obl igation to m arkers of epistemic stance and future tense in Romance l anguages, Fleischman's crucial i nsight was that mor phology, syntax, and seman tics a l l interconnect i n ways that lead to patterned,
2.4
Recent trends ill research
33
non-random change. Traugott ( 1 982) suggested that there are semantic/pragmatic factors in grammaticali zation that lead to unidirectionali ties of change, specifically tendencies leading from concrete mean ings to more abstract ones, and i n particular to ones grounded in the speaker's assessment of con nections between propositions. Concurrently work went forward in cognitive l inguistics, drawi ng attention nol only to semantics but also to (largely synchron ic) metaphors. Sweetser's book From Etymology to Pragmatics ( 1 990) suggested ways in which relationships among polysemies of modals (especial ly may) and connectives (especially but and if> might be conceptualized metaphorical ly. In 1 99 1 , Heine, Claudi, and Hiin nemeyer published Grammaticalization: a Conceptual Framework ( 1 99 I a). The data are pri marily African, but the focus is on pragmatic and cogni tive factors that mo tivate grammaticalization, and the meaning changes that forms may undergo as they grammaticali ze. For example, the authors regard "metaphorical abstraction" as one of the means by which we organize the world around us. Various relatively abstract concepts such as time, cause, manner, personal quality, and relationsh ip need to be expressed; more concrete l i nguistic "vehicles" are pressed into service to express them. Thus, time concepts are typically expressed in terms of more readily apprehensible space concepts (a "long" time, a "short" time, etc.), men tal activities l i ke thinking are expressed in terms of physical activities (to think "hard", etc.), and so on. Those metaphorical abstractions that lead to the emer gence of new structures, e.g., adverbials or prepositions, are among the main k i nds of cognitive processes that lead to grammaticalization according to Heine, Claudi, and Hiinnemeyer ( 1 99 1 a: 4 1 -5). As mentioned at the end of Section 2.3, Bybee's large-scale project on mor phology, primarily verbal, has been a prom i nent and highly influential source of ideas and data. She and her collaborators crucially see grammatical ization as both semantic and formal in nature. Among impl ications for cross-l i nguistic work on grammaticalization is the claim that· grammatical morphemes or "grams" can be studied not only as language-specific phenomena, but also as "gram-types" that are substantive u niversal categories analogous to "voiceless dental stop" i n phonetics (Bybee, Perkins, and Pagliuca 1 994: 1 49). 4 They tend to be polysemous in similar ways across languages, and to u ndergo similar paths of development as a result of human discourse and i n teractio n : "they reflect the metaphorical processes that are based on human cognitive make-up, and they reflect the inferences that humans commonly make when they communicate" (p. 302). A two-volume collection of papers arising out of a 1 988 conference organized by Givan, and edi ted by Traugott and Heine under the title Approaches to Gram maticalization ( 1 99 1 ), addresses a wide spectrum of themes, many of which are stil l of current concern to a wide range of l i ngu ists working from the perspective
34
2 The history of grammaticalization
of gram matical i zation, and many of which w i l l be elaborated in the foll owing chapters , for example: (a) (b) (c) (d) (e) ( I) (g )
Can diac hronic and synchronic approaches to grammalicalization be rec onci led , or is a new approach requ i red ? Is gram matical i zalion a continuous or discon ti nuous process? To whal extent is grammatica l i zation the result of discourse pragmat ic forces? What constrai nts are there on the choice of concepts and forms serv i ng as the i nput to grammatical i zation? When can i nc i p ient gram matica l i zation be recogn i zed ? Is gram matical i zation a u n i d i reclional phenomenon? What phenomena in language are not examples of grammatica l i zation ?
O f these, question ( f) abou t unidirectional i ty came to take center stage al the turn o f the centu ry. As a testable hypothes is, u n i d i rect ional ity attracted e fforts to provide counterexamples. A chapter in Newmeyer ( 1 998) enti tled "Decons truc ting gram maticalization" laid out some of the theoretical assumptions o f this en terprise . A spec i a l number of Language Sciences (Campbell 200 I a) was devoted to auack ing the entire premises of gram matical i zation in artic les with such provocat ive titles as "What's wrong with gram matical i zatio n ?" (Cam pbel l 200 I b), and "Is there such a th i ng as gram matical ization?" (Joseph 200 I ) . The bulk of the argu ments was devoted to the d i scussion of a smal l n u m ber of cases in which a reversal of unidirec tional i ty can be argued, some of these, h owever, such as the Engl ish possessive 's, bei n g themsel ves q u i te con troversial . On the other side, strong support for the hypothesis of u n i d i rectional i ty was put forward i n a special number of Linguistics ed ited by van Kemenade ( 1 999a), and in papers such as Haspel math ( 1 999a). the t i t l e of which was "Why is grammatical i zation i rreversi ble?" We return in Chapter 5 to the controversies engendered by these papers . During the period un der revi ew, a new word gai ned promi nence: "exaptation," the deployment of bits and pieces of lex ical and morphological material pressed i nto service as grammati cal formatives (Lass 1 990, 1 997 ; Vincen t 1 995 ). Did exaptation present a chal lenge to the theses o f grammatical ization and u n i d i rectional i ty, or was it si mply what wou ld be ex pected from the usual pre m ises of gram matical ization - the creation of grammatical out ? f previously non-gram matical forms? The related concept of "phonogenesis" (Hopper 1 99 1 , 1 994), the creation of new phonological mate rial out of the random exti nction of morphemes, also entered the vocabul ary o f grammatical i zation. We will here i ndicate some addi tional signi ficant theoretical trends, the more i m portan t of which w i l l be taken u p later. As various theories of syn tax have
2.4
Recent trends
ill
research
35
devel oped , and i nterest in h istorical morphosyntax and espec ially grammal ical i za l i on has i ncreased , the rel ationsh i p of gram matical ization to reanalysis has been d i scussed at length - are they the same thi ng, are they d i fferent, and if so does gram matical ization " fall out from" reanalysis (see, among many others, Abraham 1 99 3 ; I. Roberts 1993a; Tabor I 994a; A . Harris and Cam pbel l 1 995 ; Haspel math 1 99 8 ; New meyer 1 99 8 ; van Kemenade I 999b)? What is the relat ionship between morphosy n tactic change and lexical izat ion (C. Lehmann 1 989a, 2002 )? What arc the d i ffering roles of seman tics and pragmatics in language change (Traugol l and Dasher 2002)? W ha t are the boundaries of grammatical ization ( G i acalone Ramat and Hopper 1 998)? A particularly strong movement i n recent years has been the i n tegration or gram matical ization w i th the notion or "usage-based" structure . The term is owed to Langac ker ( 1 987), and found echoes i n the thought of severa l li ngu ists at that l i mc who were growi ng u ncomfortable w i th sentence analyses based on the l i nguist's priv.ate i ntrospection. Thus, Hopper ( 1 987) envisaged gram mar as an emergent property of usage rather than a static autonomous entity. Hi mmelmann ( 1 992. 1 997) drew on this idea in characterizi ng the consti tuency of nom i nal groups and of deictic systems. Du Bois ( 1 985 , 1 987) argued for a discourse origin of certain wel l-establ ished morphological mean i ngs such as the ergati ve/absolutive case. Haiman ( 1 994) drew a parallel between actions that become ritual ized through repetition and l i nguistic expressions that lose their novel ty and become autom ated Ihrough frequent use, thus pointing out close s i m i lari ties between language change and change i n other aspects of human communicative behavior. Croft (2000) high l ighted the replication of "utterances, more precisely the replication of l i nguistic structures in utterances i n language use" (2000: 7 ) across speakers and commu n i ties through time. Many of the papers in Bybee and Hopper (200 I ), from a conference held at Carnegie Mellon i n 1 999, i n vestigated frequency of forms and juxtaposition of forms (the "prox i m i t·y" model) as a causal mechan ism of change. These studies typically focused on the col locations of particu lar forms and the for mation of constructions with particular lexical i tems. In simi lar vei n, Krug ( 1 998 ) argued spec i fically for atten tion to "stri ng frequency" as a pri me motivating force in change and the emergence of categories. Scheibman (2000) descri bes the corre lation between freq uency and combi nations of don 't in conversation, as in dl/wID 'don ' t know. ' One outcome of t h i s direction has been a tendency to see gram matical izat ion (and grammar) i n terms of collocations of specific items rather than general ized changes, i n other words to iden t i fy the possible beg i n n i ngs as wel l as the ends of grammatical constructions. C. Lehmann ( 1 99 1 ) noted the emergence o f new prepositions i n German through prepositional constructions that occurred in
36
2 The history ojgrammaticalization
specific textual contexts, e.g., im Vorfeld 'on the eve of' always suggested an im mi nent newsworthy event or journey. Hopper ( 1 994) l i kewise d iscussed incipient grammati c a l i zation i n specific idiomatic contexts. Bybee, i n a number of publi cations (e. g . , Bybee 200 1 , but also going back as far as Hooper [Bybee] 1 976), reasoned that much change, especially that relevant to grammaticali zation, was a strictly mechan ical matter, the sl urri ng together of frequently juxtaposed forms. S i nce it i s particul ar words, not abstract classes, that undergo thi s development, Bybee too saw change, and therefore structure itself, as beginning in specific con texts. Much of this work was made possible by the burgeoning of corpus l i nguistics i n the 1 990s, which i n turn owed its impetus t o n e w technology - computer-aided pho netic analy s i s replacing mechan ical sound spectography, massi ve electronic cor pora w i th h igh-speed search software and funded projects based on them , which allowed l arge bodies of texts to be scanned for repetitive patterns and gradual grammatical changes. Historical corpora such as the Helsinki Corpus of histori cal Engl i s h texts (cf. Rissanen et al. 1 993) and the Toron to Old Engli � h lexicon and its associated corpora (Cameron et al. 1 986-) were significant tools. In ad dition, the construction of many corpora of written and spoken modern English gave insights into the role of frequency i n contemporary usage. Among writ ten corpora are the Brown Corpus of American English (Frances and Kucera 1 96 1 ), and the Lancaster-Oslo/Bergen Corpus of British English (see Johansson, Leech , and Good l uck 1 978); these and the Helsinki Corpus are among many databases combined i n the International Computer Archives of Modern English (lCAME). Precisely transcribed conversational corpora such as the Santa B arbara Corpus of S poken American English (see Chafe, Du Bois, and Thompson 1 997) and the B ritish National Corpus (of spoken and written British English) (BNC) have provided crucial insights into conversation. A good many studies of the past decade h ave drawn upon such synchron ic corpora to argue for the "emergence" of types of constructions that become visible only through the close quantitative study of u sage rather than through i ndividual i n trospection . Among many such studies can be mentioned Laury ( 1 997) on the emergence of a definite article in Fi nnish and Krug's work on emergent auxiliaries in English (Krug 2000, 200 1 ). A significant body of research on the narrowing gap between "grammar" and conversati onal usage that has characterized recent l i nguistics is represented by Ochs, Schegloff, and Thompson ( 1 996). Several proposals were i n the air also concern i ng an emergent relationship of broader syntactic phenomena to natural conversational discourse, such as relative c lauses (Fox and Thompson 1 990) and constituency itself (Lerner 1 996; C lancy et al. 1 996; Ono and Thompson 1 994). The relationship of such studies to the nature of grammaticalization remains to be worked out.
2.4 Recellt trends in research
37
The general problem of characterizing language change - especially. but by no means exclusively. change in meaning - continued to occupy researchers. This direction has meant that researchers have had to rely increasingly on languages whose histories are wel l attested and understood. such as English (and some others. such as Chinese. French . German. and Japanese). Because there is so much material on the h istory of Engl ish. and much of the considerable body of research on the details of the h istory of English has implicit and expl icit relevance for grammati cal ization. theoretical points have been reasoned for with detai led historical data especially from English. To mention just a few of very many studies on Engl ish: Traugott ( 1 995) postulated. on the basis of diachron ic data. a cline of grammati cal ization of the type: Cl ause-internal Adverbial > Sentence Adverbial > Di scourse Particle
arguing (as does Brinton 1 996 with respect to different data) that discourse mark ers l i ke clause-i n i tial ill fact and indeed are members of grammatical categories because they are operators on discourses and serve grammatical fu nctions akin to topic and focus (see also Fraser 1 988 on the grammatical category status of dis course markers). Van Kemenade ( J 999a) was devoted to many theoretical and sub stantive questions of historical Engl ish morphosyntax. as was Fischer. Rosenbach. and Stein (2000). which presented several case studies of grammaticali zation in the history of English. The introduction to the l atter (Fischer and Rosenbach 2000) is a useful and j udicious summary of the state of thinking in the field. Other areas of the world have. however. not been neglected. and i n recent years too. many l inguists have presented studies of individual languages. language fami l ies. and l i nguistic areas. and the closely related topic of creoles. from the perspec tive of grammatical i zation. e.g Fernandez-Vest (2000) for Balto-Fi nnic; B isang ( 1 996) for Southeast Asia; Bruyn ( 1 995). Baker and Syea ( 1 996). Turchetta ( 1 998). and Romaine ( 1 999) for creoles; Fraj zyngier ( 1 996a) for Chadic; Hook ( 1 99 1 ) for Indo-Aryan ; Aikhenvald (2000) for South A merica; Sun ( 1 996) and Shi (2002) for Chi nese; Dasher ( 1 995) and Ohori ( 1 998) for Japanese; Clancy et al. ( 1 996) for Japanese and Chinese. among many others. Many of these continued the com parative trad i tion begun by Heine and Reh in the 1 980s for Africa of studying trends and parallel developments in l inguistic-cultural areas from the perspective of grammaticalization . A number of longer monographs based on grammatical ization also appeared. such as Haase ( 1 994). presenting universals of honorifics as the "grammaticali zation of pol iteness." and Kilian-Hatz ( 1 995). a description of the grammar of the African language Baka as the product of grammatical iza tion. The latter book could be seen as an application of the earl ier method of "h istorical grammar" of Indo-European languages to the structure of an unwri tten language. .•
38
2
The
history of grammaticalization
As this s urvey has suggested, the past decade has seen a rich expansion of i n terest i n grammaticalization with both a wide range of descriptive studies and a v igorous debate about questions of theory and pri nciples. Inevi tably, "grammatical ization" has been u nderstood and theorized i n a number of d i fferent ways. Our aim i n the remainder of this book is to develop a synthesis of current thinking on gralll mat ical ization that will provide the basis on w h ich further work can be bu i l t. It is to this task t h at we now turn. ,
3
M e ch a n i sms : rea n a lys i s a n d a n a l ogy
3. 1
I ntroduction
We turn now to some central concerns i n any discussion of languagl! change, w i t h focus on those that are particularly important for an understand ing of grammatical i zation. In particular, we attempt to answer the questions: what moti vates c hange, what mechanisms lead to gram matical ization, what are i ts prob able "paths" of progression through time, and what are i ts end results? Particu lar changes do not have to occur, nor do they have to go through to complet ion. though some degree of change is i nev itable. As elsewhere in this book, therefore. we w i l l be referri n g to phenomena that make change possible or facilitate i t , some ti mes s i ngly, someti mes together, not to factors that are absolute or obl igatory. In t h i s chapter we consider two general mechanisms by which grammatical ization lakes place: reanalysis pri mari ly, and analogy secondari ly. In Chapter 4 we will dis cuss speaker/hearer asym metries and processes of mean i ng production and percep l i o n that motivate the operation of these mechanisms, and also some seman tically moti vated mechan isms i ncluding metaphor and metonymy. The unidirectional ity o f paths of change w i l l be the subject of Chapters 5 , 6, and 7. In Chapter 8 we w i l l d i sc uss grammatical ization i n t h e context of t h e development o f creoles. Reanalys i s and analogy have been widely recogn i zed as significant for change in general, most especially morphosyn tactic change. In reanalysis, the grammatical syn tacti c and morphological - and semantic properties of forms are modified. These mod i fications comprise changes i n i nterpretation, such as syntactic bracket ing and mean i ng, but not at first change in form . Reanalysis is the most i m portant mechanism for grammatical i zation , as for all change, because it is a prerequisite for the i mplemen tation of the change through analogy. Analogy, strictly speaki ng, mod i fies surface man i festations and in i tself does not effect rule change, I although it does etfect ru le spread either with i n the l i nguistic system i tself or with i n the com m u n i ty. For a very si mple example of the ditference between the two mechanisms, consider the d i fference between the compounding in Old Engl ish of the phrase cild ' c h i l d ' + had ' person, cond ition, ran k' i n to childhad, 'childhood ' or biscop
40
3
Mechanisms: reanalysis and analogy
' bishop' + had to biscophad 'bishophood ' on the one hand, and the extension of hood to n ew environments, such as falsehood in Middle English. Cildhad and biscophad i l lustrate at least two instances of reanalysis: first two independent nouns come to function as one (compoundi ng), then the second comes to be used as a derivative morpheme representing an abstract state (semantic and morphological c hange). Falsehood i llustrates analogy : once -hood had come to be used as a derivative morpheme it no longer required association with a word referri ng to a person, and could be extended to new contexts, i n this case an adjective referring to an abstract concept. In cases l i ke th is, the word boundary of the root, e.g.false, is reanalyzed as a morpheme boundary to allow derivation.
3.2
Some backgro u nd ass u m ptions about cha n ge
This is not the appropriate context for discussing principles of l anguage change in detail . For ful ler accounts of these principles, see Anttil a ( 1 989 [ 1 972]), Hock ( 1 99 1 [ 1 986]), McMahon ( 1 994), and more specifically on syntactic, mor phological change, A. Harris and Campbell ( 1 995), and on phonological change Kiparsky ( 1 988). However, before we proceed, some i n i tial comments on lan guage c hange w i l l be helpful i n clarifying certai n assumptions behind the material to fol low. First, when we speak of change, what is thought to be changing? We speak loosely of "language change." But thi s phrase is misleading. Language does not exist separate from its speakers. It is not an organism with a l i fe of its own ; rather each speaker has to learn that language anew. Change is replacement (Hoenigswald 1 966), on the u nderstanding that "replacement" does not entail strict identity of an earlier function or category with a later one (see discussion at the end of Section 1 .2.3). However, in so far as language is characterized by an abstract set of rules i ndependent of language users, the rules (or set of rules) can be said to change. D i fferent models of rule change have been suggested. The one most influential i n the last four decades has been the generative model. Th is model privileges rule change in terms of high-level global organization and of the whole set of ru les (the "grammar") over individual rule changes. Furthermore, it assumes that i n general, or as an i deali zation, major changes (called "restructurings") can occur only i n the discontin u i ty of transmission from one generation to another, in particu lar during the process of child language acquisition in a homogeneous speech community. The factors that enable th is transmission are twofold: universal capaci ties for language and u n i versal reasoning processes that language users bri ng to the output of the earl ier grammar.
3.2 Some background assumptions about change
41
LEA RNER
I I I -------,------------I I r----------f---------. , , I
:
Grammar !
D
:
,- - - - - - - - - - - iI - - - - - - - - - �,
I Figure 3 . 1 Model of language change (Anttila 1989: 197)
An early characterization of such ru le change was modeled i n Andersen ( 1 973: 778), and modified by Anttila ( 1 989 [ 1 972] : 1 97); i t is shown i n Figure 3 . 1 . In this model, Grammar l is the internal ized set of rules i n an i ndividual . This speaker's verbal output (Output l ) is determ ined by Grammar l . In a later generation the l anguage learner, endowed with certain universal capacities for language, hears Outpu t l . Using the universal l i nguistic capacities or "Laws," and universal rea soning processes, the learner i n fers an internalized grammar which may be dif feren t from that of the earlier speaker, i n which case it is termed Grammar2 (for the differences among the inferences, and the types designated by I, D, and A, see Section 3 .2. 1 immediately below). This internalized grammar is verbalized by Output2 which is different from Output l because i t is the verbal i zation of a d ifferent grammar. The model is a useful one for conceptualizing change, and will serve our pur poses in th is book provided it is understood in the l ight of assumptions about grammatical ization rather than the more rigid generative ones to which it has l argely been adapted. However, attention should be drawn to some of the assump tions that were origi nal to the model or that have been made about it in subsequent years. We focus on issues regarding the types of i n ferences i n language acquisition, who the l anguage learner is (ch ild or adu l t), what needs to be learned (how much is genetically endowed), and how i nnovation spreads.
3.2. 1
Induction, deduction, abduction
In this section we consider some basic logical principles of reasoni ng, known as i nduction, deduction, and, most i m portantly for change, including gram maticalization, abduction. An ideal ized artificial l anguage, for example, a com puter language, can be thought of as a cod i ng device in which ready-made ideas are converted into symbols that serve one and only one fu nction. Here a principle
42
3
Mechanisms: reanalysis and analogy
of "on e form - one meani ng" operates, and every "utterance" conveys an unam biguous message. Such transparency is not fou nd i n h u m an language. Th is is partly because i n real-world languages a smal l set of u n i ts and constructions must serve a much l arger set of functions, owing to memory and parsing l i m i tations. Moreover, language is a social institution, and one of its important functions is to mainta i n social n etworks a n d sustain interest i n a verbal i n teraction . Therefore i ndirectness (such as is fou nd in pol iteness phenomena), metaphor, and other non-li teral mean ings are an essential part of language. "One form - one mea n i ng" wou ld i n these circumstances be dysfunctional. For example, Do you mind not smoking in here ? can serve as a request for information, o r a command to stop smoking i n the gu ise of an i nq u i ry. After extensive use as an indirect command it can be felt as too "routi ne," hence too close to Please stop smoking, and therefore can be substi tu ted in some circu mstances by a lengthier paraphrase l i ke Would you mind awfully if I were to ask you not to smoke in here ? Part of the hu man abi l i ty to u nderstand and use language i s the abi l i ty to reason from the form of what is said to the in tent of what is said, as wel l as from the string of sounds that occurs as i nput .[0 the structure behind that input. Logicians focused until recently on two types of reason ing: induction and de duction. If h uman language were an arti ficial language then these logics m ight suffice. However, neither of these l ogics accou n ts adequately for indirectness, ex pressivity, or change. For this a third type of reasoni ng, "abduction," first identi lied by C. S. Peirce ( 1 965 [ 1 93 1 )), needs to be considered . The importance of abduction for language change has been stated particularly clearly by Andersen ( 1 973). The following is based on Andersen's main poi nts (especially 774-86; see also Anttila 1 989 [ 1 972] : 1 96-8). Types of reasoning are exempl i fied by three proposi tions that constitute a syl l ogism: The Law (e.g., All men are mortal) The Case (e.g., Socrates is a man) The Res u l t (e.g., Socrates is mortal) Deductive reason i ng appl ies a law to a case and pred icts a result (e.g . , A ll men are mortal, Socrates is a mall, therefore Socrates is mortal) . S trictly speak i ng, the concl usion asserts nothing that is not given in the prem ises ; furthermore, if the premi ses are true, then the conclusion is also. Inductive reason i ng proceeds from observed cases and results to establish a law (e.g., Socrates is a man, Socrates is mortal, therefore all men are mortal). Abductive reasoni ng is different, although it is often con fused w i th inductive reasoning : "Abduction proceeds from an observed result, i nvokes a law, and i n fers that somet h i ng may be the case. E.g. given the fact that Socrates is dead, we may
3.2
Some backglVund assumptions about change
43
relate this fac t to the general law that al l men are mortal, and guess that Socrate� was a man" (Andersen 1 973 : 715 ). Even if the premi ses are true, the conclusion need not be so: one may match the wrong result with the l aw. Perhaps Socrates is not a man bu t a l i zard, a wrong conclusion but nevertheless one that is compat ible with the other two prem ises. The law may be an establ ished truth, or i t may he a ten tative general ization. Peirce was i nterested in abduction because, although he saw it as a weak form of reason i ng ( i ndeed, it can lead to logical fallacy), he also saw it as the basis of human perception and as the only k ind of reasoning by wh ich new ideas cou ld originate. Andersen, and many l i ngu ists after h i m , have regarded abduction as essential to developmen t of cul tural patterns, i nc l ud i ng language. Of the process i tself, Andersen says : "In acqu iring his [sic] language, a learner observes the verbal activity of his elders, construes it as a ' result' - as the output of a gram mar and guesses at what that grammar m ight be" ( 1 97 3 : 716). The guesses are processes of reasoning based on universal pri nciples, the basic goal being the constru ction of a grammar (the case) that in some way conforms to the ob served data (the result). Abduction is the predomi nant mode of reasoni ng i n language learn i ng (Anttila 1 989 [ 1 972] : 1 97 ) . It is constantly tested out b y the process of induction (the matc h i ng of a hypothesis to the data) and by deduc tion (the production of new u tterances based on the hypothesis). In Figure 3 . 1 the c urved arrow from Output I through Universals models abduction (A). The straight arrow ti·OJTI Universals through Grammar2 to Output2 models deduction (D), and the curved arrow from Output l through Output2 to Grammar2 models i nduction (1). 3.2.2
Who is the language learner?
A ndersen writes throughout h is ( 1 97 3 ) article of "language learners," wi thout specific com m i tment to the age of these learners. Many researchers have in terpreted "language learners" as c h i ldren, most especially children i n the first two or three years of l i fe. This interpretation goes back a long way. In the early part of the twen tieth century, Hermann Paul ( 1 920) was particularly concerned with developi ng a theory of the relationsh ip between ch i ld l anguage acquisition and "evol u tive" change, that is, change that is regarded as only min imally affected by outside factors, such as conquest, demographic changes, or migrations. In the early I 960s c h i ld language acq u i s i tion was accorded a central theoretical place in generative theory, whether synchro n i c or d i achronic, because it was seen as the potential locus for insights i nto learnabi l i ty, that is, i nto the human-specific cogn itive factors that make language possi ble (see especially Chomsky 1 965 ; and , with respect to language change, Hal l e 1 964; Lightfoot 1 99 1 , 1 999). According
44
3 Mechanisms: reanalysis and analogy
to this view, the discontinuity between adul ts and children enables major changes, but the discontinuities with i n a person's l i fe do not. However, it is becoming i ncreasi ngly widely accepted among sociol i nguists and researchers on language acquistion that people continue to develop language skil l s throughout their l i ves, and also to innovate. As early as 1 982, Bybee and Slobin , studying children's acquisition of verb forms such as send-sent, sing-sang sung, and their i nnovations, such as think-thunk, concluded that: "There is noth i ng particularly special about the relation between small children 's i nnovative forms and morphophonemic change. The i nnovations of older children and adults, al though perhaps rare, where they can be elicited, may also serve as predictors of change" (Bybee and S l obin 1 982: 36-7). This position has been confirmed and elaborated on in e.g. Labov ( 1 994), and Ravid ( 1 995). Furthermore, there is i ncreas ing awareness that it would be "very difficult to demonstrate, beyond reasonable doubt, which of the many innovations observed i n child language . . . will actually be accepted by speech communities and become l i nguistic changes" (Milroy 1 992: 204). Thi s i s particularly true in the case of h istorical data from the past, because it is written and does not reflect child l anguage directly. Although children may in part play a role i n language change, there is growing evidence that young adults play a significant one as well . B oth groups innovate and the spread of innovations appears to occur at any age. However, the role of the "devel opmental impera tive" among adolescents to d isplay knowledge of and use the l i nguistic market place appears to be especiall y important in maintai n i ng and replicati ng innovations across communities (see Eckert 1 988, 1 997, 2000; M i l roy 1 992; Chambers 2003 ; Labov 200 1 ). The hypothesis that child language acquisition is the crucial factor bringing about change has been l i n ked with a tradition of cal l i ng change within a rela tively homogeneous commun i ty that is brought about by child l anguage acquisition "internal c hange," as opposed to "external change" brought about by contact, but the fi rst is u ltimately no more "internal" than the l atter - i t does not happen "in" the l anguage, or "in" the grammar, only in transmission (see Section 3 .2.4). It has also been l inked with a tradi tion at least si nce Hal le ( 1 964) that the child is pri mar ily an i n terpreter, making hypotheses abou t the l i nguistic system, rather than an active producer of l anguage. However, since it is only from evidence of production that we can assess what may have been innovated, it is crucial to conceptualize the language acquirer as an aClive producer as wel l as passive processor of lan guage. L i ke the hypothesis that child not adult language acquisition drives change, the sharp disti nction between "internal" and "external" change has recently been called i n to question as i ncreasing attention is paid to variation and l anguage users' access to strategic use of multiple styles and possibly grammars (Kroch 1 994). It seems preferable to refer to change that arises out of contact and affects multiple
3. 2 Some background assumptions about change
45
subsystems of a language as "contact-i nduced," and to other changes as "natural" (Thomason and Kaufman 1988) or "evolutive" (Paul 1 920; Andersen 1 973). 3.2.3
The question of genetic endowment
Although the m odel in Figure 3 . 1 does not force the issue, it was designed to characterize a grammar of relatively fixed structure at any one period, and uniformi ty of i nput. Such assumptions, as we have seen, are chal lenged by the study of grammaticali zation (and of sociolinguistics). The model does crucially claim that there are universal laws of some kind, but, as a model , it does not specify what ki nds of laws they are. Andersen speaks of them as: "the properties of [the learner's] constitution that completely determine the nature of li ngu istic structure, and hence the relation between a grammar and its output" (Andersen 1 97 3 : 776). The key phrase here is "completely determ ine." The hypothesis is that human beings are born with a set of constraints on what possible language structures can be, and ways in which they can vary. Refinement of this hypothesis has been the major focus of much recent generative theory. One widely accepted model that has been proposed is that all human beings are genetically endowed w i th Universal Grammar (UG) (see, e.g., Chomsky 1 98 1 ). This UG is conceived as consisting of two components: u nchanging "pri nciples" that characterize the fundamental structure of language and restrict the class of attainable grammars, and "parameters" that define the space of possible variation and are fi xed by experience. Ditl"erences between languages across geographic and social space or across time are conceived as bei ng the result of different set tings of the parameters in the process of l anguage acquisition . Lightfoot ( 1 99 1 ) elaborated on the idea of principles and parameters for change, and argued that changes from one generation to another are the result of the fact that different learners select different possibilities fro m among a restricted set of structures that are genetically encoded. Specifically, he hypothesized that children contribute to language learning (and hence to change) at least a "disposition to learn ." Th is disposition was conceived as a selective one: "an organism experiences the sur rou nding environment and selects relevant sti muli accord ing to criteria that are already present i n ternally" ( 1 99 1 : 2). Such a selective disposi tion was contrasted to an "instructive" one, w hich is essentially flexible and modifiable by outside stimul i . In this account of motivations for language change, the child is conceived as an LAD (language acq u isition device), a processor of systems, a ki nd of passive logic machine with a very rich language-specific genetic endowment . In recent years the doctrine o f i n nateness (nativism) has come under attack from several quarters. Sampson ( 1 997) surveys critical l y the entire range of evidence that has been put forward for innateness. Tomasello ( 1 999) presents
46
3
Mechanisms: reanalysis and analogy
a case for l anguage as a cultural artifact elaborated through i ncreas ingly i n tri cate social i n teractions. Deacon ( 1 997) sees language as constantly in (l ux and adapting i tsel f for optimal learnabi l i ty to "ch i ldren 's spontaneous assumptions about com m u n ication, learn i ng, social i n teraction, and even symbol ic re ference" (Deacon 1 99 7 : 1 09). Typical ly, anti-nativists reason that language i s not an isolated and spec i fi c neu ral capaci ty but is derivable from more general hu man cogn itive endowments. From this perspective the u n iversal component, far from bei ng seen as one that will "completely determine the nature of li nguistic structure," is seen as ch arac teri zing broader properties of the h u man constitution (see Green berg 1 990). It can be explai ned by reference to h uman cognition and the h uman commun ica tive goals that language serves (G iv6n 1 989). If there is a structural res idue, it consists at the most of a broad propensity to distinguish the categories noun and verb, but even this d ivision probably reflects a practical comm u n icative necessi ty. Tom asel lo argues ( 1 999: 4 1 -5) that structural complexity i n language emerged through grammatical ization as a response to the grow ing sophi stication of the human social environment during the modern period of Homo sapiens. Language evol ution went in step with the cog n i tive expansion brough t about by the neces sity to expla i n , pred ict, and control the behav i or of conspec i fics (Tomasello 1 999: 24-5 ) . Such a view is consistent with the idea of the language learner as engaged in strategic i n teraction as a producer of language, a negotiator seeki ng to get people to do th i ngs with words, not j ust a purveyor of i n formation. This is the approach we adopt i n this book. 3.2. 4
Innovation versus spread
When cons idering Figure 3 . 1 i n the light o f the c l a i m that il is ru les Ihal change, not languages, a d isti nction needs to be made between c hange and spread of the change, understood as repl ication or spread of i n novations from Ihe i ndiv idual to the group (Wei nreich, Labov, and Herzog 1 968; Croft 2000; see Lichtenberk 1 99 1 a, for discussion from the poi nt of view of gram matical ization). When an innovated form B en ters the grammar alongside of an older form A, it does so abruptly: an Ewe language user either does or does n ot use be as a complemen lizer (see Secti o n 2.2). However, the spread of the complemen tizer analysis across verbs of locu tion and cogn ition is gradual ; this kind of spread through the l i ngu istic sys tem is cal led "generalization" and w i l l be discussed in fu ller detail in Seclion 5 .2. Spread across l i ngu istic contexts is to be further d istingui shed from spread across genres and social grou ps. For example, each indi vidual reanalysis of a verb o f locution or cognition could potential ly have i ts own trajectory through soc ial space, though often there will be cumu lative effects from one change to another.
3. 2
Some background assulI1ptio/lS about change
47
As i nd icated above, Andersen's model has been understood as designed to re flect changes in the abstract grammars of individual language users of d i fferent generations. The problem is that "one swal low doth not a summer make," and one change in the gram mar of an individual does not constitute what we th i n k of as a change in "a language." From the viewpoint of generative grammar, there is no such thing as "a grammar of Old Engl ish," or "a grammar of Present- Day Engl ish," only grammars of indiv idual s; therefore, when we use such expressions as "change in the gram mar of X" we are essentially using "a convenient fict ion permitting the statemen t o f certai n generalizations and ignoring certain types of variation" (Lightfoot 1 99 1 : 262). But this leaves the quest ion o f how to th ink about the someti mes sign i licant d i fferences that can be observed over time. The answer from the generative perspective is that, however abrupt a change may appear to be in models such as that in Figure 3 . 1 , once the change has occurred, it is the aggregations of gradual changes across time that give the impression of "changes in the language." Sometimes these aggregations spread rapidly across a commu ni ty, lead i ng to what appear to be "major changes" (for exam ple, rad ical shifts in word order, loss of case morphology, the rise of a new category such as syntactic aux i l i ary verb, all of which are d i scussed later in this book). I n an effort to refocus attention away from "major changes" and onto breaking down d i achronic development into its "smal lest appreciable constituent steps," Andersen poi nts out that: each and every step i n such a development is an innovati on. not only the initial act. t h ro ugh wh ich a new linguistic entity comes into being. It is through innumerable individual acts o f i nnovati on - o f acceptance. adoption. and acqu isition - that any new e n t i ty gains currency and enters into competi tion w i th trad itional entities in the usage o f a lingu i stic community.
(Andersen
1 989: 1 4)
Th is approach is highly consonant with grammatical ization . Another way to think of what constitutes a change is to think of grammars not of the indiv idual bu t of the speech com m u n i ty : "The grammars i n which l i nguistic change occurs are grammars of the speech community" (Wei nreich, Labov, and Herzog 1 968: 1 88). Th is approach too ulti mately leaves us with unresol ved ques tions such as what is the status o f "grammar of the speech community'''! M ore importantly, though, studies of language use i n communi ties and spread through them have h ighl ighted an i m portant disti nction "between evidence of social vari ation among children that may reflect s i mple exposure, as in cl ass and eth nic ditTerences, and ev idence o f the socia/ use of variation" (Eckert 1 999: I I , i tal ics added ). We need well-coordinated long-term stud ies of language acquisi tion by chi ldren duri ng pre-puberty and by adults of all ages that pay attention to those areas of
48
3 Mechanisms: reanaLysis and anaLogy
l i nguistic change of i nterest to h istorical researchers before the empirical ques tions raised by hypotheses about the relationshi p between language acquisition and language c hange can be adequately u nderstood. In any event, the nature of historical records generally makes it difficult to pinpoint where change originated and how it spread. Clearly, historical records are in writing, and therefore not a direct reflection of what small chi ldren (or even most adu l ts) did with language in the days before widespread l iteracy. Study of tape recordings over the l ast cen tury should help in this endeavor, but when we are deal ing w i th older periods of the language we are necessari ly dealing with changes that have come down to us i n written form , even if we look to personal letters, drama, and other texts types that are l i kely to be mini mally i nstitutional (and therefore "standardi zed") i n character. A crucial q uestion is, when can we say that a change has taken place? Al though caution is ideal ly al ways exercised, in practice m any l i nguists tend to see a single example of a change that l ater spreads to other texts and other construc tions as a "first example of change x." Th is approach naturally fol lows from thinking of c hange in terms of differences in the grammars of individual lan guage users. It must be remembered, of course, that writers and speakers ex pect to be u nderstood . A first attestation therefore may not mean a first use, but rather an early example of a form that has already gained some social accep tance in the speech community. Some innovations catch on and are reproduced by other users of the l anguage, and may eventual ly be recognizable as rule changes. Others are n ot found in subsequent documents, and are identified as "nonce" forms or even scribal errors. Methodological ly it is convenient to have some cri teria by which we can conventionall y say that a rule change has occurred. We will say that a ru le change has occurred if (a) it has evidently spread from the individual and has been accepted by a group, and (b) the constraints of the former lingu istic environment are no l onger obl igatory. The following i l l ustrative exam ple concerns the use of the verb wiLL- as a tense aux i l iary in n inth-century Old English (OE). In ( I ), waLde, the past tense of wiLL- ' want,' occurs i n a context that suggests it can only have been meant as a marker of later time (equivalent to Present-Day Engl ish wouLd for wiLL in reported speech): (I)
l>a Darius geseah pa:t he overwu nnen beon wolde, p a wolde w hen Darius saw that he overcome be would, then wanted he hiene sel fne on- 6a:m gefeohte forspil l an he h i m sel f in that battle kill-INF ' When Darius saw that he would be overcome, he wanted to commit suicide in that battle.' (c. 880, Orosi us 3 9. 1 28.5)
3. 2 Some background assumptions about change
49
Does ( 1 ) exempl i fy a change at least in the grammar of the language user who wrote thi s passage, if not in the "conven ient fiction of the grammar of Old English"? The two criteria we have suggested poi nt to rule change as having occurred . First, the rate of use of wolde in the sense of ' would' increased in Old Engl ish. Secondly, the meaning change exempl ified here is consistent with a ru le change. As will be d iscussed in more detail later, will-, the ancestor of Present-Day Engl ish will, as in She will run /or Governor, was a main verb mean ing ' i ntend,' as in S/le willed herself to succeed. As such, will- was origi nally followed by a vol itional verb (one denoting an activity carried out del iberately). In the example, however, overwunnen beon 'to be defeated ' is clearly not the intention of the agent. So a former obligatory constraint on the use of will- is no longer operative. Therefore ( I ) appears to be a legitimate early example of a structure that signals a rule ch�nge at least in the individual writer even though it appears only rarely el sewhere at this time (the n i n th century). Simi larly we know that let 's « let us) has begun to be grammatical ized when the l i m i tation to the permission context (i.e., 'al low us' ) no longer holds. When th is constraint was removed, the paradigmatic relationsh ip of the first-person-plural pronoun to other pronouns and nouns no l onger held, and the stress on us in let ItS could be reduced. The assumption that Grammarl and Grammar2 are relatively fixed has some undesirable consequences. For i nstance, it is often assu med that a rule or form A is replaced directly by a d i fferent rule or form B. Consider Ewe be. From the fixed grammars model it might appear that a later generation abruptly replaces the earl ier generation's lexical V mean ing 'say ' with a particle mean ing 'that' (along with accompanying changes i n syntactic structure) and that, for the l anguage learner, the earlier meani ng and structure have disappeared al together. But, as we have seen in connection with let 's, older and newer forms coexist for individual speakers as well as for communities over time. Indeed; A probably never "becomes" B without an intermediary stage in wh ich A and B coexist: 2 (2)
Such coexistence, which Hopper ( 1 99 1 ) has cal led "layeri ng," may last several hundred or more years, as in the case of Ewe be or Engl ish be going to. Al terna tively, it may be quite short, as in the case of the brief development and dem ise during Middle English of "regressive" aspectual verbs stint and fin (meaning ap proximately ' leave off V- i ng,' 'stop V-i ng ' ) (Brinton 1 988: 1 5 1 ). We w i l l discuss the phenomenon of layeri ng in greater detail in Section 5 . 5 .
50
3 Mechanisms: reanalysis and allalogy
One fi nal poi n t abou t the assumptions behind the model worth men tioning is that the focus o n universals privileges the uniform i ty o f rule types and reasoning types across languages and times. Indeed, what has come to be cal led the "uni for m i tarian pri nc i p l e" (Labov 1 974; Romaine 1 982) is an essential ingredient of most work in historical l i nguistics. According to this pri nci ple, the l i nguistic forces that are evidenced today are in pri nciple the same as those that operated in the past. O perationally, t h i s means that no earl ier gram mar or ru le may be reconstructed for a dead language that is not attested i n a l iving one. There is no reason to believe that grammatica l i zation did not occur in languages spoken ten thousand years ago in much the same way as it does now. Whatever our m odel for change, we need to consider the ways or "mechan isms" by which change takes place and the factors that enable them to occur. In the re mai nder ofthis c hapter we foclls on the principal ways i n which grammatical ization may occur.
3.3
Rea n a lysis
In reanalysis, the hearer understands a form to have a structure and a mean ing that are different from those of the speaker, as when [Hamburg] + [er] ' i tem (of food) from Hamburg' is heard as [ham] + [burger] . Sooner or later someone substi tutes the word cheese or beef for ham; but this substi tution is merely the symptom of a change that has already occurred si lently. The reanalysis i tsel f is covert u ntil some recogn izable modification i n the forms reveals it. The hamburger exam ple i l l ustrates reanalysis in a si ngle lexical i te m ; but sy n tactic sequences may also be reanalyzed. In current English, for example, the sequence tly and VERB h as under some circumstances been reanalyzed as Auxiliary + Verb, as I 'll try alld contact hel: 'Try ' in this use is disti nct from ' try ' in They have tried alld fa iled to contact her, as well as from I 'll fly to contact IleI: In I 'll t ry and cOlltact her, there is evidence that try and is stored as a single word : (i) (ii) (iii)
The and is i ntonationally and phonetical ly bound t o try ( ' try- ; m ' ) . Only try, n o t tried, trying. tries, is poss ible (e.g. , not ' H e tries and contacts her). Adverbs may n o t i n tervene between tly and and (e.g . , I 'll fly hard to contact her, but not * I 'll fly hard and contact her) .
Moreover, t h e mean ing of try and is more modal-like t h a n tly to. I t signals the agent's inabi l i ty to achieve the complement verb and the speaker's lack of con fi dence i n the agent's success (Hopper 2002).
3. 3
Reanalysis
51
In a major paper on syntactic change, Langacker defi ned reanalysis as: "change in the structure of an expression or class of express ions that does not invol ve any im mediate or intrinsic modification of i ts surface man i festation" ( 1 97 7 : 58). From this perspective, reanalysis i nvolves a change i n constituency, hierarchical structure, category labels, grammatical rel ations, and cohesion ( type of bou ndary ) (A. Harris and Campbel l 1 995 : 6 1 ). Very often a si ngle instance of reanalysis w i l l show several of these characteristics correlated with one another, a s i s the case with fry and in the preced ing paragraph . The examples of grammatical ization in Chapter I are all examples o f reanalysis that i nvol ve changes i n constituency (rebracketing of elements in certai n constructions), and reassign ment of mor phemes to differen t semantic-syntactic category labels: be going fo from be + main verb + progressive aspect + purposive preposition to tense marker; let liS from main verb + object to modal particle; and Ewe be from main verb to complementizer. 3 A n other example of several types of change is the reanalysis of a construction consisting o f a head noun and a dependent noun (3a) as a (complex) prepos ition and head noun (3b): (3 )
a. [[back] of the barn] [back of [the barn]]
>
b.
The change from (3a) to (3b) probably did not happen in one step, bu t rather is the outcome o f a set o f smal ler changes. The poi nt here is that the change i l l ustrates the fi rst three of the fi ve characteristics mentioned above. The rebracketing is an instance of constituency change (what goes with what). The change in head noun status is an i nstance of hierarchical structure change (what is dependent on what). The reinterpretation of the noun back as an adposition in a complex preposi tional construction is an instance of category label change. Changes i n grammatical relations are i l l ustrated by the development of subject out of topic mentioned in Section 2.3 and by the requirement in E nglish that clauses have gram matical subjects. An example of the latter is the change from (4a) to (4b) (multi ple negat ion was the norm in Old Eng l i s h ; the many intermediate steps between (4a) and (4b) are omi tted): (4)
nan god don a. Donne l)am menn n e Iyst when that- DAT man- DAT not wishes no good do-INF (c. 1 000, A!LS (Memory of Sai n ts) 2 9 7 : cited i n A l len 1 995: 86) b. when the man doesn ' t wish to do any good
Changes in degree of cohesiveness have been illustrated by be going to > be gOflna, let li S > let 's > lets. In both cases, a formerly separable morpheme has become fused with the one that preceded it. Such changes al ways invol ve rebracketing
52
3 Mechanisms: reanalysis and analogy
( i .e., change i n consti tuency), but not all changes in rebracketing i nvolve changes in cohesiveness. The type most often associated w ith grammatical ization is fusion. In every i ns tance of reanalysis we can posit that i t i s the resul t of abduction. In some contexts two i n terpretations were possible, that is, there was at least the potential for ambigu i ty (also called "opacity") that allowed for the structure to conti nue to be analyzed as before, and for a new analysis to be i nnovated, and then to coexist w i th the earl ier analysis. 4 For example, given a reanalysis such as was illustrated i n (3), the abduction account of what has happened here i s as fol lows. A hearer has heard the "output" (3a) (the "resul t"), but assigns to i t a different structure (3b) (the "case") after matching it with possible nominal structures (specified by the "laws"). The conclusion is not identical with the original structure of which (3a) is a man ifestation, but is nonetheless compatible with (3a) i n that the surface string is the same. The structural differences provide the potential for different subsequent developments. Both analyses continue to exist, but with different meanings. The abduction accou nt of the reanalysis i l l ustrated in (4) is considerably more complex because it requ ires i n termediary stages i nvolving a variety of factors, among them word-order c hange and case l oss, and w i l l not concern us here (for detailed accounts of word-order changes from Old to Middle English, see, e.g., Fischer 1 992; Allen 1 995). Below w e give rather more detailed examples of reanalysis, with focus on the ki nds of reassignments that occur. B oth examples i nvolve morphosyntactic change, although the first, the development of the Romance future, pertains primarily to morphology, and the second, the development of English modal auxiliaries, to syntax. 3.3. 1
The French inflectional future
The history of the Romance future has been much discussed (for fuller accounts, see especially Fleischman 1 982; Pinkster 1 987; I. Roberts 1 993b). We will be reviewing speci fically the development from Latin of i nflectional forms i n French s u c h as je chanterai 'I will sing.' As mentioned i n Chapter 2 i n con nection with Meillet's views on word order as a kind of grammatical ization, Lati n was a l anguage of essentially object-verb word order structure, but allowed a range of orderings to convey different rhetorical strategies (e.g., the three orders cited by Mei l Iet of Petrus Paulum caedit). It had verbal inflections for past, present, and future, as well as other temporal relations. As mentioned i n Chapter I , the future was an i nflection that combined person, number, and tense:
3. 3 Reanalysis
(5)
53
cantabo si ng- I SG : FUT 'I will sing'
The question is how phrasal constructions like (6), consisting of an infini tive and a form of the verb habere 'to have,' came to compete with and eventually replace constructions l i ke (5) : (6)
cantare. Haec habeo these have- I SG: PRES sing-INF ' I have these thi ngs to sing.'
It was constructions l i ke the one i n (6) which were reduced, i n various ways in the various Romance languages, to form the new inflectional future i l l ustrated by French je chanterai. The verb habere 'to have' in Lat i n was a verb of possession and belonging. It was a transiti ve verb and could originally introduce only a nominal object. In many contexts i t did not have the strict meaning of possession, but rather had a more general l ocative meaning of 'belonging, being i n presence of, ' etc. (for the cross-l i nguistic i n terrelationship of locative-possessive-existential, see Lyons 1 968; Clark 1 978; Hei ne 1 997 : Chapter 5). In some contexts, especially those i n which the object was modified b y a gerundive, for example (7), this verb acquired Ii sense of obligation, or at least future orientation, presumably transferred from the geru ndive, which itself once expressed obligation. (7)
Aedem habu i t tuendam. house had look: after-GER 'He had a house to look after. ' (c. 40 BC, Cicero, Ver. 11. 1 , 1 30; cited in Pinkster 1 987: 208 )
Thus if I have a house to l ook after, I may have obl igations to look after it, and I may have future purposes, such as l iving in it, passing it on to my descendants. etc. Pinkster ( 1 987) suggests that habere + infinitive originated as an alternative to habere + gerundive, most particularly in contexts of verbs of speaking: (8)
a. Quid habes dicendu m? what have-2SG say-GER 'What do you have to say?' b. Quid habes dicere? what have-2S G say-INF 'What do you have to say?'
The first i nstance, accord i ng to Pinkster, of habere with an infinitive is i n the context of a verb of speaking that introduces a sentential complement:
54
3
(9)
M u l tos many dicere tell-INF
Mechanisms: reanalysis and analogy ferro, dagger- I NST, quem . . . someone . . .
multos veneno (occi d i t); habeo eni m many poison-I NST (killed); have-ISG even de ponte in Ti beri m deicerit. from bridge in Ti ber th rew
'Many he k i l led by the dagger, many by poison ; I can even give you an example of one man whom . . . he threw from the bridge into the Ti ber. ' (c. 40 BC, Cicero, S. Rosc. 1 00; ci ted in Pinkster 1 987: 206)
In (9) and several other examples l i ke ( 1 0), the ' have' -verb precedes the i n fi n i tive, and is separated from it: ( 1 0)
De re publ ica n i h i l habeo ad te scri bere nisi . . . about m atter public nothing have- I SG: PRES to you write-INF except . . . ' I have noth i ng to write to you about the commonwealth � except . . . (c. 40 BC, Cicero; ci ted i n Fleischman 1 982: 1 2 1 ) '
But later a differen t order is also attested, in which the ' have' -verb follows the infinitive directly : (I I)
Haec cantare habeo.
Although the c h a nges are assumed to have occurred between the th ird and sixth centuries AD, most of the attested examples come from later texts. Examples i nclude: ( 1 2)
a . . . . et quod sum essere habetis . . . and what be- I SG be-INF have-2PL 'and what 1 am, you have to/wi l l be' (seventh-century i nscri ption; cited in I. Roberts 1 993a: 234) b. Et si i nterrogatus fueris, quomodo dicere and i f asked be-2SG : PERF: S U BJUNCT, how say- I N F Veritatem dicere habeo. habes? have-2S G : PRES:S UBJUNCT? truth say-INF have- I S G : PRES 'And you, if you are asked, what do you have to/w i l l you say ? 1 wi l l have the truth to say/l w i l l speak the truth . '
(7 1 5, Cod. Dip/. Long. Siena; ci ted in Flei sch man 1 982: 59. I . Roberts 1 993a: 234)
From the perspective of reanalysis, the important fact is that constructions l i ke ( 1 2) contain a mai n verb hab- and an infinitive complement, i n a structure of the type: [[dicerel habeol, and i n contexts that can be understood to be obligative or at least future orie nted . If one is asked what one can say, the i n ference can be that one ought to say it. In such contexts, provided the forms are adjacent, a language user could be led by abduction to interpret the input stri ng not as representing two underlying clauses, but rather as bei ng bracketed together i n a structure of the
3. 3
Reanalysis
55
type [dicere habeo ] . The result is a hierarchic change such that dicere i s no longer subordi n ate to habeo . Once this reanalysis had occurred, further changes were possi ble. These in clude fusion across morpheme boundaries, phonological attri tion, and semantic reanalysis to a future- tense marker, 6 as i l lustrated by ( 1 3) : ( 1 3)
l usti nianus dicebat: ' Daras. ' J usti nian said : 'give:2Sa : FUT' « dare habes) (seventh-century Fredegario; ci ted in I. Roberts 1 993a: 234)
We may therefore pos i t a development in French of the kind sketched in Section 1 .2 . 3 : ( 1 4)
Cl assical Latin [ [cantare] habeo] Late Latin [cantare habeo] French [chant-e-r-ai]
>
>
S i m i lar changes occurred in some other Romance languages including Span ish, but at d i fferent periods . Some contemporary varieties of Romance either show no ev idence of the kinds of changes mentioned here, e.g., some Southern dialects of Italian, or else show d i fferent individual h istories. For example, the Sardinian future appears to have developed directly out of word-order structures of the type lIabeo calltare. In this language the verb ' to have' is aere, and the first person form is appo: ( 1 5)
a fakere L' appo I t-aere- I sa to do-INF ' I w i l l do it'
( I . Roberts 1 993a: 235, citing Jones 1 993)
The Sard i n ian developmen t demonstrates clearly that no change has to occur. However, given other changes i n the language, i f i t occurs, there are certa i n l i kely ways i n which the change w i l l proceed . 3.3.2
The English modal auxiliaries
We turn now to an example of reanalysis with far w ider-reaching ram i fi cations than the development of the French i n tlectional future. The development of the Engl ish aux i l i aries was one of the first topics to draw the attention of generative l i ngu ists working on syntactic change (see Traugou [Closs] 1 965 ; Lightfoot 1 979). It has been the focus of numerous studies since then, among them Plank ( 1 984), Den ison ( 1 993), Warner ( 1 993), Krug (2000). Originally conceived as a prime example of syntactic change, it is clearly also an instance of grammatical i zation. I t concerns change in the status of lexical verbs such as may, can, must, do such lhat they become aux i l i aries, in other words, recategorizalion.
3 Mechanisms: reanalysis and analogy
56
In Middle Engl ish around 1 3 80 (as represented by texts by Chaucer and Wycliffe) and in the fifteenth century (as represented by the Paston Letters and other texts) the fol l owing kinds of constructions were available: (i) Question i nversion and negation without do : ( 1 6)
a. ' Felistow ' , quod sche, 'thise thynges, and entren thei aughte in thy co rage?' 'Do you feel' , she said, ' these thi ngs, and do they enter at all i nto your feel i ngs?' (c. 1 3 80, Chaucer, Boethius, I . i v. l ) b. i t aperteneth nat to a wys man to . . . ' i t does not suit a wise man to . . . ' (c. 1 3 80, Chaucer, CT, Melibee 2 1 70)
(ii) Transitive c l auses consisting of verbs like call or may followed by an object NP, as in ( 1 7), or a to-infinitive complement as in ( 1 8): ( 1 7)
She koude muchel of wandryi ng by the weye. (c. 1 390, Chaucer, CT, Pro! . A. 467) 'She k new a lot about travel . '
( 1 8)
any man tJe whiche hadde mow3t to scapen tJe deth 'any man who had been able to escape death ' (c. 1 382, W. Bible 2 Par. 20.24 [MED mouen lib])
(iii) Modal verbs i n past participle form, like mow3t i n ( I 8). (iv) Sequences of modal verbs: ( 1 9)
No-tJing to hafe is su m-ty me of need, bot n03 t to may will haue is of grete vertew. 'To have nothing is sometimes a necessity, but to desire [lit. to be able to will to have] noth ing is a great virtue. ' ( 1 434, M isyn M L 1 28/8 [MED mouen l Oa])
By Early Modern English of the early sixteenth century, constructions like those in ( 1 7)-( \ 9) had become almost non-ex istent, and do-constructions were rapidly replacing those in ( 1 6). For detailed studies of the development of do and how to model it, see Kroch ( \ 989a,b). One widely accepted way of thi nking about the changes at the t i me of writing is as fol lows. In Old Engl ish all verbs, including the precursors of can, could, may, might, must, shall, should, will, would, do, and did behaved similarly with respect to properti es such as the fol l owing: they were negated by a precedi ng ne, and they inverted to clause-initial position in questions. However, some verbs were morphological ly d istinct; these were in the main the premodals and be-verbs. For example, the negative fused with forms of several of the premodals and be, e.g. ne wille 'not intend' appeared as nille (see PDE willYllilly < will I, nill l), ne wres ' not was' appeared as m£s. The premodals were also for the most part preterit-presents, which means that at an earlier stage in Indo-European the present tense had been formed with the morphology of past tense; semantically these verbs expressed completed action resulting in present state ( ' have come to be X ' ). During the
3. 3 Reanalysis
57
Middle Engl ish period several changes occurred, including the development of a new negative, not « na wiht 'no thing'), wh ich appeared after the verb as i n ( I 6b), a n d the u s e o f past tense premodals like would, might. could. must with present tense meani ng. B y the early sixteen th century a radical change occurred with respect to most verbs other than the premodals do and be; their use in negative and i n terrogative sentences l i ke ( 1 6a,b) began to decline rapid ly, or they came to be used in stylistical ly restricted contexts. Furthermore, as far as the premodals were concerned, a sufficient number of individual changes had occurred that transitional constructions of the type ( 1 7)-( 1 9) were also disappearing. In terms of reanalysis, what was origi nally one category of verbs had been reanalyzed as two: main verbs and aux i l iaries. By the eighteenth century a further change had occurred : do, did became obl igatory i n i nterrogative sentences like ( 1 6a) and favored in negative sentences like ( 1 6b) (Kroch 1 989b). This had the effect, at least in standard varieties of Engl ish, of maximizing the distinction between the new categories of modal (and also other auxiliaries l i ke be and have in passive, perfect, and simi lar constructions) on the one hand and main verbs on the other. Together the changes, most especially the devel opment of auxiliary do, had consequences for the texture of English that make it very differen t n ot only from earl ier stages but also from several other European languages, including French and German. An important aspect of the development of the modals (and all the auxiliaries) in Engl ish i s that there was a cluster of factors that set the scene for the reanalysis (the special morphology of the verbs in question, the meaning of the modals, which had to do with states of mind such as intention, desire, permission and ability), word order changes, etc. Another is that it demonstrates wel l how different degrees of detail in analysis can engender differen t ways of i n terpreting data. When Lightfoot first publ ished work on the development of modals i n 1 979 only the broadest outl i nes of the changes were understood . The characterizations he proposed were at a level so general that they obscured many·of the more fi ne-grai ned properties that a perspective from grammatical ization would focus on. For example, he initially saw the sixteenth-century changes in the modals (and other auxi l iaries) as being part of the same change as the later one i nvolving do, partly because the first change, al though far advanced, was not entirely entrenched at the time the second was gai n i ng ground. This led him to write of "a sudden, cataclysmic, wholesale restructuring" (i.e., reanalysis) (Lightfoot 1 979: 1 22). However, when a close look is taken at i ndividual verbs, we soon discover that the changes occurred i n different verbs at differen t times (a poi nt accepted i n Lightfoot 1 99 1 ). Furthermore, some of the changes are sti l l ongoing. Consider, for example, the set of verbs known as "quasi-modals": be to, dare to, need to and ought to, some of which do and some of which do not require do in negatives and questions, e.g., You needn 't go, Do you need to go, • Need you go, • You don 't ought to leave, ? You oughfll 'f fo leave, Ought
58
3 Mechanisms: reanalysis and analogy
you to leave ? (see Krug 2000). Extreme positions are rarely right; this is clear i n the case of the modals. It is true that each had i ts own history ; but it i s also true that some fairly rad ical c h anges occurred in the sixteenth and seven teenth cen turies. Small changes acc u m u lated and, i n teracting w i th other changes going on elsewhere i n the system, such as word-order changes, led to l arge-scale shi fts (called "parametric changes" in the generative l iterature, e.g., Ligh tfoot 1 99 1 ) . Grammatical i zation was i nvol ved at all stages: erstwhile lexical i tems (premodals that were main verbs and do) in certa i n l i ngu i stic constructions acquired gram matical status as auxil iaries. The c hanges i nvolved reanalysis of consti tuent, h ierarchy, and category status. To some extent they also i n vol ved analogy, as wi 11 be d i scussed in Section 3 . 5 below. Before turni ng t o analogy, however, we pause t o emphasize that not all reanalysis i s a case of grammatical i zation.
3.4
The i n depe nde nce of rea na lysis and g ra m m atica l ization
Mei llet appears to have identi fied reanalysis with gram matical i zation . However, al though many cases of reanalysis are cases of gram matical ization ( i ncluding those d i scussed above), not all are. Consider, for example, compound ing, a reanaly s i s i nvol v i ng the weakening and often loss of the boundary between words or morphemes . S ometi mes the result is a derivative morpheme l i ke -hood; often a relatively analyzable form arises, such as bo 'sull from boat + swain 'man,' hussy from house + Wife ' woman,' fishwife from fish + Wife ' woman, ' sweetmeat from sweet + meat 'food' (Anttila 1 989 [ 1 972] : 1 5 1 ) . Swain, Wife, meat have not been reanalyzed as grammatical morphemes, nor do they seem desti ned to be. The effect seems to be pri marily on the lexicon, not the gram mar, and is called "Iexical ization." Here then, we have a case of reanalysis wi thou t necessary gram matical ization. Sometimes reanalysis results in a change that has gram matical effects, but nev ertheless i nvol ves a sh i ft from grammatical to lexical structure, rather than from lexical to grammatical structure (the norm for grammatical izati on). Examples are the use of up, down, allte, etc. as verbs or nouns, cf. to up the ante, to ante up, what a downer. The c hange w hereby a non-lex ical form l i ke up becomes a fu lly refer ential lexical i tem is called "conversion." It is relatively u ncommon, but i n stances can be found i n most l anguages. A rather different instance is the development i n Engl ish of bus, a borrowed Latin dative plural that has been detached from the adjective stem omni- (omnibus 'for al l ' ) and promoted to nomi nal status. S i nce the form deri ves from a borrowi ng, and the Lati n paradigm of case i n flections is virtually i naccessi ble to most English speakers, the development of an i n tlection i nto a noun i l l ustrated by bus has status only as a u nique i n novation, not as a regular type of change.
3.4 The independence of reanalysis and grammaticalization
59
Another case of reanalysis leading to the autonomy of an earl ier affix, this time one that resul ted from sound change, is that of the emphatic part icle ep in Estonian (Campbell 1 99 1 : 29 1 ). At an earl ier stage the particle was a bound c1itic, c f. Fi nn ish -pa, -pa. By regular phonological change, the final vowel disappeared . leav i ng -p, cf. piiiilla 'on (top ot)' > paall, and paalla-pa 'right on (top ot) ' > paalla-p. The vowel of the c 1 i tic -pa had originally required vowel harmony ; with the loss of the vowel of the c1 itic, the vowel harmony rule no longer applied. and the emphatic form became peaUep. The emphatic peallep no longer had any transparent relationsh i p to the non-emphatic piitilla. Pealle-p was reanalyzed as peal-ep. Later -ep was rei nterpreted as an autonomous particle, and came to precede the word it emphasized . Reanalysis here led to the development of new i ndependent partic les, which themsel ves then cou ld become subject to grammatical ization . We w i l l discuss issues of this kind further i n Section 5.6. More widely attested cases of reanalysis that call i n to questi on the identi fica tion of reanal ysis with gram matical ization i nclude word-order changes, which we discuss im med iately bel ow. These can have major etl'ects on the morphosyntactic orga n i zation of a language, but do not exemplify the unidirectional ity typical of gram matical ization. It is best, then , to regard grammatical ization as a subset of changes involved in reanalysis, rather than to iden tify the two (Heine and Reh 1 984; Heine, Claud i , and Hiin nemeyer 1 99 1 a; I. Roberts 1 993a; A. Harris and Campbell 1 995). Whereas grammatical ization al ways i nvol ves reanalysis, many clear cases of reanalysis do not result i n grammatical i zation.
3.4. 1
Word-order change
Langacker's m ajor paper on reanalysis ( 1 977) focuses on boundary cre at ion, s h i ft, and loss, but does not incl ude discussion of word-order changes. However, the latter i nvol ve changes in constituent order. As we will see below, word-order changes can have far-reaching effects on grammatical rules as wel l as on the texture of a language. As mentioned i n Section 2 .2, Meil let, at the end of h i s path-breaki ng article ( 1 9 1 2), suggests that words are not the only sources of gram matical expression : word-order changes may be too. He compares word orders that signal nuances of meaning (what we wou ld call pragmatic mean i ngs), such as al ternative word orders in Lati n, w i th gram matical word orders that signal the sy ntactic cases subject and object, as exempl i fied by Present-Day Engl ish. Meil let therefore i ncluded word order changes among instances of grammatical ization i n the sense of reanalysis. Others have suggested that word-order changes are the outcome of grammatical i zation (e.g., Claudi 1 994). The question for us here is whether word-order changes, which exemplify a kind of reanalysis, also exemplify grammatical ization, as
60
3 Mechanisms: reallalysis and analogy
Meil let suggests, or whether they are to be considered as types of reanalysis that do not necessarily i nvolve grammaticalization. To anticipate, word-order changes may be the outcome of, as well as the enabling factors for, grammat ical ization in the nalTower, prototypical sense used in this book of the change by which lexical i tems and constructions used in certain contexts come to mark grammatical relations. Word-order changes are not u n idirectional . Therefore, they should not be identified with grammatical ization in the narrower sense. How ever, given a broader defi n i tion of grammaticalization as the organization of grammatical, especially morphosyn tactic material, they cannot be excluded from consideratio n . For our purposes it i s important t o stress that word-order changes c a n have a profound effect on the grammatical structure and the morphological texture of the language, because different constituen t orders are typically associated with VO and OV languages. VO languages include those with the order VSO (verb-subject-object), e.g., Hebrew, Masai , and Welsh, and SVO, e.g . , Engl ish, Malay, and Swah i l i . Among the OV (verb-final) languages are Basque, Japanese, and Quechua (for more combi nations and discussion of word-order typologies, see Greenberg 1 966a; Vennemann 1 97 5 ; W. Lehmann 1 978a; Hawkins 1 983; Dryer 1 99 1 , 1 992; and papers in Li 1 97 5 ; van Kemenade and Vincent 1 997). VO lan guages tend to be prepositional ; adjectives, relative clauses, and possessives fol low the noun; the aux i liary precedes the main verb, and the question particle marking yes-no questions occurs i n initial posi tion i n the clause. By contrast, verb-fi nal languages tend to show the order in reverse: they are postpositional ; adjectives, rel ative clauses, and possessives precede the nou n ; the aux i l i ary follows the main verb, and the question particle tends to appear in fi nal position in the clause. Some sample constructions are shown i n (20) : (20)
vo
OV
saw h i m in house man old that hat or man has been killed whether he left?
h i m saw house in that old man man's hat killed been has he left whether?
There is no "ideal" OV or VO order language. Instead, there are languages which may h ave predom i nant OV or VO order, or which may exhibit properties of both. This is because coding is constantly in fl ux, and because there are competing motivations in creating discourse (see Section 4. 1 ). For example, "topical ization" typically moves material to the beginn i ng of a clause, bri nging information to attention and derouti nizing it. On the other hand, routine word orders serve as "normative structures" in the everyday flow of com mun ication. Useful discussion
3.4 The independence of reanalysis and g rammaticalization
61
can be found i n Vincent ( 1 979) on "iconic" versus "symbolic" orders, and Haiman ( l 985a: Chapter 6), o n three confl icting pri nciples: (i) what i s old i n formation comes first, what is new i n formation comes later in an utterance; (ii) ideas that are closely connected tend to be placed together; ( i i i ) what is at the moment uppermost in the speaker's m i nd tends to be the first expressed . More recent work on "information packagi ng" includes E. Pri nce ( 1 98 1 ), Vallduvf ( 1 992), Chafe ( 1 994), Lambrech t ( 1 994), Kiss ( 1 995), Birner and Ward ( 1 998). In some languages, OV order favors the development of inflections, though by no means all languages with OV order are inflectional (Li and Thompson 1 974). When they arise, i nflections tend to be derived from prior lexical items. An example is provided by the development of the French future, illustrated above. When VO order arises from OV order, the change will often be accompanied by the i n novation of new phrasal ("periphrastic") ways of cod i ng what at an earl ier stage was coded inflectionally. The history of English modals illustrates among many other th i ngs the replacement of certai n subj u nctive i n flections by periphrastic expressions. We suggested i n Section 1 .3. I that the development of lets i n place of a subjunctive expression may also be an i nstance of the larger change of English from OV to VO. If i nflections develop i n OV languages, they typically do so via reanalysis of encl itics or bound forms through boundary loss, fusion, and phonological attrition of already bound forms. By contrast, when new periphrastic constructions arise i n the shift from OV t o VO, they typically develop through reanalysis o f lexical i tems as grammatical ones. They are examples of what Mei llet called "renouvellement" renewals of old functions (at first possibly more expressive ways of sayi ng the same thing). These periphrastic constructions may themselves in turn become i nflections (prefixes rather than postfixes). Because they derive in ditl"erent ways, and at differen t times, the resources used in the development of OV and VO orders may look very different from a relatively sy n chron ic point of view. For example, there is no form-meaning, i .e., "cognate," relationship between the inflectional or clitic geni tive -s in English and the preposition of that partly replaced it. Nor is there any cognate relationship between the OE i nflectional subju nctive (typically -e(n» and might, should, etc. The relevan t factors for the selection of lexical forms as grammatical ones are semantic suitability, i n ferences (both "logical" and "conversational") from context, and potential constructional ambiguities ari sing from such i n ferences . Such factors will be discussed i n the next chapter. Cross-linguistic studies suggest that there are no constraints dependi ng solely on word order that delimit the lexical resources that can be used in the development of grammatical items. This argues against word-order change as an example of grammaticalization in the narrower sense of reanalysis of lexical forms as grammatical ones.
62
3
Mechanisms: reanalysis and analogy
We give here a n example of the same lexical i tem giving rise to both i n llection and to peri phrasi s (but i n l ocal constructions with d i fferent word orders). We turn again to Romance. As we have seen, the Late Latin verb habere ' to have' was reanaly zed in postverbal (OV) position as a future i n fl ectional marker. As Romance languages developed, a new periphrastic complex perfect construction emerged alongside of the future i n flection, replacing the earl ier perfect i n llection - l'- ; e.g., probavi 'I have tried' was replaced by habeo probatum. Th is complex perfect, l i ke the future, arose out of a habere construction, but in this case it origi nated i n a construction consisting of an i n flected form of habere ' to have' and a past participle that agreed w i th the object of habere (see, with somewhat d i fferent i nterpretations, B enven iste 1 968; M . Harris 1 97 8 ; Fleischman 1 982 ; Vi ncent 1 982; Pin kster 1 987). In Late Lat i n both the future and the perfect occur i n both OV and VO orders. Thus we find: (2 1 )
a . cantare habeo '" habeo cantare (OV '" VOl b. probatum habeo '" habeo probatu m (OV '" VOl
The type cantare habeo has been i l l ustrated in ( 1 2), the type habeo cCllltare (with interve n i n g material) in (9), ( 1 0), and ( 1 5). The type probatlllll habeo may be i l l ustrated by (22a,b) and habeo probatum by (23): (22)
a. Promissum habeo . . . nihil sine eius pro m i sed-N EUT: SG(?) have- I SG . . . noth ing:NEUT:SG without his consi lio agere. ad v i ce do- INF ' I h a ve promised to do nothing without his advice.' (sixth century. Gregory of Tou rs ; ci ted i n Fleischman 1 982: 1 20) b. Quae cum i ta si nt, de Caesare satis hoc w h i c h si nce thus be-SUBJ UNCT, about Caesar enough this tempore dictum habeo. t i me said have- I SG
(23)
' Under the circu mstances, I sha l l regard what I have said of Caesar as s u fficient at present.' (c. 40 BC, Cicero, Phil. 5,52; cited i n Pinkster 1 987: 204) Metuo eni m ne ibi vos habebam fatigatos. Fea r: 1 SG for lest there you have- I M PF- I SG ti red ' For I fear that I have tired you . ' (early fi ft h century, Augustine; cited i n Fleischman 1 982: 1 20)
Both the fu ture and the perfect eventually became fi xed units and i nvolved reanalysis of a n i n llected form of the independent verb hab- as dependent on the non-finite verb with which they occurred. They di ffer in that the path from habere to the future was via an obl igative or future-oriented sense of the verb, whereas
3. 5
A nalogylrule generalization
63
the path from habere to the perfect was via the locati ve-possessivc-existential in transit ive contexts of cogni tive and sensory states. Fu rthermore, in French the fi rst became an i n tlection and the second remai ned as a peri phrasis ( though as we saw i n con nection with ( 1 5 ), the future remai ned a periphrasis in Sardin ian). It appears that i n French the future was grammatical ized while OV was sti l l the chief word order for this construction, and that the perrect was gram maticalized later when the shift to VO had already taken place (Fleischman 1 982: 1 2 1 ), but in Sardinian the fu ture was grammatical ized after VO had become the chief word order. So far, we have discussed only shifts from OV to VO, both at the general level of verb phrase constituent structure and at the more local level of individual 1110r phosy ntactic changes . Before leav ing the subject of word order, it is important to poi nt out that a shift from OV to VO or vice versa never occurs independen tly of other factors, both l i nguistic and historical . Some of the l i ngu istic factors in vol ved have been noted i n Mithun ( 1 995); she shows how in an origi nal ly SOV fam ily that includes Caddoan, S iouan, and Iroquoian, divergence i n word order came about through a variety of means. These i ncluded, according to the language or language group: the development of third-person pronomi nal prefi xes, the rise or case marking, and prol iferation of noun i ncorporation, each of which served to dislodge a once rigid verb-fi nal word order. Of historical factors, by far the most important is language contact, which often results in the adoption or new word-order patterns and changes in typological affi liation. An early study of this phenomenon was Bach's ( 1 970) analysis of verb-fi nal word order in Amharic, an A fro-Asiatic language that could be expected to show VO word order. Bach argued that certai n l i nguistic ru les of Amharic sti l l required the positing of un derl y i ng VO word order, and attributed the superficial verb-fi nal word order to the infl uence of neighboring Cushitic languages. Smal l-scale changes or this type can often be directly observed, as for examp i e the shift in Estonian compou nds from modi fier-head to head-mod ifier order through Russian i n tl uence on the me dia (Hint 2000); Russian is an Indo-European SVO language, while Estonian is a Uralic language in transition between an earl ier SOV and a newer SVO type.
3.5
Analogy/rule generalization
As we have seen, Meil let made a disti nction between the development or new gram matical forms and arrangements on the one hand, and analogy on the other. The first, which he called grammaticalization, is the result of what we now cal l reanalysis. As we have defi ned it, reanalysis refers to the replacement of old structures by new ones. It is covert. Analogy, by con trast, refers to the attraction of
64
3 Mechanisms: reanalysis and analogy
extant forms to a l ready existing constructions, for example, the attraction of Ewe verbs of locution and cognition to the complementizer construction , modeled after be. It is overt. I n essence reanalysis and anal ogy i nvolve i n n ovation along differ ent axes. Reanalysis operates along the "syntagmatic" axis of l inear constituent structure. Analogy, by contrast, operates along the "paradigmatic" axis of options at any one consti tuent node (Jakobson and Halle 1 956). When Meil let was writing, there was a rather narrow, local i n terpretation of anal ogy, which was defi ned as a process whereby irregularities in grammar, particularly at the morphological level , were regularized . The mechanism was seen as one of "proportion" or equation. Thus, given the s i ngular-plural alternatio n cat-cats, one can conceive of a nalogizing child-children as child-childs (as i ndeed occurs i n child language) : (24)
cat: cats ch i ld: X X c h i lds =
=
Or, as actual l y occurred in the history of English, given stan-stanes 'stone-stones,' shoe-shoen 'shoe-shoes' was analogized to the form now used i n PDE: (25)
stone: stones X shoes
=
shoe: X
=
The difficu l ty w i th the formula of proportion is that it gives no account of why one member of the pair is selected as the model. S i nce Meillet's time, a wide range of analogical processes has been identi fied (see Anttila 1 977, and, for a summary, Kiparsky 1 992). Kurytowicz ( 1 945-9) poin ted to some tendencies regarding selection of the model , for example, the tendency to replace a more constrai ned w i th a more general form, not vice versa. Two decades later Kiparsky ( 1 968) sought to redefine analogy in phonology as rule extension, thereby giving a formal acco u n t of the fact that analogy is not random i n language change. He views analogy as generalization or opti mization of a rule from a relatively limited domain to a far broader one. Of course, neither analogy as originally conceived nor rule general i zation are required to go to completio n : we still have foot-feet, mouse-mice alongside of stone-stones, and also run-ran alongside of love loved. Only reanalys i s can create new grammatical structures. However, the role of analogy should not be underestimated in the study of grammaticali zation. For one, the products of analogy, si nce they are overt, are in many cases the prime evidence for speakers of a language (and also for l i nguists !) that a change has taken place. Consider the development of the Romance perfect again. I n (23) (repeated and reglossed here for convenience as (26)), accusative plural agreement is overt and determinable (vos . . . fatigatos):
3.5 A nalogy/rule generalization (26)
65
Metuo enim ne ibi vos habebam fatigatos. fear- I SO for lest there you : ACC:PL have- I SO tired-ACC:PL ' For I fear that I have ti red you . '
However, i n (22a, b ) there is i ndeterminacy whether there is o r is not agreement, si nce zero neuter s ingular (nihil ' nothing' in (22a), satis 'enough ' in (22b)) is the "defaul t" gender/number marker in Lati n . With these constructions there is potential for reanalysis, but we recognize that the perfect has arisen only when there is overt and therefore determi nable lack of agreement between object and participle (PART) as i n : (27)
Haec omnia probalum habemus. those: ACC: P L all-ACC-PL tried-PART(?) have- I PL ' We h ave tried all those things.' (sixth century, Oribasius; ci ted i n Fleischman 1 982: 1 20)
So long as constructions occurred wh ich were ambiguous between adjectival par ticipials and perfects, e.g., (26), i t was not possible to tel l whether reanalysis had occurred or not, except perhaps by inference from the context. Specifically, the agreeing participial, which originated in a passive adjectival form , permits the un derstood subject of the participial to be the subject of either the sentence or of some other entity. For example, in (26) the agent of the act of tiring cou ld either be the subject ' 1 ' , as the translation 'I fear that I have tired you ' suggests (i .e., perfect), or some other, u nspec ified, individual(s), as in 'I fear I have/see you tired' (i .e., participial). B y contrast, the perfect requires that the u nderstood subject of the par ticiple is the subject of the sentence (Vincent 1 982). It is only when clear instances of non-agreement, e.g. , (27), occur, that we can fi nd defi n i tive overt evidence for the structure change. These unambiguously non-agreeing forms presumably arose by analogy ( rul e generalization) from neuter si ngular contexts to other contexts . A well-know n example of the cycl ical i n teraction of reanalysis, analogy ( generali zation), and reanalysis is the development of negation in French. The sequence of changes must have been as follows (Hock 1 99 1 [ 1 986] : 1 94; Schwegler 1 988): =
=
I. II.
(28)
Negation was accompl ished by placing the negative particle ne before the verb. A verb of motion negated by lie could optionally be reinforced by the pseudo-object noun pas 'step' in the context of verbs of movement: I I ne va (pas). he not goes (step) ' He doesn't go (a step). '
66
3
III.
T h e word pas was reanalyzed a s a negator particle i n a structure o f the type l i e Vlllovelllent (pas) . Pas w a s extended analogical ly t o new verbs having noth ing t o do w i th movement; i .e . , the structure was now lie V (pas) :
I V.
Mechanisms: reanalysis an.d an.alogy
(29)
I I n e suit pas. he not knows not ' H e does n ' t know. '
V.
The particle pas was reanalyzed as an obl igatory concom itant of lie for general negation: lie V pas. In the spoken vernacular pas came t o replace l i e via t w o stages: (lie) V pas (reanalysis of lie as optional), V pas (reanalysis by loss of lie), resulting in:
VI.
(30)
II sui t pas. he knows not ' H e does n ' t know. '
In the case of the French negator pas, we wou ld not know that reanalysis had taken place at stage III w i thout the evidence of the working of generalization at stage IV The reanalysis at stage VI would not have been possible without the generalization, si nce pas would have been too constrained by its origi nal semantics of ' step . ' Although analogy is best viewed a s generalization of a ru le o r construction, i n practice it i s often usefu l t o mai ntain the term "analogy" when referri ng t o certain local surface developments . For example, M i kola ( 1 975 : 1 70-2) describes the development in S amoyedic (Uralic) o f locati ve postpos i tions out of older locational nouns, which were themsel ves preceded by a noun i n the gen i tive, as i n : (3 1 )
Proto-S amoyedic
* miito-n + i n tent-GEN + top 'the top of the tent'
The suffixed - 1/ of the Uralic gen itive came to be reanalyzed as an in itial consonant on certai n postposi tions which were bei ng gram matical i zed out of nouns with mean ings such as 'upper surface' : (32)
mUlo + nin tent + 01110 'onto the tent"
This change began as a typical case of reanalysis of morpheme boundaries: [mato-#n##in ] > [mato-##nin] . The reanalysis i n turn yielded entire fam i l ies of postpositions with an i n itial 11 - , the cognates of wh ich may have i n i tial vowels in other Ural ic languages. We may speak of the generalization o f 1/ - here, but i t is not
3. 5
Table 3 . 1 Grallllllaticalizatioll of VO word order in English between AD and AD 1500
Accusative object before verb Accusative object a fter verb
67
A nalogy/rule generalization 1 000
c. I OOO
c. 1 200
c. DOO
c. 1 400
c .I
52.5% 47 .5%
52.7% 46.3 %
40+% 60- %
1 4.3% 85.7%
98. 1 3 %
500
1 . 870/"
Source: based on Fries ( 1 940: 20 I )
a case of rule genera l i zation, only of spread of n- i n word formation (for a s i m i lar example from Maori , see Section 6.2.4). So far we have considered analogy from the poi n t of view of generu l i zation of types of l i nguistic structure. There is, however, another im portant perspective on anal ogy : that o f generalization through patterns of usage, as rell ected by the freq uency with which tokens of these structures may occur across time. We w i l l be c i t i ng several recent examples o f stud ies o f frequency i n subseq uent chapters. Here we discuss an older, wel l -known example to i ntroduce the method : Fries's ( 1 940) study o f word-order change i n Engl ish in which the establishment of verb o�iect word order was traced through text coun ts at i n tervals of one hundred years . Among the relevant s tatistics concern i ng the position of the accusative object for the peri od AD 1 000 to 1 500 as presen ted by Fries are the figures in Table 3 . 1 . Th is method of analysis i s a quantitative one. Quantitative analyses can be done taking vari ous variables i n to account, such as spread across comm uni ties, or styles, or genres. The analys is by Fries that we h ave quoted, however, addresses only the vari able of object before verb versus verb before object. In any quantitative analysis the l i nguist ideal ly takes a representative sample of texts at regular i n tervals over several cen turies and traces the changes in form and mea n i ng of a particular construction as a fu nction of frequency of use in d iscourse. The kind of change characteri zed by the formula A > AlB > B is viewed not from the poi n t of view of types o f construction (e.g., OV > VO, or periphrastic fu ture > aff i xal future), but from the poi n t of view of tokens (how often are OV and VO used over ti me, how often are periphrastic and affi xal fu ture used over time?). The quantitative diachronic method captures the progressive aggregation of instances of the newer B construction at the expense of the older A construction . In the case of Old Engl ish word order, the A construction is verb-fi nal word order and the B construction is verb- i n itial word order. Typical ly, as here, the i n i t ial stage is already one of variation, and the fi nal exempl i fied stage may s t i l l be in variation . Such quanti tative studies highl ight the gradual ness of the spread of changes. It shou ld be mentioned that the gross numbers res u l t i ng from si mple cou nts of pre- and postverbal objects such as are i l lustrated by Fries's figures conceal com plex word-order adjustments i nvol ving differences such as those between pronou n
68
3 Mechanisms: reanalysis and analogy
and nou n , defi n i te and i ndefi n i te NP, heavy and light NP, i ndependent and depen dent clause, and so forth . A more complete explanation of word-order change i n Old and Middle English wou ld i nclude accoun ts of the structure o f the clause as a whole, includ i n g the ki nds of subjects that occur i n the clause and where, the kinds of object that occur after or before the verb, whether the verb in preobject position happe n s also to be in V2 position or not, and so forth (see Bean 1 983; Pintzuk 1 999; papers on Engl ish i n van Kemenade and Vincent 1 997, for some representative stud ies).
3.6
The diffe rentia l effects o f rea na lysis and a n a l ogy
From the perspective outli ned here, reanalysis and analogy (general iza tion) are distinctly different mechan isms and have d i fferen t etfects. Reanalysis es sentially i nvol ves l i near, syntagmatic, often local , reorgan ization and rule change. It is not directly observable. On the other hand, analogy essentially i nvolves paradig matic organi zatio n , change in surface collocations, and i n patterns of use. Analogy makes the u nobservable changes of reanalysis observable. The interaction of re analysis and analogy can be represented for the developmen t of be going to from directional phrase to future as i n Figure 3.2. Stage I is the s tage of the progressive with the directional verb and a purposive clause. Stage II i s that of the future aux iliary with a verb of activity ; it is the result of reanalysis. S tage III is that of the extension via analogy of the directional class of verbs to all verbs, i ncludi ng stative verbs. And S tage IV is the stage arising out of reanalysis of the complex auxiliary to a single m orpheme gonna. Stages I, III, and IV all still c oex ist in PDE. In the next chapter we will discuss some further extensions of the disti nctions between reanalysis and analogy, specifically with respect to mean i n g changes. While much current research makes the type of distinction outli ned here, it should be noted that it is most useful at the macrolevel, highl ighting major shifts such as the OV > VO word order, or the development of aux i l iaries discussed above. As work has progressed on defining the small steps of change that lead to such radical changes, and model s of syntax using networks rather than rules have been developed, the sharpness of the distinction has been brought i n to ques tion (e.g., Tabor 1 994a,b). One of the problems has already been alluded to evidence for reanalysis is largely found because of analogical general ization. Another issue i s that analogy i n the sense of rule generalization is itself a type of reanalysis, si nce u nder rule general ization the l inguistic contexts in which a rule may operate are extended or reanalyzed. This is covert i n the sense that structural contexts are highly abstract. Yet another issue is that where we have
3. 7 Conclusion -----
69
Syntagmatic axis
Mechanism: reanalysis
Stage I
going
[to visit B i l l ]
PROG
Vdir
[Purp. clause]
[be going to]
visit B i l l
TNS
Vact
[be going to]
l i ke B i l l
TNS
V
be
Stage II (by reanalysis)
Stage III (by analogy)
Stage IV
[gonna] l i ke/visit B i l l
(by reanalysis)
Paradigmatic axis Mechanism: analogy Figure
3.2 Schema of the development of alLrilialY be going
to
rich textual records, as in the case of the history of English and other European languages, or of Japanese and Chi nese, corpus research reveals often minuscule differences between texts across time. Ulti mately one might want to ask whether everything is not reanalysis. Nevertheless, the distinction is a useful heuristic for th inking about i nnovation (reanalysis) versus spread across the linguistic system (analogy). From this perspective we can say that reanalysis and analogy are the maj or mechanisms in language change. They do not define grammaticalization, nor are they coextensive with it, but grammatical ization does not occur without them . The subset of changes that are particular to grammatical ization are those that over time i nvol ve reanalysis of lex ical items and constructions as functional categories. We will discuss this unidirectionality of change more fully in Chapter 5 .
3 .7
Co nclusion
In this chapter we have discussed the mechan isms of reanalysis and anal ogy, and have shown that both play a crucial role i n grammaticalization, though neither is coextensive with i t. Furthermore, reanalysis i s the dominant mechanism driving it. We have also outl i ned some fundamental assumptions about language
70
3 Mechanisms: reanalysis and analogy
change, most particularly that it arises as a result of language acquisi tion by ad ults as wel l as children , and that i t occurs because of abduction, the reason i ng by which learners guess at systems. Much of the focus of this chapter, then , has been on per ception. A dom i nant theme in work on grammatical i zation s i nce the 1 970s has been the role of production in language change, most especially of ways in which speakers and hearers negotiate d iscourse strategies, and i t is to this issue that we now turn .
4
Pra g m atic fa cto rs
4. 1
Introduction
A l though it is possible to descri be change i n terms of the operation o f successive strategies of reanalysis (rule change) a n d analogy (rule general i zation), the important q uestion remains why these strategi es come about - i n other words, what enables the mechan isms we have out l ined, most especially those i nvol ved in gram maticalization. 1 It is tempting to think in terms of "causes" and even of "explanations" in the sense of "predictions." However, the phenomena that give rise to language change are so complex that they w i l l perhaps never be understood in enough detail for us to state precisely why a spec i fic change occurred in the past or to pred ict when one will occur and i f i t does what i t will be (Lass 1 980). Rather than referring to "causes" or "explanations," we speak more cautiously of motivations or enab l i ng factors, understandi ng always that we are referring to potential and statistical ly preferred, not absolute, factors (see, among many others, Greenberg 1 978b; Romai ne 1 982; Croft 2000; Maslova 2000). As mentioned previously, among motivations for change three h ave been widely discussed i n recen t years. Of greatest i n terest w i t h i n generative l i nguistics has been the role of language acquisition , especially c h i ld l anguage acq u i s i tion. Sociol i n guists, b y contrast, have tended t o focus attention on t h e role of com m u n i t ies and d i tl'erent types of con tact with i n them. Of special i n terest to those work i ng on gram matical ization has been the role of speakers and hearers negotiating meani n g i n communicative si tuations. Here we put forward arguments for the view that there are a number of competing motivations which can all i n some sense be said to be examples of max i m i zation of economy or "simplicity": basical ly they can be summarized as max i m i zation of efficiency via m i n i mal di fferentiation on the one hand, and m ax i m i zation of i n formati veness on the other. On th i s view, hearers play a major role i n change because they process i nput i n ways that may not match the speaker's i n tentions. B u t speakers also play a maj or role in enabl ing change, becau se in produci ng speech they have com mun ication as their goal, and therefore are al ways i n search of ways to guide the hearer i n i n terpretation. In an ideal commun i cative s i tuation,
72
4 Pragmatic factors
speakers take responsibility for success i n communication and seek to meet hearers' attempts "to i ntegrate new i n formation with i n formation that is already accessible" (Blakemore 1 990). However, differences i n what is actual ly accessi ble i n the com municative situation based on d ifferences in age, social background, culture, at tention, or other factors may over time cumulatively lead to change. Furthermore, the motivations of simpl icity and informativeness are i nevitably in competition i n the individual language user (see, e.g., Du Bois 1 985; Bates and MacWhinney 1 989), and therefore the development of language i nvolves conflict and problem solving (Dressler 1 985). There are great difficulties in defi n i ng the notion of economy in anything l ike rigorous terms. We know very l ittle about what does and does not take "effort" in producing or i nterpreting utterances, and stil l less about what would constitute economy of mental effort on either speaker's or hearer's behalf, although we prob ably know more about simplicity of perception than of production. Nonetheless, there seem to be useful, if sometimes intu itive, notions i nvolved. In considering the hearer's role, i t is usually assumed that the hearer w i l l seek the most u nambiguous i n terpretation, and furthermore that the hearer is actively engaged in i n terpreting (usually abducing) input, whether as a child or an adult i n the process of language acqu isition . It is therefore not surpris i ng that work on hearer motivations focuses on perception, and on meanings i nterpreted from the l i nguistic data which is the i nput to the acquisition process. However, it is also possible to thi n k of hearers as the targets of speakers' output. From this perspective, hearers motivate speakers' i n tent to be i n formative and clear. In considering the speaker's role, i t has been customary to thi n k of the tendency to reduce the speech signal , e.g., via rapid speech, a process resulting in "signal simplicity." The reduction of be going to to be gonna is an example. S ignal sim plification typicall y resul ts from the routi n i zation (idiomatization) of expressions. Rather than find different ways of saying approximately the same thing, speakers wi l l repeat expressions they have heard before, even if they are in competition with other expressions, perhaps i n the interest of sounding "with it." Wel l-known recen t .examples include the use of you know, be all, etc. Such routinized, or id iomatized, expressions can be stored and used as simple u nits. Naturally they are more frequent in discourse than expressions created and used "on the fl y," which may i ndeed be novel, once. Idiomatization of expressions tends to lead to reduction and simplification of the signal. With this process in m i nd, Langacker has said : "It would not be entirely i nappropriate to regard languages in their diachronic aspect as gigantic expression compacting mach ines" ( 1 977: 1 06). However, compacting, obl i teration of bound aries, and reduction of redundancy is balanced i n normal language situations by the introduction of new and innovative ways of saying approximatel y the same thi ng.
4. 1 Introduction
73
These new and innovative ways of say ing things are brought about by speakers seeking to enhance expressivity. This is typically done through the "deroutinizing" of constructions, in other words, through finding new ways to say old things. Ex pressivity serves the dual fu nction of improving i n formativeness for the hearer and at the same time allowing the speaker to convey attitudes toward the situation, including the speech si tuation . This very process of i nnovation is itself typically based on a principle of economy, specifically the economy of reusing extant forms for new purposes (Werner and Kaplan 1 963 ; Slobin I 977). To return to our exam ple of the extension of be going to to the intentional future: the d i rectional phrase has been reused; it is more substantive (phonolog ical ly longer) and therefore more accessible to hearers than, e.g., 'Il or even will. As a future it is also more based in the speaker's subjective attitude and perspective on what is bei n g talked about than is its locational counterpart. While the competition between motivations has been a major topi c of research , individual researchers have tended t o b e i nterested i n different subaspects of the complex balance between creativ i ty on the one hand and routin i zation on the other. Therefore studies of grammaticali zation have taken differen t paths. A focus on hearers' tendency to reanalyze abductively may correlate with work on mor phosyntactic changes i n grammatical ization (e.g., Langac ker 1 977). A focus on speakers' expressive use of language to get a point across may correlate with work on lexical origins of grammatical i zation, especially on pragmatic enrichment of lexical items i n the early stages of grammaticalization (e.g., Heine, Claudi, and Hiinnemeyer 1 99 1 a,b; Traugott and Konig 1 99 1 ). A focus on speakers ' tendency to economize may correlate with work on routinization, and frequency (e.g., G iv6n 1 979, 1 99 1 b; Hopper 1 987; Krug 1 998, 200 1 ; Bybee and Scheibman 1 999; Bybee 2oo2). These approaches are, of course, not necessarily contradictory, although they do sometimes lead to different views of what should or should not be con sidered a case of grammaticali zation. Any comprehensive study of a particular set of grammaticalization changes that lasts over long periods of time and i nvolves continued grammaticalization must ideal ly take cognizance of all three kinds of approach. Before concluding this section, we should note that the claim that grammatical ization (and indeed language change i n general) is motivated by speaker-hearer interactions and communicative strategies is a claim that change is goal-directed. For example, Heine, Claudi, and Hiinnemeyer claim that "grammaticalization can be i nterpreted as the result of a process that has problem solv i n g as i ts main goal" ( I 99 1 a: 29). There has been a tendency i n American linguistics to distrust arguments about goal-directedness, partly on the grounds that they cannot be em pirically proven, partly because the important research question has until recently been considered to be how language is understood out of context. Arguments
74
4 Pragmatic factors
based on mechan isms for change, and on (passive) capaci ties for language, have been priv ileged over arguments based on purposes to wh ich language is put. For example, Bybee has argued that the development of morphology is spurred on by spontaneous processes whereby semantic functions that are s i m i lar are sub su med under closely related grammatical functions (Bybee 1 985 : 204 ; see also Bybee, Perkins, and Pagl i uca 1 994 : 297-300). These processes are i n Bybee's view cognitive rather than commun icative. Part of the problem with the concept of goal-orien ted ness i s that i t is often discussed in terms of "need" or set goals, i n other words, in "teleological" terms. Clearly "communicative need" is not a plausible motivation i n most cases of gram matical ization, si nce not al \ l anguages express the same grammatical fu nctions, and even less do they express the same grammatical functions i n the same way. Furthermore, to assume that as an older system becomes eroded it may cease to function at an adequate level of commun icative coherence and therefore must nec essarily be revi tal ized (as suggested by G iv6n 1 982: 1 1 7, c i ted in Bybee 1 985 : 202) is to posit a stage of language such as is unknown, i n other words i t violates the uniform itarian pri nciple and is not emp i rical\y supported by the data. Most emphatical \y, languages are not goal-oriented. We agree that "the view which as cribes l anguage a will of i ts own, a sort of conscious control over i ts own future, seems to us gratuitous and untenable. It remains true, however, that language is a communicative tool at the d isposal of its speakers, to whom the attri bu tion of an i ndependent will and vol i tion is considerably less controversial" (Vi ncent 1 97 8 : 4 1 4). The position w e take i s that users m a y b e consciously o r unconsciously goal-oriented (see Kel\er 1 994 for a detailed accou n t of goal-oriented ness i n lan guage change that is not teleol ogical i n the sense mentioned above). In speaking of communicati ve strategies and problem solv i ng i n the course of speaker-hearer i nteraction, we refer not to filling gaps, but rather to strateg ies used by speakers and hearers in producing and u nderstanding the flow of speech as it is created . These speakers have intentions, and their i ntentions may lead to change over time. How ever, barring cases of "language engineeri ng," e.g., attempts to legislate against use of the derivative -mall, or exclusive use of he as a generic pron� un, speakers for the most part do not i n tend to change the language. On the con trary, many would like to preven t change i f possible.
4.2
I nfe rencing a n d m e a n i n g change
The discussion i n the precedi ng section has made reference to various motivations such as economy, efficiency, clarity, expressi v i ty, and rout i n i zation. Such motivations are issues of usage and speaker-hearer purposes, and can be
4. 2 Inferencing and meaning change
75
cal led "pragmatic," that is, they have to do w i th the relationship between language and the contexts i n which it is used, including other instances of language, most especially the mean i ngs that arise from th is relationship (Levinson 1 983, 2000). S i nce pragmatics by defi n ition deals with meani ngs beyond s tructure, many l i ngu ists working in the tradition of formal grammars for a long time excluded pragmatics from consideration in accounting for motivations for change. Many also excl uded the meani ngs usually treated by lexical semantics, that is, components of sense and their relation to the objects to which they refer. For example, in discussing the syntactic changes i nvol ved in the development of the modals, Lightfoot ( 1 979) argued that syntactic change was autonomous, i.e., i ndependent of semantic o r pragmatic motivations. He attempted to show that the modals "underwent very many changes in their syntax and in their mean ing but . . . these changes seem to have proceeded q u i te i ndependently of each other" ( 1 979: 1 00). Indeed, he went on to say : "it does not seem possible to define a class of modals (and therefore of preterite-presents) on semantic grounds" ( 1 979: 1 03). Such an approach is a natural conseq uence of conceptual izing significant change as resulting from chi ldren 's attempts to seek cues to grammar as a computational system (see, e.g., Lightfoot 1 999), rather than as a strategic one. The approach from "autonomous sy ntax" has been cal led i n to question by many l i ngu i sts working on grammatical ization . Th is fol lows i nevitably from the i n terest in changes whereby a lexical i tem becomes a grammatical one, because a lexical item by defi ni tion has semantic as wel l as syntactic, morphological, and phono logical properties. But it also fol lows from the concern that a theory which regards semantic change as i ndependent of morphosyntactic change provi des no reasoned accou nt for the extensive evidence that grammaticalization affects s i m ilar classes of lexical i tems in s i m i lar ways across a wide number of languages. This poi nt was made especially cogently by Bybee and her col leagues i n various studies of verbal morphology (e.g., Bybee 1 985 ; Bybee and Dahl 1 989) and by Heine and his colleagues i n various studies of nominal morphology (e.g. , Heine, Claud i , and Htinnemeyer 199 1 a); val uable resources for putative examples (not all with empir ical hi storical su pport) of cross-l i nguistic and cross-categorial grammatical ization include Hei ne et al. ( 1 993) and Heine and Ku teva (2002). Indeed, much work on grammatical ization si nce the early I 980s has focused ex tensively on the ki nds of mean ing changes i nvol ved in grammatica l i zation and the cognit ive motivations beh ind them. For some, the mean ing changes are regarded as semantic, e.g., "semantic change leads to the development of grammatical mean ing" (Bybee and Pagliuca 1 985 : 59). For others it is mainly pragmatic (Traugott and Konig 1 99 1 ). For some it is thought to be motivated pri mari l y by metaphor ical processes (Claudi and Heine 1 986; Sweetser 1 990), for others i t is thought to be motivated by associative or "metonymic" as wel l as metaphorical processes
76
4 Pragmatic factors
(e.g., Traugott and Konig 1 99 1 ; Heine, Claudi, and Hiinnemeyer 1 99 I a) . The position we take here and will elaborate on in subsequent chapters is that mean i ng changes and the cognitive strategies that motivate them are central in the early stages of grammaticalization and are crucially linked to expressivity. Furthermore, the meaning changes are initially pragmatic and associative, arising in the context of the flow of speech. At later stages, as grammaticalization continues and forms become routinized, meaning loss or "bleachi ng" typically occurs, but even so, older meanings may stil l continue to constrain newer, "emptier" ones. Before we proceed, a few words about some of the widely accepted ideas con cerning semantics and pragmatics may be helpful. 4.2. 1
Semantics and pragmatics
There is a vast literature on semantics and pragmatics but as yet l i ttle consensus on exactly where the bou ndaries between the two areas lie, or even whether there are indeed boundaries. Nevertheless, there is a pre-theoretical sense in which it is clear that a distinction needs to be made between the sentence (semantic) mean ing of Can you pass the salt? ( 'Are you able/willing to pass the salt?'), the expected response to which would be Yes or No, and the utterance (pragmatic) mean ing ( 'Please pass the salt'), the expected response to which is the non-li nguistic action of passing the salt. For purposes of this chapter, it must suffice to note that we believe that l i nguistic theory should eventually provide an integrated account of semantics and pragmatics. It is useful to distinguish between that part of semantics that concerns lexical , phrase, and sentence meaning, and that part of pragmatics that concerns inferences about l inguistic meaning based on contextual assumptions such as the coopera tiveness of participants in a conversation. The view of the relationship between semantics and pragmatics we adopt here is as follows (Levinson 1 983, 2000; Green 1 996 [ 1 989]). Semantics is primarily concerned with meani ngs that are relatively stable out of context, typically arbitrary, and analyzable in terms of the logical conditions under which they would be true. Pragmatlcs, by contrast, is primarily concerned with the beliefs and inferences about the nature of the as sumptions . made by participants and the purposes for which utterances are used i n the context of communicative language use. It concerns both speakers' indirect meaning, beyond what is said, and also hearers ' i n terpretations, which tend to enrich what is said in order to interpret it as relevant to the context of discourse (Sperber and Wi lson 1 995 [ 1 986] ). Many of the bel iefs and inferences which are the subject of pragmatics are thought to be cognitively universal, in other words, not learned or arbitrary. Furthermore, they are rarely if ever subject to analysis in terms of truth. =
=
4. 2 Illjerellcillg and meaning change 4.2.2
77
Relationships between senses of a form : homonymy and polysemy
An issue on which there is l i ttle agreement across various subfields of li nguistics is exactly how to characteri ze the relationsh ip between the various senses of a form. One approach is to maximize d i tl"erence and homonymy (same form, unconnected different meanings). Rejecting the traditional l iterary practice of grouping together all related meani ngs that can be associated w i th a single phonolog ical form, McCawley ( 1 968: 1 26) suggested that there is n o a priori reason for grouping items together in a dictionary : one could take the notion "lexical item" to mean the combination of a single semantic reading w i th a single underlying phonological shape, a single syntactic category, and a s i ngle set of specifications of exceptional behavior with respect to rules. He argued that sad i n the two meanings displayed i n ( I ) should be analyzed as two separate i tems: (I)
Sad , : 'experiencing sadness, said of a l i ving bei ng' Sad2 : 'evoking sadness, said of an esthetic object'
Likewise, from this perspective the call of abi l i ty and possibility and perm ission are unconnected semantically. Another approach is to maximize s i m i l ari ty and monosemy (same form, same meaning). For example, Groefsema ( 1 995) argues that the various mean i ngs of the modals are so indeterm inate that they should be analyzed as shari ng a unitary underspecified mean ing; from th is perspective, there is one can. A third approach is to argue that certai n forms share conceptually related meanings (polysemy) (e.g., G. Lakoff and Joh nson 1 980; S weetser 1 990). Bybee, for example, suggests that there is one form can in PDE with related meanings: (2)
a. b. c. d.
Mental abi lity: e.g., I can read German. Physical ability: e.g., I can swim a mi le. Root possibi l i ty : e.g., Thi s word can be used in many contexts. Permission: e.g., I can take books out for two weeks. (Bybee 1 988: 256)
It is often argued that the fi ne, sometimes min imal ly discrete, mean i ng distinc tions between various stages of grammatical i zation or between focal clusters on a cline call for a theory i n which different meani ngs may be closer or more distant (see G. Lakoff 1 987 on the concept of "networks" of polysemies ; and, from a different perspective, Kemmer 1 993b on semantic maps of related terms within the domain of voice). For example, mental and physical abil i ty are more closely related to each other than to permission among the polysemies of can. In general, from the perspective of grammatical ization it is methodological ly essential to assume polysemy if there is a plausible semantic rel ationship, whether or not the forms belong to the same syntactic category, because otherwise relation ships between more and l ess grammaticali zed variants of the same form cannot be
78
4 Pragmatic factors
establ ished, either diachron ically or synchron ical ly. What constitutes a "plaus ible seman tic relationship" has until fairly recently been a matter of considerable de bate. Drawi ng on Haiman, especially ( I 985a), Croft ( 1 990) discusses a usefu l heuristic for the disti nction between the homonymy o f two, fO, too (al l [tu]), and the polysemy of the directional and recipient mean i ngs of to in I dro ve fo Chicago, I told the story fo my brother: He suggests that a m aj or cri terion is evidence fro'!1 _ cross- typological comparison: "if many diverse languages independently have the same pattern of 'homonymy, ' then the meani ngs are closel y related" (p. 1 66). Two, to, too do not tend to be expressed by the same form cross- l i ngui stical ly; this, in addition to their l ack of historical relatedness, provides ev idence that they are homony mous i n Engl ish. However, directional and recipient to are freq uen tly, indeed typical ly, expressed by the same form cross-l i nguistical ly (cf. also -lie in Fi nnish). Th is, together with their h istorical related ness, provides ev idence that they are polysemous. There has been a long h istory in l i nguistics of concern about the notion of polysemy. For example, Boli nger has said: "the natural cond ition of l anguage is to preserve one form for one meani ng, and one meaning for one form" (Boli nger 1 977: x, ci ted in Hai man 1 985a: 2 1 ). This concern presumably stems from the "scienti fic" approach to language which is the fou ndation of l i nguistics. To overs i m p l i fy, from this poi nt of view the optimal language would be one i n which every mean ing was distinct, just as every numeral is distinct (the "ideal i zed l anguage" we referred to in Section 3.3). However, such "opti mal i ty" would clearly in actual fact be dysfunctional si nce there are far too many mean i ngs for the brain to remem ber i ndividual expressions for them . "One form - one mean ing" is an ideal on the di mensions of choice of form and the motivation to max i m i ze i n formation. It is balanced and offset by another optimal ity, that of associating l i ke forms with l i ke meanings, in other words, of developing polysem ies (Hai man 1 985a). Si nce grammaticalization comprises the set of changes whereby structural relationsh i ps and associations among them are given grammatical expression, it is not surprising that it typically i nvolves polysemy. 4.2.3
Con versational and con ventional inferencing
With regard to pragmatics, we are particu larly i n terested in those infer ences that are made in l i nguistic contexts from one clause or constituent to another, or even from one utterance to another. These are i n pri nciple impl icational infer ences ( i n the l i nguistic jargon cal led "impl icatures") of the type characterized by Grice ( 1 975) as "conversational ." Grice suggested that such i n ferences are com putable on the basis not of lexical mean i ngs alone, but of lexical meani ngs together with i m p l icatures arising from speech act maxims such as "Make your contribution
4. 2
Inferencing an.d mean.ing change
79
as i n formative as is requi red ( for the current purposes of the exchange)" (the first max i m of Quantity), "Do not make your contribution more i n formative than is req ui red" (the second max i m of Quantity), "Try to make your contribution one that is true" (the max i m o f Qual ity), "Be relevant" (the max i m of Relation), and "Be perspicuous" (the maxim of Man ner) (Grice 1 97 5 : 45-7). There i s considerable debate whether this i s the right set o f maxims. Indeed , i t has been suggested that a pri nciple of Relevance alone. defi ned in such a way as to i nclude i n formativeness, is sufficient to accou nt for pragmatic meani ng (Sperber and Wi lson 1 995 [ 1 986] ; for other views, see Atlas and Levi nson 1 98 1 , Horn 1 984, updated in Horn 1 996 and Levi nson 2000). As we w i l l show, grammatical ization changes seem to draw primari ly on the second maxim of Quantity, in the form "Say no more than you must and mean more thereby" (see Levi nson 1 983) and Relevance. Another point of debate has been w hether "max im" is the best term for the reasoning processes that people bring to the speech event, since it evokes imperatives and pri nci ples. We prefer the term "heu ristic" proposed by Lev i n son (2000), which evokes strategic choice in the comm u n i cative situation. Most conversational i mpl icatures are strictly speaking interpreted abductively (given an Ullerance. hearers may relate it to a general heuristic, and guess the speaker's intent). The guess may be wrong because the heuristics can al ways be fl outed, e.g., it is possible for speakers to be u n i n formative or to l ie. Furthermore, impl icatures are "cancelable" ei ther by the speaker (in wh ich case an explanation is given), or by hearers ' inferences from the s i tuation. An exam ple i s : (3)
a. J o h n h a s three cows.
The implicature from the fi rst heuristic of Quantity is that: (3 )
b. John has three cows and no more.
But this can be canceled by a vague statement such as (3 )
c . John has three cows. i f not more.
and even den ied: (3)
d. John has th ree cows. in fact ten.
The pragmatic effect of conversational impl icatures across u llerances and their cancelabi lity can be i l l ustrated by the example of conjoi ned clauses wi thout any con nective, that is, without any structural marker of coherence. These are l i kely to be interpretively enriched as hav i ng some coherence, that is, relevance to each other, simply because they are uttered in sequence. (4)
a. The earthquake hit at 8 a.m. A fou r-car crash occu rred.
80
4 Pragmatic factors
Typically the relationship inferred w i l l be that of temporal sequence and even causal connection if the clauses are action/event clauses and connectable in terms of encyclopedic or world knowledge, as i n (4a); however, no such relationship i s likely t o b e i n ferred i f the sequence i s i n coherent i n terms o f world knowledge, as in (4b), where inference of a causal con nection is u n l i kely, and even close temporal connection may be in doubt: (4)
b. A four-car crash occurred. The earthquake h i t at 8 a.m.
If a grammatical form is present, e.g., and, because, you see, this element w i l l further "constrai n the relevance o f the proposition i t i ntroduces" (B lakemore 1 987 : 1 30). But this constraint sti l l leaves i ndetermi nacies. For example, the presence of and i n (5a) implies only that a connection is i n tended by the speaker and that the hearer should compute one; it does not require that the i mplicature is a causal one, though that is what hearers would typically assume u nless causali ty is canceled, as it is in (5b): (5)
a. The earthquake hit at 8 a.m. and a four-car crash occurred. b. The earthquake hit and a four-car crash occurred, but actually the cause was the fog, not the earthquake.
Conversational impl icatures are typically contrasted w i th "conventional" ones (in the next section we will show that conventional implicatures typically arise out of conversational ones). Conventional implicatures are unpredictable and arbi trary, that is, they must be learned as part of the polysemies of the word, and are not cancelable. Thus i n (6a) the verb manage conventionally implicates (6b): (6)
a. John managed to sol ve the problem. b. John sol ved the problem.
Neither (6a) nor (6b) are cancelable by, e.g. : (6)
c. but he didn't sol ve it.
S i nce they must be specially learned along with the phonological " and syntactic characteristics for the i tem i n question, conventional meanings can, at least for our purposes, be i ncluded among the semantic polysemies of a form. For a classic case consider the temporal and causal meanings of since, as in (7): (7)
a. I have done quite a bit of writing si nce we last got together. (temporal) b. Since I have a final exam tomorrow, I won ' t be able to go out tonight. (causal)
With since, when both clauses refer to events, especially events in the past, the reading is typical ly temporal, as in (7a). When one clause refers to a non-past
4. 3 The role of pragmatic inferencing in g rammaticalization
81
event or to a s tate, the reading i s typically causal , as i n (7b). The causal meani n g is conventional a n d not cancelable, a s illustrated b y (7c): (7)
c. · S i nce I h ave a fi nal exam tomorrow, I won ' t be able to go out tonight, but not because of the exam !
In (7a) the first clause contains a past tense. In (7b) neither clause does . In other words, d i fferen t mea n i ngs of since can be associ ated w i th d i fferent structural con texts. The d i fference between these mean i ngs is someti mes sy ntactically obscured, and then there can be ambigu i ty, as i n (7d) : (7)
d. S ince Susan left him, John has been very miserable. (temporal or causal )
These facts allow us to conclude that since i s semantically ambiguous (polyse mous). We turn now to the question of whether there are pragmatic as wel l as semantic polysemies. Consider, for example, after i n (8a) : (8)
a. After we read your novel we fel t greatly i nspired.
Thi s may be i nterpreted as a l i teral statement of temporal sequence, or it may impl icate: (8)
b. Because we read your novel we fel t greatly inspired.
The implicature in (8b) strengthens i n formativeness because it enriches the relation between After we read your novel and the rest of the utterance, thus providing an interpretation of why the speaker thought it was relevant to i nclude these temporal facts. However, there are no regular structural correlates for thi s relationship, such as completed versus ongo i ng eventhood, and there are no regularly assoc iated, uncancelable causal implicatures. Rather, example (8a) suggests that there can be pragmatic ambiguities/polysemies as well as semantic ambig u i ti es/polysem ies. For fuller discussion of the i mportance of recognizing both pragmatic and semantic ambiguities, see Horn ( 1 989) and S weetser ( 1 990).
4.3
The role of p ragmatic i nferencing i n gra m m atica l ization
With regard to the question of what role pragmatic i n ferenci ng has i n grammaticali zation, toward the end o f h i s sem inal article "Logic and conversation," Grice tentatively stated: "it may not be i m possi ble for what starts l i fe, so to speak, as a conversational implicature to become conventional i zed" ( 1 975 : 58). Th i s idea had been explored very briefly i n Geis and Zwicky ( 1 97 1 ) i n connection with the development of causal since, and was expl ici tly followed up i n Cole
82
4 Pragmatic factors
( 1 975) i n con nection with let 's. Dahl hypothesi zes that many of what he caI ls "the secondary mean i ngs" of tense and aspect, e.g., the "present relevance" of the perfect, are derived by conven tional i zation of i mpl icatures. Below we w i l l show that in early stages of grammatical ization conversational implicatures frequently ,, become "semanticized, 2 that is, become part of the semantic polysem ies of a form. Dah l 's characterization of the process is as foI l ows: if some condition happens to be ful fi l led frequently when a certai n category is used, a stronger association may develop between the condition and the category in such a way that the condition comes to be understood as an integral pal1 of the meaning of the category. (Dahl 1 985: I I )
These i nsights have been elaborated i n connection with seman tic change i n general, not speci fical ly grammaticali zation, in Traugott ( 1 999) a n d Traugott and Dasher (2002). Here it must suffice to note that the ki nds of semantic change undergone by lexical i tems or constructions in grammatical i zation are a subset of semantic changes in general, and that for inferences to play a sign i ficant role i n gram matical i zation, they must be frequently occurri ng, si nce only standard inferences can plausibly be assumed to have a lasting i m pact on the mean ing of an expression or to function cross-l i nguisticaIly. Among stereotypical i n ferences we may i nclude the i n ferences of causality from temporal seq uence that we have already discussed i n connection with since and after, the well-known l ogical fal l acy (abduction) characterized as post hoc ergo propter hoc 'after th is, therefore because of this. ' By contrast, we would not expect grammaticalization of such strictly l ocal and idiosyncratic, h ighly contextual ized, i n ferences as are exemplified by : ( 9)
a. What on earth has happened to the roast beef? b. The dog is looking· very happy. (understood to i mpl icate that perhaps the dog (Levi nson 1 98 3 : 1 26) has eaten the roast beet)
One question is when we can recogn i ze conventional i zation to have occurred . A brief look, before we proceed , at some early examples of the con tex ts i n which the change of since (originally sipfJan) came about will be useful i n serv i ng as a methodological cau tion, and show that i t is essential to look beyond individual sentences to larger contexts before reaching too hasty conc l usions that change has occurred . In OE texts before AC 1 050 sipfJan as a preposition was used al most exclusively to mean 'from the time that, after. ' The standard causal was fo r Pre/ll pe ' for that that,' originally a deictic expression. As a con nective sipfJan meant ' from the time that,' that is, it marked the lower temporal boundary of the event in the main clause, and signaled an overlap with some poi nt in an earl ier event.
4. 3 The role of pragmatic inferencing in grammaticalization
83
In certain contexts, however, the modern reader may detect a causal impl icature. For example, M i tchel l ( 1 985 (2) : 352) ci tes ( 1 0) as a putative example of causal sijJpan : ( 1 0)
micel pa si ppan he i rre wres & gewundod. he ofslog then, after/si nce he angry was and wounded. he s l aughtered mllch pres folces. of-that troop
(c. 880. Orosius 4 1 . 1 56. 1 I )
When we read this sen tence out of i ts larger context, the partici ple gewundod, bei ng perfective and invol ving change of state, seems to favor a temporal read i ng, but the adjective i,.re favors a causal one, s ince it appears to express state. Never theless, a temporal read i ng with the adjective is plaus i ble if we assume that was is inceptive-resu ltative or perfective, i .e., ' had come to be' rather than ' was i n a state. ' In other words, the adjective i n th i s context can be i n terpreted as express i n g a conti ngent rather than a general state (the latter would m ore probably have been expressed by bid). Indeed we fi nd that this sentence occ urs i n the con tex t of a narrative concern ing the legendary Pyrrh ic victory. I n the battle a Roman sol d ier wounds an elephant i n the navel ; this elephant, hav i ng become enraged and wounded , wreaks mayhem on the army. SijJpan in ( 1 0), then, is best i n terpreted as a temporal (i ndeed it partially translates the Latin adverbial postquam ' afterward ' ) al lowing secondary causative impl icatures. I t i s not a conclusive example o f the conventional izi ng of causative i n ference in Old English. A more convi nci ng example is: (I I)
Ac ic pe wille nu giet getrecan pone weg si ppan i) u ongi tst purh but I thee w i l l now sti l l teach that way si nce thou seest through mine lare hwret sio sot')e gesreli) bii) , & hwrer hio bit'). my teaching what that true happi ness is, and where i t is ' But sti l l I will now teach you the way si nce you see that true happi ness comes through my teachi ng. and where it i s . ' (c. 8 8 0 , Boeth ius 36 1 04.26)
Here sijJpan translates the Lati n causative quoniam ' because. ' But even wi thout the Lat i n origi nal we can assume it is causative s i nce the context is non-narrative: the stat i ve perception-mental verb ' see, understand' i ntroduces an aspectual generic clause signaled by the verb bi(} i nstead of the contingency verb is. A l though the causal inference i s detectable in Old Engl ish in examples such as ( I I ), so many other examples are undecidable that we cannot establish that the causal i n ference had truly become conventional i zed at this period. The change appears to have occurred i n the fi fteenth century, when the form is attested frequently i n stative and other non-completive environments where the temporal read i ng i s blocked , as in ( I I ).
84
4 Pragmatic factors
A second question is what types of inferences are most characteristic of these early stages. Two different though mutually i n tersecting kinds, metaphorical pro cesses and metonymic processes, have been much d iscussed i n the l iterature, and it is to these that we now turn . 4.3. 1
Metaphorical processes
Metaphorical i nnovation i s one of the most widely recogn i zed processes in meaning change. S tandard examples of long-established metaphors i nclude such utterances as ( 1 2) and ( 1 3): ( 1 2) ( 1 3)
Sally is a block of ice. The sentence was fi l led with emotion.
(Searle 1 993 [ 1 979] : 97) (Reddy 1 993 [ 1 979] : 288)
Although defin itions of metaphor vary, most have certai n concepts in com mon, especially understand ing and experienci ng one ki nd of thing in terms of another, and directional i ty of transfer from a basic, usually concrete, meaning to one more abstract (see, for example, 1. D. Sapir 1 977; G. Lakoff and lohnson 1 980; Claudi and Hei ne 1 986; Hei ne, C1aud i , and Hiinnemeyer 1 99 1 a; Kovecses 2002). Metaphorical processes are processes of inference across conceptual boundaries, and are typically referred to in terms of "mappi ngs," or "associ ative leaps," from one domain to another. The mapping is not random, but motivated by analogy and iconic relationships. These relationships tend to be observable cross-l inguistical ly. Some have been thought of as "image schemata" w i th very concrete sources that are mapped onto abstract concepts. In the lex ical domain one i m age schema that i s well known is that of seeing and knowing, grasping and u nderstanding (as in I see/grasp the point of your argument). In this particular case the relatively concrete concept has been said to be mapped onto the relatively abstract one i n a relationship called the "mind-as-body metaphor": bod i l y experience is' a source of vocabulary for psychological states (Sweetser 1 990: 28-48). Another well-known relationship i s "force dynamics" (Talmy 1 988; lackendoff 1 990; S weetser 1 990), the relationship of forces and barriers fou nd in such expressions as ( 1 4) and ( 1 5) : ( 1 4) ( 1 5)
The crack in the stone let the water How through. I have a mental block about sports.
Metaphoric processes have traditionally been regarded as semantic. Recen tly, however, i t has been suggested that, not being truth condi tional, but rather being based i n communicative use, they are more appropriately considered pragmatic (e.g., Levi nson 1 983). We accept the view that metaphor is pragmatic, and argue below that in so far as metaphor is primarily analogical i n character, i t is different
4. 3 The role of pragmatic inferencing in grammaticalization
85
from the kinds of conversational processes based on heuristics mentioned above, which operate primarily in l i near, syntagmatic ways. A further d i fference identi fied by Green ( 1 996 [ 1 989] : 1 22) is that, at a superficial level , metaphors often i nvol ve propositions that are intended to be recogn ized as literal l y false (for example, it is false that a person can be a block of ice), but conversational impl icatures do not. Most examples of metaphorical processes in language change have been dis cussed with respect to the lexicon. However, arguments h ave been put forward that early grammaticali zation is also strongly motivated by metaphoric processes. Typical of early claims along these l i nes is: "Rather than subscribe to the idea that grammatical evolution is driven by communicative necessity, we suggest that human language users have a natural propensity for making metaphorical extensions that lead to the i ncreased use of certain items" (Bybee and Pagliuca 1 985 : 75). Probably the most appeal ing examples of metaphoric processes i n grammati calization are provided by the development of spatiotemporal terms. Claudi and Heine ( 1 986) and Hei ne, Claudi , and Hiinnemeyer ( 1 99 1 a,b) discuss the devel opment of body part terms into locatives, of spatia Is into temporals, etc. in terms of metaphors such as S P A C E I S A N O B J E C T , T I M E I S S P A C E (capitals indicate abstract, cross-linguistic mean i ngs, as opposed to language-speci fic lexical items). For example, spatial terms such as B E H I N D can be derived metaphorical ly from a body part (an example of the shift from O B J E C T > S P A C E ) , and subsequently temporal terms can be derived metaphorically from the spati al term (via S P A C E > T I M E ) , e.g., We are behind in paying our bills. Spatial terms abound cross li nguistically as temporal particles, auxil iaries, etc. (see Traugott 1 978, 1 985a; Bybee and Dahl 1 989; Bybee, Pagl iuca, and Perkins 1 99 1 on expressions of the future). A few examples from English which have been regarded as metaphorical i n origin i nclude be going to (future), in the years ahead (future), drink something up/down (completive), drink on (continuative), come to believe that (ingressive). Extensive examples from African languages can be found in Heine, Claudi, and Hiinnemeyer { 1 99 1 a,b), and from Oceanic languages i n Lichtenberk (I 99 1 b) (e.g. , G O for continuative a n d future, C O M E for i ngressive and future). Svorou ( 1 993) and Haspelmath ( 1 997) provide detailed cross-linguistic evidence for S P A C E > TIME.
Another domain of meaning change among grammatical categories that has been widely regarded as metaphoric is the development of modal meani ngs, particu larly the development of meani ngs relating to obligation into meanings relating to possibil i ty and probability (known as "epistemic" mean ings). For example: "The obl igation sense of have to predicates certain conditions on a w i llful agent: X i s obl iged t o Y . The epistem ic sense is a metaphorical extension o f obligation to
86
4 Pragmatic factors
apply to the truth of a proposi tion: X (a propos i tion) is obl i ged to be true" (Bybee and Pag l i uca 1 985 : 73). In this v iew (modified in B y bee 1 990), the process envi sioned appears to be strictly speaking the schematic mapping of one concept onto another. B u i l d i ng on Talmy ( 1 976, 1 988), Sweetser takes a d i tIerent approach to the modals, that of "sociophysical concepts of forces and barriers" ( 1 990: 52). The may of permission is, accord ing to Sweetser, understood i n terms of "a poten tial but absent barrier," obligative must in terms of "a compel l i ng force directing th e subject towards an act." The force of must i s "di rectly appl ied and i rresistible," whereas that of have to i s resistible under certain c i rcumstances, cf. : ( 1 6)
I have tol??must get this paper in. but I guess I ' l l go Lo the movies i nstead. (Sweetser 1 990: 54)
Sweetser regards the epistemic mean i ngs of these modals as deriving from the tendency to experience the physical, social, and epislem ic worlds in partially similar ways. Thi s s i m i l arity in experience, she suggests, allows the mapp i ng of sociophysical potential i ty onto the world of reason i ng. For example, with respect to may, she says: "In both the soc iophysical and the epistemic world, nothing prevents the occurrence of whatever is modal ly marked w i th may; the chain of events is not obstructed" (p. 60). With respect to must, she gives the fol lowi ng analysis (p. 6 1 ): ( 1 7)
a. You must come home by ten. (Mom said so.) 'The direct force (of Mom's authority) compels you to come home by len. b. You must have been home last night. 'The avai lable (direct) evidence compel s me to the concl usion that you were home. '
She goes on to show that y et another metaphorical mapping is possi ble: of potential barriers to the conversational world (what i s often cal led metal i nguistic expression). Thus, i n a hypothetical s i tuation where Mondale's advisor i s giving directions to a speech writer, the fol lowing m ight be i magi ned (p. 7 1 ) : ( 1 8)
Reagan wil l/must be a nice guy (as far as the content of the speech is concerned. even if we cri ticize his policies).
In other words, "the i nterlocutor is being allowed to treat a certai n statemenl as appropri ate or reasonable." Can all possible types of metaphor be drawn on in grammalical ization ? The answer appears to be no. Talmy ( 1 983, 1 988, 2000) has suggested thal only cerlain lypes of spatial concepts are used cross- l i ngu istically i n gram matical i tems: speci f ically, topolog ical concepts. Thus precise distances between poi n ts on a scale, or precise angles, do not gram matical i ze. Indeed , angles i n general (e.g . , corner ill
4. 3 The role of pragmatic inferencing in grammaticalization
87
time) typically do not appear to grammatical ize. 3 However, topological relations on a l i near parameter frequently do so, e.g., flv nt back up-down. Sweetser has suggested that when a lexical item expressing a spatial concept is gram matical ized, only the topological concept is transferred. The concrete image assoc iated with the lexical i tem is replaced by a more schematic one, and the mean ing transfer "is to a fai rly abstract, topological domain . . . so there i s less fleshi ng-out of meani ng" (S weetser 1 988: 393). -
4.3.2
,
Metonymic processes
There is l i ttle doubt that metaphor is one process at work i n grammatical ization. However, si nce reanalysis, not analogy, has for long been recogn i zed as the major process i n grammatical ization at the structural, m orphosyntactic leve l , it would b e surprising i f metaphor, w h i c h is analogical, were the prime process a t work pragmatical l y a n d semantical ly. In t h i s section we show t h a t other processes, wh ich depend on contiguity and ·association in the flow of speech , also play a ma jor part, and that some i nstances of grammaticalization that h ave heretofore been regarded as metaphorical can be seen to arise out of semantic contiguity rather than or as well as out of semantic analogy. The overrid i ng importance that metaphor was given in many d i scussions o f gram matical ization during the 1 980s and early 1 990s seems t o have derived i n part from the tendency t o t h i n k i n terms o f "lexical i tem > grammatical item," i.e., i n terms o f form , relatively i ndependently of context rather than i n terms of "use of lexical item i n discourse > grammatical i tem," i.e., i n terms of form i n utterance contexts. 4 For example, when the lexical i tem go is considered out of context and is said to grammatical ize to an auxi l i ary, metaphor is naturally i nvoked with respect to its spatial properties. B u t i n fact it was not go that grammatical ized; the phrase be going to did, presumably only in very local contexts, e.g., that of be going ill order to V. The contigu ity with to in the purposive sense m u s t have been a major factor i n the development of the future meaning in be going to as an aux i l i ary (we discuss this poi n t more ful ly below). Another reason for the focus on metaphor was presumably also that the term metonymy had unti l that time been thought to be rather i nsigni ficant. For example, Di rven speaks of metaphor as a "maj or associative leap" but of metonymy as a "mi nor process" ( 1 985 : 98). Furthermore, in the tradition deri v i n g from lakobson and Hal le's ( 1 956) classic distinction between metaphor as choice functioning on the parad igmatic axis versus metonymy as association and sequence fu nctioning pri mari ly on the syntagmatic axis, metaphor was thought to l ead to homogeneity and coherence, metonymy to juxtaposition and potential i ncoherence (J . D. Sapir 1 97 7 : 4). The term was used pri mari ly for changes arising out of contigu i ty i n
88
4 Pragmatic factors
the non-linguistic world, cf. such examples as Lat. coxa ' h i p ' > Fr. cuisse ' thigh' (the parts of the body are spatially contiguous i n the physical world), and boor 'farmer' > 'crude person ' (association of behavior w i th a certai n person or class of persons). One of the most famous examples is the transfer by association of the term for 'prayer' (OE gebed) to the objects by which a series of prayers was counted, the beads of a rosary (and ultimately, by general i zation, to any beads). However, contigui ty in the utterance, often resulting in ellipsis, had also bein used as an example of metonymy, cf. French foie ' liver' < Latin iecur fica tum 'liver fig-stuffed. ' Neither of these senses of metonymy i s useful for the study of grammaticalization. However, conceptual assoc i ation also i nvolves metonymy, and in this sense (often known as "conceptual metonymy") the term is valuable. This sense can already be fou nd in Stern, who, in speaking of "perm u tation," says i t results from "a word [being] used i n a phrase where a notion i n some way connected with its meani ng is liable to form an element of the context" ( 1 93 1 : 353). He goes on to list u nder examples of permutation the development of the logical meani ngs of considering, supposing, and of concessive while. More recently Anttila suggested that "[m]etaphor is semantic transfer through a similarity of sense perceptions," and is analog ical and iconic, while metonymy is semantic transfer through contigui ty and "indexical" ( 1 989 [ 1 972] : 1 4 1 -2). In one of the first works i n the last few decades to recognize the importance of metonymy i n grammatical change, Bri n ton ( 1 988) argued that the development of the English aspect m arkers, i ncluding have, is metonymically rather than metaphorically motivated. Recently the fundamental importance of conceptual metonymy in language i n general has been widely recognized. Indeed, there has been a major shift i n think ing and i t i s coming to be i ncreasi ngly recognized as "probably even more basic [than metaphor] to cog n i tion" (Barcelona 2000: 4). As a cogni tive process in which "one conceptual entity . . . provides access to another conceptual entity within the same domain' (K6vecses and Radden 1 998: 38), metonymy points to ("indexes") relations i n contexts that i nclude i n ter�ependent (morpho)syntactic constituents. In an utterance such as ( 1 9) the verb go i nvites the conversational inference that the subject arrived at a l ater time at the destination, and the purpo sive to, i n troducing a subordinate clause, invites the conversational i n ference that someone i n tended the marriage to occur: ( 1 9)
1 was/am going to be married. (in the sense ' I was/am going for the purpose of getting married ' )
However, this i mpl icature c a n b e canceled: (20)
1 was goi ng/on my way to be married, but on the plane 1 changed my mind and decided to joi n the Army.
4. 3 The role of pragmatic inferencing in grammaticalization
89
We hypothesize that the future mean ing of be going to was derived by the se manticization of the dual i n ferences of later time i ndexed by go and purposive to, not from go alone. Indeed, we hypothesize that the i n ference from purposive to must have played a signi ficant role i n the grammatical i zation of be going to given that the major syntactic change i nvol ved in the development of the auxil iary i s the rebracketing of [ [ . . . be goi ng] [to S]] as [ . . . be goi ng to V X] (Section 1 . 1 ). The progressive be -ing i ndexed activity i n process, and so motivated the tendency for be going to to be i nterpreted as a purposive that was relevant to the reference time of the clause and l i kely to be i m m i nent (see Bybee and Pagliuca 1 987 ; Perez 1 990, who differ from the analysis presented here mainly in treating the change as a case of metaphorization). To appreciate the importance of the relationship between to and go, in the developmen t of aux i liary be going to, consider the fol lowing possible early i nstance: (2 1 )
Thys onhappy sowle . . . was goyng to be broughte into hel le for the synne and onleful [unlawful] l u stys of her body. ( 1 482, Monk of Evesham [OED go 47bJ)
Thi s can be understood as an expression of motion i n the context of the belief that after death the soul goes on a journey with the purpose of being rewarded or punished for actions in l i fe. Note that i n this example the passive demotes the i n ference that the subject of go is vol i tional or responsible w i th respect to the purposive clause. Because the destination of the journey (hel l ) is an adj unct not of goyng to but of broughte, the directionali ty of going is also demoted, and the inference of imminent future resulting from the purposes of the j udg�s of the dead is promoted. Similarly, i n the passage i n (22) the answer to whither away is (to) a messenger, and I am going to deliver them seems best u nderstood as answering the question (why) sofast?, i n other words, i t seems more informative if it is i n ferred to answer the question in terms of purposes rather than directions: (22)
DUK E :
S i r Valenti ne, whi ther away so fast?
v A L. : Please i t your grace, there is a messenger That stays to bear my letters to my friends, And I am going to del i ver them. (c. 1 595, Shakespeare, Two Gentlemen of Verona l I I . i . 5 1 )
The ful l semanticization (and grammaticalization) of be going to is evidenced when the foll owing subject andlor the verb is i ncompatible w i th purposi veness, for example, an i nanimate subject or a verb of mental experience such as hear, or like. Once the semanticization of later time/future had occurred, the will future
90
4 Pragmatic factors
could no longer be used with be going to, presumably because i t had become partially redundant, and did not fit the aux i l i ary verb structure i nto which the construction had been absorbed. (Note, however, that the will-future can sti l l occur i n the main verb construction be going to, as in I will be going to visit Aunt Mildred tomorrow.) The metaphor account, whereby a trajectory through space is mapped onto a trajectory i n time, does not give adequate insight into why the progressive and most especially to are i nvolved in the English expression be going to. Th is becomes particularly clear when we compare the cross-linguistic grammatical ization of the verb with the abstract mean i ng G O to future markers, each of which seems to have a slightly different history. Sometimes, for example, there is no overt purposive, i n which case the future inference arises o u t o f the directi onal verb and i ts associ ated aspect alone, as in French. Someti mes, however, G O may be gram matical i zed i n to either a purposive or a temporal marker of immi nence, as occurred to · bang 'go' i n Rama: (23)
a. Ti iskama ni-sung-bang taak-i. baby I-see-SUB go-TNS 'I am going in order to see the baby. ' b. Ti iskama ni-sung-bang. baby I-Iook-at-ASP ' I am goi ng to look at the baby. '
(Craig 1 99 1 : 457)
In each case inferences from the highly local con texts of the verb i n i ts li near position within the clause appear to be the immeqiately motivat i ng factors for change, though the capacity to create metaphors of time from space may wel l provide a cognitive framework that supports the changes. We have suggested that semantici zation of conversational i n ferences played a major role in the development " of be going to. Another example is prov ided by the development of while (see Traugott and Kon ig 1 99 1 , on which the fol lowing , discussion is based). This con nective origi nated i n OE i n an ad verbial phrase translatable as ' at the time that' consisting of the accusative a [xix] winttre wile Stephne was k i ng. 'That lasted those 1 9 w i nters while Stephen was king. ' (Chro n E [Pl u m mer] 1 1 37.36)
I n the process, the precise speci fication of simul tanei ty signaled by the demonstra tive was lost, a l l ow i n g for other, less precise, conversational i n ferences to play a parl. One such i n ference is that the conditions speci fied i n the subordinate clause serve not only as the temporal frame of reference for those in the main clause. but also as the grounds for the s i tuation (the d isasters lasted n i neteen years because Stephen was ki ng). S uch an i n ference to grou nds for the situation is dominant over tempora l i ty in some examples dating from the later fou rteenth century : (26)
ThaI" mycht succed na female, Quh i l l fou ndyn mycht be ony male. 'No female was able to succeed while any male could be found. ' ( 1 375. Barbours B ruce 1 .60 [OED while 2a])
The causal i n ference from while did not become semanticized in Engl ish. However. i n some l anguages this inference to the grounds for the situation has become the main extension of W H I L E . For example, in German the temporal meaning of weil 'during' has become obsolete and the causal has become the main mean ing; i n teresti ngly, causal weil came to be used i n the latter part of the twentieth century with a concessive mean i ng (mai n clause rather than subord i n ate clauses syntax) (GUn ther 1 996). In English a differen t i n ference came to dominate, that of surpri se concern i n g the overlap i n t i m e or t h e relations between event a n d grou n d . T h i s led t o the ad versative, concessive meaning (cf. similar developments for as long as, at the same time as) . Probable i nstances of the semanticization of surprise and hence concessivity appear i n the early seventeen th century, among them : (27)
Whi l l others aime at greatnes boght with blod , Not t o bee great thou stryves. bot t o bee good. 'Whi le others aim at greatness that is bought with blood. you strive to be not great ( 1 6 1 7, S i r W. M u re, Misc. Poems xxi .23 [Ot:V while 2b]) but good . '
Th is cou ld be i n terpreted as a statemen t about simultaneous behaviors. However, there is a strong i n ference, reinforced by the inversion in the second l i ne, that i t is unusual n o t t o b e bloodthirsty. Unambiguous examples o f concessive meanings appear later, typical ly with presen t-tense stative verbs, e.g., While you like peaches. I like nectarines. The overall shift of while is from reference to a relatively concrete state of affairs (a particu lar ti me) to expression of the speaker's assessment of the
92
4 Pragmatic factors
relevance of simultaneity in describi ng events, to assessment of contrast between propositions. In other words, it demonstrates a shift to a relatively abstract and subjective construal of the world in terms of language. A similar development to i ncreased subjectivity is evidenced by be going to ; the motion verb requires that the direction of motion be anchored in the subject as well as in the speaker's viewpoint. 5 But the auxiliary can be anchored i n th� speaker's subjective viewpoint alone, not in that of the subject: (28)
An earthquake is going to destroy that town.
(Langacker 1 990: 23)
An earl ier example is: (29)
It seems as if it were goi ng to rai n .
( 1 890, Cham. I m l . [OED g o V.b] )
Other examples of subjectification i ncl ude the development of epistemic modals, for example the development of must in the sense of 'I conclude that' as in ( l 7b) from must in the sense of 'ought' as in ( 1 7a) (Traugott 1 989 ; Langacker 1 990).
4.4
Metaphor a n d m etonymy as pro b l e m solvi n g
In discussing the principle of exploiting old means for novel functions, and the recruitment of concrete for more abstract terms, Heine, Claudi , and Hiinnemeyer suggest that: grammaticalization can be i nterpreted as the result of a process which has problem-solving as its mai n goal , i ts pri mary function bei ng conceptuali zation by expressi ng one thing i n terms of another. Thi s function is not confined to grammatical i zation, it is the mai n characteristic of metaphor in general. (l99 1 b: 1 50- 1 )
In other words, semantic change i n general, not j ust grammaticalization, can be i nterpreted as problem solving (see also C. Leh mann 1 985). One problem to be solved is that of representing members of one semantic domain i n terms of another, and metaphoric strategies serve th is purpose. The second problem is the search for ways to regulate communication and negotiate speaker-hearer interaction. We have shown that this is a kind of metonymic change, i ndex i ng or pointing to mean i ngs that might otherwise be only covert, but are a natural part of conversational practice. The main direction of both types of problem solving is toward informativeness, but the two types correlate with shifts along different axes. Metaphorical change involves specify i ng one, usually more com plex, thing in terms of another not present i n the context. Metonymic change,
4. 4 Metaphor and metonymy as problem solving ----
93
Sy ntagmatic ax i s
Mechan i s m : reanalysis
Stage I
Stage II
be
going
[to v i s i t B i l l ]
PROG
Vdir
[Purp. clause]
[be going to] v i s i t Bill TNS
Vact
(by syntactic reanalysis/metonymy)
Stage III
[be goi ng to] l i ke B i l l TNS
V
(by analogy/metaphor) Paradigmatic ax is Mechan i s m : analogy Figure
4. 1 Revised schema of the development of auxiliary be going
to
on the other hand, i nvolves speci fying one meani ng in terms of another that is present, even if only covertly, in the context. It is largely correlated with shifts to meani ngs situated i n the subjective bel ief state or atti tude toward the si tua tion, i ncluding the l i nguistic one. While metaphor is correlated primari ly with solving the problem of representation, metonymy and semantici zation of conver sational meani ngs are correlated with solving the problem of expressing speaker attitudes. In summary, metonymic and metaphorical i n ferencing are complementary, not mutually exclusive, processes at the pragmatic level that result from the dual mech anisms of reanalysis l i n ked with the cognitive process of metonymy, and analogy l i n ked with the cognitive process of metaphor. Being a widespread process, broad cross-domain metaphorical analogizi ng is one of the contexts within which gram maticalization operates, but many actual instances of grammatical ization show that conventionalizing of the conceptual metonym ies that arise in the syntag matic flow of speech is the prime motivation for reanalysis in the early stages. We can now refine the model for be going to outl ined in Figure 3.2. to speci fy that syntagmatic reanalysis is accompanied by metonymic strategies, and paradig matic, analogical change by metaphorical ones. The revised model is presented in Figure 4. 1 . The competing motivations of expressivity (which underl ies metonymic and metaphorical inferencing) and routi nization, together with the mechanisms of re analysis and analogy discussed in Chapter 3, will be shown in the next chapter to motivate the u n idirectionali ty typical of grammatical ization .
94
4 Pragmatic factors
4.5
Pragmatic e n rich m ent versus "bleach i n g"
From very early ti mes researchers on issues related to grammatica l i zation have observed that i t i nvol ves l oss of semantic conten t. Th is has been descri bed by the metaphor of "fad i ng" or "bleachi ng" (Gabelentz spoke of "verbleichen" 'to grow pale,' Mei l let of "affaibl issement" ' weaken i n g ' ) . More recently, H�l!le_ and Reh characteri zed grammatical i zation as: "an evolution whereby l i nguistic units lose i n semantic complex i ty, pragmatic signi ficance, syn tactic freedom, and phonetic substance" (Heine and Reh 1 984: 1 5). Readers will have noted that i n this chapter we have, however, spoken o f pragmatic enrichment, strengthen i ng, and so forth . Th i s is because we have been discussing the beg i n n i ngs o f grammat ical ization, that is, the motivations that perm it the process to beg i n , rather than its outcomes. There is no doubt that, over ti me, mean i ngs tend to become weakened duri ng the process of gram matical ization. Nevertheless, all the evidence for early stages is that i n i tially there is a red istribution or shift, not a loss, of meaning. For example, with reference to the development of future go, Sweetser says : "we lose the sense of physical motion (together with all its l i kely background i n ferences). We gai n , however, a new meaning of future pred iction or i n tention together with its l i kely background inferences" (S weetser 1 98 8 : 392) . In speaki ng of the subjectification of be going to, Langacker draws atten tion to the loss of objective locational reference poi nts that movement entails, and suggests that this loss is replaced by realignment to the speaker's temporal perspective ( 1 990: 23). In other words, one meani n g is demoted, another promoted . As gram matical ized forms become i ncreasingly syntacticized or morphologized they unquestionably cease over time to carry sign i ficant semantic or pragmatic mean ing. Th is can most clearly be seen when former lexical items become empty sy ntactic clements, as in the·case of do, or when formerly separate morphemes be come bound and serve pri mari ly as "morphological detritus" a fter repeated fusion (sec Chapter 6). An excel lent example is provided by the developmen t of French ra ' that,' a form which i s the worn-down rel ic of several stages of express ive rei nforcement: (30)
hoc ' that' > (ecce) hoc 'behold that' > eccehoc > ,,0 > ce > ce(lli) 'thut (LUdtke 1 980: 2 1 2) there' > cel li > .. a
The individual meani ngs of hoc, ecce, and La have been lost, as has the form's distal demonstrative fu nction (M. Harris 1 978: Chapter 4). Two general working pri nci ples arise out of our understand i ng of the processes of inferenc ing in gram matical ization. One is that the mean i ngs will always be derivable from the origi nal lexical meaning by either metaphorical or conceptual metonymic inferenci ng. Therefore meaning changes in gram matical ization are not
4. 5 Pragmatic enrichmellf versus "bleaching "
95
arbi trary. Secondly, si nce the i n i tial phase of grammatical ization i nvol ves a s h i ft in mean i ng, but not loss o f mea n i ng, i t i s u n l i kely that any i nstance of grammati cal i zation will i nvol ve a sudden loss of mean i ng. One o f the most wel l-known examples of grammat ical ization i n Engl ish, the development o f "empty" do, has been treated mainly as an example of sy ntac tic change, and might be regarded as a counterexample. There has been considerable speculation abou t i ts origi n s . Causative do, as in (3 1 ), has been thought to be one source (Ellegard 1 953): (3 1 )
pe king sende efter h i m & did h i m gyuen u p 3at abbotrice of B u rc h . T h e king sent for h i m and made h i m give u p the abbey of ( Peter)borough . ' (Peterb. Chron 1 1 32 [MED don 4aJ )
Den i son ( 1 985) and Ste i n ( 1 990a) have suggested that gram malical ization oc cU lTed nol si mply via l oss of causative meani ng, but via a perfective mean i ng, which can arise via pragmatic strengthen i ng particularly in past tense causative con texls . For exam ple, we can i nterpret dide him gyuen up (Jat abbotrice i n (3 1 ) as representing not only that the king's making h i m do something occurred in the past, but also thal the action (give up h i s abbey) was ach ieved. This perfective mean ing demotes the i ssue of who u ndertook the action, and perm i ts the i n ference that the action i s the signi ficant issue. Th i s is particularly l i kely to occu r in constructions wi thout an overt subject of the non- fi n i te clause fol lowi n g do. For example, i n (32) the subject ( ' they ' ) may have caused the women to be mocked , but the absence of an overtly expressed i n termediary party between the mockers and their vic tims demotes the causal i ty and promotes the accomplish ment (perfectivity) of the action: (32)
And so thei dede bothe desei ue ladies and gentilwomen, and bere forthe di ucrse l angages on hem. 'And so they both mocked ladies and gentlewomen, and made various allegati ons agai nst them.' (c. 1 450, Knt. Tour-L, 2.24; ci ted i n Deni son 1 985: 50)
There are some problems w i th the causative analysis, however, among them the fact lhat many instances of do in M iddle Eng l i sh i n the area in which it developed (south west England) were and con tinue to be habitual i n mean i n g (Garrett 1 998). Garrett ci tes several modem examples such as this one from Somerset: (33)
The surpl us milk they did make into cheese and then the cheese did go to the di fferent markets, that's how that did work. (Garrett 1 998: 292, citing l halainen 1 976: 6 1 5 )
Whatever the final analysis turns out to be, the poi n t is that sudden emptyi n g of meaning is not expected . Furthermore, i t is i ncumbent on the researcher to seek a plausible set of inferences that enable changes to occur.
96
4 Pragmatic/actors
Perhaps the most damaging evidence against the automatic association of bleaching and sudden emptying of mean ing with grammaticalization comes from evidence that later constrai nts on structure or meani ng can only be u nderstood i n the light o f earl ier meani ngs. In other words, when a form undergoes grammati calization from a lexical to a grammatical item, some traces of i ts original lexical meanings tend to adhere to it, and details of its lexical history may be reflected in constraints on its grammatical distribution. Thi s phenomenon h as been called "persistence" (Hopper 1 99 1 ). An example is provided by Lord 's discussion of the development i n West African languages of object markers ("accusative cases") out of former seriali zed verbs like ' take.' In ali (a Benue-Kwa language of West Africa), the form kt i s an accusative case marker in sentences such as: (34)
E kc wblb IJ rne-sT. she ACC book lay-down 'She put down a book.'
(Lord 1 993 : 1 1 8)
It was originally a verb meaning ' take' (see Lord 1 993: 53-6) w hich has become grammaticali zed as an accusative case marker, and sentences such as these are historically of the type ' He took a book [and laid [it] down. ' Certain restrictions on the distribution of this case marker can only be understood from the point of view of i ts origin i n the meaning ' take.' Let us for a moment consider a language which has a fully developed accusative case marker, Latin . In Latin , accusative is a general marker of direct objects. The Latin accusative case occurs (to a large extent) irrespective of the semantic relationship between the verb and i ts object. Accusative objects in Latin may be perceived, produced, ordered, imagined, and so on. With a few exceptions, any noun which has the role of an object is marked as an accusative. ali is quite d i tl"erent from L atin in this respect. Consider, for example, (35a) and its ungram matical counterpart (35b): (35)
a. E IJrne w :' I :' . she lay egg 'She laid an egg.' b. ' E kC w :, l :, IJ rne. 'she ACC egg lay
(Lord 1 993: 1 20)
The semantic relationship between verb and object in 'laid down the book' is quite different from that in 'laid an egg.' In the first, the object is changed (moved, grasped, etc.) through the action of the verb; the object is "affected" (Lord 1 982; Hopper I 986a). In the second, the object is produced or brought about by the action of the verb; the object is "effected." The accusative case marker in ali can only be used if the object is "affected ." For th is same reason, the accusative case marker kt is inappropriate if the verb is one of experiencing:
4. 5 Pragmatic enrichment versus "bLeaching " (36)
a. Tete na K :' kil. Tete saw Koko 'Tete saw Koko.' b. 'Tete ke K:, kil nll. 'Tete ACC Koko saw
97
(Lord 1 993: 1 20)
These k i nds of constraints exist because G a retains the restriction o n accusative case mark i ng which derives from the historical antecedent of th is grammatical morpheme in the lexical verb ' to take' : only objects which can be ' taken' are marked morphologically as accusatives. Persistence of old meani ngs is a common phenomenon. Some d i fferences in the meanings of the Present-Day Engl ish tense/modal auxil iaries reflect possibilities of meaning which have existed for over a thousand years. Bybee and Pagliuca show that: "the differences i n the uses of these future markers [i .e., will, shaLL, be going to, PR and ET] can be u nderstood as conti nuations of their origi nal lexical mean i n gs" (Bybee and Pagliuca 1 987: 1 1 7). As mentioned in con nection w i th example (2), there are several polysemies of the "future" will. These i nclude prediction (the "pure" future), willingness, and intention. Bybee and Pagliuca show that two of these meani ngs were already implied in Old English. (37) exempli fies w i l l ingness, (38) intention: (37)
G i f he us geunnan wi le, pret we hine swa godne gretan moton . . i f he us grant will, that we h i m so generous greet shoul d . . 'If he will/is w i l l i ng to grant that we should greet h i m who i s so gracious . (Beowulf 346-7;6 cited in Bybee and Pagl i uca 1 987: 1 1 3) Wen ' ic pret he w ille, gif he wealdon mot, in prem gu�sele thi nk 1 that he w i l l , if he prevai l may, in the war-hall .
.
. .
(38)
Geotena leode etan unforhte. of-Geats men eat unafrai d 'I bel i eve that he will, i f h e should prevail, devour the people of t h e Geats without fear i n thei r war-hal l . (Beowulf 442-4; ci ted i n Bybee and Pag l i uca 1 987: 1 1 3 )
The "predictive" future has developed out of the intention/promise use of will. Bybee and Pagliuca show that the future meaning becomes establ ished in the Middle Engli sh period when inanimates incapable of vol ition begi n to appear as the subjects of will (see also A ijmer 1 985). When this happened, i t did not result i n a n across-the-board re-semanticization o f will; the predictive future remains only one of several distinct m eani ngs of will in Present-Day Engl ish. A l l that happened was that a new meaning was added to an already polysemous form, and thus new distributional possibilities were opened up for the form. Furthermore, the fact that will and be going to do not mean the same thing (be going to expresses present orientation and a goal-directed plan), is attri butable to the fact that the latter h ad progressive and directional origins while the former did not.
98
4 Pragmatic factors
The process of demotion of some lexical meani ngs and promotion o f others i s characteristic of semantic change i n general . Those lexical mean i ngs that are pro moted i n grammatical i zation tend to be relatively abstract, and particul arly re levant to expression of temporal i ty, role relationsh ips, connectivi ty, etc . (i .e., "grammat ical mean i ngs"). Furthermore, they tend to be the o nes most sal ient i n the origi nal contexts/form ulae w i th i n wh ich grammaticalization takes place (e.g., i ntention re garding the future is sal ient in the contexts relevant to the gram matical i zaii"on 'o f be going to). These originally sal ient mean i ngs tend to persist over time and to constrain the later uses of the grammaticali zed form . "B leac h i ng" must therefore be taken to be a very relative notion, and one that pertains al most exclusively to late stages of grammatical izatio n . An important q uestion for future research is what exactly constitutes bleaching, when it can be said to set in, and how i t correlates with morphosyntactic general i zation .
4.6
Conclusio n
We have argued that grammatical i zation can be thought of a s the re sult of the continual negotiation of mean i ng that speakers and hearers engage i n . The poten tial for grammatical i zation l ies i n speakers attem pting t o b e max i mally i n formative, dependi ng on the needs of the si tuatio n . Negotiating mean i ng may i nvol ve i n n ovation , speci fical ly, pragmatic, semantic, and ulti mately gram matical enrichment. It is largely motivated � y metonymic and metaphorical i n ferenc i ng. These are complemen tary processes l i nked with the dual mechan isms discussed i n Chapter 3 : reanalysis (linked with metonymy), a n d analogy ( l i n ked with metaphor) . As in novations come to be adopted by members o f the com m u n i ty, they may be subject to max i m i zation of s ignal simplicity, and ultimately to vari ous types o f re duction, typically semantic bleach i ng, morphological bond i ng, and phonological attri tion . It is to the u n idirectional changes i n structuration that arise from these diachronic processes that we now turn .
5
Th e hyp ot h e s i s of u n i d i rectio n a l ity
5. 1
I ntrod u ctio n
Gram matica l i zation a s viewed from the diachronic perspective i s hy pothesi zed to be prototypically a u n i d i rectional phenomenon. I n this chapter we elaborate further on some general pri nciples of unidirectionali ty, with particular altention to such diachronic issues as generalization , decategori a l i zat ion, i ncrease in grammatical status, and renewal . We w i l l also discuss synchronic issues such as the resultant variabi l i ty and "layering" arising from those d i ac h ronic processes. The hypothesis o f u n i d i rectionality i s a strong one, and has been the subject of vigorous debate si nce the 1 990s ; i n Section 5 .7 we summarize this debate, and concl ude that the coun terexamples to unidirectional i ty that h ave been adduced so far are sporad ic, whereas the evidence for u nidirectional i ty is systematic and cross- l i ngui stical ly replicated . In Chapter 6 we w i l l discuss in more deta i l some wel l-known ki nds of u n i d i rectional i ty found in morphological c hange, that is, i n the later stages o f gram matical i zation. I n Chapter 7 w e w i l l suggest that s i m i lar ty pes of unidirectional i ty also occur i n morphosyn tactic change, espec ially the development of complex clauses . Once gram matical i zation has set i n , there are certain l i kely paths along which i t proceeds. One path d iscussed by Mei l let is that whereby a lex ical item becomes a gram matical i tem, summari zed as : lexical i tem > morphology
As mentioned in Section 2.2, one of Meil let's examples was the Modern Greek future particle tha, as i n : (I)
Tha telefoneso tou patera mou. FUT telephone DEF: ACC father: ACC my: ACC ' [ I ] w i l l telephone my father. '
Meil let said that the source of fha is the Classical Greek the/o hina ' I wish that . ' In the preced i ng chapters we have discussed examples that suggest this form ulation of the path of grammalica l i zation is not quite righ t . The path is not directly from
1 00
5 The hypothesis of unidirectionality
lexical item to morphology. Rather, lexical items or phrases come through use i n certain highly constrained local contexts t o be reanalyzed a s havi ng syntactic and morphological functions. Schematical ly, th is can be characteri zed as: lexical i tem used i n specific l i ngui stic contexts > syntax > morphology
The lexical items that become grammaticalized must first be semantically general and serve commonly needed discourse functions. They then become syntacticiilly fixed (they become constructions), and may eventually amalgamate morphologi cally, say, as stem and affix. The basic assumption i s that there is a relationshi p between two stages A a n d B, such that A occurs before B , b u t not vice versa. Thi s is what is meant b y unidirectionality. Before proceeding, it should be mentioned that the u n i directionali ty i n question is not the same as what E. Sapir called "drift," although it has some similarities. In a famous statement, he sai d : "Language moves down time in a current of its own making. It has a drift" ( 1 92 1 : I SO). Even if we were to emend thi s statement to acknowledge that i t is not language that changes, but rather language users internal ize different rules of grammar over time (see Section 3.2), there would still be a fundamental difference from what is meant by the u n idirectionali ty of grammatical ization . Sapir was interested i n the fact that English was losing case inflections on its pronouns (e.g., the who-whom disti nction was losing ground), and that Engl ish was also becom i ng more periphrastic, for example, the possessive gen itive was being replaced by of I Wh ile Sapir was thinking of language-specific changes, and, within languages, of highly specific phenomena, other l i ngu ists later showed how the separate phen o mena he discussed for English were in fact part of the same thing (case loss and periphrasis go hand in hand), and i ndeed part of larger typological sh ifts. R. Lakoff, for example, focused on drift "defined . . . very loosely as historical fluctuation between syntheticity and analytici ty" ( 1 972: 1 79), that is, on fluctuation between bonded ness and periphrasis, while Vennemann ( 1 975) focused on shifts from OV to VO order. Lakoff's summary of Sapir's notion of drift as "a metacondition on the way in which the grammar of a language as a whole will change" (R. Lakoff 1 972: 1 7 8) serves well to differentiate "drift" from unidirectionality. Drift has to do with regularization of construction types with i n a language (see also Malkiel 1 98 1 ), unidirectionali ty w i th changes affecting partic ular types of construction. Unidirectionality is a metacondition on how particular grammatical constructions will change.
5.2
General ization
Among characterizations of grammaticalization, the fol lowi ng statement is typical : "It is often observed that grammatical mea n i ng develops out of lexical
5. 2 Generalization
10I
mean ing by a process of generalization or weakening of semantic content [Giv6n 1 97 3 ; Fleischman 1 982; and many others]. It can be further hypothesized that . . . this semantic change i s paralleled over a long period of time by phonetic ero sion" (Bybee and Pag l i uca 1 985: 59-60). As we showed in S ection 4.5, early stages of grammaticali zation do not show bleaching. Rather there is a balance between loss of older, typical ly more concrete, meanings, and development of newer, more abstract ones that at a m i n i mum cancel out the loss. Many are the result of pragmatic strengthen ing, and i ncrease i n informativeness with respect to grammatical function. We will not repeat these arguments here. I nstead, we will focus on the notion of general i zation. General ization i s a process which can be characterized, in part, as an i ncrease i n the polysemies of a form , and i n part as : "an increase of the range of a morpheme advancing from a lexical to a grammat ical or from a less grammatical to a more grammatical status" (Kurytowicz 1 976 [ 1 965] : 69). 5.2. 1
Generalization of meaning
We start with i ssues of generalization of mean ing. Here the question is not whether the meanings become less distinct i n the process of grammaticalization (as the hypothesi s of bleaching suggests), but whether there are constraints on what meani ngs are subject to grammatical ization , and on how the mean i ngs of lexical i tems that become grammatical may change. When we thi n k of the lexicon , we assume that it includes not o n l y syntactic and phonological characterizations, but also characterization of such semantic relations as take part i n fields (e.g., color terms, or verbs of saying: say, tell, claim, assert), relational terms (e.g., kinship terms), taxonomies (hierarchies such as creature, animal, dog, spaniel, i ncluding part-whole hierarchies, such as finger-hand-arm body, keel-boat), complementaries (non-gradable pairs, w i th excluded middle, e.g., true-false), antonyms (gradable pairs, e.g., slow-fast), d i recti onal opposi tions (e.g., go-come, teach-learn), synonyms (e.g. , fiddle-violin), polysem ies (e.g., mug [of tea, usually w i th a handle] and mug [of beer, often without a h andleD, and so forth . General accounts of lexical semantics can be found i n Ullmann ( 1 962), Lyons ( 1 977), Cruse ( 1 986), Levin ( 1 993) and the reader is referred to them for details. As we have noted i n previous chapters, the lexical meani ngs subject to gram mat icalization are usually quite general . For example, verbs which grammatical ize, whether to case markers or to complementizers, tend to be superordinate terms (also known as "hyperonyms") in lexical fields, for example, say, move, go. They are typically not selected from more specialized terms such as whisper; chortle, assert, squirm, writhe. Likewise, if a nominal from a taxonom ic field grammat ical izes i n to a numeral classifier, it is l i kely to be selected from the following
1 02
5 The hypothesis of unidirectionality
taxonomic levels: begin ner (e.g. , creature, plant), life form (e.g., mammal, bush), and generic (e.g., dog, rose), but not from specific (e.g., spaniel, hybrid tea), or varietal (e.g., Cocker, Peace) (Adams and Conklin 1 973). In other words, the lex ical items that grammaticalize are typical ly what are known as "basic words." In some cases, a formerly fairly specific term can be gram matical ized, but only after it has become more general . An example is prov ided by Lati n ambulare ' walk' > French aUer 'go' > future auxiliary. As these already general lexicaf items take on grammatical fu nctions, they are generalized i n so far as they come to be used i n more and more contexts, that is, they gai n w ider distribution and more polysemies. Th is follows natural ly from the fact that former i n ferences are semanticized. To the extent that there is a difference between lexical and grammatical mean ing changes, grammatical meaning changes are a subset of lexical ones. Most notably, grammatical ization does not provide evidence of narrowing of mean ing. By contrast, although many lexical changes involve broadening (generalization) of meani ngs, there are also well-known cases of narrowing, typically of a higher to a lower taxonomic level. Examples of narrowing i n lex ical change include the restriction of hound to a special type of dog, in an avoidance of synonymy (the generic OE hund became narrowed to specific rank when Scand i navian dog was borrowed). Occasionally, narrowing may i nvol ve the restriction to a su btype, in which case the i n tegrity of the lexical i tem and i ts components may become blurred. Examples i nclude raspberry, strawberry, where rasp and straw have vir tual ly lost their origi nal i ndependent mean i ng, but together with berry identify different varietals of berries. One lexical domain i n which narrowi ng is partic ularly l i kely to occur is the domain of terms for dispreferred enti ties, espec ially those associated with taboo, social prej udice, or unpleasantness, however these are defined in a particular culture. Examples i nclude stink (origi nally ' to smel l/have an odor' ), cock (now restricted in some registers to 'pen is,' with the term for the bird replaced by rooster), mistress (originally 'head of household,' now restricted to female lover, or ' kept woman ' ) (for a fuller study, see Al lan and Burridge 1 99 1 ). One constraint on lexical change that is often noted in the l i terature is "avoidance of homonymic clash," in other words, avoidance of what might be dysfunc tional ambiguity from the perspective of "one mean ing one form" (see especially Geeraerts 1 986). Well-known examples are the loss and replacement of one lex ical item by another when two or more formerly distinct lex ical items undergo regular phonological changes that make them potentially homonymous. Proba bly the most famous case is that of the replacement of the term for ' rooster' in southwestern France by forms such as faisan 'pheasant' when Lati n gallus 'rooster' and cattus 'cat' fell together as gat due to regular sound change
5. 2 Generalization
1 03
( G i l l ieron 1 902- 1 0). I n England OE l(Etan 'prevent' and lettan ' perm it' merged in ME as let. The first was replaced by forb id and prevent, and the potential problems o f ambiguity between opposite mean i ngs were avoided (Antti la 1 989 [ 1 972] : 1 82). The majority o f examples of avoidance of homonymic clash are lexical, and even in the lexical domai n they are i n frequent. In PDE the noun sanction is used in two mean ings: ' perm ission ' and ' penalty ' , and the verb rent is used in the senses 'rent out to' and ' rent from ' , with little sign of one member of the pair being ousted by the other. In acronyms such potential ambiguities are even more l i kely to be tolerated, cf. PC ' personal computer' , ' pol ice constable' , ' pol itically cor rect.' Syntactic as well as other contextual clues tend to allow for d isambiguation. Avoidance of homonymic clash is even more i n frequent i n the grammatical do mai n . If it does occu r, it does so almost exclusively in connection w i th rel atively independen t morphemes. For example, it has been suggested that the idiosyn cratic devel opment of the Old Span ish prepos itions faza 'toward' « Latin facie ad 'with one's face to' ) and (f)ata 'until' « Arabic hatta ' u n ti l ' ) i nto M iddle Span ish (h)acia and fasta respectively may be attributed to "the powerful urge to di fferentiate homonyms despite semantic similari ty" (Malkiel 1 979: I ). Similarly, the borrow ing in Middle Engl ish of the Northern Engl ish femi n i n e pronoun form she and of the Scandi nav ian pl ural pronoun they have been attri buted i n part to the merger in ME of OE he ' masc.sg.pro.,' heo ' fem .sg.pro . , ' hie 'pl.pro. ' (For a detailed study, i ncluding possible evidence for regulation of a new homonymic clash that developed between thai 'they' and the less frequent thai ' th ough , ' see Samuels 1 972.) However, as we have seen, grammatical items are characteristi cally polysemous, and so avoidance of homonymic clash would not be expected to have any systematic effect on the development of grammatical markers, espe cially in their later stages. This is particularly true of inflecti ons. We need only think of the English -s i n fl ections: nomi nal plural, third-person-si ngular verbal markel} or the -d i n tl ections: past ten se, past participle. Indeed, it is difficult to pred ict what gram matical properties will or will not be di stingu ished in any one language. Although English contrasts he. she. it, Chi nese does not. Although OE contrasted past singular and past plural forms of the verb (e.g., he rad ' h e rode, ' hie ridofl 'they rode' ), PDE does not except in the verb be, where we fi n d she was/they were. In sum, as gram matical ization progresses, mean i ngs expand their range through the development of various polysem ies. Depending on one's analysis, these poly semies may be regarded as quite fine-grained. It is only col lectively that they may seem like weakening of mean ing. The i mportant claim should not be that bleaching follows from generalization, but rather that mean i ng changes leading to narrow ing of mean ing will typically not occur in grammatical i zation.
1 04
5 The hypothesis of ttllidirectionality
5.2.2
Generalization of grammatical function
It follows from the preceding discussion that, i n so far as grammati cal forms have meanings, they will come to serve a larger and larger range of meaningful morphosyntactic purposes. Bybee and Pagli uca ( 1 985) refer to the development of progressives into i mperfects. A clear example in English is the spread of an originally highly constrai ned progressive structure be V-ing, that was restricted to agentive constructions, first to passives (the house was being built is a later eighteenth-century construction, replacing the earlier the house was building), and later to stative contexts, where i t serves a "contingency" function, as i n There are statues standing in the park. Another example of the general ization of grammatical fu nction is the develop ment in Finnish of the geni tive case morpheme to signal the u nderlying subject of a non-finite clause, for example, a clause introduced by a verb of cognition such as uskoa 'think, believe' (Timberlake 1 977: 1 44-57). What is at i ssue here is reanalysis of the genitive i n one inflectional context, and spread of the new use to an i ncreasi ng number of new contexts. It can perhaps best be explained through an English example. In a sentence such as (2) there is a certa i n ambiguity at the surface level in the function of the word Jane: (2)
We watched Jane dancing/dance.
From one point of view, Jane is the person we are watchi ng, and is therefore the object of watch. From another perspective, the event (Jane's dancing) is the object of watch. From yet another perspec � ive, Jane is the subject of the verb dance. For example, we say : (3)
a. What we watched was Jane dancing/dance. b. Who we watched was Jane danci ng/dance.
However, in a language that has a system of cases that overtly distingui sh be tween subjects and objects, it is possible to resolve the potential surface ambiguity in different ways, and there may be a diachronic shift from the one to the other. This is essentially the kind of change that occurred i n Fi n n ish. In Finnish, i n both older and modern periods, there is no single case for objects ; instead, objects are i ndicated in several different ways (Comrie 1 98 1 : 1 25-36) : (a) (b) (c )
with the accusative case if there is an overt subject, and the matrix verb is active with the nominative if there is no overt subject with the partitive i f the verb is negated, or i f the object is partially affected.
Subjects i n non-finite clauses in Modern Fin n ish are indicated with the genitive case:
5. 2 Generalization (4)
1 05
Nlien poikien menevlin. I-see boy:GEN:PL go: PART 'I see the boys goi ng. ' (lit. 'I see the goi ng of the boys. ' ) (Antti la 1 989[ 1 972] : \ 04)
The gen itive case for subjects of non-finite clauses i n the m odern language replaces an earlier case-marki ng system in which other cases were used. The following examples are from older Finnish texts: (5)
a.
Accusative
Seurakunnan hen lupasi psysyueisen oleuan. congregation:ACC he promised 10ng-lasti ng:ACC bei ng:ACC 'He promised that the congregation would be l ong-l asting. ' b.
Nominative
Homaitan se tauara ia Jumalan Lah ia poiseleua. observed i t goods: NOM and God :GEN gift NOM bei ng-lacking:NOM 'It is observed that the goods and the gift of God are lacking.' c.
Partitive
Eike lwle site syndi oleuan. Not think this:PARTIT sin being:PART 'Nor does one think this to be a sin. ' 3
(Ti mberlake 1 977: 1 45 )
These three cases comprise the set of cases that signal objects, and suggest that i n earlier Finnish the whole event was construed a s the object. Therefore the change to constructions such as are i l l ustrated by (4) has been one of reanalysis whereby a noun that was once construed as the object of a main-clause verb c omes to be construed as the subject of a subordinate-clause verb. The participle i n the non finite clause has also changed : i n the earlier type of sentence i t is i n flected and agrees with the object case (partitive, nominative, or accusative) of the noun. But in the later type of sentence the participle is not inflected - i t remains invariant. While i t has not acquired any new verbal properties such as tense, i t has shed i ts noun- l i ke properties of i nflection (that is, it has become decategorial i zed ; see next section). How did this come about? Ti mberlake ( 1 977), fol lowing Anttil a ( 1 972: 1 03a), suggests that the reanalysis originated i n very local contexts: specifically those con structions i n which the case morpheme was ambiguous. The origi nal accusative • m and the geni tive • - n of singular nouns became homophonous as a result of a phonological change i n which word-fi nal nasals merged as n. As a consequence, those constructions with singular nouns (e.g., ' boy ' ) serving as objects w i th par ticipial attributes could be reanalyzed as partitive attributes of a verbal noun (that is, as surface subjects of non- fi n i te verbs). 4 It appears that a reanalysis that occurred i n one very local construction (with sin gular agentive nouns) was then general ized through a number of new environments
1 06
5 The hypothesis of unidirectionality
via the following historical stages : first-singular NPs and pronouns; then, in add i tion, pl ural pronouns and plural agentive NPs ; and fi nally plural non-agentive NPs. Timberlake suggests that among other thi ngs these stages reflect a spread along a functional h ierarchy from noun phrases that are more subject-like to less subject l ike noun phrases. The h ighest on the hierarchy are entities that are agen tive, and i ndividuated (a singular noun or pronoun), and therefore are prime candidates for bei ng subjects. Next come those that, i f not singular, are most likely to i n i ti a ie actions, that is, personal pronouns and nouns denoting an agent. Finally come non-singular nouns that are less likely to be subjects or to i n i tiate actions, such as inani mates. The syntactic reanalysis i s therefore driven by a hierarchy of semantic contexts rather than by strictly syntactic structure. It would be remarkable to fi nd a h ierarchy of this kind reversed. We are unlikely to find, for example, that subjects of verbs (whether fi n i te or non-finite) could origi nally comprise a broad class of all nouns and pronouns, but that the class of possible subjects of verbs progressively narrowed to exclude, first, inanimate pl ural nouns, then plural pronouns and animate nouns, and so finally became restricted to singular pronouns and nouns. Nor would we expect to find subject case marking beginning with abstract, inanimate, and plural nouns and only later spread ing to animate, anaphoric pronouns.
5.3
Decategorial ization
Another perspective on u n idirectionality presents it as a spread of gram maticalization along a path or cline of structural properties, from a morphological ly "heavier" unit to one that is lighter, that is, from one that tends to be phonological ly longer and more distinct (e.g., with stressed vowels) to one that tends to be less distinct and shorter. It is from this structural perspective that we approach unidi rectional ity i n this section, with particular attention to the issue of the correlation between i ncreased grammatical status and decategorialization. Important factors in our discussion will be the tendency for relatively prototypical mem bers of Nou n, Verb, and Adjective categories to become less prototypical i n their di stri bution, in at least one of their uses. Another will be frequency : the more frequently a form occurs in texts, the more grammatical it is assumed to be. Frequency demonstrates a kind of generalization in use patterns. In the standard view of grammatical categories, l i nguistic forms are classi fied in advance according to criteria that may vary quite widely from language to language. If morphological criteria are available, these usual ly play a role i n the assignment of a form to a category. In the Indo-European languages, for example, "nouns" are typically identified through properties such as case, number, and gender, and
5. 3 Decateg orializatioll
1 07
"verbs" through properties such as tense, aspect, and person/n u m ber agreement. If morphology is lacking, as is l argely the case i n Chi nese, categories are usually identified through functions such as (for "nouns") abi l i ty to be a topic (e.g. , abi l i ty to be referential , u navai labi l i ty for questioning), or (for "verbs") abi l i ty to serve as certain kinds of pred icates . When a form u ndergoes grammaticalization from a lexical to a gram matical form, however, it tends to lose the morphological and syn tactic properties that wou ld identify it as a full member of a m aj or grammatical category such as noun or verb. In i ts most extreme form such a change i s man i fested as a cline of categoriality, statable as: major category ( > intermedi ate category) > mi nor category
In this schema the major categories are noun and verb (categories that are rel atively "open" lexically), and m i nor categories include prepo s ition, conjunction, auxiliary verb, pronoun , and demonstrative (relatively "closed" categories). Ad jectives and ad verbs comprise an intermediate degree between the maj or and m i nor categories and can often be shown to derive straightforwardl y from (participial) verbs and (locative, man ner, etc.) nouns respectively. At least two major categories - noun and verb - are identi fiable i n almost all languages with some consistency (see Hopper and Thompson 1 984, 1 985 ; Croft 1 99 1 , however, argues for three maj or categories : nou n , verb and adjective), whereas the m i nor categories vary from language to language, bei ng man i fested often only as affixes. G iven the hypothesis of unidirectional ity, it can be hypothesi zed that diachronically all minor categories have their origins in major categories. A clear case of shift from major to m i nor category is seen in the conj unction while, as in while we were sleeping. As we saw in Section 4 . 3 .2, h istorical ly, while was a noun (OE hwil) mean ing a length of time; this mean ing is still preserved i n PDE (we stayed there for a while). As a conj u nction, however, while has d iverged from th is origi nal lex ical fu nction as a noun, and is grammaticalized as a signal of temporal organ ization in the discourse. Among the changes invol ved in the grammatical ization of while to a conj u nction i s a loss of those gra mmatical features that identify while as a noun. When it is used as a conjunction, while: (a) (b) (c) (d) (e)
cannot take articles or quanti fiers cannot be modified by adjectives or demonstratives cannot serve as a subject or as any other argument of the verb can only appear in the i n i tial position in its clause, and cannot subsequently be referred to by an anaphoric pronoun.
It will be noted that these categorical changes are here presented as negative qualities or losses. This structural characterization contrasts with the pragmatic one in Chapter 4, which focuses on the fact that while has "gai n ed" an ability to
1 08
5 The hypothesis of unidirectionality
link clauses and indicate temporal relationships i n discourse i n a way that was not possible for i t as an ordinary nou n . In ascribing "decategorial ization" to a form, we are not tracing the decay or deterioration of that form, but its functional shift from one kind of role to another in the organization of discourse. Because this new role is one that does not require overt expression of the l i nguistic properties associated with discourse reference, such as articles and adjectives, speakers cease to signal these expressions overtly, but such discontinuance of use should not be identified with simple loss, as if somehow a conjunction were a "degenerate" noun. Similarly, as they become grammaticalized, verbs may lose such verb-l i ke at tributes as the ability to show variation in tense, aspect, modality, and person number marking. In the following pair of sentences, the i n i tial participial "verb" can still show some verb-l i ke features when it is understood literally as in (6), but loses this ability when it i s understood as a conjunction (see Kortmann and Konig 1 992), as in (7): (6)
(7)
Carefully considering/Having carefully considered all the evidence, the panel deli vered i ts verdict. Considering (' having carefully considered) you are so short, your ski l l at basket bal l is u nexpected.
In (6), the participle considering can take an adverbial modifier, can have a present or past tense form, and must have an u nderstood subject that is identical with the main clause subject; it therefore in a sense has a (recoverable) subject, l i ke a verb. In (7), none of these verb-like attributes are available to considering. Two typical paths of development have been much d iscussed i n the l iterature. One is a path for nominal categories, another for verbal . These are "grammatical clines," i n the sense that they make reference to hierarchical categories relevant to constituent structure. They are also clines of decategorialization, in that the starting point for the cl i ne is a full category (noun or verb) and the intermediate points are characterized by a loss of morphological structures associated with the full category. Some caveats should be noted in any discussion of unidirectionali ty along a cline. Firstly, as mentioned i n Section 1 .2.2, clines should not be thought of as continua strictly speaki ng. Rather, they are metaphorical generalizations about l i kely functional shifts, "paths" along which certain grammatical properties cluster around constructions with "fami l y resemblances" (Heine 1 992) (e.g., constructions resembling auxiliaries, or articles, or preposi tions). These cluster points should not be thought of as rigid "resting spots." A metaphor for l i nguistic forms i n these clusters might be chips in a magnetic fi eld ; over time fewer or m ore of the chips in the clusters may be pulled magnetically to another field. Secondly, because there is always a peri od of overlap between older and newer forms and/or functions
5. 3 Decategorialization
1 09
of a morpheme, the cline should not be thought of as a line in which everything is in sequence. As we i ndicated i n C hapter I , Heine and h i s col leagues use the term "chai n i ng" to emphasize the non-l inearity of relations on a cl i ne. We prefer the term "layering" (see Section 5.5 below) because that metaphor al lows more readily for multiple origins of a grammatical form . B u t here, as in other matters, the metaphors are only partly helpful. A further caveat i s that because the particu lar course of events in any cline that is presented is not predeterm i ned, once an i tem has been reanalyzed, con tinued grammaticalization is not inevitable, but may be suspended i ndefinitely at any point. Indeed, it i s typical ly suspended at the pre-affixal stage in situations of language death (Dressler 1 988). Furthermore, we cannot logical l y work back wards from some given poi n t to a unique antecedent on the same c l i ne. Absent a historical record, we cannot, for example, un iquely conclude fro m a cline on which prepositions occur (see Section 5 . 3 . 1 ) that any given preposition must once have been a certain noun, although we can state that i t might have been. This is because other sources for prepositions, such as verbs, are possible. For example, the preposition during was once the -ing form of an obsolete verb mea n i ng 'to last, endure . ' I t i s both difficult a n d unnecessary t o i l l ustrate the whole of any one c l i ne w i th a single form. It is difficult because historical records are rarely long enough to permi t the recovery of the entire sequence of events, and so usual ly we must ei ther posit reconstructed forms for past stages, or else i ndulge i n spec u l ation about future stages. And it i s u nnecessary because what is at issue i s the d i rectional i ty between adjacent forms on the cline, not the demonstration of the c o mplete se quence of events for a given form . Furthermore, at any one stage of a language, the h istorical u n idirectionality may be obscured by synchronic evidence of renewal of old forms (see Section 5 .4.3 below). And, very importantly, different languages tend to exemplify different clusterings on a cline. In other words, not every position on a cline is l i kely to be equally elaborated at any particular stage. For exam ple, English a n d Romance languages have fairly elaborated clitic structures, and m i n i mally elaborated i nflectional structures, while some other languages, such as Slavic languages, have highly elaborated inflectional structures . Last, but not least, evidence for clines i s historical. Ultimately it is the h istorical record alone that gives us evidence for a cline. Nonetheless, certai n facts can poi n t to a cl i ne even when no direct documentation exists. Typical pathways of change identified through cross-l i nguistic diachronic study can be seen in the synchronic system, for example when we find the same verb used both as a full contentive and as an auxiliary, or a definite article that is clearly cognate with a demonstrative. Differ ences i n the speech community, such as those of dialect, register, speech tempo, etc., may also reveal forms at different poi nts along the same cli ne.
1 10
5 The hypothesis of unidirectionality
5.3. 1
A noun-to-affix cline
We w i l l first consider a cline whose starting poi nt is a fu l l noun, spec i l' ically a relational noun (to be defi ned below). The c l i ne has been presented as follows (c. Lehmann 1 985 : 304): relational nOlln > secondary adposition > pri mary adposition > aggluti native case affi x > fusional case affix These fi ve poi nts should not be taken as strictly discrete categories, but as mark i ng, somewhat arbitrari ly, cl uster points on a continuous trajectory. In other words, most forms that are locatable on th i s c l i ne w i l l not fi t unambiguously i n to one or the other of the named categories, but w i l l be seen as mov i ng toward or away from one of them i n a direction that we can cal l "from top to bottom ," following the wri ting conventions adopted above. A relational noun is one whose meaning is a location or d i rection potent ially i n relation to some other noun. Top, way, and side, and many body parts such as foot. head, and back often assume a relational mea n i ng, and i n doi ng so may enter this cl i ne (Hei ne, Claudi , and Hiinnemeyer 1 99 I b). The relational noun usually appears as the head noun of a phrase, such as side i n by the side of ( > beside), , or as an inflected noun, such as German Wegen ' ways [dative pl ural] > wegen ' because of, ' as in wegen des Wetters 'because of the weather. ' The term "adposition" is a cover term for preposi tions and postposi tions. Sec ondary adpositions are usual ly forms (words or short phrases) that define concrete rather than grammatical relationships. They are typical ly derived from relational nouns, e.g., beside the sofa. ahead of the column. Pri mary adpositions arc thought of as the restricted set of adpos itions, often monosy l l abic, that i nd icate purely grammatical relationsh i ps, such as of, by, and to. However, pri mary adpositions may themsel ves be characterized by a c l i ne i n mean i ng i n so far as some may have a relatively concrete spatial mea n i ng, for example, by in a hotel by the railway statiol/, wh i l e others do not, for example, by in arrested by a plain clothes police man. While the disti nction between concrete and grammatical mean ing is often not easy to defi ne, the spatial meani n gs of pri mary adpos itions are al ways very general. The spatial meani ngs are moreover l i kely to be recovered by some k i nd of rei n forcement, e.g., by the railway station > down by the railway station ; in fhe house > within/iI/side the house. Primary ad positions are eas i l y c 1 i ticized, and may go one step fu rther to be come affi xes. Locative suffi xes of various ki nds can often be traced back to earl ier
5. 3 Decategorialization
III
postposi tions (and, s t i l l further back, to nouns). In Hu ngarian the s u ffi x -ball , as in luizball ' house-i ness ive/i n the house, ' was once the locative case of a relational noun mean i n g ' i n teri or. ' S i m i l arly, the el ative, mean i ng ' away fro m , ' as in hazbol ' from the house, ' shows a suffi x -bol that goes back to a d i fferen t c ase of the same word . The fi nal 1111 segments o f the two suffi xes are themsel ves rel ics o f the two case end i ngs on the rel ational noun (Comrie 1 98 1 : I 1 9). The H u n garian suffi xes -ban and -bol are examples of aggl u t i n ative suffi xes: they are joi ned to the stem w i th a m i n i m u m of phonological adjustments, and the bou ndary between stem and sutTix i s q u i te obvious. By con trast, fus i onal a ffi xes show a bl urri ng not only of the stem/affi x bou ndary, but also of the boundaries among the affi xes themsel ves. In Lat i n militibus ' to/from the soldiers,' -ibm; is a dati ve/ablative plural suffi x which cannot be further analyzed, and i n miles 'soldier:NOM : S G ' the -t- of the stem * milet has been lost through assi m i lation to the nomi native si ngular suffix -so 5.3.2
A verb-ta-affix cline
A parallel c l i ne has a lexical verb as i ts starting poi n t which develops i n to an aux i l i ary and even tual ly an affix (verbal cli nes have been the subj ect of several cross-l i nguistic stud ies, most notably B y bee 1 985 ; Bybee and Dah l 1 989). There are a number of poi n ts on this c l i ne w hich can be characteri zed as fol l ows : fu l l verb > au x i l i ary > verbal c 1 i tic > verbal a ffi x On this c l i ne, w e typically fi nd that verbs hav i ng a fu l l lexical m e a n i n g a n d a grammatical status as the only verb i n their clause come to be used as a u x i l iaries to another verb. A u x i l i ary verbs typical l y have semantic properties o f tense, aspect or mood , and show special i zed syn tactic behavior (e.g., in POE, a u x i l iary will can not occur in certain tem poral and i n fi n itival clauses; hence the fol l owi n g are ungram matical : * Let 's wait till she will join us, */ would like her to will joill us). There are nu merous exam ples of the s h i ft from mai n to aux i l i ary verb. From POE we have c i ted go in be going to. Other examples include have, which is a fu l l verb in have a book, but a partial, or "quasi-auxi l i ary" in have a book to read and have to read a book, and a fu l l auxi l i ary i n have had a book (here a u x i l iary have precedes main verb have i n i ts past-participial form). A nother exa m ple is keep, wh ich is a fu l l lexical verb i n she keeps indoors on cold days, but a n auxi l i ary i n she keeps watering the tomatoes. Aux i l i aries may h istorically become c 1 i tics, l i ke Engl ish have i n we 've built a new garage. And such c 1 i tics may become affixes. As
1 12
5 The hypothesis of unidirectionality
discussed in Section 3.3. 1 , this happened in the French future-tense paradigm, as in its parleront ' they will speak, ' where -ont reflects a former cliticized auxil iary ' have. ' An additional position intermediate between main verb and auxiliary verb has been proposed by Hook ( 1 974, 1 99 1 ). Presenting data in which a clause may contain a complex of two verbs known as a "compound verb," Hook has argued that in Hindi and other Indo-Aryan l anguages there is a class of "light verb S" which, following S lavic termi nology, he cal ls "vector verbs." One of these verbs, the "mai n" or "primary" verb, carries the main semantic verbal meaning of the clause, and is non-finite. The other, the "vector" verb, i s a quasi-auxiliary which is finite, and therefore carries markers of tense, aspect, and mood . Semantically, i t adds nuances o f aspect, direction , and benefaction t o t h e clause. In modern Indo Aryan languages vector verbs are homophonous with main verbs meaning 'go, give, take, throw, strike, let go, get up, come, sit, fall , ' etc . , and are derived from them (Hook 1 99 1 ). Since Hindi is a verb-final language, the order of the two verbs i n the "compound" construction is main-vector: (8)
kar J iyaa fon aap ko das baje miiT ne AGT 1 0 o'clock you OAT phone make VECTOR/brought ' \ telephoned you at 10 o'clock.' diye. paise de mal ne use AGT h i m : OAT money give VECTOR/gave (based on Hook 1 974: 1 66-7) '\ gave h i m the money. '
\
(9)
The vector verbs are in the past tense, and, as is i ndicated by the glosses, are homophonous with past tenses of verbs meani ng 'bring' (lenaa) and 'give' (denaa). The main verb in (8) is kar ' make,' and i n (9) de 'give. ' I n (9), then, 'give' appears as both the main verb (de) and the vector verb (diye). The semantic force of the vector verb is hard to specify, but in general it expresses perfectivity. Both of the sentences could be phrased with the main verb alone, as i n : ( 1 0)
(I I)
kiyaa. fon aap ko das baje miii ne AGT 1 0 o' clock you OAT phone make '( telephoned you at 1 0 o'clock.' paise d iye. use mal ne AGT h i m : OAT money give '( gave h i m (the) money. '
However, ( 1 0) leaves open the question of whether the call was successfully put through, while (8) wou ld definitely suggest that the call was completed. (9) impl ies that all the money was given, while agai n ( I I ) leaves this open . In other words, the compound verb has all of the semantic complexities of perfective aspect, such as
5. 3 Decategorialization
1 13
Table 5 . 1 Approximate proportion of compound verbs in Indo-A ryan languages Language
Proportion
Shina (Gilgit) Kashm i ri M arathi G ujarati Bengali M arwari H i ndi-Urdu
o 3 6 7 8 9
Source: based o n Hook ( 1 99 1 : 65)
emphasis on completion, ful l affectedness of the verb's object, and i nvol vement of an agent. It should also be noted that there are certain types of construction where it is mandatory. Hook argues that we have here a m ovement toward grammaticalization of a set of verbs which are becom ing specialized as vector verbs. In his view, vector verbs therefore represent an i n termediate stage between full verb and aux i l i ary. From this perspective it is interesting to trace the trajectory of the change to vector status by looki ng both at earlier texts and at other Indo-Aryan languages closely related to Hindi i n which the change has not proceeded so far. This latter strategy is a highly convenient one because the languages are fully accessible and texts in the various cognate languages are available which are thematically similar or identical . Consider first the relative textual frequency of s imple versus compound verbs in Hindi and some of the related languages (Hook 1 99 1 : 65). Table 5 . 1 shows the approximate proportions of compound verbs in texts among various languages of the group. That is to say, in comparable texts there are about nine ti mes as many compou nd verbs in Hindi-Urdu as i n Kashmiri, and twice as many in G ujarati as in Marathi . Textual frequency is often considered prima facie evidence of degree of grammatical ization (see, e.g., Heine, Claud i , and Hiinnemeyer 1 99 1 a; Bybee, Perkins, and Pagliuca 1 994; also Section 5 .6 below). Textual frequency i s accompanied by differences i n the ki nds of main verbs which may be accompanied by one of the vector verbs. In Marathi , which repre sents a less advanced stage from the point of view of the gram matical ization of vector verbs, there is a preference for them to be used only when the main verb is inherently unspeci fied for completedness ; in other words, they add aspectual infor mation. In Hindi-Urdu , where vector verbs are more frequent, they have spread to environments in which they are redundant, that is, to i nherently completive verbs, includi ng communicatio n verbs, which, in context, tend to be completive (as, for
I 14
5
The hypothesis oj unidirectionality
Table 5 .2 Ratio oj compound verbs in Marathi and Hindi-Urdu according to semantic class oJ main verb Semantic class of main verb
Marathi
H i ndi-Urd u
Displacement or disposal Creation/change of state Change of psychic state Sensation or perception Mental action Communication
1 0% 8% 8% 4% 4% 2%
44% 30% 8% 8% 1 0% 20%
Source: based on Hook ( 1 99 1 : 68-9)
example, she said) (Hook 1 99 1 : 69-70). Table 5 . 2 shows the di fference between Marath i and Hind i-Urdu with regard to the ratio of compound verbs to the total verb forms for certain classes of main verbs. This suggests that, as grammatica l i za tion proceeds, the semantic range of the emergent grammatical morpheme expands or generalizes. The difference between Hindi and Marath i is a statistical one, not a categorical one. Chal lenging Hook, B utt (Forthcom i ng) argues that si nce l ight verbs are highly stable and are h istorically a dead end, they are not intermed iate and should not be i ncluded i n the verb-to-affi x cl ine; they arise out of reanaly sis of main verbs, but not out of gram matical i zation, si nce they do not i nvolve phonological loss, or any clear trajectory toward au x i l iaries. However, even i f they do not belong on the c l i ne, they do suggest gram matical ization - phonological attri tion is not a prerequisite, as the developmen t of au x i l i ary must and might in English demonstrate, and passage through a complete c l i ne is never necessary or expected for grammaticaJ ization. In this case, the decategori al i zation of the mai n verb and the frequency patterns suggest strongly that gram maticalization is i nvolved .
5.3.3
Multiple paths
So far our examples i n thi s chapter have arguably been of changes along a si ngle c l i ne. Not all cases of grammatical ization are of this kind, however. Some show development along two or possibly more d i fferent c l i nes. Craig has given the name "polygrammatica l i zation" to such multiple devel opments, where a sin gle form develops d i fferent grammatical fu nctions i n different constructions. Her example is from Rama. As all uded to in Section 4 . 3 . 2 in connection with exam ple (23), " bang 'go' in Rama developed i n to: (i) a temporal marker in the verbal
5. 4 Some processes participating in unidirectionality
I 15
dom a i n ; ( i i ) a purpos i ve adposition i n the nom i nal dom a i n , and then a conj u nc tion i n the complex sentence domai n (conj u nctions are analyzed as adposi tions to clauses, see Section 7.3). G i van ( 1 99 1 b) shows that relative-cl ause morphology, spec i fical ly B i bl ical Hebrew 'asher (probably derived from 'athar 'place' ), spread both i nto ad verbial clause domains such as causatives, and also i n to complemen tizer domains. Lord ( 1 976, 1 993) shows that 'say ' complementi zers genera l i ze in d i fferent languages to causal clauses i n languages such as Yoruba and Telugu, and to conditionals i n Ga. Development along such m u l t i ple paths i n to d i fferent gram matical domains con forms to u n i d i rectionali ty i n that the later forms are more grammatical (abstract, reduced, genera l i zed) than the earl ier ones . Not all m U l tiple paths show spl it, however. Just as i n phonology we fi nd spl i t and merger or convergence, so i n grammatical i zation we fi nd that somet i m es forms from several slightly d i fferent domai n s may converge on one gram matical do mai n, provided that there is pragmatic, semantic, and syntactic appropriateness. The phenomenon o f convergence from various subpaths of gram matical ization is often descri bed i n terms of the metaphor of convergence in "semantic space." One example is provided by Kem mer ( l 993b), who charts the d o m a i n of reflex ives and m i ddle voice. In her characterization of these domai n s , the reflex i ve construction ex presses si tuations where the i n itiator and endpoi n t o f the event refer to the same enti ty, but are conceived as conceptually different, as i n hit one self, see ollese lf. M iddle- voice constructions such as wash (oneself), dress, get angry, think are s i m i lar in that they express s i tuations where i n i tiator and end poi nt i n the event are the same enti ty, but they are ditl'erent i n that the conceptual di fference is less than that in reflexi ve s i tuations. Kem mer ( 1 992) shows that the general i zation of re�lexives i n to middles is very common cross- l i nguistica l l y (see a l s o Faltz 1 988), b u t other sources are evidenced too, such a s passive a n d rec iprocal (the 'each other' construction). Other semantic maps w i th m u l tiple sub paths have been suggested for evidentials (L. Anderson 1 986) and conditionals (Traugott 1 985b).
5.4
Some p rocesses p a rticipating i n u n i d i recti o n a l ity
Several processes typical of grammaticalization contri bute to semantic and/or structural genera l i zation and decategorial i zation . They may, however, at first glance appear to complex ify the process, and to raise questions about it. We w i l l discuss three typical processes : specialization, whereby t h e choice o f grammatical forms becomes reduced as certai n ones become general i zed i n mea n i ng and use; divergence, whereby a less gram matical form may spl i t i n to two, one variant mainta i n i ng i ts former characteristics, the other becom i ng more gra m m atical ; and
1 16
5 The hypothesis of unidirectionality
thirdly, renewal, whereby old forms are renewed as more expressive ways are found of saying the same thing. 5.4. 1
Specialization
In consideri ng textual frequency and semantic generalization of the sort discussed in connection w i th Indo-Aryan compound verbs above, we might imag i ne that this textual frequency and semantic generali zation could i n theory proceed with exactly the same set of vector verbs at each stage. However, as the semantic range of i ndividual vector verbs becomes greater and more general, the chances of overlap and ambigu i ty on the fringes are bou nd to i ncrease, and some of the vector verbs w i l l become redundant and fal l out of use. Consequently, while the text fre quency of some of the vector verbs i ncreases, the number of different vector verbs decreases. This exempl i fies special i zation, the process of reducing the variety of formal choices available as the mean i ngs assume greater grammatical general i ty (Breal 1 99 1 [ 1 882] : 1 43 ; Hopper 1 99 1 : 22). Hook ( 1 99 1 : 75) notes that: (a) In both the H i ndi-Urdu and the Marathi text samples, the most frequently occurring vector verb was the one mean i ng G O . But in H i ndi-Urdu G O accounted for 44% of all vector verbs in the sample, while in Marathi i t accounted for only 32% of all vector verbs. (b) The five most frequent vector verbs in the Hindi-Urdu text sample accou nted for 92% of the total number of vector verbs ; while in Marathi the fi ve most frequent vector verbs accounted for only 82% of the total number. (c) In the H i ndi-Urdu text sample, only 1 0 different verbs were used as vector verbs; in the Marathi sample, 14 d ifferent verbs were used as vector verbs. These statistics suggest that i n H i n d i a handfu l of verbs i s gain i ng the ascen dancy i n the competition for aux i l i ary status. Here aga i n we see a major difference between lexical and grammatical i tems. In any domai n of meaning the number of lexical i tems will vastly exceed the n u m ber of grammatical m orphemes. More over, lexical i tems form an open class, which can be added to i n defi n i tely, while the i nventory of grammatical m orphemes i s added to only very sparingly, by i tems originating i n the lexical class. If we compare, for example, the number of tense and aspect disti n ctions which are expressed grammatical ly in a given language w i th the number of ways of modifying actions and events available through lexical adverbs, we can see i mmediately that the process of grammaticalization i s a selective one i n which only a few lexical forms end up a s grammatical morphemes. However, old forms may continue to coexist (see especially Section 5 .5 on "layering" below); therefore specialization does not necessari ly en tail the elimination of alternatives, but may be mani fested simply as textual preferences, condi t ioned by semantic types, socioli nguistic contexts, d i scourse genres, and other factors.
5. 4 Some pIVcesses participating in unidirectionality
1 17
Another good example of special i zation is the Modern French negative con struction, which in the written language consists of a negative particle ne before the verb and a supportive particle, usually pas, after i t : ( 1 2)
II ne boil pas de vin. he N EG drinks N EG PARTIT wine 'He doesn't dri nk wi ne. '
As indicated i n Section 3.6, at earlier s tages of French, predicate negation was accomplished by ne alone placed before the verb. This ne was i tself a proclitic form of Latin non, Old French non. Already i n Old French, a variety of adverbially used nouns suggesting a least quantity (Gamillscheg 1 95 7 : 753) could be placed after the verb in order to reinforce the weakened negation. These reinforcing forms included; among others: pas ' s tep, pace' point 'dot, point' mie 'crumb' gate ' drop' amende 'almond' areste ' fish-bone' beloce 'sloe' eschalope 'pea-pod ' They seem origi nally to have functioned to focus attention on the negation i tself, rather than on the verb being negated; w i thout the rei nforcer, the focus of attention would fal l on the verb (Gamillscheg 1 95 7 : 755). By the sixteenth century, the only ones still used w i th negative force were pas, point, mie, and goutte, all ofthem more general terms than those which were no longer used . Even i n the s i x teenth century, pas and point predominated, and by the modern period these were the only two which were still in use. Of the two remaining, there is a clear sense i n which pas is the only "unmarked" complement to ne i n negation. It is by far the more frequent in discourse, i t participates in more constructions than point, and is semantical ly more neutral, point being an emphat ic negator. Point today denotes only emphatic negation contradicting a previous assertion (though there is some possibility that this semantic disti nction between pas and point was originally an artifact of French grammarians). In other words, point cannot be relatively negative, perhaps because of the operation of persistence (see Section 4.5) a 'point' is not relative. Therefore in a sense pas is the only form which has become fully grammaticalized out of a n array of forms which could reinforce negation in O l d French . It has also become a negative morpheme i n its own right in a number of contexts (pas moi 'not me, ' pas plus tard qu 'hier ' not later than yesterday, ' etc.), and i n the spoken language -
1 18
5 The hypothesis of unidirectionality
the ne of ordinary verbal negation is usually dropped (je sais pas 'I don ' t know ' ), leavi ng pas as the only mark of negation. This thi n n i ng out of the field of candidates for grammatical ization as negators is accompanied, as usual, by a shift of mean i ng, in this case from the lexical mean ing 'step, pace' to the grammatical meani n g of negation. There is i n this i nstance no phonological change peculiar to the gram matical i zed form, and no fus ion w i th neighbori ng words. The original noun pas lives on i n its earl ier mean i ng of 'step, pace, ' and i t remains completely homophonous with the negative particle. Before leav i ng the example of French negators, i t is worthw h i l e to cons ider i ts i mpl ications for the discourse motivations of gram matical i zation . The origins of the use of concrete nouns as reinforcers of negation cannot be documented, bu t i t is reasonable to surmise that they were once l i nked to specific verbs. Presumably mie 'crumb' was once collocated with verbs of eati ng, i .e., ' he hasn ' t eaten a cru m b,' or perhaps - i n a m i l ieu where food was scarce and bread a common means of payment for services rendered - givi ng, and so on: ' they didn ' t pay/give me a crumb.' S i m ilarly with goutte 'drop ' : 'he hasn ' t drunk a drop. ' With pas 'step,' the verb must have been a verb of motion: 'he hasn 't gone a step . ' We may com pare the vernacular English use of 'drop' and ' spot' in ( 1 3 ), where the con tex t of drop has s i m ilarly been expanded i n a way that suggests i ncipient grammatica l i zation: ( 1 3)
He didn ' t get a drop (spot) of applause.
Bit ( i .e. , a small bite, cf. German bij3chen) is of course normal in such con texts for all dialects. There is nothing strictly conceptual (semantic) about nouns such as ' peapod , ' 'crumb,' 'step,' a n d the others which woul d pred ict that they would become nega tors. They are not all i n trinsical l y " m i n imal quantities" of thi ngs, but they assu me that mean ing when combi ned i n discourse with negators. However much in ret rospect we see semantic commonal i ties in the ways i n which forms evol ve, it is i m portant to keep i n m i nd that u l ti mately their roots and motivations are i n real speech and real collocations, and that the study of how forms are distri bu ted in discourse is i ndispensable i n understandi ng grammatical i zation. 5.4.2
Divergence
When a lexical form undergoes grammaticalization to a c 1 i tic or affix, the original lexical form may remain as an autonomous element and undergo the same changes as ordinary lexical i tems (Hopper 1 99 1 : 22). Th is characteristic of "divergence" is a natural outcome of the process of grammatical ization, wh ich be gins as a fixing of a lexical form in a spec i fi c potentially grammatical environment, where the form takes on a new meaning (the same phenomenon is cal led "spl i t" in Heine and Reh 1 984 : 57-9). S i nce the context of i ncipient grammatical i zation is
5. 4 Some pIVcesses participating in Llllidirectionality
I 19
only one of the many contexts i n which the lexical form may appear, when the form undergoes gram matica l i zation, it behaves just l i ke any other autonomous form i n i ts other, lex ical , contexts, and is subject to semantic and phonological changes and perhaps even to becom ing obsolete. Consider, for example, the Engl ish i ndefi n i te article alan. In DE this word was ClIl. Its vowel was long, the same as the vowel in the word for ' s tone,' stall. It meant ' one, a certa i n , ' and was not used i n the general non-spec i fic sense that we m ight use i t i n today, as i n I caught a fish, but was chiefly used to "present" new i tems, as in There was once a prince o/ Tuscany. The normal phonetic development of this word i n PDE wou l d have been [ow n ] , rhy m i ng with 'stone. ' Wh i le i n Scottish Engl ish the two words continue t o have the same vowel ([ey n J . [steyn]), in most other d ialects a phonological development pecul iar to this word occurred yield i ng the PDE ful l form [wA n ] . The c 1 i ticized form of this same word became de-stressed, and formed a si ngle accentual u n i t w i th the following noun or consti tuen t of the noun phrase (adjective, etc.), resulting i n i t s POE form and d istri bution: t h e vowel [a ] , and retention of the [ n ] when followed by a vowel. The divergen t histories of the stressed and u nstressed forms can be seen in alternations such as the fol lowing: ( 1 4)
Wou ld you l i ke a Mai Tai ? - Yes, I 'd love aile.
We turn now to a more detai led example of divergence, from Malay. Nou n s i n certai n discourse contexts i n Malay must b e preceded b y a classi fier (Hopper I 986b). Class i fi ers occur in many languages in association with number words; they are comparable to the word ' head' i n ' ten head of cattle' (see Schachter 1 985 : 39-40). The fol lowi ng examples are from a Malay narrative text known as the Hikayat Abdullah (CL stands for (nomi nal) classifier): ( 1 5)
( 1 6)
( 1 7)
Ada-Iah kami l i hat tiga orallg budak-budak kena hukum . happen w e see three CL boy-PL get punishment (Hopper I 986b: 64) ' We happened to see three CL boys being punished. ' Maka pada suatu pagi kel i hatan-Iah sa-buah kapal rendah and on one morn i ng was: seen-PA RTICLE a:CL ship low (Hopper 1 986b: 77) 'Then one morning a CL low ship was sighted . ' Mati-Iah tiga ekor tikus. dead-PARTICLE three CL rat 'Three CL rats were killed . ' (Hopper I 986b: 1 44)
The italicized words orang, buah, and ekor are classi fiers. In Malay they indicate that the noun which is classi fied is new and relatively important to the discourse. They are not i n terchangeable: orang is used before human nouns, buah before objects of a bulky size, and ekor before nouns which denote animals of any kind. There is i n add ition a more general classi ficatory word suatu (also found as satu), used before si ngular objects (in the sense of ' t h i ngs' ) and competing w i th buah:
1 20
5 The hypothesis of unidirectionality
( 1 8)
M aka di-beri-nya hadiah akan Sultan itu stlattl kereta bogi . and he-gave as-gi ft to Sultan the a:CL carriage buggy 'And he gave a CL buggy carriage to the Sultan as a gift. ' (Hopper 1 986b: 1 66)
The classifiers themsel ves are preceded, as in these examples, by a number word such as tiga ' three, ' or the singular c l i tic sa- 'one, a.' However, suatu is only singular and is not preceded by sa- or any other number word or quantifier; the reason for this, as we shall see, is that the s- of suatu is i tself historicall y the same singular morpheme sa- that is fou nd w i th the other classi fiers when the classified noun is singular. In Malay, as in other classi fier languages, most of the classifiers double as autonomous nouns. Buah means 'fruit,' orang 'person , man,' and ekor ' tai l . ' There is thus a divergence between a lexical meaning and a grammaticalized meaning. On occasions when the two come together, a sort of haplology (contraction of adjacent identical material) occurs; while the word for a Malay person is orang Malayu, one instance of orang stands duty for both the classifier orang and the head noun orang. Consequently, ' fi ve Malay men ' is: ( 1 9)
l i ma orang Me\ayu fi ve CUmen M alay
rather than (20)
' l i ma orang orang Melayu
Similarly, with buah: a pomegranate is buah delima, but if the expression is clas si fied, in place of sa-buah buah delima ' a pomegranate,' sa-buah delima is used. 5 The constraint against orang, buah, etc. occurring with homophonous lexical items shows that grammaticalization has not proceeded so far that these classifiers and their cognate lexical noun are sensed as being formally unrelated. With suatu the situation is quite different. While this form competes with sa buah as the classi fier for bulky inanimate objects, it is often used with abstract nouns in contexts where sa-buah would not be appropriate, e.g., suatu khabar 'a piece of news, ' suatu akhtiar 'an idea. ' The form suatu is in the modern language somewhat archaic and l iterary ; it is general ly pronounced and written satu, and has, significantly, become something l i ke a strong indefi nite article. It is also the numeral 'one' when counting 'one, two, three, etc . , ' and in this sense often corresponds in the texts to Engl ish 'one' in 'one day, one morning,' etc. The older form with u (sautu) suggests a reconstruction 'sa watu 'one stone, ' with a noun ' watu 'stone' having cognates i n Javanese watu 'stone' and Malay batu 'stone. ' The phonological change from i n itial w to ini tial b before a has several
5. 4 Some processes participatin.g in unidirectionality
121
parallels i n Malay. However, batu does not serve as a classi fier i n modern Malay, although it would not be at all strange i f it did. There i s a classifier biji whose corresponding lexical nou n means ' seed , ' and which is used for small ish round objects; l arger objects, as we have seen, are classified with buah. Presumably · wattt as a classifier once covered a s i m i l ar range, classify i ng three-dim ensional objects of an i ndetermi nate size. The older form · watu 'stone' cont i nues as a frozen classifier embedded in an i ndefi n i te-article- l i ke quanti fier mea n i ng roughly 'a, one. ' It is distributed i n the texts much l i ke the complex sa- + classifier, referring to new and prom i nent th i ngs i n the discourse, but i t occurs preferentially with nouns which do not belong i n an "obvious" category for one o f the established classi fiers. Interestingly, in the texts it is often used with abstract nouns, many of them of Arabic origin, and with nouns denoti ng concrete objects which are not part of tradi tional Malay cul ture. The evolution represented by · watu 'stone' > • sa watu 'a (classi fier for smallish objects)' > ·sa watu 'a (classifier for every kind of object)' > suatu 'oneta (with abstract or non-traditional objects)' > satu ' onela (with any nou n ) ' is a paradigm case of grammaticalization. It exempl i fies persistence, that is, the grammatical ized construction is constrai ned by its origins: a real classifier is not also used if a noun is quantified with s(u)atu, e.g., ·suatu buah rumah 'one house' i s excluded : only suatu rumah or sa-buah rumah are permi tted. This constraint against • suatu buah rumah can be explai ned by the fact that suatu itself h istorical ly already contains a classifier. It is also an example of divergence. A form assumes two d istinct functions (lexical noun and classi fier). One of these functions (that of lexical noun) i s found in an environment where i t is exposed to a phonological process ( i n i tial w- > b before a) from wh ich the other function is insulated (when protected from i n i tial posi tion by the proclitic sa-, w- did not undergo the change). The result is that the two forms satu and batu are no longer fel t as cognate by speakers o f the language. For example, satu batu 'alone stone ( one mile)' is unobj ectionable, whereas, as we have seen, sa-orang orang 'alone person , ' or sa-buah buah ' alone fru it' are avoided . It should b e added that phonological (al lomorphic) spl i t of t h e k i n d w e have described for Malay batulsuatu is not necessary for this kind of d i vergence to occur, nor is it a required outcome of the process . In many i nstances the autonomous lexical i tem and i ts grammaticalized counterpart may cooccur quite happily in the same construction ; e.g., the Engl ish auxil i ary verb do frequently occurs with do as main verb (do do it!; they do do that). As mentioned in Section 3 .2.4, there has in the past been a tendency to thi n k of change in terms of "A u n i formly > B . " Given such an approach , divergence might seem to be an u n l i kely characteristic. However, as we have noted , change must always be seen in terms of variation, and the formula for change should therefore =
1 22
5 The hypothesis of unidirectionality
be A > AlB > B . Even so, i t still needs to be stated that it is by no means i nevitable that A will disappear. A and B may instead each go their own ways and con t i n ue to coexist as divergen t reflexes of a historical l y s i ngle form over many centuries, even m i llennia. An example is the development and persistence from M iddle English on of constructions of the type The more he complains, the angrier he gets. Th is construction originated i n a comparative i ntroduced by py, the i nstru men tal form of the demonstrative. The new form has coexi s ted with the demonstrative that from which it derives, and with the article the wh ich it resembles in form , but does not have the syntax of either. The formula shou ld ideally therefore be fu rther modi fied to A > BIA ( > B). 5.4.3
Renewal
In divergence existing forms take on new mean i ngs in certain con texts, while reta i n i ng old mean i ngs i n other contexts. We turn now to a process whereby existing mean i ngs may take on new forms: renewal . Renewal results pri mari ly i n al ternate ways o f say ing approximately the same thi ng, o r al ternate ways o f orga nizing l i nguistic material . Often, but not al ways, these new ways are periphrastic, i .e., phrasal . If all grammaticali zation leads to decategorial i zation and u l ti mately to m i n i mal, compacted forms, how is i t that language users can ensure that languages continue to serve their purposes of organ izing cogn i tion and ach ieving com mun ication '? This question is i n part answered by the hypothesis of com peting motivations of i ncreasing i n formativeness versus routin ization. But does this mean that u n i d i rec tional i ty is a chimera? The answer is that new structures keep being gram matical i zed through the process that Mei llet termed "renouvel lement" or "renewal," and that i nstances of renewal consistently show evidence of u n i d i rectional i ty once the renewal has set i n . A v i v i d example o f renewal is t h e recent history o f Engl ish i nten s i fiers ( words such as very in very dangerous). At d i fferent ti mes in the last two centuries the fol lowing among others have been fashionable: awfully, frightfully, fealfully, terribly, incredibly, really, pretty, truly (cf. very, which is cognate w i th French vrai ' true' ) (Stoffel 1 90 I ): Even i n the written language, very often al ternates with such words as most, surprisingly, extremely, highly, extraordinarily. Over ti me, however, we can expect the choices to be reduced, owi n g to special ization. Intensifiers are especially subject to renewal , presumably because of their markedly emotional function. They are unusual i n undergoi ng renewal espec ially frequently. But certain other categories, although not as shortl ived as i n tensi fiers, are also renewed with some degree of predictab i l i ty. Negative constructions are one example. In spoken English expressions such as no way (cr. No way we 're taking this stufJ) are replacing si mple n 't, from not, i tself a contraction of Ila willt
5. 4 Some processes participating ill unidirectiollality
1 23
'no thi ng. ' Sch weg ler ( 1 988) wri tes of a "psychol i ngu istic proc l i v i ty" for the de velopment of negative emphasizers, and shows how they have their starting poi n t in contexts o f contradiction, i n other words, i n emotional ly loaded contex ts. The example of negat i ve renewal shows that someti mes old forms ( i n t h i s case 11-) may be i nvolved i n the new stru c ture (but not i n exactly the same way as before). An other example is prov ided by the rei n forcement in Surselvan (Rhaeto-Romance) of reflex ive se by the form sesez (Kem mer 1 992). In most Romance languages the reflexive se serves both reflexive and m iddle functions. However, in Sursel van the reflexive has been rei n forced by an emphatic version of itself (a pattern that goes back to Lati n), while the original se now serves only middle functions. The renewal of one form by another may or m ay not occur i n the same constituent pos i tio n . Engl ish i n tensi fiers such as awfully and Surselvan sesez are s imply substi tutes, i nvol v i n g no new syntactic or phonological strategy. But sometimes renewal may i nvol ve a more strategic overhaul. The spoken Engl ish negator 110 way has l i ttle in common syntactical ly with the l1 't w i th which it competes. The French negative rei n forcer pas, which is assu m i n g the role of general negator, occurs af ter rather than before the verb, reflecting a change that could be represented over several centuries as : ne va 'does n ' t go' > ne va pas > va pas
S i m i larl y in English the origi nal negator lI e preceded the verb, as in (2 1 ): (2 1 )
Ne canst �u huntian butan m i d nettu m? not know you hunt-INF except with nets ' Do you not know how to hunt w i th anythi ng but netsT
(c. 1 000, !Elfr Coi l . 62) Bei ng subject to reduction through rapid speech , it could even combine w i th some verbs, e g. n e wres ' not was' > llres, ne wolde ' n ot wanted ' > 1101de. B u t the new, phonological ly fu l l er, 110t that replaced i t fol l owed the verb, as i n (22 ) : .
(22)
,
. . . that moves not him: though that be sick, i t dies not. (c. 1 600, Shakespeare, Henry IV Part 2 . I I . i L I 1 3)
Such d i lTerences i n syn tax between older forms and their replacements or renewals are often subject to word-order changes that are ongoing in the language, or may even contribute to them . Renewal by a non-cognate i tem to effect semantic expressiveness probably un derl ies most examples of the development of i n n ovative periphrasis i n the process of word-order changes. Th is appears to be true of the development of periphrastic markers of modal i ty, such as will, shall, and must, which convey m ore preci se d i fferences of mean ing than the older subjunctive i n flection, and the developmen t of phrasal case markers such a s to a n d of, which also tend t o convey more d i f ferences than the earl ier i n flectional cases. Langacker has called periphrasis "the
1 24
5 The hypothesis of unidirectionaLity
major mechanism for achieving perceptual optimal ity in syntax" ( 1 977: 1 05). One way of defining periphrasis is to characterize it as ful fi l l ing the following criteria (Dietrich 1 973): the meaning of the periphrasis is not deducible from the constituent elements; the periphrastic construction shows syn tactic unity at some level of analysis, where i t did not do so before; the new periphrasis competes paradigmat ically with other morphologically relevant categories. Once renewal occurs, the new form may itself be subject to grammaticali zation and reduction, through rapid speech and routinization, as i n the case of not > n 't. This is one factor that makes grammaticalization a continuously occurring phenomenon. The question is when this renewal is understood to occur. When the same structure is renewed, some speak of "recursive cycles" of grammatical i zation. Some think of the cycle as starti ng with reduction of a form , i n extreme cases to zero, fol lowed by replacement with a more expressive form (e.g. , Heine and Reh 1 984 : 1 7 ; Lightfoot 1 99 1 : 1 7 1 ). This kind of model is extremely problematic, because i t suggests that a stage of language can exist when it is difficult or even impossible to express some concept. Rather than replace a lost or almost l ost distinction, newly i n novated forms compete with older ones because they are felt to be more expressive than what was avai lable before. Th is competition allows, even encourages, the recession or l oss of older forms. Textual evidence provides strong support for thi s view of coexisting competing forms and constructions, rather than a cycle of loss and renewal . The periphrastic future form existed i n Late Lati n long before the eventual loss of future -b- and its replacement by -r-. In contemporary French and other Romance languages, the inflectional -r-fu ture is itself in competition with a more "expressive" periphrastic construction with aLLer, cf. j 'irai 'I will go,' and je vais aLLer 'I will/plan to go. ' Furthermore, when the syntactic structures of the older and newer forms differ, they may be used side by side in the same utterance (cf. French ne va pas, and Middle Engl ish ne might not). When the syntactic structure is the same, but the lexical i tems are different, alternate usages coexist, as in the case of very and awfuLLy.
5.5
A syn ch ronic result of u n i d i rectional ity: l ayeri ng
As we have seen i n the context of discussion of persistence and diver gence, old forms may persist for a l ong period of time. The persistence of older forms and meanings alongside newer forms and meani ngs, whether derived by divergence from the same source or by renewal from different sources, leads to an effect that can be cal led "layering" or "variabi l i ty" at any one synchronic moment in time. We turn now to some comments about this characteristic of grammatical ization.
5. 5 A synchronic result of unidirectionality: layering
1 25
Within a broad functional domain , new layers are continually emerging; i n the process the older layers are not necessarily d iscarded, but may remain to coex ist with and i nteract with new l ayers (Hopper 1 99 1 : 22). Layering is the synchronic resul t of successive grammaticalization of forms which contribute to the same domain . In a n y single language there is always considerable synchronic d iversi ty with i n one domain. Some of the most obvi ous cases are those where a ful l a n d a reduced form coexist, w i th rel ated forms and only m i n i mally differen t functions. An ex ample is the coexistence in Classical Armen ian of three demonstratives: ays 'close to first person,' ayd 'close to second person ,' ayn 'close to third person, ' and three articles -s, -d, -n (Greenberg 1 985 : 277). In such cases it is a reasonable hypoth esis that the reduced form is the l ater form. In other cases a variety of ditl"eren t forms a n d constructions may coexist that serve s i m i lar (though n o t identical) func tional purposes. A small fragment of the PDE repertoire of tense-aspect-modal indicators suggests the potential range i nvol ved : (23)
a. b. c. d. e. f. g. h.
Vowel changes in the verb stem: (Weak) alveolar suffi x : M odal auxil i aries: Have V-en : Be V-ing: Keep on V-ing: Keep V-ing: Be going to V:
take, took look/looked will take/shall take has taken is taking kept 011 eating kept eating is going to take
(There are, of course, many more.) In cases l i ke this i t is a reasonable hypothesis that the most bonded forms have the longest histories i n their present grammatical functions, and that the least bonded are the most recent. Yet another example given comes from Eston ian. Relative clauses in Eston ian may be formed in two ways, one being a construction with a relative pronoun and a finite verb (24a) and the other w i th a participial verb and n o pronoun (24b): (24)
a. Vanake silmitse-s kaua ini mes-t kes old-man observe-PAST: 3SG for-a-Iong-time person-PARTIT REL sammu-s iile Que elu maja poole. go-PAST: 3SG across courtyard:GEN residential bu ilding : I LL ' For a long time the old man observed the person who was going across the courtyard to the residential building.' b. Vanake sil m i tses kaua iile Que old-man observe.PAST-3SG for-a-Iong-time across courtyard:GEN elumaja poole sammu-vat ini mes-t. residential building:ILL go-PRES:PART person-PARTIT (Comrie 1 98 1 : 1 23-4)
1 26
5 The hypothesis of unidirectiollality
The second type i n (24b), the relative clause of which is constructed around a present partici ple, i .e., a non- fi n i te verb, is l i terally someth i ng l i ke 'The old man watched for a long time the across the courtyard of the residential bu i l d i ng going person . ' Such clauses are characteristic of a learned or archaic sty Ie (Comrie 1 98 1 : 1 34); the more usual way of form i ng relative cl auses i n Eston ian is with a relative pronoun and a fi n i te verb, as i n (24a). Typically, grammatica l i zation does not result in the ti l l i ng of any obvious fu nc tional gap. On the contrary, the forms that have been gram matical ized compete with ex isting constructions so s i m i lar in fu nction that any explanation i nvol v i ng "fi l l i ng a gap" seems out of the question - there is no obvious gap to be ti l led . We saw that in Ewe, verbs of say i ng evol ved i nto new complemen tizers at the same time as older complementizers - themsel ves gram matical i zed verbs o f say ing were sti l l available. Latin peri phrastic futures of the kind cantare habet 'he has to sing > he w i l l sing' coexisted at one stage with morphological fu tures o f the type callfabit 'he w i l l sing,' and eventually replaced them. Duri ng any phase of coexistence there are some contexts i n which the two (or more) types in question i nvolve a clear pragmatic d i fference. There are other contex ts in which the choice between them is less clear with respect to pragmatic difference. Frequently we fi nd that one of the competing forms predominates (special ization), and eventual ly extends i ts range of mean i ngs to incl ude those of the construction which it replaces. In this way, h istorical l y continuous speech communi ties may, through repeated renewals, retain categories (such as the fu ture tense) for a considerable length of time w h i l e other speech com m u n i ties have never developed them . Quite often the newer layers of fu ncti onal ly s i m i lar constructions are symp tomatic of more global adjustments. As i ndicated at the end of Section 3 .4 . 1 , Estonian, which (like the other members of the Bal to-Fi nnic branch of Uralic) is a language historically of the OV type, has become thorough ly permeated with va features from i ts Germanic and S l avic neighbors. The two d i fferent ways of forming rel ative clauses exempli fied in (24) are part of this change in type. The older type, in which a participial clause precedes the head nou n , is characteristic of OV languages. The newer type, with a finite verb and a rel ative pronoun, is characteri stic of va languages.
5.6
Frequency
As we have seen in Section 5 . 3 .2, statist ical evidence for the freq uency of forms is a valuable tool in providing empirical evidence for unidirectional ity. Textual frequency has long been recogn i zed i n formally as a concom i tant of
5. 6 Frequency
1 27
gram matica l i zation, and i t has recently assu med an i m portant place i n the empirical study of how lexical forms move i n to gram matical roles. It is customary to d i sti nguish two k i nds o t" frequency, known as "type freq uency" and "token frequency." Type frequency refers to the number of i tems that are avai lable to a particular class o f forms. For example, the number, and therefore the type, freq uency of English nouns that form their pl ural i n -s is very h i g h , while that of English nouns whose pl ural is i n -en i s very l ow; the type frequency of Engl i s h verbs that form thei r past tense w i t h a s u ffi x - e d such a s walked. stopped is very high, w h i le the nu mber of Eng l i sh verbs that form their past tense by changing a stem vowel from [ail to [0] such as drive/drove i s very low. Most attention to freq uency, however, has been focused on token frequency, the n u m ber of times a particular form such as I guess or you know occurs i n tex ts, or the changes i n freq uency o f forms or constructions over ti me, such as know not versus do not kllow. The ki nds of changes that are most deeply characteristic of the gra m m atical i za tion of lexical forms - semantic fad i ng, phonological reduction, pos i t i onal fixi ng, erasure of word bou ndaries - are i nseparable from the absolute freq uency of the forms and the freq uency with wh ich they cooccur with other forms. The repeti tion of forms m ay lead t o their " l i beration," or "emancipation" (Haiman 1 994), from their earl ier d iscourse contex ts and to i ncreased tj·eedom to associ ate with a wider variety of other forms, such as when French pas 'step' used as a negative rei nforcer widens i ts range from physical movemen t ( 'doesn ' t walk a step ' ) to a l l verbs ( 'does n ' t believe (a step)" etc.). Combi nations of forms that occur more frequently tend to be automatized , that is, they are stored and uttered as a block (Boyland 1 996), such as take a chance and, get set to . Because the con tent of these automati zed combi nations is pred ictable, they are uttered more qu ickl y ; they are "streaml i ned" (Bybee and Thom pson 1 997), the parts of the combi nations tend to be slurred and reduced in promi nence (Browman and Goldstein 1 992), as in wanna for wa'" to. betcha for I bet you. At the same t i me their semantic and fu nctional conten t becomes vaguer, that is, they can be used in a wider variety of contexts (Heine 1 993 ; Krug 200 I ; Boyland 200 I ). Forms that often occur adjacen t to one another may even become fused into a si ngle word, for example as stem and c1itic (Bybee 1 995 ; 200 1 ), such as we 'll, you 're. 5.6. 1
Frequency effects
Bybee and Thompson ( 1 997 ) identi fy two major effects o f frequency of forms (token freq uency) that are especial ly relevant to grammatical i zation. They refer to these two effects as the Reduction Etfect and the Conservation Effect. The Reduction Effect points to the fact that freq uently used forms are eroded at
1 28
5 The hypothesis of unidirectionality
a faster rate than less frequently used forms. Thi s effect i s manifested i n such famil iar Engl ish contracted forms as I 'll, won 't, you 're, etc. It is also seen in the lost autonomy of the infinitive marker to in such forms as wanna, gotta, gonna (from want to, got to, going to) as well as hafta (have to), [spost;,] (supposed to) and [yust;,]/[yust;,] (used to). All of these forms are i n varyi ng degrees of grammatical ization as modal auxil iaries (see Krug 200 1 ), and all display both a phonetic and a functional contrast with forms in which the to has remained autonomous, as in We are going to [* gonna] the market, She is supposed to (sApozd tul* [spost;,]) befabulously rich (where supposed is paraphrasable as assumed not all speakers make this di stinction, however.) The contracted forms have a higher frequency than the full forms and are also the more casual register forms. The Conservation Effect correlates frequency with the retention of irregular forms. Forms that are isolated in a morphological paradigm will tend to con form to the paradigm unless they are especial ly frequent. Obvious examples of the conservation of h ighly frequent forms i nclude suppletion, such as good/better, bad/worse, go/went, as well as irregular forms l i ke strike/struck. The highly fre quent verbs sleep and keep retain their irregular past-tense forms slept and kept, whereas the less frequent verbs creep, weep, and leap, w i th their irregular past tenses crept, wept, and leapt, tend to be ass i m i lated into the regular paradigm and to be produced as creeped, weeped, and leaped. In syntax, the Conservation Effect of high frequency is seen i n the behavior of nouns in contrast to pronouns. Pronouns, which are of very high frequency, retain the case distinctions that have been lost in nouns (he, him; she, her, etc.), and may preserve older positional features (for example, in English possessive pronouns must precede the possessed noun, as i n her uniform, whereas lexical noun possessives can precede or follow the possessed nou n : the officers ' uniforms or the uniforms of the officers). The high frequency verbs be and have and the auxiliaries simi larly keep conservative syntactic characteristics that are no longer found i n full lexical verbs. In Engl ish, be, have, and auxi l i aries, u n l i ke lexical verbs, can invert directly w i th the subject in forming questions, as in Have you forgotten anything ? (contrast the ungrammatical * Forgot you anything ?). They also form a negative with not rather than do not, as i n They are not tired, They have not left yet. Both of these characteristics that are in PDE restricted to be, have, and auxil iaries were once found in lexical verbs also. -
5. 6.2
Synchronic studies of frequency
There are both synchronic and d i achronic empirical studies of frequency. Quantitative empirical studies usually deal with percentages rather than absolute numbers, and compare either different but functionally similar forms, or the same form in different identifiable contexts.
5. 6 Frequency
1 29
Synchronic i nvestigations typical ly i nvol ve searching a corpus for expressions suspected of moving toward some kind of grammatical status. Because by defi n i tion synchronic studies d o n o t show change d i rectly, such studies ei ther support their conclusions w ith h istorical material or compare a statistical variation agai nst a well-described type of diachronic change. One of a number of recent examples of thi s kind of study is that of the use of English though as a grammaticalized discourse marker (Barth-Weingarten and Couper-Kuhlen 2002). Discourse markers are a category of words that i ndicate how the l istener is to relate the upcom ing discourse to the previous d iscourse, such as well and anyway (cf. Schiffri n 1 987). An example of this discourse-marking use of though is the following. After the speaker S has explained why she advised her son to refuse m i l i tary service, a second speaker C raises a new subtopic, the guilt feelings of the protestor with respect to his peers (capitals i ndicate stress): (25)
S. but if thIs kid makes a mistAke on TH I S one, he may not have a CHANCE to corrEct it C. hh uh LiSten anOther factor though YOU brought up, (Barth-Wei ngarten and Cou per-Kuhlen 2002: 35 1 )6
Here though is more l i ke a discourse marker that serves to manage the segments of the discourse than an adverbial particle mean i ng 'however' signal ing a concessive relationship between two utterances. Barth-Weingarten and Couper- Kuhlen show that I I % of the occurrences of though have th is u nambiguous discourse-marking function. On the other hand, the purely concessive use was represen ted by only 1 4%. The remain i ng 63 % occupy a grey area between the concessive and the discourse-marking functions. Whi l e the quantitative data cannot alone prove that a change is u nder way, the results have Lo be compared with other studies showing that semantic change typically goes from an objective semantic mean ing toward a subjective, speaker-centered pragmatic mean ing (cf. , e.g. , Traugott 1 989, 1 995), and especially that discourse markers typically arise out of forms which serve propositional rather than discourse functions (e.g. Brinton 1 996; Traugott and Dasher 2002: Chapter 4). When this comparison is made, the distribution is seen to be consistent with a change from concessive marker to discourse marker.
5. 6.3
Diachronic studies of frequency
Diachronic studies of frequency start from the assumption that i ncreased frequency of a construction over time is prima fac ie evidence of grammatical ization. A recent example is Laury's ( 1 997) study of the emergence of a def inite article i n Fin n ish through the grammatical ization of a demonstrative. By investigating texts from three different periods, the n i neteenth century, the 1 930s,
1 30
5 The hypothesis of unidirectionality
and the contemporary period , Laury ascertained that the use of the demonstrative se (and i ts case forms) is beco m i ng i ncreas i ngly obl igatory. I n lhe d i rect object role, for example, the number of lexical nouns that were accompan ied by se ri ses from 49% in the n i neteenth century, to 60% in the I 930s, to 74% in the modern language. A l though Fi n n ish is often characteri zed even i n l i nguistic descriptions as a language that lacks a defi n i te article, Laury notes that the use of se is entirely comparable with the use of the defi n i te article i n languages where i ts ex istence is universally recognized. For d i achron ic studies, access to texts of comparable genres over a fairly long period is needed. It is only in a few languages that we are fortu nate enough to have this kind of textual history. And i t i s for only a small subset of these languages that we have any statistical studies at al l of the development of grammatical i tems. Sometimes a slowly emerging construction can be seen to take over very rapidly, perhaps at the expense of competing constructions, as i n the case of do-su pport discussed by Kroch ( 1 989a,b; see also S tein I 990a). Dramatic changes of this type ' are often d iscussed in terms of S-curves, that is, gradual beg i n n i ngs, rapid spread, and gradual taperi ng off. To date too few historical studies have been cond ucted to determ ine to what extent S-curves are the product of mul tiple correlated factors contri buting to morphosyntactic change or of i ndividual loca l i zed changes, but it seems l i kely that they result from the former. There is a need for add itional rel i able statistical stud ies of a variety of phenomena i n wh ich early gram matical i zation ap pears to be i nvolved . Among such studies, in addi tion to Hook ( 1 99 1 ), there are Givon ( 1 99 1 a) on the development of relativizers, complementizers, and adver bial cl ause markers in Bibl ical Hebrew; the large number of statistical studies of gram matical i zation of case-related phenomena ( i nflections, prepositions. etc.) in Spanish by Company may be i l lustrated by her recent work on the spread of prepo sitions (Company 2002); stud ies on developments in Engl ish i ncl ude Hopper and Martin ( 1 987) on the i ndefi n i te article, Kyto ( 1 99 1 ) on the aux i l iaries may/might and call1cou/d, Hundt (200 I ) on the get-pass ive, Mair ( 1 997 , Forthcomi ng) on be going to, be V-ing, seeing that and other constructions. S tatistical stud ies of grammatica l i zation changes that appear to be occurri ng in current speech situ ations i nclude Givon ( 1 99 I b) on verb seri al i zation in four languages of Papua New Gui nea, and Thompson and Mulac ( 1 99 1 ) on parenthetical I thillk in Engl ish (discussed below i n Section 7 .5 . 3 ) .
5.7
Cou nte rexa m ples to u n i d i recti onal ity
As has been stated frequently i n previ ous chapters, there is noth i ng de term i n istic about grammatical i zation and u n i d i rectional i ty. Changes do not have
5. 7 CounterexampLes to wlidirectiollaLity
131
to occur. They do not have to go to completion , i n other words, they do not have to move all the way along a c l i ne. A particu lar gram matica l i zation process may be, and o ften is, arrested before it is fu l l y "implemented," and the "outcome" of grammatical ization is q u i te often a ragged and i ncomplete subsystem that is not ev iden tly moving in some identifiable d i rection . Th is observation is i n con trast to historical argumentation of a determ i n istic k i n d such as is i l l ustrated by : "Before a change is man i fested l i ttle by l i ttle, i ts end resu l t i s already given i n the underlying representati ons" (Andersen 1 97 3 : 788). Taken i n i ts strong sense as presupposing a predeterm i ned outcome, even a "goal" for grammatica l i zation , such statements suggest that once a change has started, i ts progress is i nexorable. However, this hypothesis is not empirical ly supported . What i s supported is the fac t that there are strong constra i n ts on how a change may occur and on the direction al i ty of the change, even though we do not yet fu l l y u nderstand all the factors that motivate this d i rectional ity. Therefore the fact that changes do not show stages that can be plotted on a gram matical ization c l i ne does not entai l that they are necessari ly counterexamples to grammatical i zation. Nichols and Ti mberl ake ( 1 99 1 ), for example, poi n t out that i n the history of Russian there have been changes in the uses to which the instru mental case has been put that are akin to grammatica l i zation i n so far as they invol ve the cod ing of gram matical relationsh i ps, but are u n l ike grammatical i zation in i ts prototypical d i rectional sense, i n so far as they si mply demonstrate a shi ft in the way relatively stable gram matical networks operate. In Old Russian, the instrumental was allowed only with nouns expressi n g status or role that could change over time (e.g., ' tsar, ' 'secu lar l eader, ' ' n un ' ), and only in contexts of enteri ng that status (i nception), or conti n u i ng i n it for a period of t i me. Later Russ ian, however, virtually requires the i nstru mental with such nouns re ferring to status or role; also quasi-status nouns (agentive nouns such as ' bribe-g i ver' ) can now al low the instru mental in contexts o f durative aspect. There is certai n l y no case of "more > less gram matical" here. However, as the authors themsel ves say : "the overal l effect has been to fix usage in one domain and develop variation in another" (Nichols and Ti mberlake 1 99 1 : 1 42). Rather than being a cou nterexample to the unid irectional i ty of grammatical ization, the Russian i nstrumental is an example of ru le general i zation over a lengthy period o f time (about 1 ,500 years). It also ill ustrates the potential longev i ty of certai n types of gram matical organ i zation, and suggests that persistence is not l i m i ted to the mean i ngs of gram matical i tems, but is also evidenced by purely grammatical i n flections. As i nd icated at the beg i n n i ng of this chapter, the hypothesis of u n i d i rection ality has been a topic of vigorous debate in the l ast decade. On one end of the spectrum are very strong claims about unidirectiona l i ty. The strongest is lhat all gram matical i zation i nvolves sh i fts i n specific l i nguistic contexts from
1 32
5
The hypothesis of unidirectionality
lexical item to grammatical item, or from less to more grammatical i tem, and that gram maticalization cli nes are basical ly irreversible (see, e.g., C. Lehmann 1 995 [ 1 982] ; Haspelmath I 999a). Robust though the evidence of unidirectionality is, nevertheless it cannot be regarded as an absolute principle. Some counter examples do exist. Their existence, and their relative i n frequency, i n fact help defi ne our n otion of what prototypical grammaticalization is. Furthermore, a potential problem for strong versions of the unidirectional i ty hypothesis is that its l ogical conclusion is that grammatical morphemes cannot arise w i thout lexical origins. To date we do not have compelling evidence that grammatical i tems arise ful l-fledged, that is, can be innovated w i thout a prior lex ical history in a remote (or less remote) past. Some grammatical i tems, such as the Indo-European demonstrative 10- , show enormous longevity, and we cannot l ook back i n to their preh istory to fi nd a lexical orig i n . Proponents of the strong unidirectional i ty hypothesis would have to argue that 10- originated i n some cur rently u nknown lexical item. We do not at th is stage of our knowledge know what that i tem was. But neither do we know that there was none, or that there might theoretically have been none. (Indeed, a lexical history of demonstratives has been proposed by Fraj zyngier ( 1 996b) for Chadic languages, and of definite articles i n Siouan b y Rank i n ( 1 976).) We must leave for future empirical study the question whether grammatical i tems can arise fully formed, and if so u nder what circum stances . In any event, we do not need to be concerned as is Lass (2000) that this logical conclusion wou ld entail violation of the uniformitarian principle, si nce it woul d require postu lating a language state without grammatical i tems. On the as sumption that l i nguistic evolu tion was gradual, such a language state is implausible at least for language at the evolutionary stage we have access to. On the other end of the spectrum from strong claims about unidirectional ity are arguments that there are so many counterexamples to u nidirectionality that i t can not be considered a defi n i ng characteristic of grammaticalization (e.g., Janda 1 995, 200 1 ; several articles in Campbell 200 1 a). In a chapter entitled "Deconstructing grammatical i zation" Newmeyer has proposed that "there is no such thing as gram maticalization," at least as a phenomenon i n dependent of other changes ( 1 998: 226). Many of the researchers who argue from this perspective are concerned that, even i f u nid irectionali ty were irreversi ble, including unidirectionali ty in the defin i tion (as we have) makes the claim of un idirectionali ty u n i n teresting from a theoretical poi n t of v iew (see Norde 200 1 for detailed discussion). Newmeyer ( 1 998), Campbel l (200 I b), Janda (200 1 ), Joseph (200 1 ), and others use th is ar gument to c l a i m that, al though there is extensive evidence for regularly recurri ng directional changes, grammatical ization should not be thought of as a "theory," in the sense of an explanation of a subject of study. Instead, they suggest, it should be
5. 7 Counterexamples to ullidirectionality
1 33
thought of as the descriptive name of a frequently occurri ng epiphenomenon that can be explai ned by other factors that occur in language change anyhow. Such other factors are variously thought of as reanalysis (I. Roberts 1 993a) or "downgrad ing reanalysis, appropriate semantic change, and phonetic reduction" (Newmeyer 1 998: 260). While such criticisms need to be taken very seriously, several important char acteristics of the study of grammaticalization usually get lost i n the discussion. One is that grammatical ization is a functional ist theory - a theory about the i nter action of language and use; the questions posed i n functional and formal theories are not identical (Croft 1 995 ; van Kemenade 1 999b). Functionalist theorists seek to accoun t for the relationsh ip between language and use, and for local , gradient phenomena in language. On the other hand, formal theorists have sought until recently to ask about i nvariant properties of the m i nd, and about structure i nde pendent of context and use; however, at the time of writing a fundamental shift is occurring here too w i th i nterest developing i n accounting for variation in grammar, as in the l i terature on Optimal i ty Theory (cf. Archangel i and Langendoen 1 997; A. Prince and S molenski 1 997), and most especially stochastic Optimal i ty Theory, which is quantitative (cf. , e.g., Boersma and Hayes 200 1 ; Bresnan, D i ngare, and Man n ing 200 1 ). Another poi n t is that grammaticalization is a theory with dual prongs: diachronic and synchronic. From the diachronic perspective, since it is a theory of the relationship between structure and use, not of change i n grammar, the fact that many of the changes discussed are tendencies, not rules that operate 1 00 percen t of the time, is irrelevant. Use is usually variable, only occasional ly cat egorical. Newmeyer, however, expl icitly says : "I take any example of upgrading as sufficient to refute un idirectional ity" (Newmeyer 1 998: 263). From the syn chronic perspective, too, i t is a theory of the relationship between structure and use, and of emergen t properties of language. Therefore, characterizing grammat icalization exclusively as an epiphenomenon of reanalysis or of other factors i n change fails to address a large subset o f the phenomena u nder consideration i n studies o f grammatical ization. Many alleged counterexamples have been i ncluded i n a discussion of terms that cover very d isparate phenomena, such as "degrammaticalization," "Iexicalization," and "exaptation." Degrammatical ization is probably the most widely used of thesc terms, and indeed is sometimes used as a cover term for the other two, and several others not discussed here (for details, see Heine Forthcomi ng). Despi te i ts name, as Heine poin ts out, "degrammatical ization" is in fact used for many prototypical cases of end-stage grammatical ization, including development i n to an only par tially or total ly unanalyzable segment of a morpheme (cf. It in Meil let's example of Iteute < OHG Itiu tagu ' this day'), and also for complete loss (see Section 6.5).
1 34
5 The hypothesis of unidirectionality
However, the term "degram matical i zation" is also used for changes that violate schematic c l i nes l i ke: phrases/words > non-bound grams > i n flection (Bybee, Perkins, and Pagliuca 1 994: 40)
and "upgradi ng" of erstwh i le i n fl ectional or derivational forms, We w i l l discuss some upgrades at the end of th is section , Here we com ment briefly on issues in lexical ization and exaptation , Two main strands of research on lexical ization are relevant here (see Brinton 2002, Traugott Forthcomi ng, for detailed discussion of the many ways th is term has been used), One strand concerns changes that are more properly cal led "con version" and what are probably the most often c i ted putative coun terexamples to grammatical i zation : changes i nvol ving the use of grammatical items, i nclud i ng derivational morphemes, categorial ly as nouns or verbs, e,g" to up the ante, that was a downer; his uppers need dental work, I dislike her use of isms (see, e,g" Ramat 1 992, 200 I ), S i m i l arly, in German and French the second-person-singular fam i l i ar pronouns du and tu are "Iexicalized" as the verbs duzell and tutoyer, re spectively, both mean i n g ' to use the fami l i ar address form . ' Changes of the type Prep. up > V up, and most especially of the type deri vational morpheme -ism > N ism, typically i nvol ve a quotation or mention of some kind. 7 These kinds of changes are i nstantaneous - one can take any element of l anguage, i nclud i ng the letter w i th which it is graph ically represented or to which it is icon ic, and use it lexical ly, e.g., F- word, T-square, bus (ultimately from the Lati n dative plural -bus in omnibus ' w i th a l \ ' ) , and use it l i ke a nou n ; given certai n semantic constrai nts, one can take any noun and i nstantaneously convert it i n to a verb (e.g . , to calendQ/; to typo); one can also take phrases l i ke forget-me-not and acronyms l i ke laser ( ' light amplification by the sti mulated emission of rad i ation ' ) (Norde 200 I : 236). Innovations of this type may or may not spread to other speakers, j ust l i ke other changes. These c hanges are i nstances of recru i tment of l i nguistic material to enrich the lexicon and have virtually nothing in common w i th gram matical ization . However, other exam ples of lexical i zation do have much i n com mon with gram matical izati o n . For example, Li pka ( 1 990) defi nes lexica l i zation "as the phe nomenon that a complex lexeme once coined tends to become a si ngle complete lexical unit, a s i mple lexeme," a process often call "univerbation." Li pka goes on to say : "Through this process it loses the character of a syntagma to a greater or lesser degree" ( 1 990: 95). Erstwhile composi tional forms l i ke garlic « gar 'spear' + leac 'leek'), halibut ( < halig ' holy' + butte ' flat fish ' ), arise « 'on' + 'rise ' ) now function as monomorphemic, non-composi tional elements. S i nce these i tems be long to the m aj or classes N, V they are considered to be lexical . Likewise already derives from all + ready, hafta from have to, and sorIa from sort of They belong
5. 7 CounterexampLes to tinidirectionaLity
1 35
to m i nor classes - aspectual markers, modals, degree words - and are therefore considered gram matical . U n i verbation has occurred i n all of them. Furthermore, in many languages what origi n ate as phonologically pred ictable alternations may even tual ly be morphologized (e.g.,foot-/eet is the modern reflex of an earl ier stage when the plural was/o f- i ; phonetically, the 0 was fron ted before the -i, and when the -i (pl ural marker) was lost for phonological reasons, the fronted vowel remained as the marker of plura l i ty). These examples and others show that there is a point at which gram matical i zation and lexical ization may i ntersect (see, e.g., Hagege 1 993; C. Lehmann I 989a, 2002 ; Wischer 2000). Indeed, as Lehmann has poi nted out, lex ical phrases such as as Long as must fi rst be lexicali zed (frozen) before gram matical i zation can set i n . In many ways lexical ization i n the sense of u n i verbation a n d grammatical ization are parallel a n d both "constrain t h e freedom of the speaker in selecting and comb i n i ng the consti tuents of a complex expression." They "are not m irror i mages" (c. Leh man n 2002 : 1 5 ). Another term that has many interpretations and has been seized on as ev idence for coun terexamples to grammatica l i zation is what Lass ( 1 990) cal led "exaptation," a term he borrowed from biology to account for what he saw as "the opportu n i stic co-optation of a feature whose origi n is unrelated or only marginal l y related t o i t s later use" (Lass 1 990: 80) a s a resul t of "bricoLage, cobbli ng, jerry bu ilding; . . . recycl [ i ng], often i n amazi ngly original and clever ways" (Lass 1 997 : 3 1 6). The "unrelatedness" is the key to notions that exaptation is a coun terexam ple to unidirectional i ty. An example he gives is the reanalysis of a Dutch adjectival num ber-gender agreement marker as a marker of a subclass of morphologically com plex attri butive adjectives. At about the same time, Greenberg ( 1 99 1 ) used the term "re-gram matical i zation" to refer to s i m i lar phenomena, includ i ng changes in the late development of demonstratives. Demonstratives frequently give rise to deli n i te articles ("Stage I"), and then expand their range to i nclude all specific nouns, whether defi n i te or i ndefi n i te ("Stage II"). At this stage the article often becomes morphologi zed as a prefix or suffi x on the noun (cf. The Mississippi), but it retai n s some of i ts article- l i ke fu nctions, i n , for example, not bei ng used i n generic ex pressions (compare Engl ish at schooL, on /oot, etc.). I n the next stage ("S tage III"), the use of the affix spreads to v i rtually al l nouns, i ncluding proper names. Th is new di stri bution leads to a si tuation in which the former demonstrative assumes new fu nctions having to do with a form's status as a member of the cate gory "nou n ," for example they can be used to derive nom inalizations, or to mark pl uralization (Greenberg 1 99 1 : 304-5 ). S tages I and II can be considered classic cases of gram matical i zation, but not so the third stage, according to Greenberg, be cause there i s renewal of an old, marginalized fu nction and "disj u nctive" semantic change ( 1 99 1 : 30 I ) . One problem with both Lass 's and Green berg's examples is that although the changes may be semantical ly and functionally unexpected, a
1 36
5 The hypothesis of unidirectionality
detailed study of the d iscourse contexts for the changes i n question is not cited, and therefore it is difficult to assess whether there was or was not semantic and functional d i scontinuity in the history of the change. Not all cases of reuse of morphological m aterial for "opportun istic" reasons or "novel" purposes have been seen to pose problems for grammaticalization . For ex ample, Vincent ( 1 995 ) analyzes the development of the Romance defin i te article and clitic obj ect (e.g . , French Ie) out of Latin ille 'distal deictic pronoun ' as an instance of both grammaticalization and of exaptation . In his view, it demonstrates grammaticali zation because there is loss of segmental structure. At the same time it demonstrates exaptation or natural selection after case loss of "discarded vari ants to ensure that the necessary functions have clear phonological expression" (Vincent 1 995 : 444). Indeed, as reconceptualized by Croft within the framework of a typology of reanalysis as "hypoanalysis," exaptation and regrammatical iza tion are shown to be far from discontinuous semantically or functional l y : "[i]n hypoanalysis, the l istener reanalyzes a contextual semantic/functional property as an i nherent property of the syntactic unit. In the reanalysis, the i nherent property of the context . . . is then attributed to the syntactic unit, and so the syntactic u n i t i n question gains a new mean i ng o r function" (Croft 2000: 1 26-7). Norde (200 1 ) has i n terestingly expanded the notion of exaptation i n a different way to include such well-known and extremely challenging phenomena as the replacement by a c l i tic of i n flectional gen i tive in Engl ish, S wedish, Danish, and the variety of Norwegian known as B okmal (an upper-class variety of Norwegian infl uenced by Dan ish), the histories of which have been widely discussed but about which there is l ittle agreement (see, e.g., Janda 1 980; Norde 200 1 , and the extensive references therein). For example, Old English i n fl ectional genitive as i n (26) shows concord with i n the possessive N P (Ecgfrith the ki ng): (26)
6res cy ni ng-es sweoster Ecgfri6-es the:GEN king-GEN si ster: NOM Ecgfri6-GEN 'the sister of Ecgfri th the ki ng' (c. 1 000, Aelfric Hom I I , 10 87, 2 1 5)
but three hundred and fifty years later we fi n d a cl itic i n s i m ilar constructions: (27)
the god of slepes heyr 'the god of sleep's heir'
(c. 1 368, Chaucer, Book of Duchess 1 68)
The use of the clitic spread gradually to i ncreasingly m ore varied contexts, and is not a case of i nstantaneous change, but of generalization across types of NP, including pronouns (cf. anyone else 's cat) (see Allen 1 997). It is also not a case of significant semantic-functional difference. From these perspectives, it is very much l i ke grammaticalization. On the other hand, the rise of the -s clitic does on most analyses show counterevidence for the assumption that there is unidirectionality
5. 7 Counterexamples to Ulzidirectionality
1 37
in grammaticalization from clitic > affi x and not vice versa (see also examples in Janda 1 995 and Luraghi 1 998). Another much-cited example of a violation of unidirectionality in grammatical ization has already been mentioned i n Section 3 . 5 : the development of an inde pendent "affirmative adverb" ep in Estonian (Campbell 1 99 1 ). Originally the cl itic * -pa 'emphatic, ' it u nderwent vowel harmony in harmony-triggering contexts, as did another c litic * -(ko)-s ' (question)-informal speech.' After a regular sound change involving the loss of fi nal vowels, the context for vowel harmony with these clitics was l ost, the clitics' morpheme boundaries were lost, and they ceased to be analyzable as i ndependent morphemes. They were reinterpreted as -ep and -es respectively and "Iexicalized" as independent grammatical words (Campbell 1 99 1 : 292). Unl i ke cases of conversion, they do not invol ve recruitment to a major class, hence we consider them to be legitimate counterexamples to independent word > clitic. Another cou nterexample to unidirectionality that has been cited is the development i n Pennsy l vania German of the rounded form wotte of the preterit subj unctive welle ' would < wanted ' into a main verb 'wish, desire' (Burridge 1 998). The l atter appears to be the only example cited to date of a mirror-image reversal. As Norde (200 1 ) and Heine (Forthcomi ng) point out, most alleged coun terexamples do not reverse prior history exactly. The h istory of wotte exemplifies one reason for occasional counterexamples to unidirectionali ty : preemption of a morphological element for the ideological pur poses of the community. Burridge proposes that the lexicalization (more strictly conversion) of wotte is a kind of euphem ism - avoidance of expressing wish too bluntly, arising from Mennonite rel igious principles. Another reason for the development of several counterexamples to u n idirectional ity of this sort is the development of "adaptive rules" (Andersen 1 973). A language user who has de veloped a new rule is l i kely to find that at certain points Output 2 does not match Output 1 . Therefore the individual may be misunderstood, or ridiculed, etc. Such an individual may develop "cover-up" rules that are not integrated i nto his or her grammar, but which i n essence permit output analogous to that of users of Output I . Hypercorrection (overuse of an item considered to be socially or stylistically sal ient) is of this kind, as is discussed at length in Janda (200 1 ). For example, the speaker who has not acquired a who-whom distinction may attempt to ac commodate to users who do make such a distinction and produce utterances such as Whom did you say was looking for me ? In a study of such rules, Disterheft ( 1 990) suggests that hypercorrections are particularly often found in wri ting. She cites Stei n's ( 1 990b) study of the replacement in the fi fteenth century of the third singular present-tense marker -th by -so The -s form spread gradually in different syntactic and phonological environments and i ncreased in frequency until c. 1 600. However, j ust before the turn of the century, -th increased in frequency, dropping
1 38
5 The hypothesis of unidirectionality
off aga i n in the seventeen th century. The resurgence of -til is ev idence, it is argued, for an adaptive ru le which led to overuse of the older form in a wri tten Stan dard developed from Chancery English owi ng to association of the -til w i th "high sty le." If so, thi s (and other socioli nguistic data discussed i n Labov 1 972) suggests that adaptive rules may for the most part be typical of adu lt rather than c h i ld lan guage users. As Disterheft poi n ts out, they make the effects of abductive change ( i .e., reanalysis) h ard to detect. Hence they may give the i m pression of greater gradual ness of c hange than was actually the case. Furthermore, they may obscure (or even divert) the natural path of change, and so may lead to counterexamples to unidirectionali ty. When we rev i ew the l iterature on coun terexamples to gram matica l i zation , a striking fact emerges. They are sporadic and do not pattern i n sign ificant ways. However, at the level of a change schema, that is, at the level of li ngu ists ' ide al i zations and general i zations over changes, u n id irectional i ty is extremely robust cross- l i nguistically (Andersen 200 1 ; see also Dahl 2000: 1 3), whether spec i fied in terms of c l i nes, or of claims such as van Kemenade's that the final stage of grammatical i zation i s "base-generation as a fu nctional head" ( 1 999b: 1 00 I ) .
5.8
The use o f u n id i rectional ity i n reco nstru ction
Counterexamples such as those c i ted i n Section 5.7 should caution us against making u n critical i n ferences about d i recti ons of grammatical ization where h istorical data are not available, si nce the poss i bi l i ty of an anomalous devel opment can never be absolutely excl uded (Hagege 1 993 ; Tabor and Traugott 1 998; Newmeyer 1 998). Proponents of the strong version of the unidirection al i ty hypothesi s have argued that one can do reconstruction of non-attested stages of a grammatical form . For example, C. Lehmann has said: Given t w o variants which are related b y t h e parameters of gram matical i zation . . we ca n always tel l which way the grammatical i zation goes. or lIIusl have gone. The significance of this for the purposes of i nternal reconstl1lction is obvious. (C. Leh mann 1995 [ 1 982] : 1 9. italics added )
However, i t is very i mportant to recogn i ze that, given the number of coun terex amples, such a reconstruction can only be a hypothesis. The only viable way of approaching reconstruction is via weaker statements such as : We wou ld . . . expect grams that are older - i .e . . that have undergone more devel opment - to be cl oser to the stem. more fused and shorter or more reduced in segmental material than younger grams of equal relevance. ( Bybee. Pag l i uca. and Perkins 1 99 1 : 3 3 )
5. 9
Conclusion
1 39
Too confident a use of assumptions about u n i d i rectionali ty can lead to wrong concl usions, even w i th respect to attested data. For example, as shown in Tabor and Traugott ( 1 998), in d i scussing the "grammatical ization scale" of verbal nouns (geru nds), c. Leh mann cites: (28)
a. John's constantly reading magazines b. John's constant reading of magazi nes
c. ' the (constantly) readi ng magazi nes d. the constant readi ng of magazi nes
(c. Leh mann 1 995 [ 1 982] : 62)
and comments "we have two stages of our grammatical ization scale embodied in the Engl ish p o s s -ing construction . At the latter stage, the nom inalized verb has assumed all the relevant features of a n ou n ; -ing- nominalizations are even pl ural izable" (c. Leh mann 1 995 [ 1 982] : 64). It i s actuall y not clear whether Lehmann is making a synchronic or d i achronic claim here, because he usually uses the term "scale" for synchronic c l i nes, but the references to "stages" suggests he is here maki ng a d i achronic claim. In any event, the prediction is diachron ical ly incorrect: ty pes (28b) and (28d) are h istorical ly earl ier than type (28a).
5.9
Conclusion
The ev idence is overwhel m i ng that a vast n umber of known instances of the development of gram matical structures i nvol ved the development of a lexical item or phrase through discourse use i n to a grammatical i tem , and then i n to an even more grammatical i tem, and that these changes were accompanied by de categorialization from a major to a m inor category. Typological ly, changes of th is kind are widespread and show systematic pattern ing. Counterexamples are spo rad ic and only rarely cross-l i nguistical ly attested ; the rise of c 1 i tic possessi ve i n English, Swed ish, a n d Norwegian Bokmal is u nusual i n t h i s regard, b u t w e shou ld note that the languages are related, and the histories are not identical. Reconstruc tions based on an assumption of unidirectional match ("isomorph ism") between cline and d i rection of change in a specific i nstance should be framed as testable hypotheses.
6 Cla u se-internal morphological changes
6. 1
I ntro d u ction
Recognizing that there are some sporadic counterexamples, we turn now to m ore specific i nstances of regularly recurri ng types of unidirectionali ty dis cussed in the precedi ng chapter with a focus on the kinds of changes that typical ly occur clause-internally. In the next chapter we consider cross-clause changes . In the first part of this chapter we look i n some detail at examples of "com pacting" - the fus i ng of erstwhile i ndependent elements with each other, most especially the development of clitics into inflections. This process is often cal led ,, "morpholog ization. 1 Then we look in some detail at examples of the develop ment of grammatical forms in two domai ns: that of the paradigm, and that of clause structure, specifically subject and object marking. Finally we consider the "end" of grammaticalization : loss.
6.2
M o r p h o l ogization
In French and most other Romance languages adverbial formations such as the following are found: (I)
lentement 'slowly ' fermement ' fi rmly' doucement 'softly, sweetly'
(Lausberg 1 962: 1 1 1 / 1 , 95-8)
For a large class of adjectives, a correspondi ng adverb is derived by adding the adverbi al suffix -ment to the fem i n i ne form, e.g., lent 'slow (masc.),' lente 'slow (fern .),' lente-ment 'slowly. ' Thi s suffix was original ly an autonomous word, Lati n mente 'mind + ablative case. ' Its beginnings as an adverbial suffix are to be sought in such phrases as clara mente 'with a clear m i n d . ' However, it is no longer restricted to psychological senses, but is a general adverb formative, as in: (2) 1 40
L' eau cou le doucement. 'The water flows softly. '
6. 2 Morp/zologization
141
Table 6. 1 Buryat Mongolian pronouns and verb endings
I si ngular 2 si ngular I plural 2 plura l
Pronoun
V ending
bi �i bide ta
-b -� -bdi
-t
Source: based on Comrie ( 1 980: 88)
In Old French (and some modern Romance languages) there are sti ll traces of the autonomy of mente, in that it tends to appear with conjoi ned adjectives : humble e doucement 'humbly and gently ' (cf. Spanish clara y concisamente 'clearly and concisel y ' ) . The history of t h e French s u ffi x -ment shows a new lexical formative coming i n to existence out of a formerly autonomous word. It has done so in a fami l iar manner, by ousting i ts competitors such as modo 'manner, ' guise 'way, fashion' (specialization), and by being assigned to a progressively closer lexical relationship with the adjective stem. Semantically, too, the Lati n word mente 'mind + ablative case' has l ost its restriction to psychological states. An affix such as French -ment which was once an i ndependent word and has become a bound morpheme is said to be morphologized, and its historical lexical source (in thi s case, Lati n mente) is said to have undergone morphologi zation . Where long written histories are available, many bound morphemes can be shown to go back to i ndependent words. Often, too, a h istorical source i n i nde pendent words can be assumed through inspection of synchronic divergent forms. For example, i n B uryat Mongol ian (Comrie 1 980: 88) person-number suffixes on the verb are clearly related to independent pronouns in the nomi native case, as shown in Table 6. 1 . B u t as we have seen in previous chapters, not every instance of grammaticalization involves morphologization. For example, modal aux iliaries in Engl ish are grammaticalized out of earl ier ful l verbs, but they have not (yet) become affi xes. The beginnings of morphologization must be sought in repeated use of syntactic constructions. Some l ingu ists, among them Chafe ( 1 970), Watkins ( 1 964), and Hymes ( 1 956), have suggested that units of d iscourse - clauses and sentences are structured w i th the same kinds of rules as those by which words are i n ternally structured, that is, that 'syntax ' i tself is only morphology writ large. The study of grammaticalization to some extent supports such a view, i n that the conceptual bou ndaries separating constituents such as sentence, clause, phrase, and word often
1 42
6 Clause-internal morphological changes
seem somewhat arbi trary, and there is a cont i n ual movement among them. Whi Ie at any synchronic stage there m ay sometimes be reasons for setting up such d iscrete constituent types, from a h istorical perspective the relationsh i p bet ween a stem and an affix can only be considered in the contex t of the phrasal and even hi gher- level syntax from which they are derived . Virtual ly by defi n i tion, morphologization i s that part of grammatica l i zation that pri mari ly i nvol ves the second and third parts o f the c l i ne: lexical item > c l i tic > affi x
S uch a cl i ne i s of course a gross overs i m p l i fication of the h ighly detai led fac ts of language. At the very least, we need to say : l exical i tem in a specific syntactic context > c1itic > a fli x
For various detai led hypotheses about how to approach some of the h istorical phenomena encom passed by this c l i ne, see Bybee ( 1 985), Dressler ( 1 985), Bybee and Dahl ( 1 989), Sch wegler ( 1 990), Haspelmath ( 1 993), Givon (2000), M i thun (2000). Wh ile there is not always ev idence of a c 1 i tic pre-stage in the grammat ical i zation of affi xes out of autonomous lexical words, the fi x i ng or "freezi ng" and loss of lexical autonomy i nvol ved i n the process presupposes a c 1 i tic stage. In the example of French -mellt, Spanish -mente which we d iscussed above, and in other examples of derivational affi xes such as Engl i sh -hood, Iy etc. out of fu l l nouns, it may be assu med that at one stage the eventual affi x was attracted to what came to be its fu ture stem and came to form an accentual unit with it. C l i tics obv iously have a central rol e i n establ ishing the sorts of structures that undergo morphologization. I t is the frequent syntactic col location of a particular word class, such as a noun, with a particular type of c 1 i tic, such as an adposition, that most typically leads to morphol ogi zation (e.g ., as a noun w i th a case affi x). -
6.2. 1
,
Som e choracteristics of ditics
As mentioned in Chapter I , the word "c 1 i tic" is usually used to refer to a set of unaccen ted forms that tend to be found attached to a more heav i Iy accen ted form (known as the "host") . The attachment may be so close that the c1itic becomes a Il'i x l i ke, for example, English lI 't in don 't (see Pu l l u m and Zwicky 1 983 for argu ments why n 't behaves i n i ts distribution more l i ke an i n tlectioll than l i ke a c1i tic form of lIot). Or the attachment may be q u i te l oose and more l i ke an autonomous word, such as French Ie i n apportez-le ' bring i t i n ! ' I n many languages there are d i sti nct sets of c 1 i tic and "ton ic" (stressed) forms of the same word . Th is is especially true of pronouns; the clearest example in
6. 2
Morphologizatioll
1 43
Engl ish o f such a contrast is in the th i rd-person-pl ural them (ton ic) versus 'em (c1 itic), where the c 1 i tic and tonic forms probably have d i fferent origins ( 'em is, in one view, from OE heom, while them is a ME form ultimately of Scand i navian origin). More often the two forms are simply accented (tonic) and unaccented (c1i tic) varieties of the same word, e.g., youlya. Prepositions and postposi ti ons (the class o f "adpos i t i ons") are often c l i ticized vari ants of adverbs. Aga i n , this is clear i n English and some other Indo-European languages, where the di fference between an adverb and a preposition res ides basically in that prepositions precede an NP and ad verbs fol low a V (c l'. prepos itional up in up a tree versus ad verbial up in she got up early). Au x i l iary verbs and verbs of hav i ng and being are freq uen tly c1itics, and may l i kew ise have c l i tic and tonic variants (e.g., I 'm the head waiter versus I AM the head waiter) . The fu nctional characteristics of c l i t ics are consistent with their status as u n i ts that are al ready i n part grammatical i zed . Compared w i th their fu l l forms, c l i t ic forms are more con text-dependent and more general i n mean i ng. Often they have fu nctions whose closest cou nterparts i n other l anguages are clearly gram mati cal , such as aspect, modal i ty, case, and participant reference (e.g., to person and number). Other c l i tics, for i nstance those which are connectives, pronouns, or in terrogative markers, have a primarily d iscourse fu nction. 6.2.2
Positions of ditics
C l i tics are typical ly restricted to certai n posi tions i n the clause. One o r these is next to a spec i fic host; for example, possessive pronouns may form an accentual group with the possessed nou n , aux i l i aries may be constrai ned to occu r ring adjacent to the lexical verb, determi ners must be placed next to the nOlin, and so on. In these examples, the host belongs to a spec i fic word class and the c 1 i tic has a fu nctional a tli n i ty for j ust that class and no other (au x i l i aries generally do not go with nouns, etc.). Such c l i tics are called "phrasal c l i tics," because they have a grammatical affi n i ty for a particular type of phrase. Other ki nds of c1i tics are not restricted i n this way and are known as "sen tential c l i tics." Some occupy what can broad ly speaking be cal led the " first slot" i n the clause, and are "procl itic," that is, they are attached to the following element or "host," as i n Fr. j 'arrive I SG-come 'I am com i ng . ' Others are "enc l i tic," that is, they are attached to a host that pre cedes . In Lat i n -que served to conjoin two phrases, as in Sellaflls populus-que Romani 'The Roman senate and people. ' Many senten tial c1i tics fu nction as conj u nctions, senten tial ad verbs, comple menti zers, and question words ( Kaisse 1 982). In Homeric Greek of the eighth century BC, for example, an unaccented word de (appearing as d ' before vowels) served to l i n k together main clauses, especially in narrat ive, as in the fol l owing
1 44
6 Clause-internal morphological changes
passage from the Iliad: (3)
Hos ei pon proeiei, krateron d 'epl mtithon etellen. thus say ing, sent:forth: he, harsh:ACC de-upon word: ACC enjoined To d ' aekonte baten para thin ' halos atrugetoio, they : D UAL de-unwilling went along shore ocean:GEN restless:GEN, M u rmid6non d ' epf te klisfas kal nGas hikesthen. Myrmedons:GEN.PL de-upon both tents and ships came: 3PL 'Saying this, he sent them forth , adding some harsh inj unctions. So they went rel uctantly along the shore of the restless ocean and came to the tents and ships ( Homer, Iliad I : 326-8) of the Myrmedons. '
The particle de is enc l i tic; it forms a prosodi c unit with the precedi ng word, as is shown by t h e accentuation krateron de, where the normal accentuation krateron has been changed by the presence of the c 1 i tic. In Latin , enc l i tic -que could serve a simi lar function of,j o i n i ng main clauses: (4)
Omnibus copi is provolaverunt i mpetumque in equites nostros al l : with forces Hew:forward : they attack: ACC-que on cavalry our feceru nt. made: they 'They hurled all their forces forward and launched an attack on our cavalry. ' (c. 60 BC, Caesar, De Bello Gallico I I : 20)
Examples (3) and (4) i llustrate a common constraint on sentential c 1 i tics. It i s often known as Wackernagel's Law, after Jacob Wackernagel, w h o n oted that en c l i tics in Indo-European languages usually occur i n second position (Wackernagel 1 892) ; the phenomenon is now known to be widespread and not restricted to Indo-European (see, e.g., S. Anderson 1 993 ; Halpern 1 995). Sentential encli tics mean i ng 'and, but, so, ' etc., have a tendency to occur in the second pos i tion i n the sentence, fol lowing the first ton ic element (such as krateron i n (3) and impe tum i n (4» . B u t other c1 itics may occur i n that position too, for example, c 1 itics w i th determ i n er or aux i l i ary verb character. The "second pos ition" tendency may be related ·to the topic-com ment structure that spoken sentences typical ly have: in many utterances there is an i n itial phrase (the topic) that, as it were, sets the stage for what is to be said about it (the comment). Thus, i nterrogative markers may serve to focus on one item bei n g questioned, as in Indonesian , where -kah is a c1iti c attached to the first word or phrase, and this phrase is the one being questioned : (5)
Menari k-kah pilem itu? i n teresting-kah film that ' Was that film interesting?'
6. 2 Morphologization
1 45
At the same time it should be noted that the second position does not necessarily focus attention on the first word; often it is si mply the established position for sentential particles. In Saami (Finno-Ugric), for example, the interrogative particle bat i mparts a surprise attitude toward the entire utterance, not merely toward the first word : (6)
a. Don bat ledjet doppe okto? You bal were down-there alone ' You mean you were down there on your ownT b. Dus bat maid lea oO a biila? you-two bal also is new car 'You actually have a new carT
(Fernandez-Vest 1 994: 59)
As was noted by Wackernagel, pronouns and verbs may also favor the second position in the sentence. Morphologization i nvolves the creation of a bound morpheme (i .e., an affix) out of an independent word by way of c1iticization . The final stage of th is process, the un iting of the affix with i ts stem, is referred to as "univerbation ." Although univerbation can in theory i nclude the u n i ting of the two parts of a compound into a single lexical i tem (e.g., boat + swain > bo 'sun, cup + board > cupboard), the term is most often used i n reference to a later stage of morphologization, as in examples such as Latin clara mente 'with a clear mind' > French clairemem 'clearly, ' where the second element has become a derivational affix. A particularly i nstructive example of univerbation, and of morphologization in general, has been described by Andersen ( 1 987) for Polish. During the recorded history of the language, a copular verb has come to be suffixed to a participial verb stem to form an inflected past tense. The earl iest stage of the textual record (Polish prior to 1 500) shows a copular verb existing i n both c1itic and tonic forms. The c1itic typically occurs in second (Wackernagel's) position. In Table 6.2, the c1itic form of the verb 'to be' i n the th ird col umn is the ancestor of the Modern Pol ish suffixes i n the fourth column. The tonic forms in the second col umn drop out of use as copulas at an early date, but the th ird-person-si ngular jest lives on as an emphatic marker. In (7), -m is the clitic first-person-si ngu lar form of the copula, and ogla gala i s the "verb," h istorical l y a past participle: (7)
a. To- m jest ogl a Qaf a. that- I SG EMPH saw 'That I did see. ' b. Bo-cie-m sie, cafa darowafa. for-thee- I SG REFL entire gave ' For I gave myself wholly to thee.'
(Andersen 1 987: 28)
The c1itic -m and the verb are separated from one another, with position in the sentence and the verb at the end.
-m
in the second
1 46
6 CLause-internaL morphoLogical challges
Table 6.2 Polish tonic and clitic forms of the copula Old Pol ish
I si ngular 2 si ngular
3 si ngular 1 plural 2 plural
3 pl u ral
I dual 2 dual
3 dual
TOllic
Clitic
M odern Pol ish
jesm jes jest/jesc/je
-(e)sm/-(e) m -(e)s -0
-(e)m -(e)s -0
jesm(y) jesce s\l
-(e)smy -(e)sce -0
-(e)s my -(e)sce -0
jeswa jesta jesta
-(e)swa -(e)sta -(e)sta/-O
Source: based on Andersen ( 1 987: 24)
Table 6.3 Bonding of clitic copuLa to verb stem ill Polish. A D J 500 to the present Dale
Nu mber
Percentage
1 500s 1 600s 1 700s 1 800s 1 900s (exposi tory prose)
1 30 649 994 1 395 28 1 7 525
49 68 80 84 92
23
Source: based on Andersen ( 1 987: 29). Andersen's figures are based on work i n Pol ish by Theodora Rine! ( 1 975) (for fu l l reference see Andersen 1 987: 50).
After about 1 500, however, changes begin to occur. Sentence stress on any ' element in the comment part of the sentence may attract the c1 itic. Furthermore, the c 1 i tic i ncreas i ngly appears a fter the verb, regard less of the verb 's position, especially if the verb is an I-form preterit. Here i t is suffi xed (enc l i ticized) to the verb. The movemen t toward suffixal status is stri k i ngly i l l ustrated by the stat ist ics cited by Andersen ( 1 987: 29). Table 6.3 shows the percen tage of occu rrences of the c1 itic copula that appears as a suffix on the preteri t verb (whatever its pos ition in the clause), expressed as a percentage of the overall number of i nstances of the c1itic copul a i n texts of different centuries from AD 1 500 on.
6. 2
MOIphologizatioll
1 47
Table 6.4 Differential ulliverbatioll of preterit verb 'speak ' and persoll-Illllllber suffix ill Polish dialects Southern
Standard
Northern
I singular 2 singular 3 singular
m'6wH -em m'6wjo( -eS m'6wjo(
m6w' H em mow'B es m'6wjo(
m6w' H em m6w' H es m'6wH
I plural 2 plura l 3 plural
mow' i l i -smy mow' i l i -scie m6w' i l i
m6w' i 1 i -smy m6w' i 1 i-scie m6w' i l i
m6wil'ismy m6wil'iscie m6w'i
Source: Andersen ( 1 987: 32)
In the modern language, as can be seen from Table 6.3, the morphologi zation of the copula as a suffix on the preteri t verb i s sti l l not complete. Al though u n i verbation of the verb and the c 1 i tic is very general, there are accen tual retlexes i n both modern standard Polish and in the modern d ialects o f the former c 1 i t ic status of the verbal su lli xes. These d i fferences in accent suggest univerbation has pro gressed at d i fferent rates in various pariS of the preterit paradigm and in di fferent dia lects. They also suggest relatively fine disti nctions among levels or degrees of "compacting." Pol ish has generalized a penu l t rule for stress wh ich puts stress on the next-to-Iast syl lable i n the word . The d i alects d i ffer from the standard language in the degree to which they recogn ize the suffix as part of the word for purposes of assign i ng stress. In some forms, the stress (marked with before the vowel) is where it should be if the "c 1 itic" is a relatively un morphologized , separate suffi x or "word . " I n these forms t h e suffix appears i n Table 6 . 4 w i th a hyphen . B u t i n others i t is where it should be i f the "c1 i tic" is a ful ly morphologi zed suflix, that is, i f verb and c 1 i tic have u ndergone univerbat ion. The left-to-right arrangement of the table retlects the progress of univerbation : i t is almost non ex istent i n the southern di alects, the standard has carried it through in the si ngular but not i n the plural (except for the th ird person), and it i s complete in the northern dialects. I t is i m portant to note that the accentual change here is not simply a morphophonem ic (phonological) adaptation of a fu l l lexical item to a neighbori ng c1itic; the change affects only the verb in the preterit, not other forms to which the copu la is c 1 i ticized . Consider the fol lowing example: I
(8)
a . Wcz' oraj-em prz' zysed-i . yesterday- I sa arrived 'I arrived yesterday. ' b. Wcz' oraj przysz'e doJ -em. yesterday arrived- I sa 'I arri ved yesterday. '
(Andersen 1 987: 33)
1 48
6 Clause-internal morphological changes
The verb adapts its stress to the new suffix, whereas the adverb ignores the clitic for purposes of stress. The c1itic does not suffix i tself to any random sentence element, but specifically to the verb, and it "seeks out" the verb in a way that suggests that the original, verbal nature of the c l i tic m ay still be constraining its current use. The process of morphologization whereby i ndependent words become c1i tics and eventual ly affixes resul ts in a fixed order of morphemes with respect to the stem. Whereas the ancestral i ndependent word may have had a certain amount of positional freedom, univerbation removes any flexibility of position with respect to the stem. It has often been suggested (e.g . , G iv6n 1 979: 239-45) that morpheme order may reflect earl ier word-order tendencies of the language at the time when the morphologization in question was occurring. G i v6n cites data from Amharic, a Semitic language of Ethiopia. In this language the original word order was va, but through a strong substratum of Cushitic, a non-Semi tic language family w i th OV word order, verb-fi nal patterns have permeated the language. The original va word order continues to be reflected i n the morpheme order i n inflected words, however: (9)
Kassa borsa-w-n la-M u l u sana-at. Kassa wallet-the-OBJ to- M u l u gave-IOBJ ' Kassa gave the wal let to M u l u . '
(Giv6n 1 979: 244)
Assuming that the suffixes - W, -n, and -at were once autonomous words, presumably demonstratives and pronouns, we can by hypothesis derive the suffixed forms from older sequences of noun + demonstrative + pronoun, and verb + indirect object pronoun. B oth these word orders are compati ble w i th a va type of l anguage, and the hypothesis is that Amharic has preserved the original word order in its order of suffixes even though the syntactic word order has radically shifted . On the other hand, French, a modern va language in which obj ect lexical nouns must fol low the verb, shows object pronouns preceding the verb, consistent w i th i ts origins in a strongly OV language (Lati n): ( 1 0)
a.
Lexical nouns
Le boulanger donna Ie bijou a la jeune fi l le. The baker gave the jewel to the girl 'The baker gave the jewel to the girl . ' b.
Clitic pronouns
II Ie lui donna. He it to: her gave 'He gave it to her. '
In Latin the verb would normal ly have been placed at the end of the sentence in each case. Describing similar phenomena cross-l inguistically, Giv6n coined
6. 2 Morphologization
1 49
the memorable phrase "Today 's morphology is yesterday 's syntax" (Giv6n 1 97 1 : 4 1 3). While not a novel i nsight, the concept has been the subject of much recent discussion. How general is it? Can the synchronic ordering of morphologized affixes be u sed to reconstruct the prehistoric order of words in the sentence? Certainly with Amharic and French the reconstructions are in accord with what we either know or can surmise about origi nal word orders on i ndependent evidence. However, it is clear that the order of morphologized affi xes can at best only i n form us about the local order of the c l itics at the time their positions became fi xed, not about the word order in general. This fact alone is not sufficient to i nvalidate Giv6n's observation, for it is at least theoreticall y possible that when clitics move toward being affixes they adopt the place that they would have if they were ful l lexical i tems. In such a scenario, auxiliaries would fol low the verb i n OV l anguages, possessive pronouns would fol low the noun in VO languages, and so on. However, such reconstructions must be approached with caution. Comrie ( 1 980), for example, shows that in languages with variation in their basic word order (and many seemingly quite rigid languages show such variation) even phrasal clitics may appear in a position with respect to their host that is different from that normall y occupied by correspondi ng lexical i tems. In Classical Mongolian, for example, an OV language in which one might expect that possessive adjectives would precede the noun, in fact both orders were possible: (I I)
a . minii mori n my horse b. mori n minii horse my
Here, the (a) phrase was more usual, and d i ffered from the (b) phrase in some such n uance as MY horse versus my HORSE. In later Mongolian dialects it is morin minii, the less usual order, that u nderlies clitics which, eventuall y, become morphologized as suffi xes i ndicating possession, cf. Kalmyk: ( 1 2)
a. more-m horse- I SG : POSS ' my horse' b. mini more-m my horse- I SG ' my horse'
(Comrie 1 980: 90)
Comrie suggests three reasons why such a development might occur. One is that if the usual morphological process in a language is suffixation , newly emerging affixes will conform to the general pattern already avai lable. Another possible reason is prosodic: in Mongolian languages the head of a construction is never preceded by an unstressed element. Morpholog i zation of clitics as prefixes would
1 50
6 CLause-internaL morphoLogicaL changes
provide exceptions to this otherwise quite general pri nciple. Finally, there may be a syntactic reason . In languages of the OV type, Comrie argues, there is a general principle that the head of a construction can al ways be separated from one of i ts attributes by some other word. For example, the object of the verb, which wou ld be considered such an attribute, does not have to stand immediately next to the verb, but there can be adverbs or other words that intervene. For the same reason, in such l anguages (and in VO l anguages such as English that place the adjective before the noun) a possessive adjective that precedes a noun can al ways be separated from the noun by an adjective: my horse, my strong horse, etc. B u t this separation tends not to occur if the possessive fol lows the noun, i .e., in examples such as Classical Mongolian morin minii (horse my) 'my horse' there wou ld al most never be an adj ective between the two words (i.e., * ' horse strong my ' ) . The position of the c l i ticized possessive, then, conforms to that order i n which the c1itic and the host noun were i nvariably adjacent to one another. Even more damaging for the hypothesis that l ocal morpheme order necessar i l y reflects earl i er general word-order patterns are examples (admitted ly rare) of reordering of morphemes with i n a parad igm. One such example is from Pengo, a Dravidian language. Bybee ( 1 985 : 40), citing Burrow and B hattacharya ( 1 970), shows that in Pengo the perfect was origi nal l y formed by add ition of aux i l iary na to the past-tense form, after person-number inflections. So, for example, the fol l owing older forms can be found: ( 1 3)
viit alJ viit-alJ-na
'I came' 'I have come'
In other words, the perfect is formed by V:TNS + person/number + na. But the more common, and more recent, formation of the perfect is by repeating the person-nu mber i nflection after that template, as in ( 1 4a), resu lting in the structure V:TNS + person/number + na + person/ number, or even , in more stream l i ned fashion, simply V :TNS + lla + person/number, as in ( I 4b): ( 1 4)
a. hurtalJ nalJ ([hurta-alJ-na-lJJ) ' I have seen' b. hurtanalJ ([hurta-na-IJJ) 'I have seen'
Haspel math ( 1 993) similarly notes cases where the add ition of derivational mor phemes after the inflections has resulted in the creation of new stems, to which the same i n fl ectional morphemes are added (see, too, the cases and discussion in Mithun 2000). It seems, then, that while morpheme order may, and often does, reflect earl ier lo cal word-order patterns, Givon's pri nciple is of l i m i ted usefu l ness, si nce we cannot be sure which earlier word is reflected, the basic one or a secondary one. Factors of morphological type, prosody, and syntactic typology as well as general tendencies
6. 2
Morphoiogization
151
such as second position may all i n tl uence the order in wh ich c1 itics are placed with respec t to the host. A further factor for which Bybee shows overwhelming evidence is relevance to the meaning of the stem. In the Pengo example, person and n u m ber, which agree with the arguments of the verb, are less relevant to verb meaning than temporal ity. We now turn to a brief discussion of Bybee's hypothe sis of the sign i ficance of "relevance" for understanding the facts of morphological bond i ng. 6.2.3
Semantic "relevance " as a fador in fusion and morpheme order
It is a truism that i n a language which exhibits affixal morphology, not all grammatical categories w i l l be affixally expressed. Some w i l l be relatively free (st i l l lex ica l , or c1itic), others will be tightly bound and intlectional . Some will be ex pressed by a phrase, others by a word with affixes. If this were a random happenstance of when which form started to change, then no general patterns of relationsh ip between affixal and non-affixal expression would be expected in a language. However, in an exploratory cross-li nguistic survey of fifty languages, Bybee ( 1 985) showed that: (a) (b)
(c)
Meaning elements that are d i rectly relevant to verb mean ing are more l i kely to be fused or bound than those that are not. The order in which they occur is partly correlated w i t h their degree of relevance to the verb. Among meani ngs relevant to the verb, the most general are l i kely to be expressed i n flectionally.
To avoid term inological confusion, i t is important to note that Bybee is using the term "relevance" to refer to the extent to which the mean ing of a gram matical category (e.g., aspect or tense) affects the i nherent mean ing of the lexical item with which it is associated . Th is is different from the pragmatic "relevance" that we discussed as a motivation for mean ing change i n Chapter 4. The pragmatic maxim o f Relevance has to do with relevance to the participants i n the commun icative act. Verbs express events or states of being. A causative situation is wi thout question semantically relevan t to the verb, si nce it atfects the event or state Qf being directly. Causative meani ngs are often signaled by bound morphemes, e.g., redden 'cause to be red . ' However, a causal situation is often understood rather differently from the li teral combination of V + causal . The relationship between die and kill ( 'cause to die') is a classic instance of thi s kind of difference. Bybee shows that causal rela tionships are often expressed by derivational forms. These are bound forms, which, although iden tifiable as separate morphemes, nevertheless combine with a base to add new, rather spec ific, meani ngs, or change l i nguistic categories, and form a
152
6 Clause-internal morphological changes
stem to which other affixes, such as inflections, can be attached. 2 Redden can, for example, have tense attached (e.g., reddened). Even more frequently, causal rela tions are expressed by lexical ly different forms, that is, totally fused forms, rather than by i nflectional forms, because they are at least partially idiosyncratic. 3 By contrast, tense, aspect, and mood tend to be expressed inflectionally because they are highly general and can apply to most event and state types. Of the languages B y bee investigated, 72% had inflectional tense, mood, and aspect. By contrast, only 56% had inflectional person-nu mber-subject agreement (Bybee 1 985 : 33). Although number can be expressed derivationally and even in lexically different ways, person-number agreement tends to be less frequently bound than either causal rel ations or tense-aspect-mood . This, she suggests, reflects the lesser rel evance of person-number to the verb: its prime function is not to express aspects of the situation, but to express distinctions among arguments of the verb. Tense, mood, and aspect themselves have ditl'erent likelihoods of ordering with respect to each other. Aspect refers to the way in which the internal constituency of the event is viewed, that is, according to whether it is seen as a whole from the outside and completed (perfective), or from with i n and incomplete (Comrie 1 976). Tense places the situation in time with respect to an established point i n time, either the time of speech (deictic tense), or some other poi nt in time (relational tense). Mood refers to the way the speaker presents the truth of the proposition, whether as probable, possible, or certain (Bybee 1 985 : 28). Given the hypothesis of relevance, aspect is most relevant to the verb, tense less so, si nce it relates the time of the si tuation to some other time, and mood least so since it expresses the speaker's point of view on the situation . If that which is most relevant is that which is most l i kely to be close to the verb, then we would expect aspect to be the most likely of the three categories to be ordered next to the stem (or even be part of it, as a derivational form), tense next, and mood l ast. The relative positions of aspect and tense are wel l established for languages in which both are prefixes or both suffixes, as i l lustrated by : ( 1 5)
a.
(Australia) O:l-ru-unti o-apa. cook-PAST-DUR-eat iT was eating it. ' b. Kewa (New Gui nea) fra-paa-ru. cook-PERF- I SG:PAST 'I fi n ished cooking it.' Tiwi
(Foley and Van Val i n 1 984: 2 1 0)
As Bybee notes, the claim that the natural order is mood-tense-aspect-V (or, i n OV languages, V-aspect-tense-mood), may seem counterintuitive to l i nguists who speak of T-M-A (tense-mood-aspect) in that order, and are fami liar with the
6. 2 Morphologizatiotl
153
earl ier work o f Chomsky i n which i t was suggested that Engl ish constructions such as would be going were tense-mood-aspect-V. As she notes, English does not of course have bound forms. However, recent analyses of the same construction in English actually do reflect the order she predicts, si nce PDE would (also might,
could, should) is best treated as a mood marker (i.e., would is not will + Past Tense, but a single univerbated form). Normally present tense is in zero form after mood, but is expressed in many quasi-modals (e.g. , has to), and past is expressed by have after may, might, can, could, etc., but i nflectionally with many quasi-modals : would
have been going, had to be going (Bybee 1 985: 1 96-200). Bybee's hypothesis has provided results that are in keeping with independent studies of other issues in verbal morphology, such as the exten t to which predictions can be made about which verbal affi xes are l i kely to be phrasal , which affixal (see Bybee and Dahl 1 989). B u t a few cautions should be noted . These include Lhe fact that in so far as the data base is founded on extant grammars, it is subject to the difficulty that different l i nguists have made different analyses of such basic properties as morpheme boundaries, status as phrilse, word, bound morpheme, etc. (for a detailed account of various defin i tions of tense and aspect, see B innick 1 99 1 ). They have also had different definitions in m i nd of the categories in question (most notably tense and aspect are often confused). As Bybee notes, other difficul ties include the fact that languages often have portmanteau morphemes (morphemes combining two or more categories in segmentally indistinguishable ways; see the next section). Also, the morpheme order predictions naturally do not hold when affixes are not in sequence; for example, mood affi xes may be prefixed while tense-aspect affixes are postfi xed, as i n Cay uga: ( 1 6)
a · -yak6 · -nyo · - ? OPTATIVE-FEM : PATI ENT-ki l l-PUNCTUAL 'she would get killed '
(Mithun 1 99 1 : 1 77)
Furthermore, Bybee's own statistics do not always provide exactly the predicted facts. Given that aspect is the most relevant to the mean ing of the verb, mood the least so, one would predict that aspect wou ld be bound most frequently, mood least. Also, one would predict that aspect would be more likely than the other two categories to be expressed deri vationally, given that it defines the internal structure of the situation . In Bybee's fifty-language sample, the statistics are as shown in Table 6.5 ("bound forms" incl ude both derivational and in flectional affixes). Aspect and tense fol low the predictions. Aspect is most frequently found as a bound form (in 74% of the languages); in 22% of these languages its form is derivational rather than i nflectional . Tense is less frequently found as a bound form (in 50% of the languages); only 2% of the occurrences are derivational . Mood, however, does not follow prediction (a) since it is more frequently infleCLional than the other
1 54
6 CLause-internaL morphoLogicaL changes Table 6.5 AffixaL aspect-tense-mood forms
Aspect Tense Mood
% bound forms in sample
% inflectional forms in sample
74 50 68
52 48 68
Source: based on Bybee ( 1 985: 30)
two categories, and more frequently bound than both tense and aspect, despite bei ng least relevant to the V. However, it is always inflectional , probably because it is so general (the category mood includes evidentials and "hear-say" expressions which express the speaker's atti tude toward the likeli ness of the truth o f the proposition).
6.2.4
Phonological concomitants of morphologization
The fusion of a lexical item and a clitic as stem and a ffi x that typi fies morphologization is accompanied by phonological changes of various sorts . Most often these changes are characteri zable as reductions: vowels and consonants are dropped, a stress or tone accent is lost causing an accen tual readjustment over the newly formed word, and adjacent phonological segments are ass i m i lated to one another. If the l oss of the word boundary that once separated the two elements is included, some phonological adj ustment is by defin ition always invol ved in morphologizatio n . Often, a s m i g h t b e expected from the divergence a n d the res u l tant coexistence of both unreduced (tonic) and reduced (cl itic) forms, the autonomous lex ical form will u ndergo a different set of phonological changes from the bound form . The result is that someti mes the morphologized form actually preserves someth i ng cl oser to the older state of affairs. A good example of this is the vocalism of Engl ish affixes that have been protected from the effect of the Great Vowel Shift, which operated on tonic vowels. Thus we have manLy [-Ii] bes ide Like [ I ayk], and because [bi - ] compared with by [bay ] . The d i vergence of the article a [ ;! ] from numeral one [w/\n] has already been mentioned in Section 5 .4.2. I n the process o f phonological attri tion and selection that accompanies mOl' phologization , we can identify two tendencies : (a)
A quantitative ("syntagmatic") reduction : forms become shorter as the phonemes that comprise them erode.
(b)
A qual itative ("parad igmatic") reduction : the remain i ng phonological segments in the form are drawn from a progressively shri nking set.
6. 2 MOIphoiogizatioll
1 55
Th is smal ler set of phonemes resul t i ng from tendency (b) reflects the universal set of u n m arked segments. They tend especially to be apical (tongue-tip) consonants such as [ n ] , [t] , and [s] , the glottal consonants [1] and [ h ] , and common vowels such as [a] , [u], [ i ] , and [a ] . The result is that from a synchronic perspective gram matical morphemes tend to be composed of "unmarked" segments. As will be elaborated on i m med i ately below, "un marked segments" are those that are textually frequent, fou nd across a wide range of different languages (indeed, may be u n i versal), are learned early by chi ldren, and are targets of neutralization of con trast. (The concept of markedness bei n g invoked here was developed by the Prague School phonologists in the I 920s and 30s. Hyman ( 1 975 : 1 43-56) is an excellent general treatment.) One aspect o f the tendency toward unmarked segments is that morphologization is usually accompanied by a reduction in prom i nence. Prominence i s a function of spec ial accentuation, length, or some sort of posi tional privilege such as i n i tial sy l l able in the word (Trubetzkoy 1 929: 58). In an environment of lessened prom i nence, there is a general neutralization of segments, that is, a loss i n certain o f the phonological d i s t i nctions found in ful l lexical i tems. Haiman ( 1 972) poi nts out that i n this loss of phonological contrasts characteristic of non-prom i nent syllables, there is a movement toward an unmarked set of phonemes, in the sense that we have just presented it. Ad m i ttedly th is neutral ization belongs to non-promi nent syl l ables in general (a poi n t taken up again below). B u t because a reduction in prom i nence is characteri stic o f forms that are becom i ng morphologized, one outcome of morpho 1 ogizalion is morphemes that typical ly consist o f si mple, unmarked, phonological sets. In Turkish, Haiman noles, there are strategies for avoiding the vowel [0] in non-pro m i nent sy llables; this vowel is marked because it i s simultaneously low and rou nded . S ign i lican tly, the only gramm atical suffix that contains [0] is -yor, a progressive verbal form which was once an autonomous (copular) verb and "has only recently degenerated to the status of a suffi x" ( 1 972: 367); in other words, i t is newly morphologized . A second aspect o f the appearance of unmarked segments i n morpholo gized forms i s that the analogical spread of one allomorph at the expense of others is ai ded by the sheer textual frequency of the successfu l allo morph, and textual frequency i s associated w i th unmarked segments (Greenberg 1 960). Let us consider a well-known example from the Polynesian language Maori. In Maori there is a passive suflix -ia which has attracted the fi nal consonant of the verb stem , this consonant being dropped when it occurs word-fi nally (K. Hale 1 973). The result is the emergence of at least as many allomorphs of the passive su fli x as there are consonants, as shown i n the examples i n ( 1 7):
1 56
6 Clause-internal morphological changes
( 1 7)
Verb
Passive form
Allomorph o f passive suffix
hopu 'catch' aru ' follow' tohu ' point out' maatu 'know'
hopukia arumia tohulJia maaturi a
-kia -mia -lJia -ria, etc.
Some of these allomorphs have spread to verb forms where they did not originally occur, replaci n g the historically "correct" allomorph, so that the present-day dis tribution of the allomorphs does not always reflect the hi storically expected one. Especially -tia has fared wel l, and appears to be on the way to becoming the norm for the passi ve suffix in all stems (Krupa 1 965: 70- 1 ); -kia is also well represented. The tendency for these two allomorphs, especially -tia, to oust the others reflects the numerical preponderance of the two consonants [t] and [k] . In Maori texts, [t] and [k] , which are about equal to each other in frequency, occur considerably more often than other consonants ( Krupa 1 966: 22). These two factors - textual frequency and the selection of members of the unmarked set as targets of neu tralizations - are closely interrelated, and in fact are simply aspects of the same phenomenon of phonological markedness. Many of the phonological changes that accompany morphologization are not peculiar to thi s process but are simply part of the same processes of assimilation, at trition, and other kinds of reduction that are found more generally i n non-promi nent syllables and across j unctures (Hei ne and Reh 1 9S4 give detailed examples of these in grammatica l i zation in African languages). For instance, the loss of the final -ns in the French first-person-plural future (nousfinirons ' we will fi n ish, ' pronounced [finir5)]), is part of the general loss of fi nal consonants in Modern French, not a peculiarity of the morphologization of the auxiliary from Latin . A more complex example is the following. The Sahaptian language Nez Perce can code the goal of a verb of motion either with a locative suffix on the goal noun as i n ( I Sa), or with a transitivizing suffix on the verb that then treats the goal as an object as in ( I Sb): ( 1 8)
a. 'aayato-na paa-'naxpayk-a 'nli t-pe. woman-OBJ 3SUBJ:30BJ-bring-PAST lodge-LOC ' H e brought the woman to [his] lodge. ' b. 'aayat pa- 'naxpayk-60-ya miy60xato-na. Woman 3SUBJ:30BJ-bring-LOC-PAST chief-OBJ ' He brought the woman to the chief. '
(Rude 1 99 1 : 1 88)
For our purposes the significant morphological difference between the two sen tences is that ( I Sb) contai ns in the verb a (transitivizing) locative suffix -60 - , which consists of the frozen past-tense marker -e- and the copula wee, presumably in the sense of 'be [in a place] . ' By a regular Nez Perce phonological process,
6. 2 Morphologization
1 57
the resulting sequence becomes first -LltI- and then, by vowel harmony, -60-. Thus the seem ingly arbi trary erosion accompanying the morphologization of the copula verb as a transitivizing suffix is in fact the result of well-established changes in the general phonology of the language. The development of Nez Perce -60- described above i nvolves erosion (of 11') and morpheme boundary loss, or "fusion," of the tense marker and the copula. Erosion is the loss of phonological segments as the process of fusion continues (Heine and Reh 1 984: 2 1-5). This loss typically occurs at boundaries, such as at the end of a word or morpheme. Although examples are ci table from virtually any language, French and English are especially striking because of the conservatism of their orthographies, in which spellings with "silent letters" abound (one does, however, have to distinguish between orthography that reflects actual changes from orthography that reflects false etymologizing by scribes and grammarians; for example, French po ids derives from pensum, not as, the spel ling would suggest, from pondus). Latin calidum ' hot' had lost both its suffixes (-id-um) by the time of Modern French chaud [so] , and even the I of its stem, cal-, has been absorbed. Erosion may or may not bring about morphemic loss . An example of erosion that has merel y reduced a morpheme wi thout el i m i nati ng it is the Lati n ablative singu lar suffix -0, e.g., lupo ' from the wolf, ' from an earl ier lupOd. Extreme cases of fusion are easily identified. One is that in which two or more morphemes fuse as a "portmanteau" morph (Hockett 1 947 [ 1 966] : 229) without there being a one-to-one semantic/functional match between any morpheme and any set of phonological segments. French du [dii] 'of the (masc.),' i.e., de + Ie, , and aux [0] ' to the (masc.lfem. plur.) (a + leila + s) are examples of this. In many tone languages, fusion may result in a portmanteau morph which has segmental material from one morpheme and a tone from another, the tone being all that remains from the second morpheme. Matisoff ( 1 99 1 ) picturesquely refers to this process as "Cheshirization," from Lewis Carrol l's Cheshire Cat, which disappeared leaving only i ts smile. Matisoff ( 1 982: 32-4) gives the followi ng example. In Lahu, a Lolo-Burmese l anguage of northern Thailand, an original causative prefix, probably *s-, underlies alternations between voiced and voiceless i n itials in such pairs as : ( 1 9)
a. d:S 'drink' : l:l 'give 10 drink' b. de 'come 10 rest' : Ie 'pUI down'
There is a lonal change, generally from a lower 10 a mid or high tone, which is phonetically (albeit indirectly) l i n ked to the voici ng change. Where the ini tial consonant is one that does not show a disti nction in voicing, such as the nasal [m] or the affricate [c] in the next example, the tonal difference is the only remai ning trace of the former prefix:
1 58
6 Clause-internal morphological changes
(20)
a. m :, ' see' : 111 :> 'show' b. di 'eat' : eli ' feed '
Many possible examples of fusion depend crucially on unam biguously defin ing a "phonological u n i t." Indeed , Heine and Reh define fusion as the disappearance of the boundary separating two morphemes, "these morphemes thus bei ng red uced to one phonological u n i t" ( 1 984: 25). Affi xes normally form a phonological unit with their ste m . Yet fusion is someth ing more than mere affixation ; it is a stage in wh ich the phonological substance of an affix (or of the subordi nate part of a compound) and the stem start to become indistinguishable from one another. There is much curre n t d isagreement about the levels resulting from various degrees of fusion, much of it resulting from attempts to define a synchronic system in which every item has a disti nct structure. From the poi n t of v iew o f gram matical ization, the issue is not so much what the structure i s at some moment in time, but what the d i rection of change is and how far along the continuum some particular form has moved (see Giv6n 2000). Fusion, then, is a characteristic of the right-hand side of a con t i n u u m at the left end of wh ich are discrete morphs and at the right end a si ngle morph, possibly w i th some purely phonological residue o f a prev ious second morph o In morphologization, as i n all grammatical i zation, we must ask whether there are any rules characteristic of morphologization that are not part of the general or h istorical phonology of the language. It w i l l be recal led that this same question arose in semantic change, where there appeared to be no evidence that the mean ing shi fts that accompan i ed grammatical i zation were anyth ing other than SU btypes of mean i ng s h i fts affecting lexical i tems i n general . Si nce morphologization neces sarily i nvol ves the emergence of new morpheme boundaries and other junctural phenomena, and the juxtaposition of segmental clusters in ways not found inter nal to words or across "older" m orpheme bou ndaries, and si nce usually there is a prosod ic reduction o f the new affi x, any special phonological changes are to be attributed to these S Ubtypes of phonological change rather than to any intrinsic change ti·om "lexical" to "grammatical." All the � ame, given that gram matical ization occurs i n h ighly local contexts, and in later stages o ften i nvolves u n i verbation of a new affix w i th a stem, u n usual , even unpreceden ted , sequences of segments may occur, which in turn may set thi ngs up for special phonological changes . For example, Lati n has an i m perfect tense formed synchronical l y from a verbal stem, a suffix -ba-, and a personal ending such as - 1/1 ' first-person s i ngular, ' e.g. , from ama- ' l ove' can be formed amabam 'I used to love.' The d i achron ic source of the i mperfect is likely to have been a present par ticiple ( i n th i s case *amants) fol lowed by a form of the copu lar verb (*b " wam 'I was ' ) . The combination *amants b" wam presumably gave rise to amabal1l
6. 3 The development of paradigms
1 59
through a ru le whereby the com bi nation * nts + b" w even tual ly yielded b. B u t , as Bald i ( 1 976: 846-7) notes, such a sequence o f events can not be proved o r dis proved on purely phonological grou nds, s i nce i t is only i n this very collocation that the combi nation of segmenLs in question is ever l i kely Lo have occurred in Lat i n across a morpheme boundary.
6.3
Th e devel o p m ent of para d igms
Someti mes the coalescence o f two parts of a periphrastic construction as stem and ani x remains isol ated, and has no further consequences. Consider, for example, the second-person-pl u ral y 'all found in some Engl ish dialects . The form is transparently deri ved from the periphrastic you + all; yet -all has not i n these dialects spread a s a general plural morpheme t o other words, ei ther nouns or pronouns. We do not, therefore, see a "parad igm" emerging o f the k i nd shown in (2 1 ). (2 1 ) you he she
" I -all you-all " he-all " she-all
Nor do we see any real signs Lhat -all i s becom ing a plural suffix i n Engl ish; y 'all appears at least i n POE to be parad igmatically isolated, the result of the neutral ization of si ngular and plural in the origi nal second person . O ften , however, later stages of grammatical ization i nvolve a process of emergen t parad igms, i n which a set of related affixes emerges based o n a s i ngle form . With verbs, this basic form is often the th i rd-person s i ngular. With nouns and pronouns it is often a non-nom i n ative case. We i l l u strate the "paradigmatization" of a nominal marker w i th the devel opment of the early Scandi n avian (Old Icel andic and Old Norse) reflexive pronoun sik i n to an affi x . Originally the accusative of the th ird-person (si ngu lar and plural) ref-l ex ive, it spread to other persons and cases and came to mark voice as wel l . We i l l ustrate with Old Iceland ic forms. In this language, sik coex isted with iLS grammatical ized form, the encl i tic -sk: (22)
a. Hann He b. Hann He
bau� sik. offered h i msel f bauzk (zk < " �sk) o ffered-h i m sel f
(Heusler 1 92 1 : 1 42)
The development o f Lhis pronoun as a suffix in Old Icelandic and in Dan ish is a classic example of gram matical ization. With c1iLicization comes:
1 60
6 Clause-internal morphological changes Table 6.6 Old Icelandic present indicative reflexive verb forms
I 2 3
(a)
Singular
Plural
fi nnomk fi nzk (z [-t+s-]) fi nzk
finnomsk finnezk fi nnask
=
Phonological assim ilation. The pronoun and i ts host merge phonologi cal ly. For example, if the stem ends i n an apical such as t or p the combi nation is pronounced [tsk] (spelled zk). The form bauzk
(b)
-mk) has spread to the plural . In addition, the first-person plural has assumed the third person, yielding a complex suffix -msk. Therefore of the fi ve potential possibilities for autonomous reflexive pronouns only two remain. We turn now to the second kind of syncretism, that of case. The autonomous reflexive corresponding to -sk is the accusative sik. B u t -sk is found i n environments where a genitive or dative would be expected. For example, i n (23a) the autonomous pronoun ser is i n the dative, but its encl itic form is -sk i n (23b) : (23)
a. Hann eignaOe ser rfke. He appropriated to-hi mself ki ngdom 'He appropriated the ki ngdom to hi mself. ' rfke. b: Hann eignaOesk he appropriated-hi msel f ki ngdom 'He appropriated the ki ngdom to hi mself. '
(Heusler 1 92 1 : 1 4 1 -2)
Both kinds of syncretism are exempli fied in (24a, b): (24)
a. Er hefneO yOuar Ii honom. you revenge you rsel ves:GEN : PL on him ' You revenge yoursel ves on him.' b. E r hefnezk Ii honom . You revenge-yoursel ves on him
(Heusler 1 92 1 : 1 4 1 -2)
6. 3 The development of paradigms
161
The plural second-person object pronoun ycJuar i n its autonomous form is i n the genitive because the verb hefna ' to avenge' requires that case for its object. The c1itic version, however, is -sk, formerly the si ngular third-person accusative. Cliticization and morphologization of the reflexive sik is accompanied by se mantic generalization (see Section 5.2. I ). The generalization i nvolves expansion from reflexive to reciprocal, to middle voice and even passive, giving meanings such as: (25)
a.
Reciprocal
Spyriask deir tfOenda. ask: RECIP they of-news 'They ask one another for news. ' b.
Passive
Skip Mask. ships build:PASS. 'The ships are being bui lt.' c.
Experiencer
Henne hugnadesk pat vel. her: DAT pleased-sk' that wel l 'She was pleased at that. '
(Heusler 1 92 1 : 1 42)
These and other meani ngs fou nd as -sk grammaticalizers are typical of the de velopment of reflexives (see Kemmer 1 993). The c1iticized form of sik became an -s suffix in the Eastern Scandinavian languages such as Dan ish, where it has occasional passive uses, as in (26a), although its more usual function is to express middle (26b), and reciprocal (26c) meani ngs: (26)
a.
Passive
Df/lren abnedes af en tjener. door: DEF open-s by a servant 'The door is opened by a servant. ' b.
Middle
Jeg har lrengtes efter dig. I h ave longed-s after you 'I have been longing for you . ' c.
Reciprocal
Vi har mf/ldtes Here gangc. We have meet-s several ti mes 'We have met several times.'
The "passive" construction with -s is now largely confi ned to the written lan guage; the more usual passive is formed periphrastically with the verb blive 'stay, remai n ' or with vrere ' to be' and a participle, e.g . , df/Jren blev malet ' the door was pai nted . ' The history of O l d Icelandic -sk, first building a paradigm and then smoothing out its irregularities, lead i ng to a uniform stem, is a common one. Many examples
1 62
6 Clause-internal morphological changes
Table 6.7 Pre-Sanskrit noull inflection
Nomi native Gen i t i ve I n strumental Locative
S i ngular
Plural
yak vacas vaca vaci
viicas viikiilll vagbhis viiksi
Source: based on Jeffers and Leh iste ( 1 979: 59)
of this kind of parad igmatization have been c i ted in the l i terature. I n pre-Sanskrit the i n tlection of vak ' voice' contai ned stem consonants which someti mes were and someti mes were not phonetical ly motivated, as shown in Table 6.7. In the ( i talicized ) nominative and gen itive pl ural the al ternation elk (caused by a still earl ier al ternation of 'el*a i n the vowel of the s u ffi x) must have seemed arbitrary, and the pre-S anskrit gen itive plura l vakam was replaced i n h istorical Sanskri t by viiciim (Jeffers and Leh i ste 1 979: 59-60). The res u l t is a movement toward a s i ngle form of the stem in which variation in the stem is ei ther leveled out or, as here, is directly moti vated by the phonetic surrou ndi ngs. For another example, consider again the Maori passive su ffi x -ia discussed in Section 6.2.4. In this i nstance, i t w i l l be recalled, stem-final consonants were reanalyzed as part of the suffix, g i v i ng a variety of allomorphs of the s u ffi x ; but u n i form i ty w i th other stems is com ing abou t through the generalization of - t . Derived verb forms such as the causative (with the prefi x whaka- ) seem to be especially s u sceptible to this kind of level i ng (K. Hale 1 973): (27)
a. hopuk-ia 'be caught' whaka-hopu-tia « * whaka-hopuk-ia) 'cause Lo be caught' b. maur-ia 'be carried' whaka-mau-tia « * whaka-maur-ia) 'cause to be carried '
The picture drawn by such examples as these is of ragged and irregular parad igms bei ng pu l led i n to shape by analogy and general ization. Yet there are other forces at work too that lead to dispersal and d i s i n tegration. Purely phonological factors may con tribute to this, as i n the fol lowing example from Pal i , a later form of Sanskri t (Hock 1 99 1 [ 1 986]). Between Sanskrit and Pal i the copular verb as ' to be' first underwent leve l i ng, as shown by the data in Table 6.8. In Sanskrit a vowel al ternation of a and zero in the stem characteri zed s i ngular and plural forms, reflecting the Indo-European ful l/zero grade of ablaut. In pre-Pal i th i s a�fs al ternation was partly leveled, y ielding the vowel a i n al l plural forms except the th ird person .
6. 3 The development of paradigms
1 63
Table 6 . 8 Sanskrit and pre-Pali fonns of the copula
I singular 2 singular .3 singular
I plura l 2 plural .3 plural
Sanskrit
pre-Pal i"
as-mi asi " as-ti s-ma� s-tha s-anti
" as-m i 'asi ' as-ti ' as-ma "as-tim ' s-anti
SOUfce: based on Hock ( 1 99 1 [ 1 986] : 1 7 1 ) " The forms identi lied as pre-Pal i are reconstructed. hence the asteri sks. " Accord i ng to Hock. the second-person form is i rregular and has no morpheme bou ndaries .
Table 6.9 Pre-Pali aird Paliforms of the copula
I s i ngular 2 s i ngular .3 si ngular I plural 2 pl ural .3 plural
pre-Pal i
Pal i
'as-mi 'asi 'as-ti * as-ma ' as-tha Os-anti
amhi asi atthi amha attha santi
SOUfce: based on Hock ( 1 99 1 [ 1 986] : 1 7 1 )
Sound changes i n pre-Pali such as ass i m i lation then brought abou t more irreg ulari ty than ex isted even before the stem level i ng, as shown in Table 6.9. The Sanskri t-Pali development of the i rregular parad igm of the verb 'to be' can be ex plai ned in terms of u nderstood phonological developmen ts. B u t i r regulari ty may come abou t i n other quite obscure ways. Andersen ( 1 980: 1 7) shows that i n Bu lgarian d i alects l i ke that of Macedonia the conjugation of the verb ' to see' has undergone an apparently u nmotivated shift in i ts aorist tense. In Table 6. 1 0, dialect A is Western B u lgarian, dialect B is Macedonian, and d ialect C is Southern Serbian. Here di alect A has the older vocalic stem vide- while B has taken on the characteristics of a consonantal stem, changi ng vide- to vid-. Dialect C has gone even further and in add i tion to this change has i n novated new
1 64
6 Clause-internal morphological changes Table 6. I O Differential inflection of the aorist
in Bulgarian dialects
I singular 2-3 singular I plural 2 plural 3 plural
Dialect A
Dialects B and C
vide-x vide vide-x me vide-xte vide-xa
vid-ox vid-e vid-oxme vid-oxte vid-oxa
Source: based on Andersen ( 1 980: 1 7)
Table 6. 1 1 Differential inflection of past participles in Bulgarian dialects
Mascul i ne Feminine
Dialect B
Dialect C
video vide-I-a
visel visla
Source: based on Andersen ( 1 980: 1 7)
stems of the participle that destroy the transparency of the relationship to the stem
vid( e). Compare the masculine and fem in ine forms of the participle 'seen ' in dialect B (which in this respect is conservative) with those of d ialect C, as shown in Table 6. 1 I . The changes i n question have one th ing i n common: they bri ng the paradigm of the verb ' to see' closer to that of the verb ' to go,' whose forms are uniquely irregular. B u t what do 'see' and 'go' have in common that would bring about such a development? Why is the perfectly regular vocalic inflection of the verb ' to see' abandoned in dialects B and C, and recreated along irregular, even s uppletive, lines? And why should the change be restricted to past tenses (aorist and participle)? Andersen' hypothesizes that the explanation is to be found in the frequent use of certain syntagms in which 'go and see' figured together, such as idoxme i vidoxme ' went and saw, ' prisele i visel 'he has come and seen . ' It is thus not a "paradigmatic" similarity of sound or of meaning that has conditioned the change, but a "syntagmatic" discourse col location of the two verbs. The examples presented here show that while the tendency to conform to a paradigm may appear to be a potent formative force in the ongoing grammati calization of forms, grammaticali zation is not reducible to a uniform process of
6. 4 A rgument-structure marking
1 65
paradigmatization. Rather, it involves the disintegration and dispersal of forms as welI as their assembly into regular paradigms. Grammaticalization again tends to undermine the picture of stability, of clear categorial boundaries, and of structured groups of forms, showing these to be at the most temporary way-stations between different kinds of dispersal , emergence, and fragmentation. This is in fact to be expected, given the approach to grammaticalization developed in th is book: that it emerges out of processes of reanalysis in the syntagmatic domain of language, constrained by speaker-hearer negotiation. We now turn to functional-semantic hierarchies that guide the development of markers of subject and object argument structure.
6.4
Argu ment-structu re m a rking: fu nctio nal-sema ntic h i e ra rchies and morphologica l general ization We have discussed some examples of unidirectionality of segmental form
as "compacting" occurs . . We have also seen how grammaticalization can proceed along other dimensions too, such as generalization of paradigms. In the case of the development of Old Icelandic sik we also saw spread along a functional-semantic hierarchy in the syntagmatic domain (reflexive, reciprocal, passive, etc.). We i llus trate this kind of phenomenon in more detail from object marking i n Persian (also known as Iranian), and then go on to suggest how evidence from such generaliza tions can be used to develop research questions concerning languages for which we have only synchronic data, such as Sacapultec.
6.4. 1
Object marking in Persian
The development of object marking i n Persian n icely exempl ifies both the unidirectional cline which comprises lexical word > postposition > suffix and generalization along two other dimensions: the animacy hierarchy and the definiteness hierarchy. According to the animacy hierarchy, human nouns are more likely to be i ncl uded in li nguistic rules than animates in general (e.g., animals), and animates are more likely to be included than i nanimates : 4 human > ani mate > inani mate > abstract
(For the relation of this hierarchy to personal pronouns, proper nouns, common nouns, and other nomi nal types as wel l as to case marking and thematic rela tions, see Greenberg 1 974 ; S i lverstein 1 976; Dixon 1 979: 85 .) According to the definiteness hierarchy, defi nite (referential) nouns are more likely to be incl uded in linguistic rules than indefi n i te nouns. Given a three-way distinction between
1 66
6 CLause-internaL morphoLogicaL changes
referential defi n i tes (e.g., the), referential i ndefi n i tes (e. g . , some/a in Some/a mall came by trying to seLL The Tri bune this morning), and non-referential i ndefi n i tes (e.g., I need a vacation), the h ierarchy i s : +def/+ref > -detl+ref > -detl-ref
(Croft 1 990: 1 1 6)
For example, definite nouns are read i l y subjects i n Engl ish, i ndefi n i tes less read i l y s o . Pass i ve m ay b e used t o avoid i ndefi n i te su �jects (whether referential or not); and a pseudo-definite subject-slot filler there is used when a referential i ndeli n i te is the subject of an exi stential copu la sentence, as i n There is a mall at the doOl: Greenberg ( 1 978b) discusses how defi n i te articles may become i ndefi n i te, but not vice versa. It has been suggested that the motivation for hierarch ies such as these is the fact that people are more l i kely to talk about hu mans than other th ings, about referential thi ngs than about non-referential ones, in other words, factors known as "empathy" and "attention fl ow" (see Ku no and Kaburaki 1 977 ; DeLancey 1 98 1 ). These h i erarc hies capture many organ i zational phenomena in language, rang i ng from such relatively obvious properties of discourse as the l i keli hood of certai n nouns occurring i n subject position t o complex phenomena such a s i n teraction with case and aspect. The ani macy h ierarchy has already been mentioned i n con nection with the general ization of gen itive case marking to non-fi n i te clause subjects in Finnish (see Section 5 .2.2). Here we show i ts operation i n the development of object m arking. The data and m uch of the i n terpretation are taken from Bossong ( 1 985 : 5 8-79). At issue is the h i s torical background to the Modern Persian su lli x -ra i n sentences such as (28) : (28)
Ketlib-rli mi-xlin-ad book-ACC CONTI N-read-3SG ' He's reading the book. '
(Bossong 1 985: 63 )
In such sentences, the object of the verb receives a suffix wri tten as -ra. but pro nounced [(r)3.] , the [r] being dropped after stem-final consonants (Bossong 1 985 : 59). The suffix -ra i s found only on the d i rect ( accusati ve) object, not on indirect (dative) objects, and only u nder certai n semantic and d i scourse circumstances, which we d i scuss below. The path toward an object-marking ( i .e., accusati ve) suffix on the noun began in Old Persian (c. 600 Be) w i th a noun radiy 'goal, purpose' used as a postpo sition. By the M iddle Persian period this form had become reduced to - ra d and had become a postposition for dative-benefactive objects. only occasional ly used for defi n i te accusative objects and never w i th i ndefi n i te ones . The earl iest docu ments of New Persian (from the n i nth century AD on) show a suffix -ra used as a defi n i te accusative morpheme, but the dative-benefactive use sti ll Il ourishes . By
6. 4 A rgument-structure marking
1 67
the Classical Persian period (twel fth-fourteenth centuries AD) the grammatical ization of - ra is complete for all types of defi n i te objects: it is u sed with all dative and dative- l i ke objects (benefactive, possessi ve, experiencer) as wel l as with all ac cusative objects, provided they are defi n i te. The non-accusative uses are i l l ustrated in the following Classical Persian examples i n (28) : (29)
a. Hakim-i pesar-1in-r1i pand hami-d1id. wise-man-a son-PL- DAT advice CONTI N-gave 'A wise man was gi ving h i s sons ad vice. ' b. M a-r1i dar sahr dust-an besy1ir-and. I-POSS in town friend-PL many-3PL 'I have many friends in the town . ' (Bossong 1 985: 6 1 ; the spel l i ng i s modern)
There appear to be, then, three attested stages between the ni n th and fourteenth centuries: Stage I Stage II Stage III
(Middle Persian): postpos i tion -ra used for dative-benefactive objects. (Early Classical New Persian): suffix -ra used for dative benefactive and defi n i te accusative objects. (Classical New Persian): suffix -ra used for dative-benefactive objects, and extensions of the dative-benefactive use such as possessor and experiencer and for defin i te accusative objects .
The change starts with highly spec i fic, i ndividuated objects that are most capable of being affected, namely ind ividual humans. It spreads to all kinds of noun and pronolln objects provided they are i nd i v iduated (referential). Finally it incl udes IlU man objects that are only i n d i rectly affected by the action of the verb (possessors and experiencers). Th is is also a h ierarchy of discourse topical ity: - ra spreads to items down a h ierarchy of poten tial d iscourse topics, from h ighly ani mate partici pants to ord i n ary inani mate objects, al ways provided that they are actual ly present i n the discourse context, that is, that they are referential. We move now to the Modern Persian period, i n which -ra has on the one hand been extended to a wider range of NPs, but on the other hand has been restricted with respect to its use with thematic roles. The expansion is motivated by pragmatic discourse strategies that have to do w i th a foreground i ng of the referent. Thi s can be seen clearly in (30), where 'Turkish' is in contrastive focus: (30)
Arabi-O balad- i ? Torki-ra balad- i ? Arabic-ACC fam i l i ar-2SG? Tu rkish-ACC fam i l i ar-2SG? 'Can you speak Arabic? And Turkish - can you speak that?' (Bossong 1 985: 67)
1 68
6 Clause-internal morphological changes
The range of -ra may even be extended to include i ndefi n i tes : (3 1 )
DliUine derlize tlirik-i-rli peymud. corridor long dark-INDEF-ACC passed-through 'He passed through a long dark corridor. '
(Bossong 1 985: 66)
At this stage, -ra functions to focus on a prominent NP that is the object of a verb, regardless of its animacy and definiteness. Such discourse conditioning of a form as it becomes m ore grammatical ized is a very important general phenomenon. At the same time as -ra's range has been expanded on the animacy and defi n iteness h ierarch ies, its range has actually been contracted on another hierarchy to be discussed in Section 7 .4.3, that of thematic roles. With the exception of a few reli c phrases, it is now never used i n anything but a strictly accusative context, that is, it is used for direct objects only, and is no longer used for dative l i ke indirect obj ects. The causes of this contraction of range with respect to the matic roles are not completely understood (see Bossong 1 985 : 5 8-79), but may possibly h ave somethi ng to do with a tendency to special i ze case markers to the most syntactic, as opposed to semantic, cases, that is, subject and object (see H. Smith 1 996 for discussion of the interaction of case markers and syntactic case).
6.4.2
Ergative case marking: a statistical perspective
The study we have j ust outlined i l l ustrated morphological generalization over time. In some language fami lies historical data are available from which changing frequencies and discourse environments of forms can be documented. Quite often, however, written h istorical data are lacking, and trajectories and motivations for grammatical ization can only be surmised from the study of the synchronic distribution of grammatical forms in discourse. Among a number of well-known synchron ic studies of this kind are DeLancey 's ( 1 98 1 ) and Du Bois's
( 1 987) hypotheses about the clausal marki ng of case roles. Although the specific forms, and the precise way i n which they have emerged, cannot be known, this work emph asizes the fact that grammatical forms do not exist in a functional vac uum, but reflect general strategies by the speakers of languages for putting together discourses. Languages that mark subjects and objects with case morphology may present a distinction between an "absolutive" case, the category for objects and intransitive subjects, and an "ergative" case, the category for transitive agentive subjects (see, e.g., Plank 1 979; Dixon 1 994). In B asque, the absolutive case suffix is zero, and the ergative case suffix is -( e )k:
6. 4 A rgument-structure marking (32)
a. M artin ethorri da. M artin:ABS came AUX :3SG ' M arti n came. ' b. M artin-ek haurra igorri duo M artin-ERG child:ABS sent AUX:3SG ' Martin sent the child.'
1 69
(Comrie 1 978: 333)
Ergative case-marking systems like that of B asque are widely distributed among the world's languages. Ergative languages in fact often agree with one another down to such details as marking the absolutive case with a zero morpheme. To a speaker of a standard European-style "nominative/accusative language" such a system of cases may seem unmotivated . Why do not al l "subjects," whether transitive or intransitive, behave as a single grammatical class? Why should objects be marked in the same way as some subjects? In order to answer th is question, we may ask what common functions link the object of the verb with the intransitive subject (absolutive), and set these apart from the transitive subject (ergative) . Du Bois ( 1 987) investigated texts in an ergative language, the Maya language Sacapul tec, and determined that "new" information, that is, reference to newer things or persons in the discourse, was often presented in the object of the verb if the clause was transitive, and in the subject if the clause was i n transitive. Consider the following two clauses in SacapuJtec (spoken in succession by a single speaker): (33)
a. s-e:-pe: e: iseb' al"10: m, CMP-3PL: ABS-come PL three boy-PL 'Three boys came,' b. s-O-a:-ki -siky'-a1 pe:ra CMP-3ABS-MVT-3PL: ERG-pick:up-MVT the pear (Du Bois 1 987: 824) '(They came) and picked up the pear. ' =
(Here, CMP is the completive aspect prefix and MVT is a morpheme meaning 'movement. ' ) In (33a), the three boys are introduced i n to the narrative with the verb 'come. ' In Sacapul tec, the argument roles are signaled by affixes on the verb rather than by affixes on the NPs. Since this verb is intransitive, its subject is 'boys' in the absolutive case; the agreement prefix e: on the verb reflects this. In (33b), the newer item is the pear; it is the object of the transitive verb siley' meaning 'pick up, ' and is likewise i n the absolutive case. There is a verbal prefix, which happens to be zero, and which reflects the absolutive case of its object, the pear. The verb has, in addition to the zero aspectual prefix and a ' movement' prefix, a second agreement prefix, ki-. This ki- is a third-person-plural ergative and agrees with an unexpressed ergative agent (the situation is roughly the same as the English 'Three boys came and 0 picked up . . . ' ) .
1 70
6 Clause-internal morphological changes
In these texts (and it turns out i n texts from other languages too) agents are i ntroduced as new enti ties pri mari ly in i n trans itive events. New enti ties in the discourse i n troduced in the role of transitive agent are much more rare. One tel l i ng statistic i s the d istribution of the three basic roles (trans itive agent, i n trans i tive subject, and transitive object) over the appearance of fu l l nouns, si nce there is a high correlation between newness i n the d iscourse and reference through lex ical nouns. In Engl i s h sentences such as (34a, b), all three roles - transitive agent and transitive object in (a), i n transi tive subject in (b) - are represented by full lexical nouns: ( 34)
a. The paper-boy del ivered the magazines. a. The letter arrived late.
Yet sentences such as (34a), w i th a lexical noun in the role of transi tive agent, are found to be rare in actual discourse. More common ly, transitive agents are represented by a pronoun (i.e., He delivered the magazines). It must be emphasi zed that the notion of d i scourse d istri bution is critical here. It is not that sentences such as (a) are u ngrammatical, or sound strange, or are difficult to el i c i t i n isolation from native speakers; q u i te the contrary. Rather, empirical evidence i n the form of quantitative studies of spoken d i scourse reveals a marked skew i ng toward the representati o n of transi tive agents by pronouns rather than by lexical nouns. In Sacapultec d iscourse, Du Bois found, a total of 56.5% of all lexical mentions were i n the absolutive case category, that is, i n transi t i ve subj ect or transi tive object (Du Bois 1 987: 827). With i n the absol utive case, the two roles were fairly evenly distributed : i n transitive subjects comprised 32.8% of lexical noun mentions, and transitive objects 23.7%. By con trast, only 3 .4% of fu l l lexical nouns referred to transitive agents. (The remai nder of the lexical nouns were fou nd i n other sen tence roles.) Evidently Sacapul tec has grammatical i zed i n i ts case mark i ng not some c1ause level system of roles i nvol v i ng "agency" and "patienthood," nor even a semantic disti nction of animacy, as has sometimes been suggested for ergative languages, but a h igher-level function i nvolv i ng i n formation flow, that is, the d i fferent l i keli hoods that new i n formation w i l l be presented in one position in the clause rather than another. Accord i n g to this analysis, case-marking systems seem to emerge as a result of some general d iscourse tendencies. One of these is to have only one piece of signi ficant new i n formation per clause. Transi tive clauses, which contain two arguments, must therefore "manage" their argument structure so as to have at the most one of these two arguments as a lexical nou n . Another tendency is to keep the transitive agent anaphoric or "old" (what Du Bois cal ls the "Given Agent Constrai n t" - Agent bei ng the designation for the transitive agent) and to ass ign
6. 4 A rgument-structure marking
171
new i n formation preferentially to the object of the verb. I n languages that signal the ergative case only i f the ergative is a lexical noun, then, and have a d i tTerent system of mark i ng for pronouns, the ergative case can be seen as the case that marks the transitive agent when i t is, contrary to the general tendency, new to the discourse. The gram matical i zation of ergative case mark i ng therefore may consist of a spread of the ergative case to all transitive subjects, both lexical nouns and pronouns. By con trast, nom i n ative-accusative systems appear to have grammatical ized in thei r case m arki ng the syntactic argument roles subject and object, perhaps via a discourse strategy that aligned agents, whatever their status with respect to trans itivity or to old or new i n formation (Du Bois 1 985 ). I f this is so, it may be that at least some ergative systems arise out of mark i ng of new information, whereas some accusative systems arise out of marking of agency. Such quan t i tative stud ies as Du Bois ( 1 987) of the synchronic relationships between forms and d i scourse fu nctions have signi fican t impl ications for the study of gram matical ization, in that they suggest explanations from actual usage for the emergence o f a gram matical function. It shou ld be stressed, however, that the question of which form or set of forms comes to express this function is a separate one. While the grammatical ization of ergative case morphology may "fall out" from d i scourse pressure to d istri bute arguments i n certain ways, the source of the forms themsel ves varies . In Malay, for example, an agentive-ergative preposi tion aleh seems to have i ts ori g i n in a verb of a separate clause; a sentence such as (35) wou ld then have originated from somethi ng l i ke 'The letter was wri tten (and) my brother did ( i t) ' : (35)
Surat itu ditu l i s oleh abang saya. letter t he 3AG: write ERG: PREP brother my 'My brother wrote the letter. '
Here the former verb aleh, now the ergative case preposition, once had a range of meani ngs apparently encompassing 'get, obtai n, do, manage, return . ' In other languages, ergative constructions emerge from q u i te d i fferen t sources, such as the passive with an agent i n the i nstrumental case. As we wou ld expect from gram matical ization in general, the type of source is constrai ned by discourse strategies operating on pragmatically and seman ticall y relevant structures. Among histor ical studies, see S. Anderson ( 1 977), Garrett ( 1 990), A. Harris and Campbell ( 1 995 ); Chung ( 1 977) on Polynesian languages, Shibatani ( 1 99 1 ) on Phi l ippine languages; B u tt (200 I ) shows that where h istorical evidence is available, as it is for Indo-Aryan, a ful l accou nt must address not only ergative-absolutive and nom i native-accusative morphology, but use of other cases as wel l , such as dative for subject.
1 72
6 Clause-internal morphological changes Table 6. 1 2 Old English stronga adjective
singular inflection
Nomi native Genitive Dative Instrumental Accusative
Masculine
Neuter
Feminine
god godes godum gode godne
god godes godum gode god
god godre godre godre gode
The strong declension was used in the absence of a demonstrative. a
6.5
Loss
As we saw in connection with the Russian i nstrumental (Section 5 .7), and the Persian object marker (Section 6.4. 1 above), morphemes can often remain stable for very long periods, shifting their function i n broadly predictable ways but persisting in their shape. Nonetheless, at the extreme end of the history of a particular form as a grammatical marker we may find loss, either of form alone or occasional l y of both form and function . Examples of the loss of a form alone occur whenever two or more competing forms exist for the same function, and one is eventually selected at the expense of the others. We have cited many examples of this phenomenon, including the specialization of the French negative pas from among a wider set of possibi l i ties, or the selection of periphrastic tenses and aspects over inflectional ones in Late Latin and early Romance. Similarly, whole inflectional paradigms can pass out of general use, as has happened wi th the French "passe simple" such as elle s 'evanouit 'she fainted . ' The same h a s almost happened with t h e German "imperfect" ( i . e . , preterit), where
er las 'he read' is fast receding. In both these instances the older paradigm remains i n written and formal registers, but is essentially dead i n the colloquial registers. "Renewal," that is, the replacement ofa dying form by a newer, usually periphrastic, form with a similar meaning, is common in such cases . Thus the French "passe simple" has been replaced as the ordinary past tense of the verb by the periphrastic perfect (elle s 'est evanouie 'she fainted' ) , and the German imperfect also is giving way to the perfect (er hat gelesen 'he read ' ). More unusual is loss of both the morphological function that a form once served and loss of the form itself (or absorption into the stem as a meaningless component). In Old English, for example, adjectives still had case, number, and gender suffixes; thus the singular of the adjective meaning 'good' was as shown in Table 6. 1 2.
None of these suffixes has survived into POE as a productive morpheme; how ever, in the modern adverb seldom we find a rel ic of the dative plural in -um of the adjective seld 'strange, rare. ' The suffix -om in POE-seldom is said to have become "de-morphologized," that is, to have lost its morphological value (for some dis cussion of de-morphologization, see Joseph and Janda 1 988). From another, more positive, perspective the process of de-morphologization can be seen as one of "phonogenesis" (Hopper 1 990, 1 994), whereby "dead" morphemes become sedi men ted as phonological segments and over long periods actually create and repair the phonological bulk of words, rather l ike the way the shel ls of dead molluscs create geological formations. The Modern Irish verbs in the left-hand column in
(36) derive from simple verb stems to which were attached one or more adverbial prefixes with directional or locative: (36)
Modern Irish
Early Old Irish
tag- 'come' i migh 'go, leave' friotaigh 'resist' fog- ' leave' faigh 'get' abai r 'say '
to-theig 'to-go' i mb-theg 'about-go' frith-to-theg 'against-to-go' fo-ad-gab 'under-toward-take' fo-gab 'under-take' ad-ro-ber 'toward-for-bear' (data from Michael Noonan [p.c.])
But the Modern Irish forms are no longer synchronically analyzable as having prefixes - the earlier prefixes are now simply part of the phonology of the verb stem. In this way phonological segments can often be seen to consist of old morphemes; the -nd of English friend, fiend is a rel ic of the Germanic present participle -ende (cf. German freuend 'rejoicing'), and these two nouns derive from verbal roots meaning, respectively, ' love' and 'hate. ' De-morphologization can have a real effect on phonology because it may bri ng about phonotactic changes, that is, changes in canonical syl l able shape (introduc ing, e.g., new consonant clusters) and word length, and this in turn may affect tone and stress . An instance of the effect of de-morphologization on phonotac tics has been described by Dixon in the Austral ian language Olgolo (Dixon 1 982 [ 1 969]). In Olgolo many word-initial consonants were lost through erosion. As a result, a considerable number of words began with vowels, an "unnatural" situa tion which speakers of the language appear to be in the process of remedying by creating noun prefixes out of old demonstratives. These ex-demonstratives marked semantic classes; consequently the new initial consonants still roughly reflect such semantic divisions as animals and insects (nh-); fish, oysters, and eels (y-); and a broad class of inanimates that include trees, grasses, sun, fire, and language (w-). The most important effect has been a phonol ogical one: to restore to the language many more instances of word-initial "natural" (i.e., CV) syllables.
1 74
6 Clause-internal 17lorphological changes
The end prod uct of grammatical i zation is thus phonology in the very l i teml sense of phonological segments. Phonogenesis plays the v ital role of ensuring that the attri tion w h ich occurs in the natural course of change is compensated for by accreti o n . De- morphologization i n i ts end stages i s therefore not reduci ble to loss. but rat her i nvol ves a kind of "phonological strengtheni ng." There is an i n teresting parallel here to the pragmatic strengthening that we saw al ways accompan ies seman tic loss i n earl ier stages of gram matical i zation ( see Section 4 . 5 ) . Someti mes de-morphologization res u l ti ng i n phonologization will n o t b e com plete. but w i l l result in the emergence of a new gram matical form . Th is is what has for the most part been analyzed recently as "exaptation" (or, in Greenberg's terms "regrammatical i zation") (see Section 5.7).
6.6
Conclusion
In th i s chapter we have i l l ustrated a variety of changes that i nvol ve vari ous degrees o f fus i o n over time, as well as of pattern reorgan i zation and structumtion. As we have seen i n other chapters, there is a constant tension between changes that pertai n to the now of speech (sy n tagmatical ly) and those that pertai n to the choices i n any one pos i tion ( parad igmatical ly). How these develop is best understood in terms o f discourse strategies.
7
G ra m m atica l izati o n a cross cla uses
7. 1
I ntro d u ction
Ord i nary discourse does not cons ist of isolated, con text- free ulterances, but of l i n ked d i scourse un its compri s i ng reports, orders, comments, descri ptions, and other k i nds of l i nguistic activity. These u n i ts, usually expressed by clauses, Iypical ly consist of a verb and i ndicators of the arguments of the verb, in the form of lex ical nouns, pronouns, or pronom inal affi xes. A l l languages have devices for linking cl auses together i nto what are cal led complex sen tences. These tend to be class i fi ed i n gram mars accord i ng to fu nctional-semantic pri nci ples, for exam ple, whether a clause fu nctions as an NP (complemen ts, or "noun cl auses," thai are arguments of the clause), modifies an NP (relative clauses), or has ad verbial fu nctions (e.g . , temporal , causative, or cond i tional clauses). However, the form of a "complex sen tence" may di ffer q u i te radically among languages and among speakers and occas ions o f speech i n one and the same language, from fai rly sim ple juxtapos i tions of relatively i ndependent clauses characteristic of casual speech, such as ( I ), 10 complex dependent rhetorical constructions typical ly ari s i ng in the context of tradi ti ons of wri lten grammar, such as (2): (I) (2)
With i n the decade there w i l l be an earthquake. I t is l i kely to destroy the whole town. That there wi l l be an earthquake with i n the decade that will destroy the whole town i s l i kely.
I t has been customary to discuss the devel opment of markers of clause l i n kage such as the two i nstances of that in (2) in terms of grammatical ization. For example, the development o f complementizers, condi tional conj u nctions, relativizers, and so forth are standard topics exempl i fying the gram maticalization of lexical items or the i ncreased grammat ical i zation of already grammatical items in spec i lic con tcxts (in this case, the context of clause combini ng) . So are syntactic changes whereby initially separate clauses may become total l y i n terlaced such that the boundaries between clauses may become obscured at the surface level (e.g., (It) seems that he is right > He seems to be right), or at least a clause may become attached to a consti tuent i n s ide the matri x c lause.
1 75
1 76
7 Grammaticalization across clauses
Unti l recently, however, less attention has been paid to the possibility of in cluding the process of clause combi ning itself within grammaticalization, except notably in the work of Giv6n. As mentioned in Chapter 2, he proposed ( 1 979: 209) a path of gramm aticalization of the type: di scourse > syntax > morphology > morphophonemics > zero
By "d iscourse" here he meant the loose, unplanned, i n formal mode of communi cation i n language. Giv6n illustrated the first three stages of the path (discourse > syntax > morphology) by such phenomena as shifts from topic into subject, and topic sentences i n to relative clauses, finite clauses in concatenated structures into non-fin i te complementation structures (e.g. , the serial type I want I go > I want to
go). If grammaticalization is defi ned broadly so as to encompass the motivations for and development of gram matical structures in general, then processes of clause combi n i ng clearly fall squarely within its domain, as Giv6n suggested. In this chapter we show first that synchronic clause combining can be considered from the poin t of view of a unidirectional cline from relatively free juxtaposition to syntactic and morphological bonded ness within the framework of grammatical ization broadly construed. We then focus on a few examples of the development of clause combin i ng across time where the theory of grammatical ization may either help us understand the facts of complex sentence structure, or may suggest a dif ferent way of thi nking about it than has been customary.
7.2
A cl i n e of cla use-co m b i n i ng co n structions Many studies of complex sentence structure suggest a sharp distinction be
tween coordinate and subord inate clause structure. This tradi tion has been based i n part on evidence from Indo-European languages and especially the written records that give insight i nto their history. In the 1 980s the distinction was repeatedly called into question as a result of the study of non-Indo-European and of spoken languages (see especially Haiman and Thompson 1 984 on the problem of defining subordi nation cros � - l i n guistically). Here we have space to sketch only one framework for the study of c lause combi ning. More thorough treatments suggesting alternative viewpoints are to be found in Foley and Van Val i n ( 1 984), Shopen ( 1 985, vol . 2), Haiman and Thompson ( 1 988), Austin ( 1 988, on Austral ian languages), and Bybee, Haiman, and Noonan (200 1 ). A complex sentence, syntactically defined, is a u n i t that consists of more than one clause. A c lause that can stand alone can be referred to as a "nucleus" (Longacre 1 985). A complex sentence may consist of a nucleus and one or more addi tional nuclei, or of a nucleus and one or more "margins," relatively dependent
7. 2 A cline of clause-combining constructions
1 77
clauses that may not stand alone but nevertheless exh ibit difl"erent degrees of dependency. Among clauses which form margins, three types can be semanti cally disti nguished: "those which function as noun phrases (cal led complements), those which function as modifiers of nouns (cal led relative clauses), and those which function as modifiers of verb phrases or entire propositions (called ad verbial clau ses)" (Thompson and Longacre 1 985 : 1 72). As will be discussed below, adverbial and apposi tive relative clauses may be less dependent than re strictive relative and complement clauses in some languages, or at some stages of a language. From the point of view of evolution and of language acquisition, the initial formation of a complex clause would appear to i nvolve the combi ning into one in tegrated structure of two separate and autonomous nuclei that are mutually relevant (or presented as mutually relevant). The act of combining the clauses and signal ing this combination l i nguistically is grounded in rhetorical production strategies. We have no historical textual evidence of a stage of a native language wi thout complex clauses, fol lowed by the emergence of complex ones. In other words, to our knowledge human languages have had complex sentence structure avai lable throughout recorded history. B u t reorgan ization of complex combinations is wel l evidenced, as we will see below, as is the association of certain complex sentences type with certain genres, especially of plan ned discourse. Complex sentences range in type from multiple nuclei that are j uxtaposed under one intonation contour but have no segmen tal (overt morphological or syntactic) indication of a grammatical relationship between them , to combinations of nucleus and margin in which this rel ationship is highly compressed. To simplify, we can think initially of a cline with three "cluster points" as follows (the cline is based on discussion by Matthiessen and Thompson 1 98 8 ; C. Lehmann 1 988, 1 989b; Langacker 1 99 1 ) : (a)
"Parataxis," or relative independence, except as constrai ned by the prag
(b)
"Hypotaxis" or i nterdependency, in which there is a n ucleus, and one or more clauses which cannot stand by themselves, and are therefore relatively dependent. However, they are typical ly not wholly i ncl uded with i n any constituent of the nucleus. "Subordination," or, in its extreme form, "embedding," in other words,
matics of "making sense" and relevance.
(c)
complete dependency, in which a margin is whol ly included within a constituent of the nucleus. These cluster points can be characterized by a "cline of clause combini ng": parataxis > hypotaxi s > subordi nation
1 78
7 Grammaticalization across clauses
Th is can further be elaborated by speci ficati on in terms of combinations of the features ±dependent, ±em bedded : I parataxis -dependent -embedded
>
hypotaxis +dependent -embedded
>
subordination +dependent +embedded
In estab l i s h i n g these three cluster poi n ts, we preempt and redefine the term i nology of two trad i tions, and expand two overlapping pai rs i n to a three- way d isti nction. One pair - paratax is versus hypotax i s - derives from a pri mari ly n i neteenth-cen tury trad i tion i n w h ich parataxis was understood to incl ude all ki nds of j uxtapos ition, and hypotax i s to i ncl ude al l ki nds of dependency. The other pair - coordi nation versus subord i nation and especially embedd i ng - deri ves from more recent tra ditions, in w h i c h coord i nation and embedd i n g are defi ned formal ly in terms of constituent structure. The m i n i mal process in cl ause comb i n i ng is u n i fication and bond i ng, at least pragmatical l y. S uch bond ing is often, most especially in the case of subordi nation, accompan ied by hierarch ical downgrading and desentential ization (c. Leh mann 1 988), hence decategorialization of one mem ber of the complex structure into a marg i n . A n u cleus canonical ly con tains a fi n i te verb. Therefore decategorialization typically enta i l s reduction of the fi n i teness of the verb. Of parLicular i n terest to us is the extent to which the c l i ne of dependency m atches up with a c l i ne o f grammatical integration, for example, fi n i teness on the left and non- fi n i teness on the right o f the c l i ne, expressed by clausal rem nants such as i n fi n i tives and participles. A spec ial case of i n tegration i s that in which bond ing bri ngs about what C. Leh mann cal ls "interlac i ng": the sharing of participants (e.g., same subject), or of tenses and moods, and a l so the i n terweavi ng of originally separate clauses into the surface structure of the matr i x (in She seems to be smart, she appears in the matrix but is actually the subject of the embedded clause, as shown by It seems that she is smart). The question is whether the ditl'eren t types of clause combi n i ng are moti vated, and if so, by w hat. G i v6n has suggested that there is a cogni tive form-funct ion parallelism of the fol l owing type: "The more two events/states are i n tegrated se mantical ly or pragmatically, the more will the clauses that code them be i n tegrated grammatical ly" (G iv6n 1 990: 826). This is a statement about diagram matic icon ic i ty as it perta i n s to the overt form that a clause takes, not i ts covert, abstract structure, and can be i l l ustrated by the various forms of complementation in Engl ish. Under most curren t syntactic analyses, the four sentences in (3) i nvol ve a "matrix" and a "subord i n ate" clause at some level of syn tactic abstraction; however, they also show i ncreasi ng degrees of overt morphosy ntactic i n tegration. From a discourse perspective they can also be seen to represent i ncreas ing degrees of connected ness
7. 2 A cline of clause-combining constructions Paratax i s
------
(relmive i ndependence) nucleus
hypotaxis
------
(interdependence)
subordi nation (dependence)
------
minimal i ntegration
------
maxi mal overt l i n k i ng
-------
1 79
marg i n
maximal i n tegration
minimal overt l i nking
Figure 7. 1 Properties relevant to the cline of clause combining
between states of affairs (or, according to Langacker, i ncreased grou nding in the subject's i m mediate experience of the event): (3 )
a. b. c. d.
We real i ze that you have to make a profit. H i s wife only pretended to bel ieve his i mplausible story. Poni a rea lly enjoys wal k i ng along the beach. Nu merous witnesses heard the bomb explode. (Langacker 1 99 1 : 439)
I n other words, the hypothesi s i s that the more overt and i ndependent devices for signal i ng cl ause l i n kage (e.g., c1i tics such as that in (3a)) are correlated with m i n i mal semantic-pragmatic i n tegration, and the least overt (in some languages an i n flectional affi x , in others l i ke Engl ish no marking at a l l ) are correlated with maximal semantic-pragmatic i n tegration . I n t h i n k i n g about the c l i n e of clause com b i n i ng from the poi nt of view of gram l11atical ization, then , we need to think of the con-elations among at least the prop erties in Figure 7 . 1 . 2 We il lustrate some correlations, both weak and strong, between the various factors i m m ediately below, and then go on to d i scuss some examples of the develop ment of com plex sentence structures that i l lustrate the properties in Figure 7 . 1 .
Z2. 1
Parataxis
The si mplest kind of rel ationship between two cl auses is juxtaposi tion i n which two o r more nuclei occur next t o one another a n d the semantic relationship between them i s by i n ference on l y : (4)
Fort Su mter has been tired on. My regi ment leaves at dawn.
The nature of the i n ferential rel ationship between such j u x taposed clauses has been discussed in connection w i th examples (4a) and (4b) in Section 4 . 2 . 3 . Two juxtaposed clauses of this k i n d w i th i ndependent i ntonation contours and without any overt signal of l i n ki ng do not constitute a si ngle complex, and are therefore
1 80
7 Grammaticalizatioll across clauses
not of concern here. However, when juxtaposed clauses are l i nked in some way, such as by i n tonation , or by virtue of shari ng a single lexical noun subject, there is reason to th ink of the two clauses as united grammatically into one sentence by parataxis. Caesar's putative decl aration, already cited in Section 2.3 as an example of diagrammatic icon icity, is a well-known example of parataxis without overt clause linkage: (5)
Ven i , vidi, vici . 'I came, 1 saw, 1 conquered. '
(c. 1 46, Suetonius, J u I . 37)
Each of the three clauses is autonomous. Each constitutes a nucleus. But (5) is considered to constitute one sentence with mUltiple nuclei because the punctuation represents a single unit, that is, a single overarching i n tonation contour. Paratax is is i n many languages a normal way of forming complex sentences. In Chinese, for example, paratactic clauses may function exactly l i ke clauses which in other languages are overtly marked as marg i n and nucleus: (6)
Ta mei nian shu, ta da qiu Ie. he NEG study book he hit bal l ASP ' H e didn't study, he played bal l . ' (Thompson and Longacre 1 985: 1 75)
This sentence would be the usual way in Chinese of say i ng 'Instead of studying, he played bal l ' (in this English version, ' i nstead of studying' i s expressed adverbially, and therefore as a marginal element). In some registers of English, simple juxta position over a single intonation contour is a common way to express conditional sentences: (7)
You keep smoking those cigarettes, you're gonna start coughing again.
and even for relative clauses: (8)
That guy [0] just wal ked out the store remi nds me of the photo in the post-office wi ndow.
Paratactic clauses of this type are often said not to show any overt l inkage mark ers. If this is so, then there is clear evidence for the independence of parataxis and maximal overt l i nking-as characterized in Figure 7 . 1 . A question that still needs to be resolved is to what extent the different intonation contours i nvolved in parat actic clause combin i ng may function exactly l ike overt grammatical markers. For example, (7) is interpreted as a conditional only if the first clause ends in a rising intonation . Otherwise, the construction is considered i ncoherent. 3 It has been cus tomary to treat only explicit con nective words such as and, if, who, or segmental morphology such as cl itic -que, 'and' in Latin, in discussions of clause combining. However, recent interest i n information fl ow and differences between orality and
7. 2 A cline of clause-combining constructions
181
literacy and especially the close study of spoken language i n conversation have led to important developments in understanding the role of intonation as a mor phosyntactic phenomenon (see, e.g . , Bolinger 1 984 ; Chafe 1 98 8 ; Couper-Kuhlen
1 993, 1 996; Ford 1 993; Couper-Kuhlen and Selting 1 996). We w i l l not discuss this issue further here, however, because insights into changes in such contours are hard if not virtual ly impossible to achieve from written texts. Constructions consisting of adjacent nuclei under one sentential i n tonation con tour and having an expl icit connective word such as 'and,' are considered to be more "grammatical ized" than those without, that is, they mark the grammatical relationship. Such clauses are said to be "coordinated": (9) ( 1 0)
I came and I saw and I conquered. Emily is training to be a speech therapist, and Joel works for a law fi rm in Phil adelphia.
Overt clause l inkage markers i n coordinate sentence structures (equivalents of and,
but, or) tend to be developed relatively l ate in languages, or to be borrowed from "high" registers (Mithun 1 988). They are unstable and tend to be renewed fre quently (Meillet 1 9 1 5- 1 6 [ 1 95 8]). The development of plus (originally borrowed from Latin , and recently from mathematical discourse) for and is an example in contemporary American English.
Z2.2
Hypotoxis
The equivalents of the "coordi nated clauses" of modern European lan guages are in many other languages not structured with overt coordinating con junctions, but are ei ther presented as simple paratactic clauses, as i n the case of
(9) and ( 1 0), or as a margin with a nucleus, as in the following examples from Japanese: (I I)
a. Koto 0 nui-de hanga n i kaketa. coat OBJ take: off-de hanger on hung ' I took my coat off and hung it on a hanger. ' b. Wai n 0 nomisugi-te atama ga i taL wine OBJ drink:too: much-te headache S UBJ have ' I drank too much wine and have a headache. '
In ( 1 1 ) the suffix -tel-de on the verb stem signals that the clause in question is interdependent and more marginal than an i ndependent nucleus. However, it is not fully dependent. In other words. it is hypotactic. In many languages of Africa, the Americas, New Guinea, and Asia a construction simi lar to the Japanese -Ie construction is used to l i n k clauses together over long stretches of discourse. This phenomenon is known as "clause chaining." In such chains, usually only one of the
1 82
7 Grammaticalization across clauses
clauses is a nucleus conta i n i ng the fu l l range of verbal markers for tense, aspect, mood, and so on. If the language i s verb-final (OV), the fully marked verb is the last i n the series. (And at least i n some verb- i n i tial African languages, this verb is lhe first in the series; see Longacre 1 990.) Often the verbal i n l1ections in the chai ned clauses (marg ins) are restricted to carryi n g i n formation aboul the participants, not temporal i ty. Thi s i s especially true of New Gui nea languages, where the i n lerior verbs (known as "medial" verbs) are suffixed w i th i nd icators of the person and number of the subj ect of the fol lowing verb but not for tense, aspect or mood . The following is an example from Hua w i th a medial (MED) ending -galla lhat signals switch reference (SW) ( i .e. that the subject of the second verb is not the same as that of the first): ( 1 2)
M i naroga rmu-gana baie. down: lhere go:down-MED:SW stay : 3 SG : NON- FUT '\ went down there and he stayed After/because \ went down there. he (Hai man 1 984: 68) stayed. ' �
Hypotaxi s i s not l im i ted to clause chai n i ng . Among other hypotactic construc tions are apposi t ional relatives in Engl ish . These are semantically and even syn tactically equ i valent to apposi tional (parenthetical, c1arificational) structures, and even to coordi n ate clauses. For example, ( 1 3a) is equ i valent to ( 1 3b): ( 1 3)
a. B i l l Smith, who is our president, would like to meet with you. b. B i l l Smith would like to meet with you. By the way, he i s our president.
Furthermore, the m atri x NP and the relative can be reversed in order, with only m i nor pragmatic d i tTerences: ( 1 3)
c. Our president, Bill Smith, would l i ke to meet with you.
Appositive relatives can even carry their own i I I ocutionary force. that is, they can function as speech acts i ndependent of the speech act of the matri x clause. Th is means that they can fu nction as questions or i m perati ves with i n statements, some thi ng which canon ically embedded clauses cannot do. An example from Lati n i s : ( 1 4)
sunt. quos legite quaeso. Xenophonti s I i bri Peru tiles highly-usefu l Xenophon's books are, which read- I M P/PL ask-ISG. studiose ! studiously ' H ighly u sefu l are Xenophon's books; please read them thoroughly. ' (c. 4 0 B C . Cic .Cat.M.59; ci ted in C . Lehmann 1 988: 1 94)
Other hypotactic constructions include adverbial clauses, i nclud i ng tem po rals ( ' when ' -clauses), causals ( ' because' -clauses), conditional ( ' i f' -clauses), and
7. 2 A cline of clause-combining constructions
1 83
concessives ( ' al though ' -clauses). Example (7) contains two nuclei, and is para tactic. In other, more formal or " l i terate," registers, the same relationsh i p may be indicated with ex plicit morphology : ( 1 5)
I f you keep smoking those cigarettes, you 're going to start coughing agai n.
I n ( 1 5 ) the if-clause cannot stand on i ts own . Trad i tional grammars have treated adverbial clauses as subord i nate or embedded. But as Matthiessen and Thompson ( 1 988) show, the rel ationship of dependency is different from that of the prototyp ical cases of embeddi ng. Specifical ly, there is a semantic difference. For example, in ( 1 6), before leaving is equ ivalent to before his departure, a nominalization that presents an even t as an enti ty, not to before noon, which is an adverbial phrase with a nou n : ( 1 6)
Before leaving Krishnapur, the Collector took a strange decision. (Matthiessen and Thompson 1 98 8 : 1 80)
As we w i l l see below, ad verbial clauses actually arose out of the reanalysis o f adverbial phrases a s adverbial clauses; however, they have not reached the level of i ncorporation that, for example, complements have done. Ad verbial clauses themsel ves show a continuum of looser-to-tighter i n tegration, a con tinuum that correlates w i th their function. Thus an explanatory causal can have i ndependen t i l locutionary force, as in ( 1 7a), wh i le a causal giving the grounds for the conseq uent cannot do so ( I 7b): ( 1 7)
a. The Knicks are goi ng to win. because who on earth can stop Bernard? (G. Lakoff 1 984: 474) b. 'The K n icks are going to w i n because do they have the best players?
Z2.3
Subordination
S u bord i n ate clauses are dependent on their matri x clauses i n various ways. For one, they cannot have d i fferent i I l ocutionary force from the matrix . For another, they are equ i valent to the constituents they express. In addi tion, they typical ly are interl aced i n some way. In the fol lowing example of an Engl ish restrictive rel ative, who just walked out of the store is structurally equ i valent to nom i nal mod i fiers such as the. Furthermore, it is surrou nded by material ti'om the matrix clause (/ think the guy . . . resembles the photo): ( 1 8)
I think the guy who just wal ked out of the store resembles the photo in the post office wi ndow.
In the foll owing exam ple of a Lati n conj unct participial construction, the predica tive adjecti ve phrase patria pulsul1l is equ i valent to a modi fier of A ristides. The
1 84
7 Grammaticalization across clauses
event of Aristides' expulsion is expressed as hierarchically dependent on the object argument (Aristides): ( 1 9)
Aristidem patria pulsum videru nl. Aristides-ACC country-ABL expelled-ACC saw-3PL 'They saw Aristides, who had been exiled.' (Hai man and Thompson 1 984: 5 1 5)
Note that ( 1 9) contrasts with the considerably more hypotactic ablative absolute construction in (20); here there is no relationship between the absolute clause and the verb: (20)
Aristide patria pu lso, Persae Graecos Aristides-ABL country-ABL expelled-ABL, Persians Greeks-ACC aggressi sunl. attacked AUX 'Aristides having been exiled, the Persians attacked the Greeks. ' (Hai man and Thompson 1 984: 5 1 5 .)
Subordinate clauses on the right-hand s ide of the clause-combining cline in Figure 7. 1 function both semantical ly and syntactical ly as expressions of a con stituent. This is particularly obvious in late-stage complementation types, where a clause functions as an argument of the matrix. In (2 1 ), that the Titanic sank is equivalent both to the nominalization the sinking of the TItanic and to the noun phrase the disaster: (2 1 )
That the Ti tanic sank was u nexpected.
7.3
The g ra m m atica l izatio n of cla use l i n kers
The world's languages display a wide variety of techniques for linking clauses into tighter amalgamations. These techniques range from forms and ex pressions that are i ndistinguishable from lexical items, such as ' time,' 'place,' to affixes indicating subordination whose origins are completely u n known. Still, significant generalizations about the origi n s of such forms can be made. We have seen that grammatical ization does not involve special principles that make reference only to morphosyntax, but is a subset of the types of language change in general. The historical development of connectives tends therefore to be similar to that of other lexical items, though their positional and suprasegmen tal features may m ake them subject to special kinds of changes . Clau se li nkage markers have their sources in nouns, verbs, adverbs, pronouns, case morphemes (including prepositions and postpositions), derivational affi xes, and in phrasal com binations of these. Fairly recently, for example, in spoken English the prepositional
7. 3 The grammaticalization of clause linkers
1 85
phrases on the basis (of) and in terms of have come to be used to link clauses together: (22) (23)
He's asked for the special retirement package 011 tile basis he's been with the firm over twenty years. They 're a general nui sance in lerms o/ they harrass people trying to enjoy the park.
Once they are recruited as clause l inkage markers, the originally lexical construc tions typical ly undergo the same types of changes as they would if they were members of a noun-to-affix or of a verb-to-affix cline. Typical of hypotactic developments (though by no means necessary or diagnostic of them) is the recruitment to connective function of deictics and other demon stratives . The motivation here is the extension of deictic reference from entities referred to in the non-li ngui stic world to anaphors and cataphors of NPs and then to anaphors or cataphors of propositions (clauses). In other words, deictics may be used for textual functions i nvolving clause reference in order to achieve overt linking of clauses. For example, in Gunwinggu, an Australian aboriginal language, we find clauses occurri ng where the l inkage is signaled by deictics anaphoric to the preceding clause as wel l as by a single overarching i n tonation contour. In (24)
gunu ' that' refers back to 'our language we write' : (24)
. dja m i : n bu lJadman gadbere gunwo:g gari bi ' bi : mbun, . . . and not i n : regard : to oursel ves our language we-write, . .
gunu gari ' wagan that we-don 't: know but we don ' t know at all how to write our own language' (Berndt and Berndt 1 95 1 : 37; cited in M i thun 1 984: 498)
Such textual functions are de dicto, that is, they refer to parts of the discourse as linguistic forms rather than to the semantic content of the forms (de re). (For the de dicto/de re disti nction, see Frajzyngier 1 99 1 , 1 996.) Sometimes a more explicit deictic phrase is used contain ing a lexical noun rather than simply a pronoun . One example, the development of while, has been discussed in Section 4.3.2. The maj or difference from the development of prepositions dis cussed i n Section 5 . 3 . 1 is that clause linkage markers, when they function i n ways similar to prepositions, in troduce whole clauses, at least at first. In so doing they maximize the constituent with which they are associated. Clause l i n kage markers are in their origins presumably motivated by speakers' desire to be clear and informative, particularly to give directions to hearers for in terpreting clauses i n terms of their lingu istic environment (see Halliday and Hasan 1 976). Initially they serve to signal the functional relationship of the combined
1 86
7 Grammaticalization across clauses
clauses to each other (as in the case of the deictics i l l ustrated above), and to mark syntactic bou ndaries (wh ich m ay be obscured by later sy ntaclic i n terl ac i ng). The cross-linguistic s tudy of clause l i n kage markers and the observalion lhat they tend to fall i n to clearly defi nable semantic-pragmatic sets has led l i nguists recen tly to characterize somewhat more fu lly than i n the past the conceptual and rhetorical functions of many types of clause combi n i ng. For example, i t has been observed (Haiman 1 985a; Traugott 1 985b) that condi tional connectives derive from sources such as the fol lowi ng (exam ples are from Traugott 1 985b) : (a)
(b) (c)
(d) (e)
Forms w i th modal ly related meani ngs, e.g . : suppose (often ("ou nd as the only m arker of the "protasis" or if-c1ause in creoles); Mandari n Chi nese yao 'wish, be necessary, if' ; Mi nangkabau (Sumatra) kil' ' perhaps, maybe i f. ' Interrogatives: Hua -ve ' i nterrogative, topic status, i f' ; Russian esLi < est ' be' + Ii ' whether. ' Temporals expressing duration, or temporals that are am biguous be tween duration and punctuality : H i tti te man 'when, if, potential ' ; Tagalog (ka)pag(ka), kung ' i f, then, whi le' ; Indonesian jika ' i f, when . ' Copul a constructions: Swah i l i i-ki- wa ' i t bei ng that' (-w- ' be ' ) ; Japanese nara ' be,' Chickasaw (h)oo ' be.' Forms signal i ng somethi n g as known or given : Eng l i sh given that, Lat i n a n d Romance s i « sei ' s o ' < ·s 2 n d person deictic + e i localive); Sanskrit ytid ' topic, conditional ' ; Indonesian kalau ' i f, as for. '
Conditionals raise poss ibi l i ties and cast doubt on propositions; therefore the pres ence of modal i ties and i nterrogatives among the sources of cond i tionals seems natural ly motivated by the fu nction of conditionals. The fact that conditionals de rive from d urative temporal relationships and copu las can be seen to reflect the fact that conditionality presupposes an extan t (durati ve) condition. The presence of topic markers among sources for cond i tionals may seem more surpris i ng, however. The recru i tmen t of such topic markers for conditional marki ng, as well as olher factors such as clause order patterns, suggested to Haiman ( 1 978) lhat cond i ti onals are topics. However, the presence of non-topic sources, especially m odali ties, as wel l as the pragmatics of certai n uses of condi tionals, suggests lhat cond i ti onals are only partial l y topics i n function (for d i scussion, see Akatsuka 1 986). The poi n t here is essentially that grammatical i zalion of i tems, whether lexical or morphological , i s constrai ned by the gram matical fu nction to be ex pressed , and by the appropriateness of the i n ferences from the source i tems for the fu nction in question. With sufficient evidence from cross-l i nguistic studies, the researcher can extrapolate pragmatic functions from the forms recru i ted for gram matical
7. 3 The grammaticalization of clause linkers
1 87
purposes. Th is is especially usefu l when new areas of grammatical structure are studied; in recent years, for example, eviden tials have come under scrutiny (see, e.g., L. Anderson 1 986; Chafe and Nichols 1 986; Comrie 2000). Once a form has been recru i ted for a new fu nction, i t w i l l i tsel f be subject to new i n ferences from that fu nction , and these new inferences w i l l in turn perm i t further gram matical ization. Thus i t appears that temporals can be t h e sources of condi tionals (and causals), not vice versa; condi tionals can in turn be the source of concessi ves (as can temporals such as while and focus particles such as even or universal quanti fiers such as any as in anyhow) (Kon ig 1 986; M. Harris 1 988). Thus we fi nd ifbeing used as a concessi ve i n : (25)
This is an i n teresti ng, i f complicated . sol ution.
( Konig 1 986: 239)
Concessive meani ngs develop late i n the history of specific clause l i n kage markt!rs partly because the concessi ve is more abstract, partly because it is more complex logical ly. Once the pri nciple of an extremely close cognitive relationship between the form of a grammatical marker and i ts function has been appreciated, the reasons for both the range and the l i m i tations of the sources for particular clause l i nkage markers should seem quite natural. We take a more detailed look here at some of the sources of complemen tizers. As discussed i n Chapter I , one source ofcomplementizers is verbs of say i ng. The gram matical ization of a verb of say i n g as a complementizer, and the generalization to verbs of desiri ng, feari ng, and other mental states, is very widespread (see, e.g., Lord 1 976; Klamer 2000). The origins of 'say' as a complementizer can perhaps be understood from i ts constant use in discourse in which, for example, wanting, feari ng, wonderi ng, etc . are not purely pri vate matters but are to a great extent the result of public construction of the object through speech. The Greek New Testament has nu merous examples to i l lustrate this, e.g. : (26)
Hoi de esti n houtos hoti anthropoi ethau masan legontes potamos the then men wondered what kind:of is say i ng that this kai hoi anemoi kai he thal assa hupakouous i n auto. and the w i nds and the sea obey him 'A nd people were amazed. say i ng : "What kind of man is this, that the wi nds (Matthew VIII, 27) and the sea obey hi m?'' '
Here legontes 'say i n g ' is a verbal participle that is virtual ly a complementizer of ethaumasall ' were amazed . ' It is as if a person can be general ly in an amazed state of m i nd, bu t as soon as the amazement is attributed to a specific source, it must be articul ated aloud. A modern translation might wel l be: 'And people wondered to one another what kind of man this was, that the wind and the sea obeyed h i m . '
1 88
7 Grammaticalization across clauses
Another source of complementizers is closely associated with nominal forms, hardly a surpri s i ng factor when we consider that complements are essentially clauses functioni n g as NPs. One nomi nal source of complementizers is pronouns. We will discuss the history of pronominal deictic that below i n Section 7 .4. 1 . Other nom inal sources i nclude case morphemes of various types. These may be prepositions or postpositions, including case prefixes and suffixes attached to a verb phrase or nominalized form of the verb. The following is from the Tungusic language Evenk i : (27)
;l n i i - m ;I;I-c;I;I-n saa-r;l si t;ln;lw;I ;lm;l-n;l;l-w;l;l-S. mother-my NEG-PAST-3SG know-(?) you yesterday come-PART-ACC-2SG 'My mother doesn't know that you arrived yesterday. ' (Comrie 1 98 1 : 83)
The accusative case morpheme is added to the resultative participle of the verb come and the participle is seen to be nominal from both the case end ing and the second-person-singular possessor suffix. Genetti ( 1 99 1 ) provides a detailed accoun t of the development of postpositions into subordinators in Newari, and of cross-linguistic evidence of similar changes in other Tibeto-B urman languages. Often the parallelism with an ordinary noun phrase is obvious, as i n i nstances such as the Evenki one in (27), and in English: (28)
a. He left without tell i ng the guide. b. He left without his compass.
Sometimes the similarity is not so clear because the h istorical origins have become obscured. For example, in English the infinitive in to functions as the object of some complement-taking verbs, but the parallelism with the "allative" proposition
to may not seem obvious: (29)
a. We want to ask you a few questions. b. We handed the box to the officer.
Absent historical data, one m ight wrongly assume that the two instances of to were simply fortuitous homonyms. However, because the reanalysis of a dative-allative particle as a complementizer is widespread, it is methodologicall y appropriate to assume that there m ight indeed be some motivation for the form. For i nstance, in Maori ki is both a dative and an allative, and is a complementizer with the same kinds of verbs as Engl ish want, etc . : (30)
a.
Allative
E au ki te kaai nga. hoki ana to the village PRES return PROG 'I am going back to the vil lage. '
7. 3 The grammaticalization of clause linkers b.
1 89
Dative
Ka hoatu te taurekareka ki te rangatira. AOR given the slave to the chief 'The slave was given to the chief. ' c.
Complementizer
E hiahia ana raatou ki te haere. PRES want PROG they COM P the go 'They want to go. '
(Noonan 1 985: 47-8)
Similarly French a in (3 1 ) has its origins in the Latin preposition ad ' to,' which when construed with a verb required the verb to be i n the nominal ized form known as the "gerund." (3 1 )
La musique contribuait a epaissir I' atmosphere. the music contribute-IMPF to thicken-INF the atmosphere 'The music contributed to thicken ing the atmosphere. ' (Simenon ; ci ted in Gami l lscheg 1 957: 462)
In Latin , given the verb root viv- ' live,' the i n fi n i tive vivere and the gerund vivendum could be formed, for example: (32)
Ad beate vivendum virtute opus est. to blessedly I i ve-GER courage-ABL need is 'To live blessedly there i s need of courage. '
(Gamillscheg 1 957: 462)
In Late Lati n ad was already being construed with the infinitive: (33)
Quomodo potest hic nobis camem dare ad how can this-MASCINOM us-DAT meat-ACC give-I N F to manducare? eat-I N F 'How c a n t h i s m a n give us meat t o eat?' ( Vulgate; ci ted i n Gami llscheg 1 957: 462)
Later the i n fi n i tive gai ned ground at the expense of the gerund, and prior to the Old French period had replaced it altogether. The use of an allative-dative marker as a complementizer is common when (as in English want to V, etc.) the tense of the complement is determi ned by the nature of the main clause verb (Noonan 1 98 5 : 47-8), perhaps especial l y when the infinitive suggests an action which is potential or future. It is possible to see in this kind of phenomenon a common tendency toward a mental association of "prospective" space with "prospective" time. The poi nting forward of the allative-dative to the goal is inevitably l inked to the "forwardness in time" of the complement of want. Most verbs expressing commands, requests, expectations, wishes, etc. have as part of their seman tics a goal, whether another person's action, or a desired object (for a cross-l ingu istic study, see Haspelmath 1 989).
1 90
7 Grammaticalizatioll across clauses
Evidence from the history of Engl ish to does in fact i l l ustrate the kind of change i nvol ved , even though i ts origins are far from perspicuous. To originally served as a prepos i tion mean ing ' toward' i ntroducing a nom inalized verb. It was, in other words, com parable w i th the Latin geru nd. The nom i nalized noun was in DE the i n fi n i tive, that is, the verb stem suffi xed with -alll-iall, as in wyrcall 'do, perform , ' i n fl ected w i th t h e dative case marker - e , i .e., to wyrcanne : ( 34 )
ne ()e nan neodpearf ne lrerde to wyrcanne ()ret oa:l pu nor you: DAT no need not taught to perform that lhal you worhtest. performed 'nor did any necessi ty teach you to perform what you performed. ' (c. 8 80, Boelhius 3 3 . 79. 1 6)
In other words, the origins of the POE to- i n fi n i tive lie in a verbal noun in a prepo si tional phrase. The loss in the Middle Engl ish period of the case i n flection, includ i ng case i n fl ection on the i n fi n itive, allowed the to eventually to be reanalyzed as an i n tegral part of the verbal noun, which then could i tself be prefixed with a new prepos i tion for i n certai n lexical environments (see, e.g., Warner 1 982; Los 1 999; Fischer 2000), e.g . : (35)
pe hog l ouyth wei for to be i n a foul place. 'The hog wel l loves to be in a foul pl ace. ' (c. 1 400, Lavynham; ci ted in Visser 1 963-73 : 1 3 1 4)
7.4
Exa m p l es of the d evel o p m ent of co m p l ex sente n ce constructions
We turn now to sketches of a few examples of the development of com plex sentence structures, with the prime purpose of demonstrati ng the d i rectional i ty of the change along various parts of the c l i ne from j u xtapos ition or rel atively loose adjoi n i ng to embedd i ng. For detai led d iachronic studies of one such se quence of c h anges, the development of senten tial complemen tation in Germanic, see Kiparsky ( 1 995), and in the Babylonian d ialect of Akkad ian (one of the earl iest wri tten languages), see Deutscher (2000). Z4. 1
That-complem entation in English
Among standard examples of that-complementations in Engl ish are those that serve as obj ects as i n (36) and those that serve as subjects as in (37 ):
7. 4 Developmen.t of complex sentence constructions (36) (37)
191
B i l l thought that John had left. a. It was obvious that John had left. b. That John had left was obvious.
We sketch here some of the characteristics of the emergent that-com plement struc ture in Old and M iddle Engl i s h . Although t h e earliest records show that OE /)(£t behaved i n some respects like its PDE equ i valent, there are many exam ples of usage that suggest traces of a pronom inal ori g i n and a more hypotactic structure than is characteri stic now (M itchel l 1 985 ; Traugott 1 992). Consider: (3 8)
Sa cy ni nges "egnas on morgenne geh i erdu n "ret "res when/then i n morning heard-PL OEM DEM:GEN k i ng's thanes be ridon se cyn i ng ofslregen wres, ba h i m bere ftan wrerun "ret who h i m beh i nd were was then rode COM P the king slai n hie bider. they thither ' When in the morn i ng the king's thanes who had been left beh i nd heard that he had been kil led. then they rode up there. ' (ChronA (Plum mer) 755.23)
The fi rst pret is a pronoun which anticipates the later complement clause, and the second pret is the complementizer. There i s a paratactic structure to complex sen tences o f th i s kind that may help us understand the fu nctions of the two i nstances of pre(. As is usual in the older texts of OE, the relative clause is not em bedded, but is located after the whole nucleus. What looks l i ke a complement clause is sti l l in part an apposi tional clause resuming the fi rst jJret. The O E tex t i s , then, n o t such a compressed rhetorical u n i t as our translation wou ld suggest. I t is more l i ke the hypotax is i l l ustrated i n (39): (39)
Then i n the morni ng the k i ng's thanes heard th is (these thanes had been left behind earl ier) that the k i ng had been slai n . Then they rode up there.
The pattern of demonstrative pronoun + resumptive demonstrative represented by the pre t + jJret construction makes it poss i ble to l i n k the n uclear clause with i ts verb ' heard ' to the marg i n headed by P(£t, even though the relative clause ' who had been left beh i nd earl ier' i n tervenes . The stage represented by this example is already someth i ng more than merely paratactic, si nce the i n i tial jJret of 'that the king had been slai n ' already marks the clause as a complement (a tru ly paratactic version in OE word order wou ld be: Then in the morn i ng heard the ki ng's thanes this: was the king slai n , ' with adverbial - fi n i te verb - subject order in the I-i rst clause. fi n i te verb - subject in the second).
1 92
7 Grammaticalization across clauses
The pronominal status of the source of complementizer pret is particularly strik ing i n (40): (40)
l>ret gefremede Diulius hiora consul t>ret t>ret angin weaf(� DEM arranged Diulilis thei r consul COMP DEM beginning was tidlice t>urhtogen. in-time achieved 'Their consul Diulius arranged (it) that it was started on ti me. ' (c. 880, Orosius 4 6. 1 72.2)
The preverbal pret in this example is a fronted (topicalized) object pronoun antic ipating the complement introduced by the second pret. (The thi rd pret is simply the quasi-defi nite article with the neuter noun angin ' beginning' .) Such correla tive structures, especially correlatives which mark the beginning of both clauses, and their interdependency, are typical of hypotaxis in OE (and many earlier Indo European texts). Such features are rem iniscent of oral language and of strategies clarifying interdependencies in the fl ow of speech (for oral residue in early Old English prose syntax, see O ' Neil 1 977; Hopper 1 992). The majority of instances of pret-complements in OE are l i ke (4 1 ), however, and do not reveal the pronomi nal origins of the construction overtly : (4 1 )
Dysl i c bia t>ret hwa woruldlice speda forhogie for manna herunge. foolish is COM P someone worldly goods despise for men 's praise ' I t is foolish to despise worldly goods in order to win the praise of men . ' (c. 1 000, IECHom I , 4 60.32)
It appears that the complementizer pret started out as a "copy" in the margin clause of the object pronoun in the nucleus. It was reanalyzed from a pronoun which was a constituent of the matri x clause to a complementizer that had a whole clause within its scope. Example (42) shows the use of pret spreading to non-accusative object environments : (42)
And t>res us ne scamaa na, ac t>res us And DEM : GEN we:ACC not shames never, but DEM:GEN we: ACC scamaa swyt>e t>ret we bote aginnan swa swa bec trecan. shames much COMP we atonement begin so as books teach 'And we are not at all ashamed of that, but we are ashamed of this: of begi nning atonement in the way that the books teach. ' (c. 1 0 1 0, WHom 20.3 1 60)
The impersonal verb scamaiJ ' shames ' in OE requires i ts experiencer argument (the person who is ashamed) to be in the accusative case, and the stimulus of the shame (the thing of which the experiencer is ashamed) to be in the genitive. Consequently, if pret were still analyzed as an argument of the nucleus, it would have to be i n the genitive case, since it is "stimulus" for scamaiJ.
7. 4 Development of complex sentence constructions
1 93
In the following example two features show that the fxet clause is in a hypotactic construction: (43)
. . . pohte gif he h i ealle ofsloge, pret se an ne . . . thought i f he them all slew-SUBJUNCT, that that one not aetburste pe he sohte. escape-SUBJU NCf that he sought ' . . . thought that if he slew them all, the one he sought would not escape.' (c. 1 000, JECHoml 5 82. 1 0)
The reasons for thinking of the fxet-clause here as a complement i nclude: (a)
The clause introduced by fxet is dependent and not appositive, si nce it is in the subj unctive, the "irreal is" mood required by the verb /Whte
(b)
The complementizer is clearly no longer a pronoun. If it were, one would
' thought.' expect it to precede the if-clause that depends on /Whte. However, the second point also shows that the complementizer is associated di rectly with the proposi tion ' the one he sought wou ld not escape,' and is not yet a marker of the whole dependent structure. Another way of stating th is is to say that fxet is not as fully syntacticized as it is in POE. In POE the incorporation of the conditional clause i n to the complement would be favored, as in the modern translation of (43). The complement clauses discussed so far have been object complements. We now turn briefly to subject complements such as are i l l ustrated by POE (44):
(44)
a. It amazes me that they found the purse. b. That they found the purse amazes me.
It has been assumed for OE that complements in impersonal constructions i llus trated by predicate constructions such as (4 1 ) and impersonal constructions such as (42) are subject complements (Lightfoot 1 979). That is, i t is assumed that the subject of ' i s foolish' is the entire clause 'that someone despise worldly goods to win men's praise,' and that the subject of 'shames' is the entire clause ' that we begin atonement i n the way the books teach . ' There is no indisputable evidence, however, that subject complement clauses ex isted in OE. For one thing there was in OE a "heavy constituent constraint" that constituents that were long and full of content words should occu r after l ighter and shorter constituents. There are no examples of complements preceding the matrix verb. Complements in constructions such as (4 1 ), with adjectival or nominal predi cates, may simply have been hypotactic complemen ts of those constituents, just as the fxet-clause is the complement of weddes in (45) :
1 94 (45)
7 Grammaticalization across clauses �onne beo ic gemyndig mi nes weddes pret ic nelle then am I mi ndfu l my-GEN pledge-GEN COM P not-wi l l heonunforrem tune J>e se deada man on I ii) next that homestead S U B that dead man in l i es ' next to the homestead that the dead man lies i n/next to the homestead i n which the dead man lies/which the dead m a n l i es i n ' (c. 880, Orosi u s I 1 .22.2) b. · nyhst J>rem tune on J>e se deada man I i O · · next t o the homestead i n that the dead man l ies'
On a scale of i n tegration as evidenced by marking ofNPrel (as opposed to pos i tion), the personal pronouns are least grammatical i zed, the relative pronouns more so, the subord i nators yet more so. The most grammatical i zed way of marking NPrel along this continuum is zero or "gapping." A topic m uch discussed i n connection with relative clauses is the fact that not all potential NPreis can be relativized in a particular language, and it is pre dictable which NPrels may be relativized from the fu nction i n Srel and from the form of NPrel. Keenan and Comrie ( 1 977) identified a fu nctional "Accessibil i ty Hierarchy" to account for these facts, now known as "the Keenan-Comrie Accessibi l i ty Hierarchy" (for further discussion, see Keenan 1 985 ; Comrie 1 989 [ 1 98 1 )). This hierarchy has the following form in Comrie ( 1 989 [ 1 98 1 ] : 1 56) : subject > di rect object > i nd i rect object > non-direct object > possessor
where possessor (an obl ique relation), is "low" on the hierarchy, and subject (a grammatical relation), is "high" on the hierarchy. If NPrel is expressed by a gap (and/or a subord i nator) low on the hierarchy, then it can general ly relati vize al l higher positions w i th the same form (Keenan 1 985 : 1 54). By contrast, if NPrel has
7. 4 Development of complex sentence constructions
203
the form of a personal pronoun, then only posi tions lower on the h ierarchy can be pred icted . For i nstance, the fact that in Engl ish we can say The woman I took the money from (where NPrel is object of a preposition) al lows us to pred ict that all posi tions to the left on the h ierarchy may be relati vized . Thus we can say The woman I s a w (where NPrel is the object) and (with subordinator) The woman that left, or ( i n some varieties) I kllow a womall will help you . 7 The s ituation where NPrel is expressed by a personal pronoun i s the opposi te: here the fact that OE (and other languages such as Urhobo and Yiddish [Keenan 1 985 : 1 47]) allows subject NPrel pronoun forms predicts that other NPrels to the right on the h ierarchy are allowed to be rel ati v i zed . However, there are exceptions. For example, languages such as Hebrew do not allow subject NPrels to be expressed by personal pronouns, though Hebrew does allow NPrels that are objects or serve other functions on the right of the h ierarchy to be so: (65)
ha-sarim she-ha-nasi shalax otam la-mitsraim . . . the-ministers that-the-president sent them to-Egypt . . . (Keenan 1 98 5 : 1 46) 'the ministers that the president sent to Egypt'
Keenan and Comrie suggest that the accessibil i ty hierarchy "directly reflects the psychological ease of comprehension" ( 1 977: 88). We wou ld rephrase this sug gestion in somewhat d i fferen t terms. We could hypothesi ze that the accessi b i l i ty o f h ighly gram matical i zed forms of NPrel from left to right on the h ierarchy (i .e., the motivation for relativizing first subjects, then direct objects, etc.) i s a function o f the tendency to i n tegrate rel ationships that are frequen tly established, and that fu nctions on the left of the h i erarchy are frequently establ ished because they are natural discourse col locations. On the other hand, what amounts to double marking (by a relativizer and a personal pronoun) i n languages which allow or requ i re the NPrei position to be marked by a personal pronoun i n fl ected for number, case, and gender works from right to left on the h ierarchy, affecting the oblique and object posi tions before the subject, because such extra, complex marking serves the pu rposes of identify i ng relationships that might not otherwise be fully clear to hearers . Th is generalization may hold for other pronomi nal forms of NPrel as wel l . It is a well -known fact of the history of Engl ish that early ME had one relativizer, the su bord i n ator that. When the new pronom i nal relativizer wh- came i n , i t d id not start w i th subjects, but w i th indirect objects and possessives (for a detai led study of the development i n M iddle Scots, see Romaine 1 982), i n other words, on the right of the h ierarchy. Presumably the more di stinctive wh- forms served the purposes of identi fy i ng the less accessible, non-thematic, NPs precisely because the wh- forms were less general i zed in their use, and therefore more i n formative.
204
7 Grammaticalization across clauses
7.5
From com p l ex to s i m p l e clauses The examples of the development of that-complements and of relative
clause constructions have exemplified how paratactic structures may develop through hypotaxis to embedding. Throughout the changes, the clauses serving as (pre)-complement or (pre)-relative have remai ned relatively identifiable as clauses in surface structure. Over time, however, the unification and bonding that occur in this set of changes may eventually lead to further reduction i n to simple clauses, most particularly as exemplified by shifts from head to dependent status, somewhat as i n the case of changes from main verb to auxiliary verb.
ZS. l
From clause chaining to verb inflection in Lhasa
Our first example illustrates the way in which a hypotactic clause-chaining structure develops into an i nflectional one with loss of clause boundary and hence of clausal identity in surface structure. Here we have a case of the development of maxi mal integration . The data are from Lhasa, a Tibeto-B urman language. The analysis and examples are from DeLancey ( 1 99 1 ). Like all Tibeto-Burman languages, Lhasa evidences clause chaining, which expresses sequence of events, as i n (66). One or more verbs with non-final affixes are followed by a finite verb with tense-mood affixes. (66)
Khos l as = ka byas-byas zas -pa red. he: ERG work did-NF ate PERF 'He worked and ate/having worked, he ate. '
(DeLancey 1 99 1 : 9)
The "non-final" (NF) suffix -byas marks the verb as one of a series, and in this respect is comparable with the Japanese te/de suffix mentioned above. The "final" verb, zas (glossed 'ate'), is i n tlected with a fi n i te tense-aspect marker -pa red (historicall y derived from a nominalizer -pa and the copula red). Constructions of the type in (66) express multiple events. However, according to DeLancey, three verbs, 'gro 'go,' yong 'come,' and sdad ' sit, stay,' can occur as finite verbs preceded by verbs with or without the NF marker, and these con structions can express a single event, rather than a sequence of events. In such constructions 'gro expresses not 'go' but distal directionality, yong 'come' ex presses not motion but proximal directionality, and sdad 'sit, stay, ' not stasis but aspectual continuation (as in keep V-ing): (67)
a. Kho 'dir gom = pa brgyab (-byas) yongs- pa red. he here: LOC walked (NF) came PERF ' He walked here. ' (not " He walked here and came. ' )
7. 5 From complex to simple clauses
205
b. Khos las= kabyas (-byas) sdad- pa red. he: ERG work did (NF) stay PERF 'He was working/kept working . ' (not " He worked and stayed . ' ) (DeLancey 1 99 1 : 8-9)
DeLancey argues that because clause chai ning is very frequent in Lhasa, some verbs, such as the three in q uestion, may have come to be used frequently with non-fi nal verbs where a sequenced event interpretation was pragmatical ly ei ther redundant or i mplausible, and hence they came to be reanalyzed not as fu ll verbs, but as markers of aspect. For example, 'gro 'go' is informative as a motion verb i n construction with a non-motion verb such a s 'eat, ' but it is redundant with respect to motion in constructions with a verb of motion such as ' walk,' 'flee. ' In the latter context, motion is demoted, directional ity promoted. The three verbs 'come, ' 'go,' and ' stay' in their non-motion mean ing may cooccur with non-final markers, or they may not, which suggests that they are no longer ful l verbs in these uses. DeLancey calls them "serialized" verbs. Other former ful l verbs, among them tshar ' fi nish,' and myong 'taste, ' often occur as serialized verbs, and in such constructions express perfect. In vol i tional predicates with first-person actors in statements and second-person actors in questions (i.e., in h ighly local contexts based in speaker-hearer reference) they do not cooccur with non-final forms, and they do not have independent word tone. This suggests that they have gone further on the cline of gram maticalization than 'come, ' 'go, ' 'stay. ' Example (68) i llustrates tshar in its three coexistent uses : (a) main verb meaning 'finish' in an event separate from that expressed by the non-fi nal verb, (b) serialized verb meaning ' fi nish' or 'perfect' in an event expressed by the non final (or bare) verb, and (c) grammatical ized suffi x meaning 'perfect' functioning in the tense-aspect slot: (68)
a. Kho phyin-byas tshar -ba red. he went-NF finish PERF 'He went and finished it.' (main verb) b. Kho phyin tshar -ba red. he went finish PERF ' He has gone.' (serialized form, wi thout NF marker) c. Nga krom-Ia phyi n-tshar. I market-LOC went-PER F 'I've gone to the store. ' (affix) (DeLancey 1 99 1 : 1 0- 1 1 )
Yet other former verbs have been morphologized to the extent that they do not only fail to carry their own tone; they also lose nasalization . An example is song [s6] , [so] , [s] 'evidential perfective' < song "which is an old suppletive perfective stem of 'go, ' supplanted in this use in the modern languages by phyill" (DeLancey
1 99 1 : 1 1 ).
206
7 Grammaticalization across clauses
The exam ples of the gram matical i zation of the verbs show a c l i ne of dependency from hypotaxis i n clause chai n i n g to the complete dependency of i n flectional bond i ng. They also show the reduction of complex structure to s i m pler structure, and the accompany i ng total loss of boundary markers, i n other words, integration, CO Ill pression, and bond ing. Like many other examples we have seen i n this book they also i l l u strate the coex istence of d ifferent stages of grammatical i zation (layeri ng), and di vergence i n a highly local context. It should be noted further that although the particulars of the changes i n Lhasa are very d i fferen t from those o f the de velopment of aux i l i aries in Engl ish (Engl ish does not have and never had clause chai n i ng), and the fu ll path from m a i n verb to i n tl ection has not yet occurred (and may perhaps never do so), nevertheless there are s i m i larities: an erstwhile non temporal m a i n verb with a dependent verb has acq u i red temporal mean i ng, and the relationsh i p between the erstwhile main and dependent verbs has changed such that the dependent verb has become the mai n verb or head of the construction. 8 Z S.2
Two conjoined clauses reanalyzed as a single clause
The Lhasa changes renect two widespread phenomena: an erstwh i l e non temporal m a i n verb with a dependent verb has acq u ired temporal mea n i ng, and the original main verb typical l y is drawn from a set of "si mple" verbs such as ' go, come, s i t, l i e, stand. ' Furthermore, the relationsh ip between the original main verb and the dependent verb has changed such that the dependent verb has become the main verb or head of the construction. There are parallels to both the si tuations in Lhasa described in Section 7 . 5 . 1 . In some European l anguages these same verbs have assu m ed a quasi-au x i l i ary status. In the European case, however, the two verbs are arranged paratactical ly, l i n ked by a coord i nat i ng conj unction . Detailed discussion i s found i n Kuteva (2000), from which the fol lowing data and analysis are taken . The typical pattern i s 'SIT (etc.) and V: as i n Dan ish: (69)
Han ligger og k91rer rundt hele nallen he lies and drives round whole night DEF ' H e has been driving all night long' (Braunm U l ler 1 99 1 : 1 03 ; cited in Ku leva 2000: 46)
In th i s Dan ish construction and s i m i lar constructions in some other Eu ropean languages, no semantic sense of ' l ie' remains i n the verb ligger; the meaning is simply ' has been dri v i n g . ' Wh i l e there appears to be no direct evidence of phonological bond i ng or of morphologization i n combinations l i ke ligger og krjJrer ' i s/has been driving' (given that the data do not i nclude i n tonation), syntactic bond i ng is ev idenced by new constra i n ts on the pos i tions of adverbs, which i n the new construction may not i n tervene between the two verbs. Thus i n Middle Dutch,
7. 5 From complex to simple clauses
207
where the construction did not survive i n to the modern language, an adverb could sti l l mod i fy the fi rst verb : (70)
De steden staen dagelicx en de vervallen. The ci ties stand daily and go: to: ruin The cities are fal l i ng to ru i n from day to day. '
( Kuteva 2000: 69)
Here, staen ' stand ' has developed a progressive aspect mean i ng. It has dropped the req u i rement that the subject be hu man , but has not become syntactically bonded to the second verb, as is shown by the presence of the adverb dage/icx 'dai ly' immed iately after staen. Ku teva refers to this as S tage II of a grammatical iza tion chain (c l i ne) of lex ical verb > a u x i l i ary, Stage I bei ng the stage at which the subject must sti l l be h u man. In Bulgarian, which represents S tage II I of the same development, the adverbial expression must appear outside the two-verb conj u nct: (7 1 )
Tri onat lezi i rlizdjasva v mazeto. saw: DEF lie and geLrusly in cel lar The saw is getting rusly i n lhe cellar. '
(KUleva 2000: 70)
In (7 1 ), v mazeto ' i n the cel lar' which seems to belong logical ly with leii ' l ies' appears i nstead as a mod i fier of the whole sequence leii i riiidjasva 'is getting rusty. ' At this point the reanalys is of a two-clause construction as a monoclausal one appears complete, despite the con tin ued presence of the conj u nction eq uivalent to aI/d. It is perhaps no coincidence that these European languages use a general coordi nating conju nction l i ke Engl ish and that serves to join together main clauses. The global distribution of such all-purpose coordinators outside the European area is q u i te l i m i ted (Mithun 1 98 8 ) ; most languages use i n its place some kind of clause-chai n i ng construction s i m i lar to that of Lhasa described above. ZS.3
From main clause construction to sentential adverb in contemporary English
Our last exam ple i s from PDE, and i l l ustrates both integration of structure via a sh i ft from multiclause to si ngle clause structure, and also how a di scourse orien ted statistical analysis can suggest how to recogn i ze possible ongoing gram maticalization. As we have mentioned many ti mes, grammatical ization can be thought of as a form of routinization of language (Haiman 1 99 1 ). A form or a combination of forms occurs in discou rse with increasing ti'equency, and from bei ng an "unusual" way of making or rei nforc i ng a discourse poi nt comes to be the "usual" and unremarkable way to do so. The freq uency with which such ex pressions occur
208
7 Grammaticalizatioll across clauses
will be one factor that determines whether or not they come to be regarded by the speech com m u n i ty as "grammatical" (Bybee and Hopper 2002). (Other factors are, of course, more obviously social: acceptance by broadcast and pri nt media, endorsement by educational and other institutions, and so on.) Our example i s that of the emergence i n PDE of "evidential parentheticals." Thompson and Mulac ( 1 99 1 ) suggest that verbs of proposi tional attitude such as
think and guess w i th first- and second-person subjects are coming to be parenthet icals. Such verbs typically serve to introduce propositions, as in (72), where think is the main verb and the sen tence serves as an assertion that a certain bel ief is held by the speaker or a question concern ing the belief state of the hearer: (72)
a. I think that the coup was planned by the CIA. b. Do you think that the cou p was planned by the CIA?
On the other hand, they may serve to qualify an assertion ; they are then known as parentheticals : (73)
a. I think Commander Dalgleish writes poetry. b. Commander Dalgleish writes poetry, I thi nk.
Here the main verb is writes, and the sentence is a (qualified) assertion about an activity of Commander Dalgleish, not about the state of mind of the speaker. Alternatively, think and guess with second-person subjects may serve to indicate interactive communication: (74)
What's the poi nt of that, do you thi nk?
(fhompson and Mulac 1 99 1 : 322)
In such circumstances, there is no complementizer that and the parenthetical re ceives less stress than the ma.i n verb. Moreover, parenthetical //you think, //you guess have the same syntax as an adverb, in that they are not restricted to one position in the clause. A change of meaning is also noticeable. When it is par enthetical , / think is less certain than when i t is non-parenthetical ; the speaker is not staki ng out an epistemological position, but indicating the degree of vali dation of the statement by suggesti ng that he or she has no direct evidence for it. In other words, Thompson and Mulac suggest, the parenthetical is begin ning to serve the kind of function often served by special ized c1itics and parti cles expressing such modal distinctions as "witnessed," "deduced," "speculative," "hearsay," functions largely expressed in PDE by epistemic modals (e.g., They must be students 'I conclude they are students' ) and by adverbs such as evidently,
apparently, etc. From the perspective of the conti nuum of clause integration, we can see such parentheticals as i nstances of a complex sentence consisting of the nucleus with a verb of propositional attitude and a margin (e.g., that Commander Dalgleish writes
7. 6 Some counterexamples to clause combining
209
Table 7 . 1 Occurrence o/ that with think and guess versus all other verbs -
think
guess other
tha t
622 (9 1 %) 1 48 (99%) 342 (75%)
+ that
To tal
6 1 (9%) 2 ( 1 %) 1 1 2 (25%)
683 1 50 456
Source: based on Thompson and Mulac ( 1 99 1 : 320)
poetry) that has been reanalyzed as a single nucleus. The former margin is now the nucleus, and the former n ucleus has been demoted to something that looks like a sentence adv�rb (comparable with evidently, apparently, etc.). From a quantitative viewpoi nt, think and guess seem to be becom ing distinct from other verbs of propositional attitude such as suggest and believe in that the former occur more frequently without that. Thompson and Mulac's data show the statistics reproduced in Table 7 . 1 . They are also becoming distinct in that they account for 88% of all verb tokens with first- or second-person subjects. In other words, they appear to be becoming specialized (as pas was specialized into the negative in French). At the same time, they are com ing to function as elements in the subjective domain of speaker attitude (as be going to did when it became a tense marker and while when i t became a concessive). Only time will tell whether guess and think will conti n ue along this path of development, and whether they will serve as the model to other verbs of propositional attitude such as believe or suppose at some time i n the future.
7.6
Some cou nterexa m ples to u n i d i rectiona l ity i n clause co m b i n i ng There is substantial evidence that in most languages and most instances
there is a continuum of development from less to more uni fied clause combining, from parataxis > hypotaxis and sometimes > subordi nation . This occurs regardless of literacy, although l iteracy u nquestionably favors hypotaxis and subordination (see Harris and Campbell 1 995 : 308-9). However, there are some counterexamples. A contemporary example is that although is frequently used, especially by college students, as if it were however, that is, as a consti tuent of the matrix clause. Th is could be a hypercorrection resulting from literacy and learned punctuation, and therefore a temporary in novation. But a simi lar development that has withstood the test of a couple of centuries has been noted in Japanese.
210
7 GrammaticaLizatiolZ across clauses
In Modern Japanese, d iscourses of the fol lowi ng kind are possible: (75)
Taro-wa wakai(-yo). Ga, yoku yar-u(-yo). Ta ro-TOP young(-yo). but well do-PRES(-yo) 'Taro is young. But he does a good job. '
(Matsu moto 1 98 8 : 340)
Here the two clauses may have the particle -yo on the pred icate; the mea n i ng of -yo is somet h i ng l i ke ' you know' i n the col loqu ial sense of 'con trary to what you m ight be thinking.' Ga is a weakl y adversati ve conj u nction. An equ ivalent way of sayi ng the same th i ng would be: (76)
Taro-wa wakai -ga, yoku yar-u(-yo). Ta ro-TOP young-ga, well do-PRES(-yo) 'A lthough Taro i s young, he does a good j ob. '
(Matsumoto 1 98 8 : 340)
In (76) ga is not a conj u nction at the begi n n i ng of the second clause, as it is i n (75 ), but a s u ffi x on t h e pred icate of t h e first clause. Moreover, yo, a sentence- (-inul particle, may o n l y appear on the second predicate i n (76). C l early (76) is a s i ngle sen tence, whereas (75) consists of two sentences, an analysis supported by the intonation . The hypothesi s o f u n i d i rectional i ty from parataxis > hypotax is > subordi nation would predict that the type of sen tence i l l ustrated by (76) is deri ved from a construction s i m i lar to that of (75 ). Being more i n tegrative, the s i ngle sen tence wou ld be thought to resu l t from the col lapsing of two separate sen tences, and the reanalys is of ga from an i ndependent conj u nction i n the second clause to a suffi x in the first clause. However, the h istorical data in fact suggest that the reverse occurred . Earl ier Japanese texts poi nt to sen tences of the type i n (76) as an teceden ts of those l i ke (75 ), and i ndeed suggest that the type in (75 ) is relatively recent. The latter start to be recorded i n the seven teenth cen tury, especially in popular play scripts and s i m i lar texts suggesting a col loquial style. Another somewhat s i m i lar i nstance of great i n terest is what has been construed as the development i n spoken German s i nce about 1 980 of main clause (non verb-final) order for clauses marked w i th weil 'because' and obwohl 'al though' (it is, however, possible that this order d i fference existed earl ier but was simply not recorded in wri tten language, or noted by prescri ptivi sts). I n standard German these clauses have verb- final word order, are non-asserti ve, and typically occur with i n the same i n tonation contour as the mai n clause, as in (77). They are there fore considered to be subordinate clauses. In spoken German, however, they now often also occu r with non-fi nal word order, and separate i n tonation con tours ; they are assertive and non-subord i nate, as i n (7 8) (Gu nther 1 996). The d i fference is i l lustrated with weil ' because' :
7. 7 COl/elusiol/ (77)
21 I
au f der andern Sei te war der Kielmann viel leicht besser wei l der on the other hand might the Kielmann perhaps better because he gleich ein Rontgengerat da hat. ri ght an X : ray-mach i ne there has
(78)
'On the other hand Kiel mann might be better because he has an X -ray machine (GU nther 1 996: 325 ) right there. Der hat sicher wieder gesoffen. Wei I sie lauf! total he must surely aga i n been :drinki ng. Because she goes totally depri miert durch die Gegend. depressed through the neighborhood . ' ' He must have been dri nking aga i n because s h e walks arou nd l ooking tota lly (GUnther 1 996: 328) depressed . '
Gunther points out that non-subordi nate clause (paratactic) constructions l i ke (78) involve ei ther epistem ic or "speech act" meani n gs in the sense of Sweetser ( 1 990). As in the case identi fied by Matsumoto, the newer meani ngs and word orders conform to regularly occurri ng unidirectional semantic and pragmatic changes: (75) is more subjective than (76), (78) more subjective than (77), that is, they are more lirmly based i n the speaker's subjective assessment of the s ituation (Traugott 1 989; Traugott and Kon ig 1 99 1 ). These developments do, however, ru n cou nter to the expected i ncrease in clausal i ntegration. The presence of coun terexamples once more shows that the continua of gram matical ization are not exception less. Nevertheless, there is overwhel m i ng evidence for the preponderance of changes from more to less paratactic modes of clause combini ng. Probably all l anguages have paratactic structures . At certain points, given certain trad i tions and discourse uses, dependen t structures are i n novated, and later renewed (M. Harris 1 988).
7.7
Conclusion
In this chapter we have shown how complex clause structure can be modi fied over time. Typical ly the shift, as at the morphological level di scussed in Chapter 6, is from less to more bonded. The fi nal outcome o f this i ncrease i n uni fication m a y b e t h e devel opment o f s i m ple clauses o u t of complex structu res, with an orig i nal verbal construction com i ng to be downgraded into aux i l i ary-like status, or i n to adverbial status.
8
Gra mmatica lization in situations of extreme langu age conta ct 8. 1
I ntro d u ction Most of the work on grammatical ization that we have discussed so far
was conducted withi n the framework of a relatively monogenetic view of change. This approach arose out of a tradition that started with comparative l i nguistics and persisted in the very largely different context of generative grammar. However different they have been, nevertheless both of these traditions have idealized ho mogeneity of language and of transmission, whereas in fact most actual situations involve contact, at the m i n imum with speakers of other dialects, whether social, regional, or sty l istic (see Weinreich, Labov, and Herzog 1 968; Thomason and Kaufman 1 988 for landmark studies of the consequences of heterogeneity for an understanding of how language changes). We turn in this chapter to the question whether studies of contact situations raise special issues regarding grammatical ization (see also Heine and Kuteva 2002). We ignore situations of contact that entai l only partial external i n fl uence on subparts of a li nguistic system . One such situation is that of "borrowi ng," which often involves extensive incorporation of foreign elements i n only one or two areas of the language, typical ly the lexicon, with m i n imal i n fluence elsewhere. Ord inarily the kinds of i tems borrowed are i ndependent words and morphemes (Weinreich 1 953), al though very occasionally morphological paradigms may be borrowed (Thomason and Kaufman 1 988: 20 cite the unusual case of Mednyi Aleut, with Russian fi nite verb morphology but with other largely Aleut grammar and vocabulary). English is an example of a language which has borrowed lexical i tems extensively, from Scandinavian (e.g., give), French (e.g. , table), Latin (e.g., tubular), and many other languages. As a result of the influx of French vocabulary, it has developed a system of stress alternatio n such as is found i n real-reality-realistic, but nevertheless it has u ndergone l i ttle radical syn tactic or morphological change as a result of contact. Most of the seeds of radical restructuri ng in word order and morphology that occurred between Old and Middle English are to be fou nd already in Old Engl ish (though the rapi d i ty with which the changes occurred may well have been triggered by contact). In addition, there are relatively m inor pockets of morphology that show 212
8. 2 Basic characteristics of pidgins and creoles
213
contact influence; see, e.g., Kroch and A . Taylor ( 1 997) on the probable influence of Scandinavian on the spread of third-person-si ngular -s agreement (as opposed to the southern -eth). A second situation that we ignore is one of language admi xture across a wide area, such as is evidenced by the Balkan and the Dravidian languages (see Sandberg 1 968 [ 1 930] and Masica 1 976, respectively). Instead, we take a brief look at pidgins and creoles, languages that evidence extensive influence, indeed "intertwi ning" (Bakker and Muysken 1 995), of two or more languages on each other. Pidgins and creoles illustrate in rather ex treme form a number of theoretical issues, and provide insights into how the progression of grammatical ization across languages as wel l as time can be conceptualized. The study of pidgin and creole languages is especially important for historical linguistics because contemporary varieties are relatively recent in origin (three or four centuries at the most in many cases), and exemplify rapid change in non-literate situations. This is in contrast to most of the more traditional subjects of historical and comparative l inguistics, which may have m i l lennia of history, much of it wri tten and often highly l iterate. Furthermore, the study of pidgins and creoles h as chal lenged many basic assump tions about homogeneity and the role of adults and chi ldren in language change (for .overviews see Bickerton 1 988, Muysken 1 988, Romaine 1 988; for ful ler dis cussion see Hymes 1 97 1 , Valdman and Highfield 1 980, Rickford 1 987, DeGraff
I 999a). As mentioned in Section 3 .2.2, distinctions are often made in h istorical work between "internal" and "external" factors in change. Roughly speaking, "internal" change is associated with child language acquisition in a relatively homogeneous speech community, and "external" change with contact, whether with communities speaking dialects of the "same language" or other languages. This distinction is es sentially similar to the contrast regardi ng the locus of change between m ind/brain and grammar ( "internal") and social interaction and use ( "external"). As has been indicated throughout the book, such a polarization obscures the realities of language change, i n which structure and use, cogn itive and socia\ . factors contin ually interact: "speakers' mutual accommodations can draw materials from either the same l i ngu istic system or separate ones" (Mufwene 200 1 : 1 5). This is nowhere clearer than in pidgin and creole situations, where the notion of "homogeneous speech community" is typically not appropriate, and where, as we will see, second language acquisition plays a significant role in structural innovation . =
8.2
=
Basic cha racte ristics of pidgins a n d creoles A pidgin is usually thought of as a non-native contact language which
typically develops in social situations characterized by major class distinctions and
214
8 Situations of extreme language contact
by numerical disparities between these classes (e.g . , a slave economy in which the number of slaves is far greater than that of the landowners). I It is characteri zed by a relatively si mple grammatical structure. By con trast, a creol e is typica\ 1y, though, as we w i l l see, not necessari ly, a native language. It is more complex than the pidgin to which it is related, but to what extent it is "sim pler" than other types of language depends on its social function and longevity. Often several mutua\1y uninte\1 igible languages are invol ved in the emergence of pidgins and creoles, with one language being used as the socially and pol itical ly prestigious standard . The language of the socially dominant group, e.g., colon ialists or slave traders, is often cal led the "superstrate language," that of the socially subordinate group is often cal led the "substrate language." Examples where the prestige language was Engl ish incl ude West African Pidgin Engl ish, Hawai ian Pidgin Engl ish, Hawaiian Creole English, and Tok Pisin, spoken i n Papua New G u i nea. Examples where the prestige language was Spanish i nclude Palenquero, spoken in Cartagena, Colombia. Those where the prestige language was Portuguese include Cape Verde Creole, spoken in the Cape Verde islands in the Atlantic otT of West Africa. Both Spanish and Portuguese were the prestige languages for Papiamentu , spoken in the Leeward Islands off Venezuela, and French in the case of Haitian Creole (see Holm 1 989 for coverage of creoles world-wide).
8.2. 1
Some characteristics of pidgins
Although the lexicon of a pidgin is mixed , it is usually predominantly derived from the superstrate language, also known as the "Iex i fier l anguage." Pidgin studies have shown that there are several kinds of pidgins, ranging from mini mal comm u n ication devices which are highly unstable, e.g., jargons used in limited domai ns such as trade or labor recruitment, to more complex pidgins arising from relatively stable situations and used i n a wide variety of l i nguistic contexts (e.g. , West African Pidgin English), to extended pidgi n s used as general l i ngua francas (e.g., Melanesian Pidgin Engl ish) (Foley 1 988). Jargons and m i n i mal pidgi ns are characterized by slow speech and ex h i bit only m i nimal morphology. Both may be associated with speakers' u n fam i l i ari ty with the language, especially i n situations where discourse is l i m ited to basic practical affairs. Many pidg i n s are, however, spoken more rapidly, and have more com plex structure. All the same, the following characteristics of pidgins releva n t to grammaticali zation (or absence of it) are often c i ted (see, e.g Romaine 1 988): .•
(i)
a lexicon comprised largely o f the t w o major categories N and V (e.g., sik u sed for ' be sick,' and for 'disease ' )
(ii)
lack of word-formation rules in the lex icon
8. 2 Basic characteristics o/pidgins and creoles
(iii) (iv)
(v) (vi) (vii) (vi i i )
215
peri phrasis (e.g., halls sik 'hospital , ' gras bUong pisill 'feather, ' l i teral ly 'grass of bird ' ) temporal, aspectual, and modal expressions expressed by adverbs or par tic les (e.g., baimbai « by alld by) ' later, future,' pin is « finish) ' fi nished. completed, past,' atink « I think) ' m aybe' ; no consistent means of ex pressing tense, aspect, or moda l i ty absence of i n tlection and allomorphy absence of c1efti ng, topical i zation, etc . , largely resulting from absence of fi xed word order absence of em beddi ng absence o f styl istic variants
Characteri stic ( i ) demonstrates that "essential commun ication" does not fol l ow a pri nc iple of "one mean ing - one form" such as one might associ ate with attention to clari ty for hearers. Rather, speakers econom i ze by using the same lexical i tem for several syn tactic fu nctions and also by using peri phrasis (Miihlhiiusler 1 997 [ 1 986]). · Example ( I ) i s from n i neteenth-century Hawai ian Pidg i n Engl ish, as spoken by a Portuguese imm igrant who was a lighthouse keeper i n Nawi l i w i l i , Kaua ' i : (I)
When time up 1 go to boss and say you take old woman back now. He get verra mad. and make me pay all up. Marie, he good woman . but talk. tal k all ti me. make me mad then pilikia agai n. 'When the time was up 1 went to my boss and said ''Take the old woman back now." He got very mad . and made me pay everyth ing up. Mary, she was a good woman. but she tal ked and tal ked and tal ked all the ti me. and made me mad. Then there was trou ble agai n . ' ( S . Roberts 1 998 : 1 6. citing Vergne 1 898: 1 06)
Pidg i n features observable here i ncl ude (i) absence of tense marking ( When time I go to boss), ( i i ) aspect marking by repetition and adverbial (talk, talk all time), ( i i i ) absence of artic les and other determi ners (to boss), (iv) absence of gender distinction, (v) absence of copula (absence of a be-verb) « i i i )-(v) are i l l ustrated by he good woman); there is also language m i x , as pilikia ' trouble' is Hawai ian. However, most pidgi ns exh i b i t sign i ficantly more complex structure than is illustrated by ( I ). As pidg i n s stab i l i ze and are extended to new uses, they acquire more extensive morphology, most especially predicate markers (PM) which l i n k t h e pred icate t o t h e subject (e.g., i i n (2» , aspectual particles (dOli i n (2» , a n d some hypotactic structures . Th i s suggests attention to expressi v i ty and to hearers' need for clari ty. An example from West African Pidg i n Engl ish w i l l i l lustrate some of the characteristics of a stable extended pidg i n . The passage is excerpted from a narrative about the outwitting of a king by a clever boy called Sense-pass-king or 'Wiser-than-king ' : up
216
8 Situations 0/ extreme language contact
(2)
Sens-pas-ki ng i bin gow, i mas-rut fo rowt, waka Sense-pass-king PM PAST go, PM makes-foot for road, walks trong fo h i l , sowtey i rish fo king i tong. vigorously u p-and-down hills, so-that PM reaches at king his palace King i tok sey, yu don korn. Meyk yu kl i n rna het. king PM talk say, you COM PL come. make you clean my head. Biabia i don plenti tumos fo rna het. Sens-pas-ki ng hair PM COM PL grow too-much for my head. Sense-pass-king i bin don gri sey, i go bap king i het. he PAST COM PL agree say/COM P, PM go barber king his head ' Wi ser-than-ki ng began his journey. Up and down hill he went, and so finally he arrived at the king's pal ace. The king said, "You have come. S have my head (because) my hair has grown too long." Wiser-than-king agreed to cut the king's hair. ' (Traugott 1 976: 7 1 , based o n Schneider 1 966: 1 77)
Note in particular the serial-verb constructions tok sey and gri sey which are the kinds of construction we noted in Section 1 .3.2 in connection with the development in Ewe of the complementizer be. There are no clear instances of embedding, but there is some hypotaxis, as exempli fied by the purpose clause i ntroduced by sowtey 'so that,' and the indirect, third-person quotation
dOli gri say, i go bap king i het
'agreed that he would cut the king's hair' (as opposed to 'agreed "I will cut your hair" '). And, as would be expected from an oral narrative, there is parataxis, as in
king i tok sey, yu dOli kom 'the king said : "You have come" ' ) . The West African construction may have been translated l i terally (i.e. , "calqued") from Akan selsi ' that complementizer, quotative say ' ; or possibly it is a direct borrowing of Akan
se. Another n otable feature is the periphrastic possessive, as in king i tO llg , which uses the PDE order of the inflectional possessive (king 's town), not the periphrastic (town o/king) . The predicate-marker construction with i will be discussed further in Section 8 . 3 . 2 below.
8.2.2
Some characteristics of creoles
By contrast with l imited pidgins, creoles are typically more complex, especially syntactically. However, they may be relatively simi lar to extended pidgins like West African Pidg i n English. Among often-ci ted characteristics of creoles are: (i) Articles: a distinction is made between defi nite referential (3a), i ndefi nite referential (3b), and indefi nite non-referential (3c) (for the differences, see Section 6.4. 1 ): (3)
a. M i bai d i buk. ' I bought the book (that you already know about) . ' b . M i b a i w a n buk. ' I bought a (particular) book . '
8. 2 Basic characteristics a/pidgins and creoles
217
c. M i bai buk. ' I bought a book (or books). ' (Even the speaker does not know/remember speci ficall y which book[s)) (Guyanese Creole; Bickerton 1 977: 58)
(ii) Tense-modal ity-aspect (TMA) systems: periphrastic expressions are widely found. B ickerton (especially 1 984) argues that they are typically sequences of the following type: ± anterior2 tense, ± irreal is modal i ty and ± non-punctual aspect markers, in that order, e.g. : G uyanese Creole bin [+anterior] - go [+irrealis] stei [+non-punctual] V ' would have been V- ing.' Examples from two creoles, -
Saramaccan (Surinam) and Fa d' Ambu (Equatorial G u inea), both with Portuguese lexifiers, and cited in (4a) and (4b) respectively: (4)
df fisi. a. M i bu-o-ta-njan I ANTERIOR-i RREALIS-NONPUNCTUAL-eat the fish ' I would have been eating the fish.' b. Ineni bi ske xa tabaya. They ANTERIOR IRREALIS NONPUNCTUAL work 'They would have been working.' (Thomason 200 1 : 1 74)
While such examples can be multiplied, the view that this tense-aspect-mood sequence is typical of or unique to creoles has, however, been shown to be greatly oversimpl ified. Furthermore the termi nology is somewhat idiosyncratic, and is itself alleged to have contributed to arguments that creoles are very different in kind from non-creole languages. Plag ( 1 994) and Wi nford (2000), among many others, argue that TMA systems in creoles can only be fully understood if they are analyzed in terms of universal cross-linguistic categories such as have been identified by Dahl ( 1 985) and Bybee, Perkins, and Pagliuca ( 1 994), and in terms of their discourse functions in narrative and other discourse genres, rather than i n terms of the features ± anterior, ± irrealis, ± non-punctual. ( i i i ) A d istinction is made between reali zed (-irreal is) complementation as in (Sa) and unreal i zed (+irreal is) complementation as in (5 b): (5)
a. II desid al met posohladah She decided at put fish-in-it 'she decided go put fish i n-it' [inference: she did what she decided to do] b. Li ti pe ale aswar pu al bril lakaz sa he TNS M go one-evening pu at burn house that garsoh-Ia me lorsi me ban day i n fin atke I i . Boy-the b u t on path P L witch COMPL attack h i m 'He would have gone that evening t o burn the boy 's house, b u t o n the way he was attacked by witches. ' (Mauritian Creole; Bickerton 1 98 1 : 60- 1 , citing Baker 1 972)
218
8 Situations of extreme language contact
In (Sa) al «
Fr. alter 'go ' ) impl icates a realized even t in (5 b), pu al «
Fr. pOll
' for' alter ' go ' ) implicates unreal i zed events. (iv) Multiple negation: i n negative sentences, typical ly non-definite subjects, non-defi nite verb phrase constituents and the verb must all be negated : (6)
mi neva s i i notn in dat b i l i n . never see noth ing i n that bu ilding 'I never saw anything i n that bui l d i ng. ' (Guyanese Creole: R ickford 1 987: 1 48 )
( v ) Clause dependency, especially relativization : (7)
mi witnis da wid mi ai wo gaad gi mi . d i i kozn w i tness that with my eyes that God give me . . . the cousin
1
wo swi i t na waan fu guo that i ntox icated not want to go ' I wi tnessed that with my own eyes which God gave me . . . the cousi n who was i ntoxicated did not want to go.' (Guyanese Creole; Rickford 1 987: 1 53 )
Complementation also occurs . In Sranan (the Dutch- a n d Engl ish-based creo le of S urinam), there are three possible complementizers w i th partially d i fferent distributions: taki «
Eng. talk), dat ( < Dutch dat ' that' ei ther in its complementizer
function, or in i ts pronom inal function ), and fu « for unrealized events): (8)
Eng. fo r) (the latter is preferred
Pieter taigi en piki n taki/dati/fu a sa opo na dom. Peter told his child that he IRREALIS open the door (Sranan Creole; Plag 1 99 3 : 1 0 1 ) 'Peter told his child to open the door. '
(vi) Focusing by leftward movement: (9)
Eni k a i n l aengwij ai no kaen spi k gud. any-kind language 1 not can speak well 'There's n o kind of l anguage that 1 can speak wel l . ' (Hawai ian Creol e, 1 896: B i ckerton 1 984: 1 76)
Like pidgins, creoles show a considerable range of elaboratio n , specifica l l y from varieties furthest from the standard (known as "basilects") to varieties close to the standard (known as "acrolects"). This kind of range has been variously descri bed . One model is that of a conti n u u m , with individual speakers having mastery over part of this conti n u u m (see B ickerton 1 975, and , from a somewhat d i fferent view point, Rickford 1 987). The assumption in this model is that the basi lect is closest to the original language. More recently, a consensus has been bu i l d i n g that u n i dimensional continua are n o t supported b y evidence from creole com m u n i ties (see, e.g., Patrick 1 999), and that the basi lect may not represent the origi nal slate
8. 3
Implications of pidgins and creoles for language change
219
o f the language. Rather, i t i s the outcome of symbol ic practices, specifically the sociopol itical construction of identities that are in opposition to l i nguistic assim ilation, or "decreo l i zation," toward the "standard." This "standard is often the lexi lier, and therefore associated with foreign dom i nance and oppression. Speak ers' attention to d isti nctively creole features of languages was first noted by Le Page and Tabouret- Keller ( 1 985). Assu m i n g that the basi lect was the origi nal form of the language, they thought of the phenolTlenpn as refocusing on older forms and called it "recreol i zation ." More recen tly, it has been concluded that ideological ly motivated choice of divergent forms may occur i ndependen tly of the speaker's l i nguistic repertoire, whether basilectal, mesolectal, or acrolectal , (see, e.g., Carri ngton 1 99 3 ; S. Roberts Forthcom i ng).
8.3
I m pl icatio n s of p i dgins and creoles fo r langu age cha nge
In this section we discuss two issues in language change for which pidg i n a n d creole studies have much-discussed impl ications: t h e role of c h i l d versus adult lan.guage acq u isition ( for recent discuss ion see DeGraff 1 999c), and of simpl i fi cation a n d el aboration (see Valdman 1 97 7 ; Miihlhiius ler 1 997 [ 1 986] ; DeGraff 200 1 on elaboration in creoles; McWhorter 200 1 (and peer commentary thereon) on creoles' "si m ple grammars"). These issues are cen tral for language change i n general, a n d there fore are n o t surprisingly crucial for u nderstanding assumptions behi n d claims about gram maticalization in pidgins and creoles. We discuss specific impl ications for gram matical i zation in Section 8.4. 8.3. 1
Child versus adult language acquisition
As we mentioned i n Section 3.2.2, one of the topics that has been of major in terest to historical l i ngu istics has been the role of chi ldren versus adults in lan guage change. The generative position has been that language change i s a change in m i n d/brain states. There is a biologically determ i ned l i nguistic capac i ty with invari ant pri nci ples (known as Universal Grammar or UG) and a number of "param eters" or structural options which chi l dren set, based on evidence from the i nput. Fu rthermore, on this analysis, c h i ldren are the only possi ble i n itiators of reanaly sis. Work ing on Guyanese Creole and Hawai ian Creole, both of which developed in plantation si tuations. B ic kerton (e.g. 1 975, 1 984) put forward the hypothesis that speakers of the pidgi n l ost access to their native languages in circumstances of traumatic d i sconnection from their native countries and communities. such as occu rred in the context of the slave trade, when they had only fleeting access to the l i nguistic repertoires of su perstrate language speakers. The pidg i n they used was
220
8 Situations of extreme language contact
rudimentary, not a full language. Their children therefore grew up in the absence of viable native language models. They rapidly developed creoles, many of which show similar structural properties throughout the world. Si nce these children had no access to native languages and developed with exceptional rapidity, he argued, creoles provide privileged evidence for a "bioprogram" or i nnate human-specific neurological disposition that perm its children who have no extensive consistent language i nput to create a new language out of the bits and pieces of degenerate input they encounter. In an early characterization of the bioprogram, B ickerton ( \ 98 1 : 2 1 2) hypothe sized that it had the following essentially semantic characteristics: (i)
spec ific/non-specific
(ii)
state/process
(iii)
punctuallnonpunctual
(iv)
causative/non-causative
In a later version he hypothesized that the bioprogram had the syntactic charac teristics of a l i m i ted simple clause, one which assigns only subject and object, but no other case markers, and therefore no prepositions (Bickerton 1 984: 1 79). Other characteristics i nclude zero copula. The bioprogram was, however, hypothesized to include serial verbs of the type i l l ustrated in ( 1 0) : ( 1 0)
a. Dei gon get naif pok you. they go get knife poke you 'They will stab you with a knife.' ( 1 896 Hawaiian Engl ish Creole; B ickerton 1 984: 1 75) faka tjoko unu b. Dee o-tei they MODAL-take knife stab you : PL 'They w i l l stab you with a knife.' (Saramaccan Creole; B ickerton 1 984: 1 79)
Later Bickerton put forward the stronger hypothesis that "there is a s ingle set of u n i versal syntactic principles. These principles are absolute and do not u ndergo any form of v �riation, parametric or other" (Bickerton 1 98 8 : 272). A l l variation is, according to this theory, a function of acquisition of lexical i tems and of processes acting on them . In other words, the bioprogram is hypothesized to be neurologi cally far more res tricted than UG; nevertheless, it has much in common w i th that hypothesis. A particularly striking example of the operation of the bioprogram, i t was claimed, is exemplified in Surinam i n the eighteenth century, where after a revolu tion the slaves dispersed into the bush and developed S aramaccan Creole indepen dent of any access to native languages or to a l ocal creole (B yrne 1 987). However, Singler ( 1 992) and McWhorter (2000a) challenged the claim that Saramaccan
8. 3 ImplicatiollS ofpidgins and creoles for language change
22 1
Creole must be assumed to exh ibit bioprogram characteristics, on the grounds that deductions from a variety of textual and comparative evidence suggest that it had its origins in a Portuguese pidgin , mixed with Sranan Creole. The latter had itself emerged from a pidgin in Barbados, ultimately derived from contact languages that developed in Africa (with A kan and Igbo as major donor languages). If indeed a local creole did exist as a model, there is no reason to suppose that the chi ldren of pidg i n speakers had to rely on a bioprogram for the features of this Saramaccan Creole (or any other creole). Another putative situation providing evidence for the bioprogram was, accord ing to B ickerton, to be found in Hawai ' i in the late ni neteenth century, where many mutually unintelligible languages were spoken , and where there was "no preexisting language in common" out of which the creole could arise (Bickerton 1 984: 1 74). Evidence for the development of the bioprogram was alleged to be provided by Hawai ian Creole speakers who at the time of data collection were in their seventies, eighties, and nineties. However, the interpretation of their speech was made on the assumption "that the speech of ind ividuals does not change appre ciably after adulthood is reached" ( 1 974: 1 74), an assumption that appears to have seriously skewed the interpretation of the data. Another assumption was that their parents had no access to their native languages (in this case, especially Japanese or Chinese). Attractive though the concept is of a bi oprogram revealed through creoles, there does not appear to be empirical evidence to support the hypothesis that the prop erties B ickerton identifies are u nique in any extant creole to the extent that they must be assumed necessarily to have arisen out of a bioprogram (see discussion of Bickerton's 1 984 target article and later work such as Rickford 1 987; San koff 1 990; B aker 1 99 1 ; Plag 1 993 ; B ruyn 1 995 ; DeGraff 1 999b,c, 200 I ; Mufwene 200 I ). There may nevertheless be unique characteristics of creoles that arise for other reasons, such as discontinuity of transmission across language-speaking communities. B ickerton proposed that the bioprogram is evidenced most especially in plan tation situations, where work conditions were especially stressful, and access to native language use (sub- as well as superstrate) was minimal . He argued that the sugar plantations of Hawai ' i were one such site. Finding evidence about the languages that children did actual ly have access to is therefore especially important for understanding the condi tions u nder which the creole arose. Basing her work on extensive textual evidence from a variety of Hawaiian solirces including diaries, school records, and newspapers, S. Roberts ( 1 998, 2000a,b) argues that the crucial variable in the development of Hawaiian Engl ish creole is whether speakers are locally born or foreign born. She shows that the locally born group actively con tributed extensively to innovation, while the latter did not. Furthermore, con trary to
222
8 Situations of extreme language contact
B ickerton's hypothesis, it was not first-generation c h i ldren, bilt second-generation locally born child who contri buted most (and most rapidly) to the development of the creole. In other words, there is no evidence even in Hawai ' i for catastrophic, d i scontin uous change between pidgin and creole. S . Roberts (2000a) also shows that English creole i nnovations may have developed among older chi ldren , in school setti ngs, rather than at a very early age, as weI l as among young adu l ts. Studies of Tok Pis in, an Engl ish-related creole in New Gui nea that developed in a situation very di fferent from that of postcolonial plantation s lavery, were con ducted i n the 1 970s (e.g. SankotJ and B rown 1 976; Sankoff 1 980) . In this case, the locus of i n n ovation was found to be adu l ts, and therefore creol i zation was found to be a fu nction not of first but of second language acqu isition. Introduced by traders in the Pacific, Tok Pisin was used primarily as a l i ngua franca to i m prove com m u n ication, especially at a time of developing pol i tical i ndependence, among tri bes whose language varieties had become for the most part mutually i ncomprehensi ble because o f the wild terrai n i n the island and the di l'ficlJ l ty of travel from one community to another. The studies revealed that ad ults were de veloping a nativized , creole version of the l anguage, and suggested that extensive grammatical i zation accompanied this development, e.g. the emergence of relative clauses, c 1 i ti c i zation, and phonological reduction (cf. baimbai ' by-and-by' > boi ' l ater, future ' ) . Al though the rate of development was considerably faster among c h i ldren, the changes themsel ves were shown not to originate with them but with (mostly younger) adu l ts. These studies cal led i n to question the hy pothesis that change is auri butable pri mari ly to very smaIl c h i ldren. To the ex tent that creoles are the res u l t of adu l ts expand i ng the pidgi n , they provide evidence that structural i n n ovations can be made by ad u l ts, a posi tion that is com i ng to be i ncreas i ngly widely accepted . 8.3.2
Simplification and elaboration
B ic kerton's hypothesis that c h ildren develop creoles out of rudimen tary pidgins without access to native l anguages was seen to prov ide exc i t i ng su pport for the generative claim that smaI l chi ldren are the only o nes who can i n novate, reanalyze, and restructure (e.g., HaI le 1 964) . At the same time it cal led i n to question another generative claim: that structural s i m pl i fication or "opt i m i zation" is the natural and i ndeed expected result of c h i l d language acqu isition (e.g. Ki parsky 1 968). The reason for cal l i ng the latter c l a i m i n to question is that the creole is al ways more complex than the pidg i n . Theories of pidgin and creole development often refer t o "pidg i n s i m pl i fication" and "creol e el aboration." There is no q uestion that the lexicon and cod i n g of basic pidgins is relatively s i mple. However, to c haracteri ze the process of pidgi n i zation
8. 3 Implicatiolls of pidgins alld creoles for language change
223
as one of s i m p l i fication can be mislead i ng. Clearly an early stage pidgin is sim pler than any native language. Even though pidg i ns are always regarded as m i xed languages, nevertheless, they are typical ly descri bed i n terms of the lexifier lan guage (e.g. , English, French, Portuguese, German), and "pidgi n simpli fication" usual ly means simpli fication of the lexi fier language. Th is is i n part a practical matter, owing to uncertai nty about which languages were origi nally i nvol ved . I t is, however, problematic because it obscures the extent t -o which the grammars arc thought to be "mixed ." There is sti l l active debate on whether the languages were developed pri marily by speakers of the subordinate languages to fac i l i tate com mun ication among themsel ves because their native languages were often m u tually u n i n telligible (e.g . , Whin nom 1 97 1 , and in a somewhat d i tferent vei n , McWhorter 1 998), or by speakers of the superstrate languages, whether native speakers, or non-native speakers who nevertheless were fluent in the superstrate, for example house slaves (e.g . , Chaudenson 1 995 ; 200 1 ; Mufwene 200 1 ). Another problem is that those who think in terms of "si m p l i fication" of the lex i fier language often think of that language in terms of the standard, rather than the vernacular, i .e., the spoken language of everyday, which is i nevi tably the i nput to a pidg i n . Wrong compari sons may therefore be made because vernacular language is usually less complex than the standard ; the lexicon tends to be more restricted ("hypothesis" and "grammaticalization" are h ard ly vernacular words !), the morphology tends to be less complex (who pronounces al l the consonants in posts, most particu larly the plural marker -s?), and embedd i ng is d i spreferred. In some cases there may even be substantial d i fferences in bas ic clausal structure . As mentioned i n Section 1 . 3 . 3 , Lambrech t ( 1 98 1 ) d iscusses features of non standard French, i ncluding the facts that topic/anti-topic structu res abou nd, and that c 1 itic pronouns have become bou nd to verbs as agreement markers. In ( I I ) we fi nd topic ma femme and the bou nd agreement marker ii, which shows only agreement, not gender (contrast the anaphoric fem i n ine pronoun elle). In ( 1 2) we fi nd the agreement-mark i ng bou nd pronoun ii, and anti-topic (i .e., postposed topic) garrol/ : (I I)
( 1 2)
Ma fem me il est venu. my w i fe AGR has come 'My w i fe has arrived . ' I I -auend devant la porle, I e garfi:0n. AGI�-wait before the door. the boy 'The boy is wai t i ng in front of the door. '
( Lambrecht 1 98 1 : 40)
(Lambrecht 1 98 1 : 74)
Because anti-topic rightward NP sh i ft is freq uen t i n non-standard French, verb-subject order abou nds, in con trast to s tandard French which has subject-verb order, as i n ( 1 3) :
224
8 Situations of extreme Language contact
( 1 3)
Le gar�on attend devant la porte. the boy waits before the door
The so-called predicate markers of West African Pidgin (i in (2» and many other pidgin and creole languages bear a resemblance to the bound pronouns in non standard French illustrated in ( I I ) (and also in vernacular English, cf. Jane, she
called me up this morning). Where mismatches between pidgins or creoles and the lexifier language occur, they may originate in vernacular varieties of the lexifier language, or possibly in the subordinate language, or both, and considerable care needs to be taken in assessing possible origins of any construction. What we can most plausibly say regarding the emergence of a pidgin is that a restricted system is innovated based on the lexicon of the lexifier language, and some pri nciples, probably universal, of m i n i mal grammatical organization . This is basically an abductive process. It suggests that adults can simpl ify their l i nguistic systems, and so cal ls into question the assumption that only small children can achieve th is. The process of pidgin stabilization is a process of complexification of morphosyntactic organ ization as well as of the lexicon . The process of creoliza tion is a significantly more extensive process of expansion and complexification. It appears often to be initiated by older children or adults, and therefore l i kewise challenges the assumption that only small children can do this. In both cases, we need to conceptualize second language acquisition and d iffusion to other second language speakers as the pri me context for the innovations that lead to the new lan guages; however, small chi ldren do u ndoubtedly contribute to the rapid extension of innovations in the creole.
8.4
S pecific i m p l i cati o n s of pidgins and cre o l es fo r g ra m matica l ization S tudies of the m orphosyntactic developments i n pidgins and especially
creoles, in other words of grammatical ization, must be u nderstood in terms of the assum ptions that the researcher bri ngs to bear on the data. As mentioned above, the main issues are who innovates, and how discontinuous the in novation is from structures available in the languages that contribute to the language mix. If one takes the position that creole genesis represen ts a radical discontinuity from the donor languages, then the assumption w i l l be that grammaticalization takes place extensively (and un iquely) in the creoles, independently of developments i n the donor languages. If one assu mes that the lexifier is the m ajor donor, then one will tend to assume that the grammaticalization is d i rectly of the cognate form or construction in the lexifier, usually after creolizatio n . However, if neither assumption is m ade, then the locus of grammaticalization can be pushed back to the
8.4 Implications for g rammaticalizatioll
225
various donor languages, and a relatively continuous development can be posited from the donor to the creole. In this case, a kind of polygrammatical ization (see Section 5 . 3 . 3 ) may sometimes be noted, in which li nguistic forms in the creole and i n varieties of the donor language that are not in contact with the creole have diverged in different directions from a si ngle earl ier form. The approach assuming that innovations start in the creole, or in the transi tion from the pidgin to the creole, can be i l l ustrated by the pipneering work of Sankoff and B rown ( 1 976) on the development of relative clauses in Tok Pisi n. The authors highlighted the use of the marker ia as a "bracket" of syntactic slots where relative clauses, cleft-constructions, and other reference-tracking constructions can occur in hypotactic constructions. An example of the relative construction bracketed by ia is: ( 1 4)
Meri ia [em i yangpela meri , draipela meri ia], em harim istap. G i rl ia [she PM young girl, big girl ia] she listen-OM was 'This girl, who was a young girl , big girl , was l istening. ' (Sankoff and B rown 1 976: 632)
Here a nominal, meri, is presented as a referent whom the hearer is invited to rec ognize or uniquely identify in the context. Although it is separated by intonation breaks (represented by commas), it is not entirely paratactic, since it is intertwined with i n the matrix clause. The relative ends optionally with ia, which serves as bracket of the modifier. The bracketed modifier is fol lowed by a resumptive pro noun which marks the return to the matrix clause. ( 1 5) provides an example of the cleft construction, in which ia is also used : ( 1 5)
Em Iiklik barata i a [mi tok ia] He younger brother ia [I talk ia] 'It's the younger brother I' m talking about. '
(Sankoff and B rown 1 976: 638)
( 1 5) was designed to correct the previous speaker's assumption that the older brother was being talked about. It is a typical contrastive focus construction, distinguishing Liklik barata from the older brother. Sankoff and B rown suggest that ia-bracketing was an innovation among adult speakers of Tok Pisin who were begi n n i ng to develop a creole. They also suggest that it was based on the English lexifier; specifically they regard it as a derivative from here, and note its relationship to, e.g., this man here/this here man, where here also serves an identifying (and sometimes characterizing) function. It also serves as a slot that permits another speaker to in troduce relevant information as speakers and hearers coconstruct pertinent information and check reference-tracking in the course of conversation. The authors note that in the data available to them, prior to 1 9 IO no markers occur of modifier, especially relati ve clause, slots, i.e., the
226
8 Situations of extreme language contact
constructions were pri mari ly paratactic. However, ia became more widely used in the I 940s and I 950s by adu l ts acqu iring a creole. Another study of gram matical i zation in Tok Pis i n (Romaine 1 999) on the de velopment of laik 'want/l i ke/desire' « Eng. like) and klostu ' near' « Eng . close ,, to) into markers for near or "proximative 3 time meaning 'al most, nearly, be about to' , tracks evidence for their h i s tory in the last hu ndred years. Romai ne gives a detai led accou n t of the competi tion between two forms and argues that it supports Bybee, Perki ns, and Pag l i uca's ( 1 994 : 1 2) hypothesis that source meani ngs deter m i ne the semantic path grams can travel. The older of the two forms, klostu, which origi nated i n a spatial preposition or adverbial, came to be grammatical i zed as part of the verb phrase, where it can be used with both later and earl ier time expressions as i n ( 1 6a) and ( 1 6b) respectively : ( 1 6)
a. 01 tok klostu bai wanem ren pundaun nau. they talk cl ose FUT which rai n fal l now 'They said it was about to rai n now. ' b. Em tok 01 masalai kl ostu ki l i m m i . h e s a y P L spirit nearly kil led me 'He said that the spi rits nearly kil led h i m . '
(Romai ne 1 999: 337)
It has largely been replaced by laik in i mmediate future contex ts. Successive stages of laik reveal development of the meani ng ' about to' , auxi l i ation, and phonological reduction ( > lai > la) . An early example of enrichment to proxi mative 'about to' and loss of vol i tional meaning i s : ( 1 7)
Mach i ne he like die. 'The machine i s about to die.'
(Romaine 1 999: 328, citing Downing 1 9 1 9)
The spec i a l i zation of laik correlates with i ts ori g i n i n a verb mean i ng ' want,' which i s fu ture oriented . Laik i tself i s i n competition with bai, but while the latter expresses m ore general fu ture, as m ight be expected from i ts original semantics ( ' by and by ' ) , the former expresses exclusively i m mediate future. These ki nds of analyses envision grammatical i zation as ari sing within a cre ole modeled on the lexi fier. A l though toward the end of their article Sankoff and B rown note that s i m i lar constructions occur in Melanesian l anguages, and s i m i lar ia marki ngs arose in other Melanesian pidgins, nevertheless, they do not ques tion the i n dependent origin of the con"structions in each pidgi n/creole. Li kew ise, Romai ne menti ons that research needs to be done on the poss i bi l i ty 0 1" substrate i n tl uence on the mean i ngs of klostu and laik, but the argu ment concern i n g why and how the competi ng mean i ngs were kept apart assumes Engl ish seman tics. Such approaches have been countered by various versions of a claim that stable pidgins and creoles have the grammar of the subord i nate languages and the lex icon
8. 4 Implicatiolls for g rammaticalizatioll
227
of the lexi fier language. Th is claim was studied in some detail in the context of Melanesia, where severa l pidgins, of wh ich Tok Pisin is only one, have stabilized over the last hundred and fi fty years. For example, Keesi ng ( 1 988, 1 99 1 ) argued that Vanuatu ( B i slama) and S olomons Pij i n provide exceptional ly good insight into the relationship between a pidgin and the donor languages that contribute to it, because the pidgi ns are and have over a long time been in direct contnct not only with Engl ish but with Melanes i an languages. Th is, and the fact that they originated in languages used in nautical si tuations, is in contrast to the pidgins arising i n the Cari bbean duri ng colonial ti mes, which have had far less stable contact with the A frican languages, and were largely developed in si tuations of slavery. Keesing shows that many features of the pidgins can be attri bu ted to the Eastern Oceanic Austronesian l anguages of the area calqued i n to a lexicon and morphology that looks l i ke Engl ish: "from the 1 840's onward, Islanders took the lexical resources of Engl ish and nautical jargon and progressively ham mered them i n to gram matical designs common , at an abstract level , to their native languages" (Keesing 1 99 1 : 3 1 6). I n other words, an Eastern Oceanic Austronesian form is translated i n to a semantical ly roughly eq uivalent form in English, but may serve the grammatical purposes of the Eastern Oceanic Austronesian form . He suggests that there are, therefore, "form u las of equ i valence" that show that the pidgins are less i n novative than is often thought, and more dependent on the extant structure of surround i ng languages. One example is the use of serial constructions, pred icate markers, and object markers. Eastern Oceanic languages typical ly use deictics mean ing ' h i ther' and 'thi ther' in seri al constructions transl atable i nto such Engl ish lexical items as 'bri ng, ' ' take,' and 'ask.' They also use an object marker. B oth of these are i l l us trated by the Kwaio sen tence i n ( 1 8): ( 1 8)
mai. Ori-si-a ask-TRANS-il hi ther 'Ask about i t . '
( Keesing 1 99 1 : 323 )
Kwaio speakers using Solomons Pij i n express the same idea by ( 1 9) : ( 1 9)
Ask-em ka m. ask-OBJ come 'Ask about i t. '
( Keesing 1 99 1 : 323)
Here kam < Eng. come i s eq u i valent to deictic mai ' h i ther' , and em < Eng. him is eq uivalent to the object marker a. Another example is the use o f desltes < Eng. just i n precisely the position in which the local languages use an aspect marker. Kees i ng c i tes (2 1 ) as the Bis lama (pidg i n ) eq uivalent of the Vetmbao (South Malekula) expression in (20):
228
8 Situations of extreme language contact
(20)
Naji nga-mandrxa mu n . h i m he-ASP drink 'He has just drunk.' Em i tes trink. him h e ASP drink 'He has just drunk.'
(2 1 )
(Charpentier 1 979: 353; cited i n Keesi ng 1 99 1 : 326)
(Keesing 1 99 1 : 326)
Especially interesting is the use in Kwaio of aspectual bi 'i ' i m mediate past' with the irrealis/future particle ta to signal immediate future (a conventional ized meaning that cannot be computed directly from i ts parts), as in (22) (INCL is short for "inclusive," i .e. ' we including you ' ): (22)
bi ' i aga-si-a. ta-goru FUT-2PLlNCL just see-TRANS-it ' We ' l l see i t i n a while.'
This is expressed in Solomons Pij i n by bae « (23)
bae i u m i das luk-im FUT 2PLlNCL just see-TRANS
( Keesing 1 99 1 : 327)
by and by) and das/tes, as in (23) : (Keesing 1 99 1 : 327)
Keesing's conclusion is that where both the native languages and the pidgins (or creoles) are in s i tuations of long-term stabi l i ty, there is evidence that formulae of equivalence from the subordi nate language may be as important as, or more i mpor tant than, formu l ae from lexifier languages in the process of grammaticalization. Indeed, "a Melanesian Pidgin grammatical element derived from an English lexi cal form may have acqu i red this grammatical force not through the usual chain of grammatical i zation but through a direct calquing onto a substrate pattern" (Keesing 1 99 1 : 334). Rather than syntactic reanalysis within the relevant languages, there is "a semantic bridge across which a borrowed form can move through the calquing process" ( 1 99 1 : 335). In saying this, Keesi ng conceptualized a rel atively seamless relationship between pidgins, creoles, and their lexifiers, and did not exclude the possibility of i n n ovative grammatical ization. In later studies of Sranan, B ruyn ( 1 995 , 1 996) extended Keesing's conclusion to a situation where the kind of stabi lity he found in Melanesia is not in evidence. She also reconceptualized h i s conclusion in the framework of assumptions that pidgins and creoles show unique structural characteristics and are abruptly discontinuous from their lex i fi ers. For example, Bruyn ( 1 996) argued that i n Sranan preposi tional constructions like na NP baka 'in back of, behind' were probably calqued directly from Ewe and other Niger-Congo languages. She also argued that any grammaticali zation from a relational noun to a secondary adposition to a complex preposition (see Section 5 . 3 . 1 ) originated in one of the donor languages, such as Ewe, not i n the creole. Once calqued from the African language, and part of the
8. 4 Implications for grammaticalizatioll
229
structure of Sranan , it did undergo a small "internal development" to a preposition
lIa baka 'behind' (though primarily in temporal uses). Bruyn's main purpose is a methodological one. Most particularly, she argues that the relative difficulty of obtaining and interpreti ng historical data on creoles may make investigators espe cially prone to attempt to project prototypical trajectories of grammaticalization onto data that is only synchronical ly available. This is wrong, at least for Sranan, she shows, not only because, i f one does the research, �JOe can fi nd historical data going back at least to the eighteenth century, but also beC ause it underestimates the possible role of calquing. These points are very well taken. However, she also sug gests that a special kind of instantaneous grammatical ization needs to be posited for creoles because such grammaticalization as may occur in the history of one of them occurs in a very short time period. Th is claim appears to be u nwarranted because some types of grammaticalization in standard languages are also attested over a short period of time, such as the development of reduced auxiliaries (see McElhinney 1 992 on the sudden attestation of clitic forms like ' ll, 'd, for will and
would in the mid sixteenth century). The extent to which slow emergence can be tracked appears to be in part a function of the avai lability of wri tten materials, and the conservatism of those materials . Any reanalysis is i nstantaneous; gradualness is a function of spread across the linguistic system and across speech communities. The size and fluidity of the community, as well as the discourse types in which language is used will affect the rapidity of the spread. At the time of writing, opi nion appears to be shifting away from B ickerton's claim that pidgins and creoles have unique l i nguistic properties, or reveal unique properties of a bioprogram that is different from UG. Instead, evidence is growing that creolization "can be regarded as a special kind of contact-induced language change, occurring u nder very special social circumstances" (Plag 1 993: 1 5 1 ). As a result, opinion is also shifting away from assumptions of discontinuity between "fully formed" lexifier languages and "simpl ified" languages like pidgin and cre oles (see especially Mufwene Forthcoming). The result is that study of these lan guages (as well as of other contaCt languages) chal lenges the assumption that grammaticalization (or indeed language change in general) occurs in situations of relatively continuous transmission from one speaker to another, and one commu nity to another. Just as auxiliary do underwent different developments in different dialects in English, so particular form-meaning pairs may develop in different ways and at different rates in different creole situations, given different donor languages. The importance of recognizing multiple as wel l as single sources of input to grammaticalization can be seen from the development of creoles such as Sri Lanka Portuguese Creole. Portuguese, the lex ifier language, is VO. Tam i l , the substratum language, is OV. The Sri Lanka Portuguese Creole has taken the resources of the
230
8 Situations of extreme language contact
Portuguese preposition para 'for' and used it as a postposition, with phonological reduction to p' i n an OV structure. Compare: (24)
a.
Portuguese
Eu t i nha dado 0 di nheiro a/para Joiio. , have given the money tolfor John b.
Tamil
Nan call i-yay jon-ukku kutu-tt i ru-nt-an. , money-ACC John- OAT gi ve-PAST be- I -Sa c.
S r i Lankan Portuguese
Ew d i neru jon-pa jaii-dii tina. , money John- OAT give-COM PL gi ve-PAST " had given John the money. ' ( I . Smith 1 987: 8 3 ; ci ted in Romaine 1 988: 40)
The study of the development of m i xed languages demands that more attention be paid to multiple origins of grammatical structures. Contact has been an i m porta n t factor for most languages, a n d a strictly monogenetic view of gram matical i za tion is ultimately inappropriate. At the same time. we agree with Thomason and Kaufman 's cautio n , that borrowing or calqu i ng should be postulated only when a source language a n d a source structure i n t h a t language c a n b e iden ti fied. The identi fication of a source language requires the establ ishment of present or past contact of sufficient i n tensity between the proposed sou rce language and the recipient language . . . The proposed source-language structures need not be. and frequ ently are not. identical to the i n novated structu res in the recipient lan guage, but a successful claim of i n fl uence must of course provide a reasonable account of any rei nterpretation or generalization that has occurred as a result of the interference. (Thomason and Kaufman 1 98 8 : 63-4)
Reasonable interpretations w i l l be based in pri ncipled u nderstandi ngs of the mech anisms and motivations for change including grammatica l i zation.
9
S u m m a ry a n d s u ggestio n s fo r fu rth e r wo rk In this book w e have i n troduced the major theoretical and methodological issues under discussion in work on grammatical ization. As i nd icated in Chapter I , the approach we have taken is two-pronged. Spec i fical ly, we have considered gram matical ization as (i) a research framework for studying the rel ationships between lex ical , constructional , and grammatical material in language, whether diachron ical ly or synchron ical ly, whether in particular languages or cross-li nguist ical ly, and ( i i ) a term referri n g to the change whereby lexical i tems and constructions come in certai n l i nguistic contexts 10 serve grammatical fu nctions and. once gram matical ized , continue to develop new grammatical functions. The term is, however, used in d i fferen t ways by different l i nguists, and we have outl i ned some di fferences in approach, most especial ly in Chapters 2 and 5 . Chapters 3 and 4 d iscussed the maj or mechanisms o f change that c a n lead 10 grammatica l i zation: reanalysis and anal ogy at the morphosyntactic level , concep tual metonymy and metaphor at the semantic, both driven ini tial ly by pragmatic i n ferenc ing. We have argued that grammatical i zation can be thought of as the re sult of the continual negotiation of meani n g that speakers and hearers engage i n i n the con text of discourse production and perception. The potential for grammat ica l i zation l ies in speakers attempti ng to be max i mally informative, and in hearers attem pting to be maximally cooperative, depend ing on the needs of the particular si tuation. Negotiating mean i n g may i nvolve innovation, speci fically, pragmatic, semantic, and u l t imately grammatical enrichment. Th is means that grammatical ization is conceptual i zed as a type of change not limited to early child l anguage acq u i sition or to perception (as is assumed i n some models of language change), bu t due also to adu l t acq u i sition and to production . The question naturally arises how far grammatical ization diachronically con ceived is coextensive w i th language change. As presented here, grammatical ization is a subset of possible changes. It excl udes purely semantic change (e.g. the mean ing change from DE steolj- 'die' to that of POE starve, or from girle 'you th ' to that of POE girl), abru pt word formation (e.g., the nouns window, partner > the homonymous verbs), and purely phonological change (e.g., Gri mm's Law or the Great Vowel S h i ft). However, these may be precursors or by-products of 23 1
232
9
Summary alld suggestions for further work
grammaticalization . It also excludes word-order change, though it interacts ex tensively with it. Crucial ly, grammaticalization i nvol ves morphosyntactic change initiated by and correlated with pragmatic and semantic changes; the develop ment of aux i l i aries, case markers and clause con nectives are classic instances. Grammaticalization may also be associated with morphophonological reduction and bondi ng (Chapter 6), although , as we have seen, th is is not always the case (Chapter 7). S emantic-pragmatic, morphosy ntactic, and morphophonological changes typically result from habituation through frequen t use, which brings about routinization and bleaching (Chapter 5). As i ndicated by the defin ition of grammaticalization as "the change whereby lexical items and constructions come in certain linguistic contexts to serve gram matical functions and, once grammaticali zed, continue to develop new grammati cal functions," grammatical ization is typically unidirectional . It is also gradual, i n two senses: over time (i) a seq uence o f very tiny local structural changes can be seen to emerge, (ii) the frequency with which the new structure is used increases gradually across linguistic types, styles and genres, and speakers. While for the individual speaker i nnovation may be abrupt, change can be considered to occur only when innovations spread to other individuals, and this spread is necessarily gradual. Dramatic changes such as are illustrated by S-curve phenomena are not necessary outcomes of grammatical ization, which may i n fact i nvolve fairly sta ble minor restructuring over long periods of time (see Nichols and Timberlake 1 99 1 ). The so-called "paths" of grammaticalization are schemas that can be seen from the distance of time and from the l i nguist's analytic perspective (Andersen 1 997, 200 1 ). Synchronically they are accessible only indirectly through patterns of frequency, and through generational differences. In many cases these patterns of frequency may remain below the consciousness of speakers, i n others they may acquire social value, whether in identity formation, hypercorrection or some other social practice (cf. Labov 1 972). At any moment synchronically the trajectory of change may therefore look random, but when studied from the perspective of con tinuous segments lasting over the course of a century or several centuries i t may look highly patterned. Indeed, the study of grammaticali zation can be understood as an attempt to d isprove the assumption that changes resulting in grammatical forms are completely random and unpredictable. Because l i n g u i stic change is a social phenomenon, change does not have to oc cur. Whether something is innovated, and who i n novates it, is a matter of chance (a factor which may contribute to the impression of random ness synchronically). It follows that all . change, including grammatical i zation, must be thought of in terms of tendencies, not absolutes. Because speakers m ay preempt elements of language for social purposes, because most societies have complex mixtures of l i n guistic popUlations, and because patterns of grammaticali zation may be renewed,
9
Summary and suggestions for further work
233
reconstructions based on an assumption of unidirectional match ("isomorphism") between cline and direction of change in a specific i nstance should be made with caution and should be framed as testable hypotheses. While much has been learned through the study of grammaticalization in the last twenty years, much still remains to be understood . We have briefly mentioned some of the issues in quantitative studies that can shed further light on the question of exactly how gradualness is to be understood (Chapter 5). We have also looked briefly at some of the issues in creolization, most especially how the l i nguistic changes i nvolved show many of the characteristics of grammatical ization, whether initiated before or after contact (Chapter 8). Several other areas of work have not been considered here. They i nclude patterns of grammaticalization across styles and genres ; the involvement of i nstitutions such as education, language plann i ng, and l iteracy in in i tiating, establishing, or deferring change; and the role of psychological factors such as short- and long-term retention and attention . A ful ler understanding of all the issues suggests that it is necessary to coordinate historical work with sociolinguistics, psychol inguistics, and corpus studies; such coordination should lead to a better understanding of the extent to which the locus of change is to be found in changes in grammars of the speech community, to what extent in the m i nd/brain of speakers, and how changes in language use may lead to changes in the language system.
Notes
Some prel i m i n a ries I . We fol l ow Quirk
el al. ( 1 985 : 2 1 4) in analyzi ng be goillg to as a future rather t han as a "prospective aspect" (Comrie 1 976: 64-5). This is because it has distinct deictic propenies based in the speaker" s point o f view (see Section 4.3.2).
2. Punctuation i n examples from other authors and texts has been standard ized . Dates of periods. authors and works are often not known with cenai nty; dates given in the book are approx i mate only.
3. We owe this example to Tony Hurren. 4. At least the fi rst person is probably maintai ned inferentially i n all examples of lets. 5. This and all other examples from Old Engl ish in this book are cited in the form given in Healey and Venezky ( 1 980).
2
The h i story of gra m m atical izati on
I . T h i s a n d other translations i n t h i s chapter are by Paul Hopper. 2. A third kind o f i conicity mentioned i s metaphorical iconicity, in which a representat ive characteri stic of a referen t i s represented by something else. We may think here of examples like My Im'e is a mse, in which cenain characteristics of love (beauty, d i fficulty, etc .) are represented by the paral lel of a specific flower plant that has special beauty but also thorns.
3. The paniculars of Li and Thompson's analysis of ba have subseq uently been hotl y debated . The relevant issue here is the i mponance the paper had in the history of work on grammat ical ization.
4. For a related v i ew of grams, see Dahl ( 1 985, 2000).
3
Mechan isms : rea na lysis a n d a nalogy
I . In recent l i nguistic theory "rules" have largely been reconceptual i zed as "constrai nts:' " R u les" as speci fied here can also be rethought as constrai nts.
2. Hei ne, Claudi, and Hiinnemeyer ( 1 99 1 a) and Heine ( 1 992) regard the coexistence of older and newer meanings, and s m a l l differences between forms and stages as pecul iar to granullat ical i zation. They are cenai n l y panicularly characteristic of grammatical i zation, but not unique to it.
3. Lord ( 1 976: 1 82 ) s peci fies only mi n i mal reanalysis, consisting of relabe l i n g the syntact ic category : we assume a more signi ficant reanalysis along the l i nes of Heine and Reh ( 1 984: 38). 4. A. Harris and Campbell ( 1 995: 7 1 ) suggest that while there can be no reanalysis un less two anal yses are possible, opacity is not a prerequisite to reanalysis. They give two reasons : i) in some instances two meanings are not available; but the example they give is out o f context and does not account for the kinds of i n ferences that arise i n the flow of speech, ii) the old and the new form coex i s t ; but this is true of all change, at least for a whi le, includ i ng changes they would agree are t he resu l t of
234
Notes
to pages 54-/40
235
opac ity. A better reason for q uestioni ng whether ambiguity i s a prerequisite for reanalysis might be that in some sense every new analogy is a mi nor reanalysis (see Section 3.6) and therefore close s i mi larity in struct ure might be sufficient for reanalysis to take place, not strict ambiguity. 5. Fleischman's translation is 'public matters.' 6. I . RObel1s (I 993a) claims that habeo had a "straightforward future meani ng" even i n the examples ci ted in ( 1 2). However, obl igation mean ing can not be ruled out.
Pragmatic fa ctors
4
I . Wei nreich, Labov, and Herzog ( 1 968) have cal led this the "act uation problem." 2. The process by which i n ferences become polysemies is oftcn called "Iexicali zation." However, the term "Iexicalization" also refers to a di fferent process whereby meaning disti nctions that were formerly ex pressed by i ndependent morphemes come through phonological and other changes to be monomorphemic (e.g., the forms lay versus lie, or set versus sit, are said to be "Iexical ized" as a result of the loss of the former causative morpheme -i- that fol l owed the verb stem) (see Section 5 .7). It has also been used for "the systematic relations between meaning and surface expression" i n a language (Talmy 1 985). To avoid confusion, we prefer to distinguish the processes by which pragmatic i n ferences become polysemies by a d i fferent term, "semanticization." 3. However, for some ( I ndo-)Guyanese speakers, comer has been grammaticalized as a preposition, e.g . He lil'e comer olle rumshop 'He lives diagonal ly across the street from a rumshop' (John Rickford, p.c .). 4. A related use is found i n Hal liday ( 1 994 [ 1 985 ]), who regards expansions of constituents, for example expansion of NP to S (e.g., complementation), as metaphors or abstractions. 5 . The speaker's v iewpoint is m i n i mal, except when directional go i s contrasted with come (He lI' .
goillg to see his Wlll t alld thell comillg to see Ilis sister).
6. The date of composition is a matter of contin ued debate, ranging from and Obermeier 1 997). We accept an eighth-century date. 5
c.
750 to
c.
1 000 (see Bjork
The hypothesis of u n i d i rectionality
I . While the II'ho-lI'holll distinction has become al most entirely obsolete, the clitic -s gen i tive is remarkabl y stable, and i ndeed has been gain i ng ground (see Rosenbach 200 I for discussion of the extension from [+ani mate] possessors to the [-animate] domain in POE). 2. We do not incl ude t he nominal genitive because this is more appropriately defi ned as a c l itic, c f. The mall acmss tIle street 's fellce.
3. Timberlake t ransl ates as ' Does one think this to be a sin?' 4. "The gen itive subject of the participial cl ause i n a system without subject-to-object raising" (Ti mberlake 1 977: 1 47). 5 . Because the proc l i tic n u meral sa- otherwise appears only before classi fiers, such examples may wrongly suggest that it is the head noun which has been elided rather than the classifier. 6. Capitals used for consonants and vowels i ndicate pri mary accent, capitals for vowels only i ndicatc secondary accent. Other prosod ic detai ls have been omitted here. 7. Thanks to Arnold Zwicky (p.c.) for this observation . 6
Cla use-i ntern a l m o rp hologica l changes
I . The term "morphologization" i s someti mes also used for the development of phonological alterna t ions that occur as the result of phonological changes in spec i li c morphological contexts, in other words, of morphophonemic alternations, as found in the Engl i sh past-tense form -ed ([t d 3d]). �
�
236
Notes to pages 152-226
2. The distinction between derivational and i nflectional morphology is not clear-cut, however. For discussion, see B ybee ( 1 985 : Chapter 4). 3. The history of kill is interesting in this connection, since it originates in a causal form of the verb cwe/- 'die,' cwe/-j- 'die + causative.' 4. Different authors use different symbols to indicate the hierarchy, including the leftward arrow
to avoid confusion.
7
Gram matical ization across clauses
I . We use the same features as Foley and Van Val i n ( 1 984: 242), but for a d i fferent analysis. Foley and Van Vali n consider some constructions to be "co-subordinate," that is, +dependentl-embedded. These clause combinations, which are typically instances of switch reference or verb seriali zation (clause combinations which are under the scope of one operator, e.g., declamtive), are more rather than less syntactically bonded than those which are +dependent/+embedded. This is a theoretically problematic analysis, since it treats the two "plus" features as less dependent than one of the possible +I-combinations. We agree with C. Lehmann ( 1 988) that dependency and bonded ness
are
best treated as two separate criteria within the c luster of clause combining properties.
2. A slightly different "constellation" of four properties is suggested in Langacker ( l 99 1 ). Others have suggested a larger number of properties. For example, Haiman and Thompson ( 1 984) suggest seven, C. Lehmann ( 1 988) suggests six. 3. In the 1 950s some education psychologists mistakenly thought that even with the i ntonation rise the construction was i ncoherent. Labov ( 1 969) exploded then-current pedagogical assumptions that because certain A frican-A merican speakers used the intonation pattern and not the segmental if, they did not understand conditional constructions and were "verball y deprived." Clearly, rising intonation has a grammatical function. 4. As a collective, 'tribe,' though grammatically singular, can have plural agreement (cf. i n PDE dialect variants between the committee is . . . and the committee are . . . ). 5. In citing Hittite examples, we follow the transl itemtion given in the source from which the citation is taken.
6. Litemry German, which has preposed, participial relatives, may be an exception to this. 7. Such constructions were used in writing in earlier Engl ish; they have largely d isappeared i n writing, bufare sti ll found in spoken English (Bailey 1 97 3 : 1 65). 8. Zwicky ( 1 993) provides evidence that in English the auxil iary-main verb construction is actually not adequately described as dependent-head; mther, the auxiliary has properties intermediate between dependent and head; this is exactly what one would expect on the assumption that change occurs along a contin u u m in smal l mther than giant steps.
8
G ra m m aticalization in situations of extreme la nguage contact
I . The social contexts for pidgins that arise in tmde situations, are, however, more egalitarian, and commonly involve limited but sustained contact. 2. Anterior tense is also known as "relative past tense" - an event is marked as occurring prior to another, w ithou t respect to speaker ti me. 3. Romaine attributes the term "proxi mative" in this sense to Heine ( 1 994).
References
The publisher has used i ts best endeavors to ensure that the URLs for external websites referred to in this book are correct and active at the time of going to press. However, the publ isher has no responsibi l i ty for the websi tes and can make no guarantee that a site will remai n l i ve or that the content is or will remain appropriate. Abraham, Werner. 1 993. Grammatikal isierung und Reanalyse: ei nander ausschlieBende oder ergiinzende Begriffe? Folia Linguistica Historica 1 3 : 7-26. Adams, Karen L., and Nancy Frances Conklin. 1 973. Toward a theory of natural classi fi cation . I n Claudia Corum , Cedric Smith-Stark, and Ann Weiser, eds., Chicago Linguistic Society 9: Papers From the Ninth Regiollal Meeting, 1 - 1 0. Chicago: Chicago Li nguistic Society. Aij mer, Karin. 1 985. The semantic development of will. In Fisiak, ed. , 1 1 -2 1 . Ai khenvald, Alexandra Y. 2000. Areal typology and grammaticalization: the emergence of new verbal morphology in an obsolescent language. In Gildea, ed. , 1 -38. Akatsuka, Nori ko. 1 986. Conditionals are discourse-bou nd. I n Traugott et ai. , eds., 333-5 1 . Allan, Keith, and Kate Burridge. 1 99 1 . Euphemism alld Dywhemislll: Language Used as Shield and Weapon. New York: Oxford University Press. Al len, Cynthia L. 1 995. Case Marking and Reanalysis: Grammatical Relations from Old to Early Modem English. Oxford: Clarendon Press. Al len, Cynthia L. 1 997. The origins of the "group" gen itive i n Engl ish. Transactiolls of the Philological Society 95: 1 1 1 -3 1 . Andersen, Henni ng. 1 973. Abductive and deductive change. Language 49: 765-93. Andersen, Henni ng. 1 980. Morphological change: towards a typology. I n Fisiak, ed. , I -50. Anderseo, Henni ng. 1 987. From auxiliary to desinence. In Martin B. Harris and Paolo Ramat, eds., Historical Developmelll of Auxiliaries, 2 1 -5 1 . Berl i n : Mouton de Gruyter. (Topics in Linguistics. Studies and Monographs 35) Andersen, Henning. 1 989. Understanding l i nguistic innovations. In Leiv Egi l Brei vik and Ernst Hakon Jahr, eds. , Language Change: Contributions to the Study of its Causes, 5-27. Berl i n : Mouton de G ruyter. (Topics in Linguistics. Studies and Monographs 43) Andersen, Henning. 200 1 . Actualization and the undirectional i ty of change. In Henning Andersen , ed. , Actualization: Linguistic Change ill Progress, 225-48. Amsterdam: Benjamins. (Curren t I ssues in Linguistic Theory 2 1 9) Anderson , Lloyd. 1 986. Evidentials, paths of change, and mental maps: typologically regular asymmetries. I n Wal l ace Chafe and Johanna N ichols, eds., Evidemiality:
717
238
References
the Linguistic Coding of Epistemology, 273-3 1 2 . Norwood , NJ : Ablex . (Advances in D i scou rse Processes 20) Anderson. S tephen. 1 977. On mechanisms by which languages become ergative. In Li. ed . . 3 1 7-63. Anderson. S tephen. 1 993. Wackernagel's revenge: c1itics, morphol ogy. and the syntax of second position. Language 69: 68-98. Anttila. Ra i mo. 1 977. Analogy. The Hague: M outon. (Trends i n Li nguistics. State of the Art Reports 1 0) Antti la. Rai mo. 1 989 [ 1 972]. Historical alld Compa rative Lillguistics. 2nd ed ition. A msterdam: Benjamins [1st edi tion 1 972, New York: Macmillan]. (Cu rrent Issues in Linguistic Theory 6) Archangel i . Diana, and D. Terence Langendoen, eds. 1 997. Optimality TheOl:\,: all Overview. Malden, M A : Blackwel l . (Expl a i n i ng Li ngui stics I ) Atlas. Jay D . , and Stephen C. Levi nson. 1 98 1 . It-clefts, i nformativeness. and logical form. In Cole, ed .. 1 -6 1 . Austin, Peter, ed. 1 988. Complex Selltellce COllstructiolls ill Australiall Lallguages. Amsterda m : Benjamins. (Typol ogical Studies in Language 1 5 ) Axmaker, Shelley, Annie Jaisser, and Helen S i ngmaster. eds. 1 988. lJerkeley Lillguistics Society 14: General Sessioll and Parasessioll all Grallllllaticalizatioll. Berkeley: Berkeley Linguistics Society . Bach. Emmon . 1 970. Is Amharic an SOY language? Journal of Ethiopiall Swdies 7: 9-20. Bach. Emmon, and Robert T. Harms, eds. 1 968. Ulliversals ill Lillguistic TheOl)'. New York: Holt. Ri nehart and Wi nston. Bailey. Charles-Ja mes N . 1 973. The pattern ing of language variation. In Richard W. Bai ley and J ay L. Robinson. eds., Varieties of Presellt-Day Ellglis/l, 1 56-89. New York: Mac m i l lan. Baker. Phi l i p . 1 972. Kreol: a Description ofMauritiall Creole. London: Hurst. Baker, Phi l i p. 1 99 1 . Causes and effects. Joumal ofPidgill alld Creole Lallguages 6: 267-78. Baker, Phi l i p, and Anand Syea, eds. 1 996. Challgillg Meallings, Changing FUllctions: Papers Relatillg to Gralllmaticalization ill COil/act Lallguages. London : UniversilY of Westminster Press. (Westminster Creol i stics Series, vol . 2) Bakker. P and Peter M uysken. 1 995. M i xed languages and language i ntertwi n i ng. I n Jacq ues A rends, Peter M uysken, a n d Norval S m ith. eds . . Pidgills alld Creoles: all Ill1rodllctioll, 4 1 -52. A msterdam: Benjami ns. (Creole Language Li brary 1 5 ) Baldi . Phi l i p . 1 976. The Latin i m perfect i n * ba . Lallguage 52: 839-50. Barcel ona, A nt o n i o. ed. 2000. Metaphor and Metollymy at the Crossroads: a Cogllitive Perspective. Berl i n : Mouton de G ruyter. (Topics in Engl ish Lingui stics 30) Barth-Weingarten, Dagmar, and Eli zabeth Cou per- Kuhlen . 2002. On the development of fi nal though: a case of grammatical i zation? I n Wi scher and Diewald. eds., 363-78. Bates. El i zabeth. and Brian MacWhinney. 1 989. Fu nctional i s m and the competi tion model. In B ri a n M acWhi nney and El i zabeth Bates. eds. , The Crosslillgllistic Study of Sentellce-Processillg, 3-73. Cambridge: Cambridge University Press. Beals, Kathari ne. Jeannette Denton, Robert K n i ppen, Lynette Melnar, Hisami Suzuki. and Erica Zei n feld, eds. 1 994. Chicago Linguistic Society 30: The Parasessioll all \i,ri· aliOIl ill Lillguistic Theory. Chicago: Chi cago Lingu i stic Society. Bean, Mari a n C. 1 98 3 . The Developmellt of Word Order Pal/ems ill Old Ellglish. London : Croom Hel m . .•
239
References
Benveniste. Emile. 1 968. Mutations o f l i nguistic categories. I n Leh mann and Mal kiel . eds. . 85-94. Berndt. C. H and R. M. Berndt. 1 95 1 . An Oenpel li monologue: cul ture-contact. Oceania 22: 24-49. Bickerton . Derek. 1 975. Dynamics 0/ a Creole !:J)'stem. Cambridge: Cambridge University Press. Bickerton. Derek. 1 977. Pidginization and creol izatio n : language acquisition and lan guage universal s . I n A l bert Valdman. ed . Pidgin and C,:eole Linguistics. 49-69. Bloomi ngton: I ndiana Uni versity Press. Bi ckerton. Derek. 1 98 1 . Roots 0/ Language. A n n A rbor: Karoma. Bickerton. Derek. 1 984. The language bioprogram hypothesis. Belia vioral and Brain Sciences 7: 1 73-22 1 (target article with peer com mentary). Bickerton. Dere k . 1 988. Creole languages and the bioprogram. In Newmeyer. ed vol . 2: 268-84. Binnick. Robert I. 1 99 1 . Tillie and tlie Verb: a Guide to Tellse and Aspect. New York: Oxford University Press. B i rner. Betty J and Gregory Ward. 1 998. In/ormatioll Statlls and Noncanonical WOI1I Order in Englisli . A m sterdam : Benjamins. (Studies i n Language Companion Series 40) B isang. Wa l ter. 1 996. Areal typol ogy and gram maticalization: processes of grammatical ization based on nouns and verbs i n East and Mai n land South East Asian Languages. Studies in Language 20: 5 1 9-98. Bjork . Robert E and Anita Obermeier. 1 997. Date, provenance, author. audiences. I n Robel1 E . Bjork a n d J o h n D. N i les., eds., A Beowulf Handbook, 1 3-34. Li ncol n : University o f Nebraska Press. B l akemore. Diane. 1 987. Semantic COllstrailils 0 1 1 Relevance. Ox ford : Blackwe l l . Blakemore. Diane. 1 990. Constraints i n interpretation. I n Kira, H a l l , Jean-Pierre Koen ig. Michael Meacham. Sondra Rei n man, and Laurel A . Sutton , eds.. Berkeley Linguis tics Society 16: General Session and Parasession oll tlie Legacy o/ Grice. 363-70. Berkeley: Berkeley Linguistics Society. Boersma. Pau l . and B ruce Hayes. 200 I . Empirical tests of the Gradual Learn i ng A l gorith m. Linguistic Inquiry 32: 45-86. Bol i nger. Dwight. 1 975. Aspects 0/ Language, 2nd ed ition. New York : HarcoU lt B race Jovanovich. Bol i nger. Dwight. 1 977. Meaning and Form. London: Longman. ( Engl ish Language Series I I) Bol i nger. Dwight. 1 984. I n tonati onal signals of subordi nation. I n B rugman et a l. . eds . . 40 1 - 1 3 . Bossong. Georg. 1 985. Elllpirisc!,e Universaliell/orsc!wng: dijJerentiel/e Objektmarkienlllg in den neuiraniscliell Spracliell. Tiibi ngen : Giinter Narr. (A rs Linguistica. Com men tat iones Analyticae et Cri ticae 1 4) Boy l and. Joyce. 1 996. Morphosyntactic Cha nge i n Progress: a Psychol inguistic Approach. Unpubli shed PhD di ssertation. University of Cal i fornia, Berkeley . Boyland. Joyce. 200 1 . Hypercorrect pronoun usage in English'! Cognitive processes that accou nt for pronoun usage. In Bybee and Hopper, eds . . 383-404. Brau nmiiller. Kurt. 1 99 1 . Die skandinavisclien Spracliell illl Oberblick. Tii bi ngen : Francke Verl ag. .•
.
.•
.•
.•
240
References
Breal, M ichel. 1 99 1 [ 1 882] . George Wolf, ed . and translator, The Beginnings ofSemalltics; Essays, Lectures alld Reviews. Stanford : Stanford University Press. Bresnan, J oan, Shi pra Di ngare, and Christopher D. M anning. 200 1 . Soft constrai nts mirror hard constraints: voice and person in English and Lum m i . In Butt and Hol loway Ki ng, eds. , 1 3-32. Brinton, Laurel. 1 988. The Development of Ellglish Aspectual Systems: Aspectualizers alld Postverbal Particles. Cambridge: Cambridge University Press. (Cambridge Studies in Lingui stics 49) Bri nton, Laurel. 1 996. Pragmatic markers ill English. Grammaticalization and Discourse Functions. Berl i n : Mouton-de Gruyter. (Topics i n English Lingui stics 1 9) Bri nton, Laurel . 2002. Grammaticali zation versus lexicali zation reconsidered: on the "late" use of temporal adverbs. In Teresa Fanego, Marfa Jose L6pez-Couso, and Javier Perez-Guerra, eds. , Ellglish Historical Syntax and Morphology: Selected Papers
from the 1 1th Internatiollal COllference 011 English Historical Linguistics, Santiago de Compostela, 2000, 68-97. Amsterdam and Phil adelphia: Benjamins. The British National Corpus (BNC). 1 994. Oxford : Oxford Universi ty Press. (On-l i ne corpus
of spoken and British English, http://www.hcu.ox.ac.uklBNClindex.html) B rowman, Catherine P. , and Loui s M . Goldstein. 1 992. Articulatory phonology : an overview. Phonetica 49: 1 55-80. B rugman, C l audia, and Monica M acaulay, with Amy Dahlstrom, Michele Emanation, Birch M oonwomon, and Catheri ne O' Connor, eds. 1 984. Berkeley Linguistics Society 10: Proceedings of the Tenth A nllual Meeting of the Berkeley Linguistics Society. Berkeley : Berkeley Linguistics Society. Bruyn, Adrienne. 1 995. Grammaticalization in Creoles: the Development of Determiners and Relative Clauses ill Srallall. Amsterdam: Insti tute for Functional Research into Language and Language Use (IFOTr). Bruyn, Adrienne. 1 996. On identifying i nstances of grammaticalization i n Creole languages. In Baker and Syea, eds. , 29-46. Burridge, Kate. 1 998. From modal aux i l i ary to lexical verb: the curious case of Pennsylvania German wolle. I n Richard M. Hogg and Linda van Bergen, eds., Historical Linguis tics 1 995. Selected Papers from the 1 2th International Conference 011 Historical Linguistics, vol . 2: Germallic Lillguistics, 1 9-3 1 . A msterdam Benjami ns. Burrow, T. , and S. Bhattacharya. 1 970. The Pengo Language: Grammar; Texts, and Vocab ulary. Oxford: Clarendon Press.
Butt, M iriam . 200 1 . A reexami nation of the accusative to ergative shift in I ndo-Aryan. In Butt a nd K i ng, eds. , \ 05-4 1 . Butt, Miri a m , and Tracy Holloway King, eds. 200 1 . Time over Maller: Diachronic Perspec tives on MorphosYlltax. S tanford University: CSLI Publications. Butt, Miri a m . Forthcoming. The light verb jungle. In Claire Bowern and Conor Quinn, eds., Harvard Working Papers ill Lillguistics 8 . Bybee, Joan L. 1 985. Morphology: a Study of the Relation between Meaning and Form. Amsterda m : Benjamins. (Typol ogical Studies i n Language 9) Bybee, Joan L. 1 988. Semantic substance vs. contrast i n the development of grammatical mea n i ng. In Axmaker et al. , eds., 247-64. Bybee, Joan L. 1 990. The semantic development of past tense modal s in English. In Wol fgang W5Jck, Betty L. Brown, and Dan Devitt, eds. , Buffalo Working Papers in Linguistics 90-9 J , 1 3-30. Special issue for Paul Garvin .
References
24 1
Bybee, Joan L. 1 995. Regular morphology and the lexicon . Language and Cognitive Pro cesses 1 0 : 425-55 . Bybee, Joan L. 200 1 . Frequency effects o n French liaison. In Bybee and Hopper, 337-60. Bybee, Joan L. 2002. Phonology and Lallguage Use. Cambridge: Cambridge University Press. (Cambridge S tudies i n Li ngui stics 94) Bybee, Joan L., and O sten Dah l . 1 989. The creation of tense and aspect systems in the languages of the world. Studies in Language 1 3 : 5 1 - ) (.)3 . Bybee, Joan, J o h n Haiman, a n d Michael Noonan, eds. 20t)J . Complex Sentellces ill Grammar and Discourse. Studies Presellted to Sandra Thompson. Amsterdam: Benjamins. Bybee, Joan, and Pau l Hopper, eds. 2002. Frequency and the Emergence of Linguistic Structure. A msterdam: Benjamins. Bybee, Joan L., and Wil l i a m Pagliuca. 1 985. Cross-l i nguistic comparison and the develop ment of grammatical meaning. In Fisiak, ed. , 59-83. Bybee, Joan L., and Wi l l i a m Pagli uca. 1 987. The evolution of future meaning. I n Giacalone Ramat et aI. , eds., 1 08-22. Bybee, Joan L., Wi lliam Pagliuca, and Revere D. Perkins. 1 99 1 . Back to the future. In Traugott and Heine, eds., vol. 2: 1 7-58 . Bybee, Joan, Revere Perkins, a n d Wi lliam Pagliuca. 1 994. The Evolution of Grammar: Tense, Aspect; and Modality in the Languages of the World. Chicago: Universi ty of Chicago Press. Bybee, Joan L., and Joanne Scheibmann. 1 999. The etl"ect of u sage on degree of consi tituency : the reduction of don 't i n A merican English. Linguistics 37: 575-96. Bybee, Joan L., and Dan I. S iobin. 1 982. Why small chi ldren cannot change language on their own. In A nders Ahlqvist, ed. , Papersfrom the 5th International Conference 011 Historical Linguistics, A msterdam: Benjamins (Current Issues in Linguistic Theory 2 1 ) 29-38 . Bybee, Joan, and Sandra Thompson. 1 997. Three frequency effects i n syntax. In Matthew L. Juge and Jeri L. Moxley, eds. , Berkeley Linguistics Society 23: General Session and Parasession on Pragmatics and Grammatical Structure, 65-85. Berkeley: Berkeley Linguistics Society. Byrne, Francis. 1 987. Grammatical Relations ill a Radical Creole: Verb Complemelltatioll in Saramaccan. A msterdam : Benjamins. (Creole Language Li brary 3) Cameron, Angus, et al. , eds., 1 986. Dictionary of Old English. Toronto: Ponti fical Institute of Medieval Studies. (http://www.doe.utoronto.calabout.html) Campbel l , Lyle. 1 99 1 . Some grammatical ization changes i n Estonian and thei r i mpl ications. I n Traugott and Heine, eds., vol . I : 2 8 5-99. Campbel l , Lyle. 1 999. Historical Linguistics: all IlIIroductioll. Cambridge, MA: M IT Press. Campbel l , Lyle, ed. , 200 1 a. G rammaticalizati on : a critical assessment. Lallguage Sciellces 23, numbers 2-3 . Campbell, Lyle. 200 1 b. What's wrong with grammaticalization? I n Campbel l , ed. : 1 1 3-6 1 . Carri ngton, Lawrence D . 1 993. I mages of creole space. loumal of Pidgin and Creole Languages 7 : 93-9. Chafe, Wal l ace L. 1 970. Meaning and the Structure of Lallguage. Chicago: University of Chicago Press.
242
References
Chafe, Wal l ace L. 1 988. Linking i ntonation un its in spoken Engl ish. In Hai man and Thom pson, eds., 1 -27. Chafe, Wallace L. 1 994. Discourse, Consciouslless. alld Tillie: the Flow and DisplacelllellI oj Consciolls Experiellce in Speaking alld Writing. Chicago: University of Chicago Press. Chafe, Wallace. John Du Bois, and Sandra Thompson. 1 997. A proposal for a Corpus of Spoken American English. I n Kari n A ij mar and Bengt Alytenberg, eds Ellglish COl7Jl1S Unguistics: Studies in Honor of Jan Sl'artl'ik, 64-82. London: Longman . Chafe, Wallace, a n d Johanna Nichols, eds. 1 986. El'identiality: tile Unguislic Codillg of Epistemology. Norwood, NJ : Ablex . Chambers, J. K . 2003. Sociolinguistic TIleOl:V: Unguistic Variatioll and its Social Sign({1callce, 2nd edition. Malden, MA: Blackwel l. Charpenti er, Jean-Michel. 1 979. Le pidgin bislallla(n) et Ie lIIultilinguisme aux NOIlI'elles Hebrides. Paris: Societe d' etudes I i ngu isti ques et anthropologiques de France. ( Langues et civilisati ons a tradition orale 35) Chaudenson, Robert. 1 995. Les Creoles. Pari s: Presses Universitaires de France. Chaudenson. Robert. 200 1 . Creolizatioll of Language and Culture. Revi sed in col labo ration with Sal i koko S. M u fwene; trans. by Sheri Pargman et al. London, New York: Routledge. (Orig. publ . as Des fles. des homilies. des langues: essai sl/r la creolisation lillguistiqlle et culturelle. Pari s: Harmattan. 1 992) Chomsky, Noam. 1 965. Aspects oJ the TIleory of SYllfax. Cambridge. M A : M IT Press. Chomsky, Noa m . 1 98 1 . Lectures on GOl'el'll l llent and 8inding. Dordrecht: Fori s Pu bl ica tions. ( Studies in Generati ve Grammar 9) Chung, Sandra. 1 977. On the gradual nature of syntactic change. In Li. cd . . 3-55. Clancy, Patri c i a , Sandra Thompson. Ryoko Suzu k i , and Hongyin Tao. 1 996. The con versational use of reactive tokens in English. Japanese, and Mandari n. JOl//'llal of Pragmatics 26: 355-87. Clark, Eve V. 1 978. Locationals: exi stential, locative. and possessive constructions. In G reenberg et al. , eds., vol. 4: 85- 1 26. Claudi, Ulrike. 1 994. Word order change as category change. In Pagl iuca, cd. , 1 9 1 -23 1 . ClilUdi, Ulrike, and Bernd Heine. 1 986. On the metaphorical base of gram mar. Studies ill Language 1 0: 297-335. Cole, Peter. 1 975. The synchronic and diachronic status of conversational i mplicature. I n Cole and Morgan , eds 257-88. Cole, Peter, ed. 1 98 1 . Radical Pragmatics. New York: Academic Press. Cole. Peter, a n d Jerry L. Morgan . eds. 1 975. Speech Acts. New York: Academic Press. (Sy ntax and Semantics vol . 3) Company, Concepci6n Company . 2002. Gra m matical i zation and category weakness. I n Wischer a n d Diewald, eds . . 20 1 - 1 5 . Comrie, Bernard. 1 976. Aspect. Cambridge: Cambridge Uni versity Press. Comrie. Bernard . 1 978. Ergativity. In W. P. Lehmann. ed. , 329-94. Comrie. Bernard. 1 980. M orphol ogy and word order reconstruction: problems and prospects. In Fi siak, ed. , 83-96. Comrie, Bernard. 1 98 1 . Tile Languages of the SOl'iet Union . Cambridge: Cambridge U n i vers i ty Press. Comrie, B ernard. 1 989 [ 1 98 1 ] . Language Universals and Unguistic Typology: Syllla.l' ami M0I7Jhology, 2nd ed i tion. Oxford : Bl ackwel l . .•
.•
Referellces
243
Comrie. Bernard. 2000. Evidentials: semantics and history. I n Lars Johanson and Bo Utas. eds . . Evidentials: Turkic, Iraniall alld Neighbouring Lallguages, 1 - 1 3 . Berlin: Mouton de Gruyter. (Empirical Approaches to Language Ty pology 24) . Cou per- Kuhlen. El isabeth. 1 993. English Speech Rhythm: Form and Function ill E,'el:l'day Verbal Interaction. A msterdam: Benjamins. Cou per- Kuhlen. El isabeth. 1 996. Intonation and clause combining i n discourse: the case o f because. Pragmatics 6: 389-426. Cou per- Kuhlen. El isabeth. and M a rgret Selting. 1 996. Towaras an interactive perspective on prosody and a prosodi c perspective on interaction. In El isabeth Couper- Kuhlen and Margret Sel ti ng. eds . . Prosody in Conversation . I I -56. Cambridge: Cambridge University Press. Craig. Colette. 1 99 1 . Ways to go in Rama: a case study in polygrammaticalization. In Traugott and Heine. eds . . vol . 2: 455-92. Croft. Wi lliam. 1 990. Typology and Universals. Cambridge: Cambridge University Press. Croft. Wi lliam. 1 99 1 . Syntactic Categories and Grammatical Relations: the Cognitive Organization of Information . Chicago: Universi ty of Ch icago Press. Croft. Wi lliam. 1 995. Autonomy and functional ist linguistics. Language 7 1 : 490-532. Croft. Wi lli am. 2000. Explaining Language Change: an EvolutionQ/:v App roach Harlow. Essex : Pearson Education. Cruse. D. A. 1 986. Lexical Semantics. Cambridge: Cambridge Un i versity Press. Dahl .. O sten. 1 98 5 . Tense and Aspect ::':vstellls. Oxford : Blackwel l. Dahl . O sten. 2000. The tense-aspect systems of European languages i n a typological per spective. In Osten Dahl, ed . Tense and Aspect in the Languages of Eumpe. 3-25. Berl i n : Mouton de G ruyter. (Empirical Approaches to Language Typology 20-6) Dasher. Richard B . 1 995. Grammatical i zation i n the System of Japanese Predicate Honori fics. Unpublished PhD dissertation, Stanford U n iversi ty. Deacon. Terence W. 1 997. The Symbolic ::'jlecies: the Co-E,'olution of Language and the Brain. New York: Norton. DeGraff. M i chel . ed . . I 999a. Language Creation and Language Change: Creolization, DiaclllvlIY, and Development. Cambridge, M A : M IT Press. DeGraff. M ichel . I 999b. Creol i zation. language change. and language acqu isition: an epi logue. In DeGraff. ed 473-543 . DeGra ff. Michel. I 999c. Creolizati on. language change, and language acqu isition : a prole gomenon . In DeG ratr. ed . 1 -46. DeGraff. Michel. ed 200 I . On the origin of creoles: a Cartesian critique of Nco- Darwinian li nguistics. Linguistic Typology 5 : 2 1 3-3 1 0. DeLancey. Scott. 1 98 1 . An i n terpretation of split ergativity and related patterns. Language .
.
.•
.
.•
5 7 : 626-57 .
DeLancey. Scott. 1 99 1 . The origins of verb seri alization i n Modern Ti betan. Studies in Language 1 5 : 1 -23. Den ison. David. 1 98 5 . The origins o f peri phrastic D O : El legiird and Vi sser reconsidered . In R oger Eaton. Olga Fischer. Wi llem Koopman . and Frederike van der Leek. eds . . Papers /lvm the 4th Intemational Conference o n English Historical Linguistics. 44-60. A msterdam: Benjami ns. (Current Issues in Li nguistic Theory 4 1 ) Den ison. David. 1 99 3 . English Historical Syntax: Verbal ConstrllctiollS. London: Long man. Deutscher. Guy. 2000. Sylllactic Change ill Akkadian: the Evoliltion of Selltential Comple melllation . New York: Oxford University Press.
244
References
Dietrich, W. 1 973. Der periphrastische Verbalaspekt in den romanischen Sprachell. Tiibi ngen: Niemeyer. Dirven, Rene. 1 985. Metaphor as a basic means for extending the lexicon. In Paproue and D irven, eds., 85- 1 1 9 . Disterheft, Dorothy. 1 990. The role of adapti ve rules i n language change. Diachronica 7: 1 8 1 -98. Dixon, R. M . W. 1 979. Ergativity. Language 55: 59- 1 38. Dixon, R . M . W. 1 982 [ 1 969] . Olgolo syllable structure and what they are doing about it. In When� Have all the Adjectives GOlle ? And Other Essays in Semantics alld Syntax,
207- 1 0. Berlin: Mouton [ fi rst publ ished in Linguistic Inquiry I : 273-6, 1 969] . Dixon R. M . W. 1 994. Ergativity. Cambridge: Cambridge University Press. Down ing, W. H . 1 9 1 9. Digger Dialects. Melbourne and Sydney : Lothian Book Publ ishing Co. Dressler, Wolfgang U. 1 985. MOl7,!zonology: the Dynamics of Derivation. Ann Arbor: Karoma. (Linguistica Extranea. Studia 1 2) Dressler, Wol fgang U. 1 988. Language death. In Newmeyer, ed. , vol . 4: 1 84-5. Dryer, M atthew S. 1 99 1 . SVO languages and the OV : va typology. Journal of linguistics 27: 443-82. Dryer, M atthew S . 1 992. The Greenbergian word order correlates. Language 68: 8 1 - 1 38. Du Bois, John W. 1 985. Competi ng motivati ons. I n Haiman, ed. , 343-65 .. Du Bois, John W. 1 987. The di scourse basis of ergativity . Language 63 : 805-55. Ebert, Robert. 1 976. I ntroduction. I n S teever et al. , eds. , vii-xvii i . Eckert, Penelope. 1 988. Adolescent social structure and the spread of l i nguistic change. Language in Society 1 7 : 1 83-207. Eckert, Penelope. 1 997. Age as a socioli nguistic variable. In Florian Coul mas, The Handbook of Sociolillguistics, 1 5 1 -67. Oxford: B lackwell . Eckert, Penelope. 2000. linguistic Variatioll a s Social Practice: the linguistic Construction of Identity ill Belten High. Malden, M A : B l ackwell . (Language i n Society 27) Ellegard, A l var. 1 953. The Auxiliary Do, the Establislunent and Regulation of its Use in English. S tockhol m : Almquist and Wiksel l . (GoUenburg Studies i n English 2) Faltz, Leon ard. 1 988. Rejlexivizatioll: a Study in Universal Syntax. New York: Garland. Fernandez-Vest, M. M . Jocelyne. 1 994. Les par/jcules enonciatives dans la constructioll du discours. Pari s: Presses Universitaires de France. Fernandez-Vest, M. M . Jocelyne, ed. , 2000. Grammaticalisation ariale et simantique cog nitive: les langues fenniques et sames. (Areal G rammaticalization and Cognitive Semantics: the Fi nnic and Sami Languages.) Actes du Col l oque International du C.N.R.S. tenu les 9 et 1 0 avri l en Sorbonne. Paris: Fondation de la l angue estoni enne (Eesti Keele Sihtasutus). Fi l l more, C harles J., Paul Kay, and Mary Catheri ne O' Connor. 1 988. Regularity and id iomaticity in grammatical constructions: the case of let alone. Language 64: 50 1 -38. Fi negan, Edward, and Niko Besnier, 1 989. Language: its Structure and Use. San Diego: Harcourt B race Jovanovich. Fischer, Olga C. M. 1 992. Syntax. I n Norman B lake, ed., The Cambridge History of the English Language, vol . 2 : 1066-1476, 207-408. Cambridge: Cambridge University Press. Fischer, Olga C. M. 2000. Grammatical i zation: unidirectional, non-reversable? The case of to before the i n fi ni tive i n Engl ish. I n Fischer, Rosenbach, and Stein, eds. , 1 49-70.
References
245
Fischer, Olga, and Annette Rosenbach, 2000. I ntroduction. In Fischer, Rosenbach, and Stein, eds., 1 -38. Fischer, Olga, Anette Rosenbach, and Dieter S tei n, eds. 2000. Pathways of Change: Grammaticalization in English. Amsterdam: Benjamins. (Studies i n Language Companion Series 53) Fisiak, Jacek, ed. 1 980. Historical Morphology. The Hague: Mouton. (Trends i n Linguistics . Studies and M onographs 1 7) Fisiak, Jacek, ed. 1 98 5 . Historical Semantics, Historical Wo.rd-formation. Berlin: Mouton de Gruyter. (Trends in Li nguistics. Studies and Monographs 29) Fleischman, Suzanne. 1 982. The Future in Thought and Language: Diachronic Evidence from Romance. Cambridge: Cambridge University Press. (Cambridge Studies i n Lingui stics 36) Foley, Wil l ia m A. 1 988. Language bi rth : the processes of pidginization and creolization. In Newmeyer, ed. , vol . 4: 1 62-83. Foley, Wi l l ia m A., and Robert D. Van Val i n . 1 984. Functional Syntax and Universal Gram mar. Cambridge: Cambridge University Press. (Cambridge Studies in Linguistics 38) Ford, Cecilia. 1 993. Grammar in Interaction: Adverbial Clauses in American Ellglish Con versation. Cambridge: Cambridge Universi ty Press. Fox, Barbara, and Sandra Thompson. 1 990. A di scourse explamition of the of the grammar of relative clauses in Eng l ish conversation. Lallguage 66: 297-3 1 6. Fraj zyngier, Zygmunt. 1 99 1 . The de dicto domain of language. In Traugott and Heine, eds. , vol. I : 2 1 9-52. Frajzyngier, Zygmunt. 1 996a. On sources of demonstratives and anaphors. In Barbara Fox, ed. , Studies in Allaphora, 1 69-204. Amsterdam: Benjamins. (lYpological Studies in Language 33) Frajzyngier, Zygmunt. I 996b. Grammaticalizalion of the Complex Sentence: a Case Study in Clzadic. A msterdam: Benjamins. (Studies i n Language Companion Series 32) Frances, W. Nelson, and Henry Kucera. 1 96 1 . The BrowlI Corpus of Standard American English. Providence, R I : Brown University. Fraser, B ruce. 1 988. Ty pes of English discourse marker. A cta Linguistica Hungarica 38: 1 9-3 3. Fries, Charles C. 1 940. On the development of the structural use of word-order in Modern Engl ish. Language 1 6: 1 99-208. Gabelentz, George von der. 1 89 I . Die Sprachwi ssenschaft. I hre Aufgaben, Methoden, und bisherigen Ergebn isse. Lei pzig: Weigel . Gam i l l scheg, Ernst. 1 957. Historische franzosiche Syntax. Tiibingen : Niemeyer. Garrett, Andrew. 1 990. The origin of spl i t ergativity. Language 66: 26 1 -96. Garrett, Andrew. 1 998. On the origin of auxil iary do. English Language and Linguistics 2: 283-330. Geeraerts, Dirk. 1 986. Functional explanations i n d iachronic semantics. In Alain Bossuyt, ed., Functional Explanations in Lillguistics ( Belgiall Journal of Linguistics I), 67-93. B russels : Edi tions de l ' Universite de Bruxelles. Geis, M ichael L., and Arnold M . Zwicky. 1 97 1 . On invited inferences. Linguistic Inquiry 2: 5 6 1 -6. Gelb, Ignace J., and B urkhardt Kienast. 1 990. Die altakkadiscllen Konigsinschriften des Drillen Jahnausellds vor ChristLlS. Stuttgart: Franz Stei ner. =
246
References
Genetti, Carol 1 99 1 . From postposition to subordi nator in Newari. In Traugott and Hei ne. eds. , vol . I : 227-5 5 . Giacalone Ramat. Anna. Onofrio Carruba. and G i u l i ano Bern i n i . cds. 1 987. Papers fmlll the 7th International Conference 011 Historical Linguistics. Amsterdam: Benjamins. (Current Issues i n Linguistic Theory 48) Giacalone Ramat. Anna. and Paul J. Hopper. eds. 1 998. The Limits of Grallllllaticalizatioll . Amsterdam : Benjamins. (Typological Studies i n Language 37) Gi ldea. Spike. ed. , 2000. Recollstructing Grammar: Comparative Linguistics and Gralll maticalization. A msterdam: Benjamins. (Typological Studies i n Language 43) Gillic�ron. Jules Louis. 1 902- 1 0. Atlas lillguistique de la Frallce. Pari s: Champion. Giv6n. T. 1 97 1 . Historical syntax and synchronic morphology : an archaeologist's field tri p. Chicago Linguistic Society 7 : 394-4 1 5 . Giv6n. T. 1 97 3 . The ti me-axis phenomenon. Language 49: 890-925. Giv6n. T. 1 979. On Understallding Grammar. New York: Academic Press. Giv6n, T. 1 982. Tense-aspect-modal ity: the creole prototype and beyond. In Paul J. Hopper, ed. , Tense-aspect: Between Semallfics and Pragmatics. 1 1 5-63 . A m sterdam: Benjamins. (Typological Stud ies i n Language I ) Giv6n, T. 1 985. lconicity, isomorphism. and non-arbitrary coding. I n Haiman. ed . . 1 87-2 1 9. Giv6n. T. 1 989. Mind, Code alld Context: Essays ill Pragmatics. H i l l sdale. NJ : Erlbau lll Associates. G i v6n. T. 1 990. Syllfax: A Functiollal-Typological lllfmductioll. vol . 2. Amsterdam: Benjamins. Giv6n, T. 1 99 1 a. The evolution of dependent clause morpho-syntax i n B i b l i cal Hebrew . In Traugott and Hei ne. eds vol . 2: 257-3 1 0. Giv6n. T. 1 99 1 b. Serial verbs and the mental reality of "event": grammatical vs. cognitive packaging. In Traugott and Heine. eds vol . I : 8 1 - 1 27. Giv6n. T. 2000. Internal reconstruction: as method. as theory. I n G i ldea. ed . . 1 07-59. Green. Georg i a M . 1 996 [ 1 989] . Pragmatics alld Natural Language Understandillg . 2nd edition. Mahwah, NJ : Erl bau m. Greenberg. Joseph H . 1 960. A quanti tative approach to the morphological typol ogy of language. International Journal ofAmerican Linguistics 26: 1 78-94. Greenberg. Joseph H. 1 966a. Some uni versals of language with particular reference to the order of meaningful elements. In Greenberg, ed. , 73- 1 1 3 . Greenberg. Joseph H ed. I 966b [ 1 963 ] . Lallguage Universals, with Special Reference 10 Feature Hierarchies. 2nd ed ition. The Hague: Mouton. Greenberg. Joseph H. 1 974. The relation of frequency to semantic feature in a case language (Russian). Workillg Papers 011 Lallguage Universals, Stanford University. 1 6: 2 1 -89. Repri nted i n Keith Denning and Suzanne Kemmer. eds . On Language: Selected Writings of Joseph H. Greenberg. 207-26. Stanford : Stanford Uni versity Press. 1 990. Greenberg. Joseph H . 1 978a. How does a language acquire gender markers? In Greenberg et al. , eds vol. 3: 47-82. Greenberg, Joseph H. I 978b. Generalizations about nu meral systems. In Greenberg et al. . eds v o l . 3 : 249-95. Greenberg. Joseph H. 1 985. Some iconic relationships a mong pl ace. ti me. and discourse deixis. I n Hai man. ed 27 1 -8 1 . Greenberg. Joseph H . 1 990. Two approaches to language u n i versals. I n Keith Denning and .•
.•
.•
.
.•
.•
.•
References
247
Suzanne Kemmer, eds., On Language: Selected Writings of Joseph H. Greenberg, 702-20. Stanford : Stanford University Press. Green berg. Joseph H. 1 99 1 . The last stages of grammatical elements: contractive and ex pansive desemanticization. In Traugott and Heine, eds., vol . I : 30 1 - 1 4. Greenberg. Joseph H . , Charles A. Ferguson, and Edith Moravcsik, eds. 1 978. Universals of HUlllan umguage. Stanford : Stanford Uni vers i ty Press, 4 vols. Grice. H . Pau l. 1 975. Logic and conversation. In Cole and Morgan, eds., 4 1 -5 8 . Groefsema, Marjolei n . 1 995. Can. may. must a n d should: a Relevance theoretic account. Journal of Linguistics 3 1 : 5 3-79. GU nther, Susanne. 1 996. From subordi nation to coordi nation? Verb-second position in German causal and concessive constructions. Pragmatics 6: 323-56. Haase, Martin. 1 994. Respekt: die Grammatikalisierung von Hojlichkeit. Manchester, Munich: L1 NCOM EUROPA. (Edition Li nguistik 03) Hagege, Claude. 1 993. The Language Builder: an Essay on the Human Signattlre in Linguis tic MOl1Jhogenesis. A msterdam: Benjami ns. (Current Issues in Lingu istic Theory 94) Hai man. John. 1 972. Phonological targets and unmarked structures. Language 48: 365-77 . Haiman. John. 1 978. Cond itionals a re topics. Language 54: 5 64-89. Haiman, John. 1 980. The iconicity of gram mar. Language 56: 5 1 5-40. Haiman. John. 1 983. Iconic and economic motivation. Language 59: 78 1 -8 1 9. Haiman. John. 1 984. H u n : a Papuan language of New Gui nea. In Ti mothy Shopen, ed . . Languages and their Status. 35-90. Cambridge, MA: Winthrop. Haiman, John. 1 985a. Natural Syntax: Iconicity and Erosion. Cambridge: Cambridge Un i versi ty Press. Haiman. John, ed. 1 985b. lcon icity in Syntax. Amsterdam: Benjami ns. (Typol ogical Studies in Language 6) Hniman. John. 1 99 1 . Motivation. repeti tion and emancipation: the bureaucratization of language. In H. C. Wol fart, ed. , Linguistic Studies Presented to JollII L Finley. Memoire 8 . Algonquian and l roquoian Linguistics (Winni peg, Manitoba), 45-70. Haiman. John. 1 994. Ritual i zation and the devel opment of language. In Pagl i uca, ed. , 3-28. Haiman, John, and Sandra A . Thompson. 1 984. "Subordi nation" i n uni versal grammar. In B rugman et al . eds., 5 1 0-23 . Haiman. John. a n d Sa ndra A. Thom pson , eds. 1 988. Clause Combining ill Gralltlnar and Discourse. A msterdam: Benjamins. Hale. Kenneth. 1 973. Deep-surface canonical dispari ties i n relation to analogy and change: an Australian example. In Thomas Sebeok. ed., Currelll Trends ill Linguistics, vol. I I : A real and Typological Linguistics. 40 1 -58. The H ague: Mouton. Hale. Kenneth. 1 976. The adjoi ned relative cl ause i n Australia. In R. M. W. Dixon, ed . , Grammatical Categories i n A ustralian Languages, 78- 1 05 . Canberra : Austral ian Institute of Aboriginal Studies. Halle. M orri s. 1 964. Phonology i n generative gram mar. I n Jerry A . Fodor and Jerrold J. Katz. eds The Structure of Language: Readings in the Philosophy of Language, 3 34-52. Englewood C l i ffs, NJ : Prentice-Hal l . Hall iday, M . A . K . 1 96 1 . Categories o f the theory o f grammar. Wont 1 7 : 24 1 -92. Hal l i day. M. A. K. 1 994 [ 1 985]. An Introduction to Functiollal Grammar, 2nd edition. London: Arnold. Hall iday. M. A . K . . and Ruqaia Hasan. 1 976. Cohesioll ill English. London: Longman. .
.•
248
References
Halpern, Aaron. 1 995. Oil the Placemellt and Morphology of Clitics. Stanford: CLS I Publications. Hammond, M., and M. Noonan, eds. 1 988. Theoretical MOI7J/lOlogy. New York: Academic Press. Harris, Al ice C., and Lyle Campbel l . 1 995. Historical SYlltax ill Cross-Linguistics Perspec tive. Cambridge: Cambridge University Press. (Cambridge Studies in Linguistics 74) Harris, Martin. 1 97 8 . The Evolution of Frellch SYll tax: a Comparative Approach. London: Longman. Harris, M arti n. 1 988. Concessive clauses in English and Romance. I n Haiman and Thompson, eds., 7 1 -99. Harris, Roy, and Talbot Taylor. 1 997 [ 1 989] . Landmarks in Lillguistic Thought, vol . I . London : Routledge. Haspelmath, Martin. 1 989. From purposive to i n fi n i tive - a universal path of grammatical ization. Folia Lillguistica Historica 1 0: 287-3 1 0. Haspelmath, Martin. 1 993. The diachronic externalization of i nflection. Linguistics 3 1 : 279-3 10. Haspelmath , M arti n . 1 997. From Space to TIme: Temporal Adverbials in the World 's Languages. Munich: LINCOM EUROPA. (LINCOM Studies in Theoretical Linguistics 3) Haspelmath, M arti n . 1 998. Does grammatical i zation need reanalysis? Studies in Language 22: 3 1 5-5 1 . Haspelmath, M arti n . 1 999a. Why is grammaticalization irreversible? Linguistics 37: 1 043-68 . Haspelmath, Marti n . 1 999b. Review of Gilbert Lazard, ed., La typologie actancielle. Linguistic Typology 3: 36 1 -5 . Hawkins, J o h n A . 1 983. Word Order Universals. New York: Academic Press. Healey, Antonette diPaolo, and Richard L. Venezky. 1 980. A Microfiche Concordance to Old Ellglish. University of Toronto: The Dictionary of Old English Project, Centre for Medieval Studies. [Also available i n electronic form: Angus Cameron, Ashley Crandel l A mos, S haron Butler, and Antonette diPaolo Healey, The Dictionary of Old English Corpus ill Electronic Form. University of Toronto: Dictionary of Old Engl ish Proj ect.] Hei ne, Bernd. 1 992. Grammatical ization chai ns. Studies in Language 1 6: 335-68. Heine, Bernd. 1 99 3 . Auxiliaries: Cognitive Forces and Grammaticalization. New York: Oxford University Press. Hei ne, Bernd. 1 994. On the genesis of aspect in A frican languages: the proxi mative. In Kevin E. M oore, David A. Peterson, and Comfort Wentu m, eds., Berkeley Linguistics Society 20: Special Session 011 Historical Issues ill African Linguistics, 3 5-46. Berkeley : B erkeley Lingui stics Society. Heine, Bernd. 1 997. Cognitive Foundations of Grammar. New York: Oxford Universi ty Press. Heine, Bernd. Forthcoming. (De)grammaticali zation. In Kate Burridge and Barry Blake, eds., Historical Linguistics 200 1 : Papers from the 1 2th International Conference on Historical Linguistics, Melbourne. Amsterdam: Benjamins. Heine, Bernd, U l rike Claudi, and Friederi ke Hilnnemeyer. 1 99 1 a. Grammaticalization: a Conceptual Framework. Chicago: University of Chicago Press.
References
249
Heine, Bernd, Ulrike Claudi, and Friederike Hiinnemcyer. 1 99 1 b. From cognition to grammar - evidence from African languages. In Traugott and Heine, eds., vol . I : 1 49-87. Hei ne, Bernd, Tom Giildemann, Christa Kilian-Hatz, Donald A. Lessau, Hei nz Roberg, Mathias Schladt, and Thomas Stolz. 1 993. Conceptual Shift: a Lexicoll of Gram maticalizatioll Processes ill Africall Languages. University of Cologne: I nstitut fUr Afrikanistik. (Afrikanistische Arbeitspapiere 34/35) Heine, Bernd, and Tania Kuteva. 2002. Coinmon Grammaticalization Processes in the Languages of the World. Cambridge: Cambridge University Press. Heine, Bernd, and Mechthild Reh. 1 984. Gramlllaticalizatioll alld Reanalysis in African Languages. Hamburg : Hel mut Buske. Held, Warren H . , Jr. 1 957. The Hittite Relative Sentence. Balti more: Linguistic Society of A merica (Language Di ssertation no. 55). Heusler, Andreas. 1 92 1 . A ltisliindisches Elememarbuch, 2nd edi tion. Heidelberg: Wi nter. Himmelmann, Nikolaus. 1 992. Grammaticalization and Gral/ullar. University of Cologne: Institut fUr Sprachwissenschaft. (Arbeitspapier NS 1 6) H i m melmann, Nikolaus. 1 997. Deiktikon, A rtikel, Nomillalphrase: zur Emergenz syntak tischer Struktur. Tiibingen: Niemeyer. (Linguistische Arbeiten 362) Hi nt, Mati . 2000. The Estonian noun under Indo-European pressure. In Fernandez-Vest, ed. , 1 69-78. Hock, Hans Henrich. 1 99 1 [ 1 986]. Principles of Historical Linguistics, 2nd edition. Berlin: Mouton de Gruyter. Hockett, Charles F. 1 947. Problems of morphemic analysis. Language 23: 32 1 -43 . Reprinted in Marti n Joos, ed., Readings in Linguistics, vol . I : The Developmem of Descriptive Linguistics in America / 925-/926, 229-42. Chicago: University of Chicago Press, 1 966. Hodge, Carlton T. 1 970. The l inguistic cycle. Language Sciences 1 3 : 1 -7. Hoenigswald, Henry. 1 966. Are there universals of l i nguistic change? . In Greenberg, ed., 23-4 1 . Hol m, John. 1 989. Pidgins alld Creoles, 2 vols. Cambridge: Cambridge University Press. Hook, Peter E. 1 974. The Compound Verb in Hindi. Ann A rbor: Center for South Asian Studies. (Mich igan Series in South and Southeast Asian Languages and Linguistics I ) Hook, Peter E. 1 99 1 . The emergence of perfective aspect i n I ndo-Aryan languages. In Traugott and Heine, eds., vol . 2: 59-89. Hooper [Bybee] , Joan L. 1 976. Word frequency in lexical diffusion and the source of morphophonological change. In Wi l l iam M. Christie, ed. , Current Progress in His torical Linguistics. Proceedings of the 2nd International Conference on Historical Linguistics, Tucson, A rizona, /2-/6 Janumy / 976, 95- \ 05. Amsterdam: North
Hol land. (North-Holland Linguistic Series 3 1 ) Hopper, Paul. 1 986a. Discourse function and typological shift: a typological study of the VSISV alternation. In W. P. Lehmann, ed., Language Typology /985. Papers from the Linguistic Typology Symposium, Moscow, 9-/3 December /985, 1 23-4 1 . A msterdam: Benjami ns. (Current Issues i n Li nguistic Theory 47) Hopper, Paul. 1 986b. Some discourse functions of classifiers in Malay. In Colette Craig, ed., Noull Classes and Categorization, 309-25. A msterdam: Benjamins. (Typological Studies in Language 7)
250
References
Hopper, Paul . 1 987. Emergent grammar. In Jon Aske, Natasha Berry. Lau ra M ichaelis. and Hana Fi l i p , eds., Berkeley Linguistics Society 13: General Session and Pa rasession 011 Grammar and Cognition , 1 39-57 . Berkeley: Berkeley Li ngui stics Society. Hopper. Pau l . 1 990. Where do words come from? In Wi l l iam Croft. Keith Denn i ng. and Suzanne Kemmer, eds Studies ill Typology and Diachrony: Papers Pre sell ted to Joseph Greellberg on his 75th Birthday, 1 5 1 -60. A msterdam: Benjamins. (Typological Studies i n Language 20) Hopper, Pau l . 1 99 1 . On some pri nciples of grammaticization. In Traugott and Heine. eds . . vol . I : 1 7-35 . Hopper, Paul . 1 994. Phonogenesis. In Pagl iuca, ed. , 29-49. Hopper, Paul . 1 996. Some recent trends in grammaticalization. Anllual Review of Anthro· po logy 25: 2 1 7-36. Hopper, Paul . 2002. Hendiadys and aux i l iation i n Engl ish. In Joan Bybee. John Hai man. and M ichael Noonan, eds . , Complex Sentences ill Grammar and Discourse : a Festschrift for Sandra Thompsoll, 1 45-73. A msterdam: Benjami ns. Hopper, Pau l J . , and Janice M artin . 1 987. Structuralism and diachrony : the developmenl of the i ndefi n i te 31ticle in English. I n Giacalone Ramat et al. , eds . , 295-304. Hopper. Pau l J . , and Sandra Thompson. 1 984. The di scourse basis for l exical categories i n uni versal grammar. Language 60: 703-8 3 . Hopper, Pau l J., a n d Sandra Thompson . 1 985. The iconicity of t h e uni versal categories "nou n" and "verb." In Haiman, ed. , 1 5 1 -83. H orn, Laurence R. 1 984. Toward a new taxonomy for pragmatic i n ference: Q-based and R-based i mpl icature. In Deborah Schi ffri n. ed. , Mealling, Form and Use in Colllext: Linguistic A pplications , 1 1 -42. Washi ngton DC: Georgetown Uni versity Press. (GURT 84) H orn, Laurence R. 1 989. A Natural History of Negation . Chicago: University of Chicago Press. Horn, Laurence R . 1 996. Presupposi tion and i mplicature. I n Shalom Lappi n, ed . . The Hand book of Cofltemporary Semantic TlzeOlY, 299-3 1 9. Oxford : B l ackwel l . H umboldt, Wi l hel m von. 1 825. O ber das Entstehen der grammati kal i schen FOI'men und ihren Ei n H u B auf die Ideenentwicklung. A bhandlullgell der KOlliglichen Akademie del' Wissellschaften zu Berlill , 40 1 -30. H umboldt. Wi l he l m von. 1 988 [ 1 836]. Oil Lallguage: the Diversity of HUlllan Language Structure alld its Influellce on the Mell Ial Developmelll of Mankind. Translation by Peter Heath of Uber die Verschiedenheit des mellschlichen Sprachbaus und i/wen Eillfluft auf die geistige EIlIWicklullg des Menschellgeschlechts. 1 836. Cambridge: Cambri dge University Press. Hundt. M arianne. 200 1 . What corpora tel l us about the grammatical isation of voice i n get-constructions. Studies ill Lallguage 2 5 : 49-87 . H y man, Larry M . 1 975. Phonology: TlleOlY alld A llalysis. New York : H o l t . R i nehart a n d Winston. Hymes, Del l . 1 956. Na-Dene and positional analysis of categories. American An tlllvpologist 5 8 : 624-38. Hymes. Del l , ed . , 1 97 1 . Pidgillizatioll and Creoliza tion of Lang uages . Cambridge: Cambridge University Press. ICAME: International Computer Archi ves of M odern Eng l i sh. Bergen. Norway: Norwegian Computing Center for the H u man ities. .•
References
25 1
I halainen. Ossi . 1 976. Peri phrastic do in affi rmative sentences in the di alect of East Somerset. Neophilologische Milleilu/lge/l 77: 608-22. Jackendoff. Ray. 1 990. Semantic Strtlclllres. Cambridge. MA: M IT Press. (Current Studies in Lingui stics Series 1 8) Jakobson. Roman. 1 964 [ 1 960]. Closing statement: l i nguistics and poetics. In Thomas A. Sebeok. ed . Style in Language. 350-77. Cambridge. MA: MIT Press. Jakobson. Roma n. 1 966. I m plications of language universal s for l i nguistics. In Greenberg. ed . . 263-78. Jakobson . Roman. and Morri s Hal le. 1 956. Fundamentals oj Lallguage. The Hague: Mouton. (Janua Li nguarum Series M i nor I ) Janda. Richard D. 1 980. On the decline of declensional systems: the overall loss of OE nominal case i n flections and the ME reanal ysis of -es as Ilis. In Traugott et al .. eds . . 243-52. Janda. Richard D. 1 995. From agreement affix to subject "c1 itic" - and bound root: -mos > -/lOS vs. (-)nos(-) and /lOS-otros in New Mexican and other regional Spanish dialects. I n Audra Dai nora. Rachel Hemph i l l . Barbara Luka. Barbara Need. and S heri Pargman. eds . Chicago Linguistic Society 31: the Parasession on Clitics. I 1 8-39. Chi cago: Ch icago Linguistic Society. Janda. Richard D. 200 I . Beyond "pathways" and "unidirectionality": on the discontinuity of transmission and the counterabi l i ty of grammaticalization. I n Campbel l . ed . . 265-340. le iTers. Robert J .. and l i se Lehi ste. 1 979. Principles and MetllOdsjor Historical Linguistics. Cambridge. M A : M IT Press. Je lTers. Roben J .. and Arnold M. Zwicky . 1 980. The evol ution of c1i tics. In Traugott et al. , eds .. 22 1 -3 1 . Johansson. Stig, i n col l aboration with Geoffrey N. Leech and Helen Good luck. 1 978. The Lancaster-Oslo/Bergen Corpus of British English. See ICAME. Jones. M i chael A . 1 993. Sardinian Syntax. London: Routledge. Joseph. Brian D . 200 I . Is there such a thing as "grammaticalization"? In Campbel l . ed . . 1 63-86. Joseph. Brian D .. and Richard D. Janda. 1 988. The how and why of diachronic morpholo gization and de-morphologization. In Hammond and Noonan, eds., 1 93-2 1 0. Justus. Carol. 1 976. Relativation and topical ization in H i ttite. In Li, ed., 2 1 3-46. Kai sse. Ellen M . 1 982. Sentential c1itics and Wackernagel 's Law. In Daniel P. Flicki nger, M arlys Macken, and Nancy Wiegand, eds., West Coast Conjerence on Formal Linguistics, 1 - 1 4. Stanford University: Linguistics Department. Keenan. Edward L. 1 985. Relative clauses. In Shopen. ed vol. 2 : 1 4 1 -70. Keenan. Edward L., and Bernard Comrie. 1 977. Noun phrase accessibi l i ty and universal gram mar. Linguistic Inquil)' 8: 63-99. Keesi ng. Roger M. 1 988. Melanesian Pidgin and the Ocea/lic Substrate. S tanford: Stanford University Press. Kees i ng. Roger M. 1 99 1 . Substrates, calquing and grammaticalization in Melanesian Pidgin. I n Traugott and Hei ne, eds . . vol . I : 3 I 5-42. Keller. Rud i . 1 994. O/l Language Change: the Invisible Hand in Language. trans. Brigitte Nerl ich. London : Routledge. Kemenade, Ans van . ed .. 1 999a. FUllctional Properties ojMOI]lllOsyntactic Change. Special issue. Linguistics 37.6. .
.
.•
252
References
Kemenade, Ans van. I 999b. Functional categories, morphosyntactic change, grammatical i zation. Linguistics 37: 997- 1 0 1 0. Kemeilade, Ans van, and Nigel Vincent, eds. 1 997. Parameters ofMorphosyntactic Change. Cambridge: Cambridge Universi ty Press. Kemmer, Suzanne. 1 992. Grammatical prototypes and competing motivations in a theory of l i nguistic change. In Garry Davis and Gregory I verson , eds., Explanation ill His torical Linguistics, 1 45-66. Amsterdam: Benjami ns. (Current Issues in Linguistic Theory 84) Kemmer, Suzanne. I 993a. M iddle voice, transitivity, and the elaboration of events. In Barbara Fox and Paul Hopper, eds., Voice: Form and Function, 1 79-230. Ams terdam: Benjamins. (Typological Studies i n Language 27) Kemmer, Suzanne. 1 993b. The Middle Voice. Amsterdam: Benjami ns. (Typological Studies in Language 23) Kilian-Hatz, Christa. 1 995. Das Baka: Grundziige einer Grammatik aus der Grammatikali sationsperspektive. Cologne: Universitiit KOln, Institut fUr Afri kanistik. (Afrikanis tische Monographien Nr. 6) Kiparsky, Pau l . 1 968. Linguistic universals and linguistic change. In Bach and Harms, eds., 1 7 1 -202. Kiparsky, Paul . 1 988. Phonological change. In Newmeyer, ed., vol . I : 362-4 1 5 . Kiparsky, Paul . 1 992. Analogy. In Wi l l iam B right, ed. , International Encyclopedia of Linguistics, vol. I : 56-6 1 . New York: Oxford University Press. Kiss, Katali n E., ed. 1 995. Discourse Configurational Languages. New York: Oxford University Press. Klai man, M . H. 1 976. Correlative clauses and IE syntactic reconstruction. In Steever et aI. , eds., 1 59-68 . Klamer, Marian. 2000. How report verbs become quote markers and complementizers. Lingua 1 1 0 : 69-98. Klavans, Judith L. 1 985. The i ndependence of syntax and phonology in c1iticization. Language 6 1 : 95- 1 20. Ktlnig, Ekkehard. 1 986. Conditionals, concessive conditionals, and concessives. In Traugott, Elizabeth Closs, Al ice ter Meulen, Judy Snitzer Rei l l y, and Charles A. Ferguson, eds., On Conditionals, 229-46. Cambridge: Cambridge University Press. Kortmann, Bernd, and Ekkehard Ktlnig. 1 992. Categorial reanalysis: the case of deverbal prepositions. Linguistics 30: 67 1 -97. Koster, Jan. 1 987. Why subject sentences don ' t exist. I n S . Jay Keyser, ed. , Recent Transfor mational Studies ill European Languages. Cambri dge, MA: MIT Press. (Linguistic Inqu i ry M onographs 3) Ktlvecses, Zoltan. 2002. Metaphor: a Practical Introduction. New York: Oxford University Press. Ktlvecses, Zoltan, and GUnter Radden . 1 998. Metony my: developing a cognitive l ingui stic view. Cognitive Linguistics 9: 37-77. Kroch, Anthony S. 1 989a. Function and grammar i n the history of English: peri phrastic DO. In Ralph W. Fasold and Deborah Schiffri n, eds., Language Change and Variation, 1 33-72. Amsterdam: Benjamins. (CulTent Issues in Linguistic Theory 52) Kroch, Anthony S. 1 989b. Reflexes of grammar i n patterns of language change. Language Variation and Change I : 1 99-244. Kroch , Anthony S . 1 994. Morphosyntactic vari ation. In Beals et al. , eds. , 1 80-20 1 .
References
253
Kroch, Anthony, and Ann Tay lor. 1 997. Verb movement i n Old and M iddle English: dialect variation and language contact. In van Kemenade and Vi ncent, eds., 297-325. Krug, Manfred. 1 998. Stri ng frequency : a cognitive motivati ng factor in coalescence, l anguage processing and change. 10umal of English linguistics 26: 286-320. Krug, Manfred. 2000. Emerging English Modals: a Corplls-Based Study of Grallllllatical ization. Berl i n : Mouton de Gruyter. (Topics in Engl ish Linguistics 32) Krug, Manfred. 200 1 . Frequency, iconici ty, categorization: evidence from emerging modals. I n Bybee and Hopper, eds. , 309-36. Krupa, Vi ktor. 1 966. Morpheme and Word in Maori. The Hague: Mouton. Krupa, Viktor. 1 968. The Maori Language. M oscow: Nauka Publ i shing House. Kuno, Susumu, and Etseko Kaburaki. 1 977. Empathy and syntax. linguistic Inquiry 8: 627-72. Kuryt owicz, Jerzy. 1 945-9. La nature des proces dits analogiques. Acta linguistica 5: 12 1 -38. Reprinted i n Eric P. Hamp, Fred W. Householder, and Robert Austerl i tz, eds . , Readings in linguistics 2, 158-74. Chicago: University of Chicago Press, 1 966. Kuryt owicz, Jerzy . 1 964. The Inflectional Categories of Indo-European. Heidel berg: Winter. Kuryt owicz, Jerzy. 1 976 [ 1 965] . The evolution of grammatical categories. Repri nted i n J . Kurytowicz, 1 976, Esquisses linguistiques, vol . 2 : 3 8-54. M u n i c h : Fink. Kuteva, Tania. 2000. Auxiliation: an Enquiry into the Nature of Grammaticalizatioll. Oxford: Oxford University Press. Kytij, Merja. 1 99 1 . Variation and DiachrollY, with Early American English in Focus: Stud ies on CAN/MA Y and SHALUWILL. Frankfurt-am-M a i n : Peter Lang. (Bamberger Bei trage zur engl ischen Sprachwissenschaft vol. 28). Labov, Wi lliam. 1 969. The logic of non-standard Engl ish. In James Alatis, ed. , Georgetown Monographs ill Languages and linguistics 22: 1 -44. Repri nted in Wi lliam Labov, Language in the Inner City: Studies ill Black English Vernacular, 20 1 -40. Phi ladel phia: University of Pennsylvania Press, 1 972. (Conduct and Com munication 3) Labov, Wi lliam. 1 972. Sociolinguistic Paltems. Phi ladel phia: University of Pensylvania Press. (Conduct and Communication 4) Labov, Wi l l i am. 1 974. On the use of the present to expl ain the past. In Luigi Hei lman, ed. , Proceedings of the 1 1th Intemational COllgress of lillguists, 825-52. Bologna: M u l i no. Labov, Wi lliam. 1 994. Principles of linguistic Challge: Intemal Factors. Oxford: B l ackwel l . (Language in Society 20) Labov, Wi lliam. 200 I . Principles ofLinguistic Change: Social Factors. Oxford: Blackwel l. (Language i n Society 20) Lakoff, George. 1 984. Performative subordi nate clauses. In Brugman, et aI. , eds., 472-80. Lakoff, George. 1 987. Women, Fire, and Dangerous Things: what Categories Reveal about the Mind. Chicago: University of Chicago Press. Lakoff, George, and Mark Johnson. 1 980. Metaphors we live by. Chicago: University of C hicago Press. Lakoff, Robi n. 1 972. Another l ook at dri ft. In Robert P. Stockwell and Ronald S. Macaulay, eds. , linguistic Change and Generative TheOlY, 1 72-98. Bloomi ngton: Indiana University Press. Lambrecht, Knud. 1 98 1 . Topic, A ntitopic and Verb Agreement in NOIl-Stalldard French. A msterdam: Benjami ns. (Pragmatics and B eyond I I : 6)
254
References
Lambrecht, Knud. 1 994. Illformation Structure and Sentellce Form: Topic. Focus alld the Mellfal Represelllatiolls of Discourse Referents. Cambridge: Cambridge University Press. (Cambridge Studies in Lingui stics 7 1 ) Langacker, Ronald W. 1 977. Sy ntactic reanal ysis. I n L i , ed . . 57- 1 39. Langacker, Ronald W. 1 987, 1 99 1 . Foundatiolls of Cogllitive Grammar. 2 vols. Stanford: Stanford University Press. Langacker, Ronald W. 1 990. Subjecti fication. Cognitive Linguistics I : 5-38. Lass, Roger. 1 980. 011 Explaillillg Language Change. Cambridge: Cambridge University Press. (Cambridge Studi es in Li ngui stics 27) Lass, Roger. 1 990. How to do things with junk: exaptation i n language evol ution. Journal of Linguistics 26: 79- 1 02. Lass. Roger. 1 997. Historical Linguistics and Language Challge. Cambridge: Cambridge Univers i ty Press. (Cambridge Studies in Lingui stics 8 1 ) Lass, Roger. 2000. Remarks on (uni)di rectiona l i ty. I n Fischer. Rosenbach. and Stein. eds . . 207-27. Laury. Ri tva. 1 997. Demonstratives in IlIleraction: the Emergence of a Defillite A rticle ill Finnish. A msterdam: Benjami ns. (Studies in Discourse and Grammar 7) Lausberg. Heinrich. 1 962. Romanische Sprachwissenschaft. Berl i n : de Gruyter. 3 vols. Leh mann. Christian. 1 985. G rammatica l i zation: synchron ic vari ation and d i achronic change. Lingua e Stile 20: 303- 1 8. Lehmann, Christi an. 1 988. Towards a typology of clause l i n kage. In Hai man and Thompson. eds . . 1 8 1 -225. Leh mann, Christian. 1 989a. Grammati kal i s ierung und Lexikalisierung. Zeitschrijl ji'ir PllOlletik. Sprachwissenschaft und Kommuflikalionsforscllllllg 42 : I 1 - 1 9 . Lehmann. Chri stian. 1 989b. Latin subordi nation i n typol ogical perspective. In Guall iero Cal bol i , ed. , Subordillatioll and other Topics ill Latin: Proceedings of the Thinl Colloquium on Latin Lillguistics. Bologna. 1-5 April / 985. 1 53-79 . A msterdam: Benj a m i ns. (Studies i n Language Companion Series 1 7) Leh mann. Christian. 1 99 1 . Grammatical i zation and related changes i n contempora ry German. In Traugoll and Hei ne. eds vol. 2: 493-535. Leh mann. Christian. 1 995 [ 1 982]. Thoughts on Grammalicalization. M u n i c h : Li ncom Europa. (First publ ished as Thoughts on Grammalicalizatioll: a Prog n/ll/lllUlic Sketch. no. 48 in the series Arbei ten des Kai ner Universalien- Projektes. Univer sity of Cologne. I nstitut fUr Sprachwi ssensc haft, 1 982.) Leh mann, Christian. 2002. New re Hections on gram matical ization and lexica l i zation. In Wischer and Diewald, eds., 1 - 1 8. Leh mann, Wi n fred P. 1 978a. The great underly i ng ground-plans. In Leh mann. ed .. I -55. Leh mann, Wi n fred P. , ed . 1 978b. SYlltactic Typology: Studies ill the Phenomellology of Lallguage. Aust i n : University of Texas Press. Leh mann. W. P. , and Y. Malkiel, eds. 1 968. Directiolls for Historical Lillguislics: a Sym. posiulll. Austi n: Uni versity of Texas Press. Le Page, R obert B . . and A . Tabouret- Kel ler. 1 98 5 . A cts of Idelltity: Creole-Based Approaches to Language and Etllllicity. Cambridge: Cambridge U n i versity Press. Lerner. Gene H. 1 996. On the "semi-permeable" character of gram matical units in conver sati o n : conditi onal entry i nto the turn space of another speaker. In Ochs el al . . eds . . 238-76. .•
References
255
Levin. Beth. 1 993. Ellglish Verb Classes alld A lternatiolls: a Prelimillw:1' IlIveMigatioll. Ch icago: U n i versity of Chicago Press. Lev inson. Stephen C. 1 983. Pragmatics. Cambridge: Cambridge University Press. Levi nson. Stephen C. 2000. Presulllptive Meanings: the Theory of Gelleralized COllversa tiollal implicalllre. Cambridge, MA: M IT Press, B radford Book. Li. Charl es N ed. 1 975. WOld O,der alld Word Order Change. Austin: University of Texas Press. Li . Charles N ed. 1 976. Subject and Topic. New York: Acadcl.n ic Press. Li . Charles N .. ed. 1 977. Mechallisms of �»llIactic Change. Austi n: University of Texas Press. Li . Charles N and Sandra A . Thom pson. 1 974. H istorical change of word order: a case study of Chi nese and i ts i mplications. In John M. Anderson and Charles Jones. eds . Historical Linguistics I : �)'lIIax, MOl7J/t0Iogy, Internal. alld COlllparative RecolI stmctioll. 200- 1 7. Amsterdam: North-Hol land. (North-Hol land Li ngu istics Series 1 2a ) Li. Charles N . . a n d Sandra A. Thompson. I 976a. Development of the causative i n Mandari n Chi nese: i n teraction of di achronic processes i n syntax. I n Masayoshi Shibatani, ed. , The Grammar of Causative Constructions. 477-92. New York: Academic Press. (Syntax and Semantics, vol . 6) Li . Charles N . . and Sandra A. Thompson. 1 976b. Subject and topic: a new typology of language. In Li , ed .. 45 7-90. Lichten berk . Frantisek. 1 99 1 a. On the gradual ness of grammatica l i zation. In TraugoLL and Heine, eds . . vol . I : 37-80. Lichtenberk. Frantisek. 1 99 1 b. Semantic change and heterosemy i n grammatica l i zation. Language 67: 475-509. Lightfoot. David. 1 979. Principles ofDiac/llvllic Sylllax. Cambridge: Ca mbridge University Press. (Cam bridge Studies i n Li ngu istics 23) Lightfoot. David. 1 99 1 . How to Set Parameters: A rgulIlel1ls /lvm Language Challge. Cambridge. M A : MIT Press. Lightfoot. David. 1 999. The Development ofLanguage: Acquisition. Challge. alld Evollltioll. M alden. M A : Blackwel l . Lipka. Leonhard. 1 990. A l l Oll/lille of English Lexicology: Lexical Structure, Word Seman tics, alld WOld-formation. Tiibingen : Niemeyer. (Forsc hung & Studium Anglistik 3) Longacre. Robert. 1 985. Sentences as combi nations of clauses. I n Shopen. ed . . vol . 2: 235-86. Longacre. Robert. 1 990. Story l i ne concerns and word order typol ogy i n East and West A frica. Studies in A/I"icall Linguislics, S upplement 1 0. Los Angeles: Ja mes S. Coleman A frican Studies Center and the Department of Lingu istics. University of Cal i fornia at Los A ngeles. Lord . Carol. 1 976. Evidence for syntactic reanalysis: from verb to complementizer in K wa. I n Steever el al . . eds., 1 79-9 1 . Lord . Carol. 1 982. The development of object markers i n seri al verb la nguages. In Paul J . Hopper and Sandra A . Thompson, cds., Studies ill Transitivity. 277-300. New York : Academic Press. (Syntax and Semantics. vol . 1 5 ) Lord . Caro l . 1 993. Historical Change in Serial Verb Constructions. Amsterdam : Benjamins. (Typological Studies i n Language 26) .•
.•
.•
.
256
References
Los, Beuelou. 1 999. Infinitival Complementation in Old and Middle English. The Hague: Thesus, Holland Academic Graph ics. Ludtke, Hel mut. 1 980. Auf dem Wege zu ei ner Theorie des S prachwandel s. In Hel mut Ludtke, ed. , Komlllunikatiollstheoretische Grundlagen des Sprachwandels, 1 82-252. Berl i n : de Gruyter. Luraghi, Silvia. 1 998. On the directional i ty of grammaticalization. Sprachtypologische Universal-Forschungen (STUF) 5 1 : 355-65. Lyons, John. 1 968. Introduction to Theoretical Linguistics. London: Cambridge University Press. Lyons, John. 1 977. Semantics. Cambridge: Cambridge University Press, 2 vols. Mair, Christian. 1 997. The spread of the going-to future in wriuen English: a corpus-based i nvestigation i nto language change in progress. In Raymond H ickey and Stan islaw Puppel , eds. , Language History and Linguistic Modelling: a Festschrift for Jacek Fisiak, 1 537-43. Berl i n : Mouton de Gruyter. Mair, Christian. Forthcoming. Corpus lingui stics and grammatical isation theory: statistics, frequencies, and beyond. In Hans Lindquist and Christian Mair, eds., Corpus-based App roaches to Gral1lmaticalisation in English. Amsterdam: Benjamins. Malkiel, Yakov. 1 979. Problems in the diachronic di fferentiation of near-homonyms. Language 55: 1 -36. Malkiel , Yakov. 1 98 1 . Dri ft, slope, slant. Language 57: 535-70. Masica, Col i n P. 1 976. Defining a Linguistic A rea: South A sia. Chicago: University of Chicago Press. Maslova, Elena. 2000. A dynamic approach to the veri fication of distributional universals. Linguistic Typology 4: 307-33 (target article with peer commentary). Matthiessen, Christian, and Sandra A . Thompson, 1 988. The structure of discourse and "subord ination." In Hai man and Thompson, eds., 275-333. Matisoff, James A . 1 982 [ 1 973]. The Gramlllar of Lahu. Berkeley : University of Cal i fornia Press, 2nd pri nting. Matisoff, James A. 1 99 1 . Areal and universal d i mensions of grammatization in Lahu . In Traugott and Hei ne, eds. , vol . 2: 383-53. Matsu moto, Yo. 1 988. From bound grammatical markers to free discourse markers: h istory of some Japanese connectives. In Ax maker et al., eds. , 340-5 1 . McCawley, James D. 1 968. The role of semantics i n a grammar. I n Bach and Harms, eds., 1 25-69. McElhinney, Bonnie. 1 992. The i nteraction of phonology, syntax, and semantics i n language change: the hi story of modal contraction in Engl ish. In Costas P. Canakis, G race P. Chan, and Jeanette Marshall Denton, eds., Chicago Linguistic Society 28: the Main Session, 367-8 1 . Chicago: Chicago Lingui stic S ociety. McMahon, A pri l M . S. 1 994. Understanding Language Change. Cambridge: Cambridge Uni versity Press. McWhorter, John H . 1 998. Identify i ng the creole prototype: vindicating a typological cl ass. Language 74: 788-8 1 8 . McWhorter, John H . 2000a. The Missing Spanish Creoles: Recovering the Birth of Planta tion Contact Languages. Berkeley : University of Cal i fornia Press. McWhorter, John H., ed. 2000b. Language Change and Language Contact in Pidgins and Creoles. A msterdam: Benjamins. (Creole Language Library 2 1 )
References
257
McWhorter, John H. 200 1 . The world 's simplest grammars are creole grammars. lillguistic Typology 5: 1 25-66 (target article with peer commentary). MED: The Middle English Dictionary. 1 956-200 1 . Ann A rbor: University of M ichigan Press. (http: //ets.umdl.umich.edu/m/medl) Mei llet, Antoine. 1 9 1 2. L'evolution des formes gra mmaticales. Scientia (Rivista di Scienza) 1 2, no. 26, 6. Reprinted in Mei l let 1 95 8 : 1 3�8. Meillet, Antoine. 1 9 1 5- 1 6 [ 1 958]. Le renou vel lement des conjonctions. Alllluaire de I' Ecole Pratique des Hautes Etudes. Reprinted in Meil let 1 9.58: 1 59-74. Meil let, Antoine. 1 958. linguistique historique et linguistiql;e generale. Paris. Champion. (Collection Linguistique publiee par la Societe de Lingu istique de Paris 8) M ikola, Ti bor. 1 975. Die alten Postpositiollen des Nenzischen (Juraksamojedisc/lell). The Hague: Mouton. (Janua Linguarum. Series Practica 240) Milroy, James. 1 992. linguistic Variation and Change: on the Historical Sociolinguistics of English. Oxford : Blackwell. Mitchell, B ruce. 1 985. Old English Syntax, 2 vols. Oxford: Clarendon Press. Mithun, Marianne. 1 984. How to avoid subordination. In Brugman, et al. , eds., 493-509. Mithun, Marianne. 1 988. The grammaticization of coordi nation. In Haiman and Thompson , eds. , 33 1 -59. Mithun, Marianne. 1 99 1 . The role of motivation i n the emergence of gram matical categories: the grammaticization of subjects. In Traugott and Hei ne, -eds. , vol. 2: 1 59-84. Mithun, Marianne. 1 995. Morphological and prosodic factors shapi ng word order. Pamela Downing and M ichael Noonan, eds. 1 995. Word Order ill Discourse, 387-424. A msterdam: Benjamins. (TYpological Studies in Lnnguage 30) Mithun, Marianne. 2000. The reordering of morphemes. I n Gi ldea, ed., 23 1 -58. Mufwene, Salikoko S. 200 I. The Ecology of Language Evolution. Cambridge: Cambridge University Press. Mufwene, Salikoko S. Forthcoming. Grammaticalization i s part of the development of cre oles. In O sten Dahl and Velupilai Weker, eds., Applvaches to Gramlllaticalization. Miihlhiiusler, Peter. 1 997 [ 1 986] . Pidgin and Creole linguistics, 2nd edition. Oxford: Blackwell. (Westminster Creolistics Series 3) Muysken, Peter. 1 988. A re pidgins a special type of language? In Newmeyer, ed., vol . 2: 285-30 1 . Newmeyer, Frederick J . , ed. 1 988. linguistics: tlte Cambridge Survey. Cambridge: Cambridge University Press, 4 vols. Newineyer, Frederick J., ed. 1 998. Language Form and Language Function. Cambridge, M A : M IT Press, B radford B ooks. Nichols, Johanna, and Alan Ti mberlake. 1 99 1 . Grammaticalization as retextualization. In Traugott and Heine, eds., vol . I : I i9-46. Noonan , Michael. 1 985. Complementation. In Shopen, ed. , vol . 2: 42- 1 40. Norde, M uriel . 200 1 . Deftexion as a counterdirectional factor in grammatical change. In Campbel l , ed., 23 1 -64. Ochs, Elinor, Emanuel Scheglotf, and Sandra A. Thompson, eds. 1 996. Grammar alld Interaction. New York: Cambridge University Press. (Studies in Interactional Sociol ingui stics 1 3) OED: The Oxford English Dictionary. 1 989. Oxford : Cl arendon Press, 2nd edition. [Available i n electronic form on CD-ROM .]
25 8
References
Ohori . Toshio, ed . 1 998 . Studies ill Japanese Grammaticalization: Cogllitive and Discourse Perspectives. Tokyo: Kurosio Publ i shers. O'Nei l . Wayne. 1 977. Clause adjunction in Old Engl ish. General Linguistics 1 7 : 1 99-2 1 1 . Ono. Tsuyos h i , and Sandra Thompson. 1 994. Unattached NPs i n Engl ish conversation. In Susanne Gahl, Andy Dolbey, and Christopher Johnson. eds., Berkeley Lillguistics Society 20: General Session Dedicated to the Contributiolls of C/wrles .I. Fillmore. 402- 1 9. Berkeley: Berkeley Lingui stics Society. Ortony. Andrew, ed. 1 993 [ 1 979]. Metaphor alld Thought. 2nd edition. Cambridge: Cambridge Uni versity Press. Pagliuca, Wi l l i a m , ed. 1 994. Perspectives 011 Grammaticalizatioll. Amsterdam: Benjamins. (Cu rrent I ssues i n Linguistic Theory 1 09) Paprotte, Wol f. and Rene Dirven, eds. 1 985. The Ubiquity of Metap/lOr: Metap/lOr ill Lall guage and Thought. A msterdam: Benjamins. (Current Issues i n Linguistic Theory 29) Patrick, Peter. 1 999. Urball Jamaicall Creole: Variatioll ill the Mesolect. Amsterdam: Benj a m ins. (Varieties of Engl ish Around the World, No. G 1 7) Paul. Hermann, 1 920 [ 1 880]. Prillzipien der Sprachgeschicllle, 5th ed i tion. H a l l e : Niemeyer. Pei rce, Charles Sanders. 1 965 [ 1 93 1 ] . Collected Papers. ed. Charles Hartshorne and Paul Weiss. Cambridge, M A : Harvard Univers i ty Press, Belknap. Perez, Avel i ne. 1 990. Ti me i n motion: gram matical isation of the be goillg to construction in Eng l i sh . La Trobe Ulliversity Working Papers ill Lillguistics 3: 49-64. Pin kster, Harm. 1 987. The strategy and chronology of the development of fu ture and perfect tense auxiliaries i n Latin. In Martin B. Harris and Paolo Ramat, eds. , The Histor ical Developmellt of A uxiliaries, 1 93-223. Berl i n : M outon de Gruyter. (Trends in Li ngu i stics. Studies and Monographs 35) Pintzuk, Susan. 1 999. Phrase Structures ill Competitioll: Variatioll alld Challge ill Old English Word Order. New York: Garland . Plag. l ngo. 1 993. Selltential Complemelltation ill Srallan : 011 the Formatioll of an Ellg/ish based Creole Lallguage. Tiibi ngen : Niemeyer. Plag, I ngo. 1 994. Creolization and l anguage change: a compari son. In Dany Adone and I ngo Plag, eds. , Creolization and Language Change, 3-2 1 . Tii bi ngen : Niemeyer. (Linguisti sche A rbeiten 3 1 7) Plank, Frans. ed. 1 979. Ergativify: Towards a Theory of Grammatical Relatiolls. New York: Academic Press. Plank. Frans, cd. 1 984. The modals story retold. Studies in Lallguage 8: 305-64. Pri nce, A l a n , and Paul S molensky. 1 997. Opti mality: from neural networks to Universal Gra m m a r. Sciellce Magazine, M arch 1 4. Pri nce, Ellen. 1 98 1 . Toward a taxonomy of given-new i n formati on. In Cole, cd. , 223-55. Pu l l u m , Geoffrey K . . and Arnold M . Zwicky . 1 98 3 . C l i ticization vs. i n flection: the case of Engl i s h n 't. Language 59: 502- 1 3 . Qu i rk . Randol ph, S idney Greenbau m, Geoffrey Leech, and Jan Svartvik. 1 98 5 . A COllcise Grammar of Colllemporary English. New York: Harcourt B race Jovanovich. Ramat, Paolo. 1 992. Thoughts on degra mmatical ization. Linguistics 30: 549-60. Ramal, Paolo. 200 I . Degrammatica l i zation or transcategori zation? In Chris Schaner Wol les. John Rennison, and Friedrich Neubart h , eds . Naturally! Lillguistic Swdies ill HOllour of Wolfgang Ulrich Dressler Presented 011 the Occasioll of Ilis 60th Birthday, 393-40 1 . Turi n: Rosenbach and Sellier. .
References
259
Ranki n. Robert L. 1 976. From verh to au x i l iary to noun cl assi fier and deli nite article: grammaticalization of the Siouan verhs S I T , S T A N D , L I E . I n R. L. Brown. K. Hou l i han. L. Hutch i nson, and A. M acLeish, eds., Proceedings oj the 1 9 76 Mid America Linguistics ConJerence. 272-83 . M i nneapolis: University of M i nnesota. Ravid, Dorit Diski n . 1 995. Lallguage Change ill Child and Adult Hebrew: a PsycllOlinguistic Perspective. New York : Ox ford University Press. Reddy. M ichael J. 1 993 [ 1 979] . The condu i t metaphor - a case of fra me conllict i n language ahout language. In Ortony, ed ., 1 64-20 1 . Rickford. John R . 1 987. Dimensions oj a Creole Continuum: History. Texts. and Linguistic A nalysis oJ Guyanese Creole. Stanford : Stanford Universi ty Press. R i ssanen. Matti , Mer:ia Kyto, and Mi nna Palander-Col l i n , eds . . 1 993. t.a rly English in the COlllputer Age: E.'plorations through the Helsinki C0 1 1J11S . Berl i n : Mouton de G ruyter. Roherts. Ian. 1 993a. A formal accou nt of grammatical i sation in the history of Romance fu tures. Folia Lingllistica Historica 1 3 : 2 1 9-58. Roberts. Ian. 1 993h. Verbs and Diaclllvnic Syll la.x: a Comparative Hist01:V oj English and French. Dordrec ht: K l u wer. (Studies i n Natural Language and Li nguistic Theory 28) Roherts, Sarah J . 1 998. The genesis of Hawaiian creole and di ffusion. Language 74: 1 -39. Roherts, Sarah 1 . 2000a. Nativization and the genesi s of Hawai ian creole and diffusion. In John Rickford and Suzanne Romai ne, eds., Creole Genesis, A llillldes, and Dis course: Studies Celebrating Charlene J. Sato, 45-70. Amsterdam: Benjamins. (Creole Language Li hrary 20) Roberts, Sarah J. 2000h. Nati vization and the genesi s of Hawaiian creole. I n McWhorter. ed . 25 7-300. Roherts. Sarah J. Forthcoming. The role of style and identi ty in the development of Hawaiian Creole. I n Armin Schwegler and Genevieve Escure, eds . . Amsterdam: Benjamins. Romai ne, Suzanne. 1 982. Socio-Historical Linguistics: its Status and Methodology. Camhridge: Cambridge University Press. (Cambridge Studies in Li nguistics 34) Romai ne, Suzanne. 1 988. Pidgin and Creole Languages. London and New York: Longman. Romai ne. Suzanne. 1 999. The gram matical izati on of the prox i mative in Tok Pisin. Language 75: 322-46. Rosenhach. A nette. 200 1 . The Engl ish -s-genitive: ani macy, topical i ty, and possessive re lationship i n a diachronic perspective. In Laurel J. B rinton., ed. , Historical Linguis tics 1 999: Selected Papers ftvm the 14th International ConJerence 011 Historical Linguistics, Vancollvel; 9- 13 August 1 999, 277-92. Amsterdam and Ph iladel phia: Benjami ns. (Cu rrent I ssues i n Linguistic Theory 2 1 5 ) Rude. Noel. 1 99 1 . Verbs to promotional suffixes in Sahapti an and Klamath. In Traugott and Heine. eds . , vol . 2: 1 85-99. Sampson. Geoffrey. 1 997. Educating Eve: the "Language IIIStillct " Debate. London: Cassel l . Sa muels. M . L . 1 972. Linguistic Evolutioll: with Special ReJerence t o English. London: Cam hri dge University Press. (Cambridge S tudies in Lingui stics 5) Sandherg. Kristian. 1 968 [ 1 930] . La Lingllistique balkanique: l11vblemes et resultats. Paris: K l i nksieck. Sankoff, G i l l ian. 1 980. The Social Life oj Language. Phi ladelphia: Un iversity of Pen nsy l vania Press. .
260
References
Sankoff, Gill ian. 1 990. The grammatical i zation of tense and aspect i n Tok Pisin and Sranan. Language Variation and Change 2: 295-3 1 2. Sankoff, G i l lian, and Penelope Brown . 1 976. The origins of syntax in discourse: a case study of Tok Pi sin relatives. Language 52: 65 1 -66. Sapir, Edward. 1 92 1 . Language: an lntlVduction to the Study ofSpeech. New York: Harcourt Brace and Co. Sapir, J. David. 1 977. The anatomy of metaphor. In J. David Sapir and J. Chri stopher Crocker, eds., The Social Use ofMetaphor: Essays on the A nthlVpology ofRhetoric, 3-32. Philadel phia: University of Pennsyl vania Press. Saussure, Ferdinand de. 1 986 [ 1 922]. Course in General Linguistics. Translated and anno tated by Roy Harri s. La Salle: IL: Open Court Press. [The source text is Cours de linguistique generale, 2nd edition, 1 922.] Schachter, Paul. 1 985. Parts-or-speech systems. I n S hopen, ed. , vol . I : 3-6 1 . Scheibman, Joanne. 2000. I dunno . . . A usage-based account of the phonological reduc tion of dOIl 't in American Engl ish conversation. Journal of Pragmatics 32: 1 0524. Schiffrin, Deborah. 1 987. Discourse Markers. Cambridge: Cambridge University Press. Schneider, G. D. 1 966. West African Pidgin-English: a Descriptive Linguistic A nalysis with Texts and Glossary flVm the CamelVon A rea. Athens, Ohio. Schwegler, Armin. 1 988. Word-order changes i n predicate negation strategies i n Romance . languages. DiachlVnica 5: 2 1 -58. Schwegler, Armin. 1 990. Analyticity and Syntheticity: a Diachronic Perspective with Special Reference to Romance Languages. Berl i n : de Gruyter. Searle, John R. 1 993 [ 1 979] . Metaphor. In Ortony, ed. , 83-1 1 1 . S h i , Yuzhi. 2002. The Establishment ofModern Chinese Grammar: the Formation ofthe Re sultative Construction and its Effects. A msterdam: Benjamins. (Studies i n Language Companion series 59) Shibatani, M asayoshi. 1 99 1 . Grammatici zation of topic i nto subject. In Traugott and Hei ne, vol . 2: 9 3- 1 33. S hopen, Ti mothy, ed. 1 985. Language Typology alld Syntactic Description. Cambridge: Cambridge University Press, 3 vol s. S i l verstein, M ichael. 1 976. Hierarchy of features and ergativity. In R. M . W. Dixon, ed. , Grammatical Categories ill Australian Languages, 1 1 2-7 1 . Canberra: Australian Institute of Aboriginal Studies. (Li nguistic Series 22) Singler, Joh n . 1 992. Nativization and pidgi n/creole genesis: a reply to B ickerton. Journal of Pidgin and Creole Languages 7: 3 1 9-33 . Siobin, Dan I . 1 977. Language change i n childhood and i n history. In John M acNamara, ed., Language Learning and Thought, 1 85-2 1 4. New York: Academic Press. S mith, Henry. 1 996. Restrictiveness ill Case Theory. Cambridge: Cambridge University Press. (Cambridge Studies in Lingui stics 78) S m i th, I . R. 1 987. Realign ment and other convergence phenomena. University ofMelbourne Working Papers in Lillguistics 4: 67-76. Sperber, Dan , and Dei rdre Wi lson. 1 995 [ 1 986] . Relevance, Communication alld Cognition, 2nd edition. Cambridge, M A : Harvard University Press. Steever, Sanford B., Carol A. Walker, and Sali koko S. M u fwene, eds. 1 976. Chicago lin guistic Society 12: Papers fivm the Parasession 011 Diachronic Syntax. Chicago: Chicago Linguistic Society.
References
261
Stei n, Dieter. 1 990a. The Semantics of Syntactic Change: Aspects of tile Evolution of do ill English. Berlin: M outon de Gruyter. (Trends i n Linguistics. Studies and Monographs 47) Stei n, Dieter. I 990b. Functional differentiation in the emerging standard language: the evoluti on of a morphological discourse and style marker. In Henning Andersen and Konrad Koerner, eds., Historical linguistics 1 987: Papersfrom the 8th International Conference on Historical Linguistics. lille. August 30 September 4, 1987, 489-98. Amsterdam: Benjamins. (Current Issues in Lingui s�ic Theory 66) Stern. Gustav. 1 93 1 . Meaning and Change of Meaning: with Special Reference to the English Language. Bloomington: I ndiana University Press. Stoffel, C. 1 90 I . Intensives and Down-toners: a Study in English Adverbs. Heidelberg: Winter. (Anglische Forschungen I ) Sun, Chaofen. 1 996. Word-Order Change and Gramlllaticalization in the History ofChinese. Stanford : Stanford Universi ty Press. Svorou, Soteria. 1 993. The Grammar of Space. A msterdam: Benjami ns. (Typological S tudies in Language 25) Sweetser, Eve E. 1 988. Grammaticalization and semantic bleaching. In Axmaker et al. , eds. , 389-405 . Sweetser, Eve E. 1 990. From Etymology to Pragmatics: Metaplwrical and Cultural Aspects ofSemantic Structure. Cambridge: Cambridge University Press. (Cambridge Studies in Linguistics 54) Tabor, Whitney. 1 994a. Syntactic Innovation: a Connecti onist Model . Unpubl ished PhD Dissertation, Stanford University. Tabor, Whitney. 1 994b. The gradual development of degree modi fier sort of and kind of: a corpus proximity model . In Beals et aI. , eds. , 45 1 -65. Tabor, Whitney, and Eli zabeth Closs Traugott, 1 998. Structural scope expansion and gram maticalization. In G i acalone Ramat and Hopper, eds., 229-72. Talmy, Leonard. 1 976. Semantic causative types. In Masayoshi Shibatani , ed. , TIle Gram mar of Causative Constructions, 43- 1 1 6. New York: Academic Press. (Syntax and Semantics, vol . 6) Tal my, Leonard . 1 983. How language structures space. I n Herbert L. Pick and Linda P. Acredolo, eds., Spatial Orientation: Theory, Research and Application, 225-82. New York: Plenum Press. Tal my, Leonard. 1 985. Lexicalization patterns: semantic structure i n lexical forms. In S hopen, ed. , vol . 3 : 57- 1 49. Tal my, Leonard. 1 988. Force dynamics in l anguage and cognition. Cognitive Science 2: 49- 1 00. Tal my, Leonard. 2000. Toward a Cognitive Semantics. Cambridge, MA: MIT Press. Thomason, Sarah G. 200 1 . 4Jnguage Contact: all Introduction. Washington: Georgetown University Press. Thomason, Sarah Grey, and Terrence Kaufman. 1 988. Lallguage Contact, Creolizatioll, and Genetic linguistics. Berkeley: University of Cal i fornia Press. Thompson, Sandra A., and Robert Longacre. 1 985. Adverbial clauses. In Shopen, ed. , vol. 2: 1 7 1 -234. Thompson, Sandra A.. and Anthony Mulac. 1 99 1 . A quanti tative perspective on the gram maticization of epistemic parentheticals in Engl ish. In Traugott and Heine, eds., vol. 2: 3 1 3-29. -
262
References
Ti mberlake. Alan. 1 977. Reanalysis and actualization in syntactic change. In Li. ed . . 1 4 1 -80. Tomasel lo. M ichael . 1 999. The Cultural Origins of Human Cognition . Cambridge. M A : Harvard Uni versity Press. Trask. R. L. 1 996. Historical Linguistics. London: Arnold. Traugott. El izabeth [Closs]. 1 965. Diachronic syntax and generative grammar. Language 4 1 : 402- 1 5 . Traugott. Elizabeth Closs. 1 976. Pidgins. creoles. and the origins of Vernacular B l ack English. In Deborah Sears Harri son and Tom Trabasso. eds . Black English: a Sem inar. 57-93. H i l l sdale. NJ : Lawrence Erlbau m Associates. Traugott. El izabeth Closs. 1 978. On the expression of spatio-temporal relations. In Green berg et al.. eds vol . 3: 369-400. Traugott. Elizabeth Closs. 1 982. From propositional to textual and expressive meani ngs: some semantic-pragmatic aspects of grammatical i zation. I n Wi nfred P. Leh mann and Yakov Mal kiel . eds Perspectives 0 1 1 Historical Linguistics. 245-7 1 . Amsterdam: Benjamins. (Current I ssues in Linguistic Theory 24) Traugott. Elizabeth Closs. I 985a. "Conventional" and "dead" metaphors revisi ted . In Paprotte and Di rven. eds 1 7-56. Traugott, Eli zabeth Closs. I 985b. Condi tional markers. In Hai man, ed. , 289-307. Traugott, Elizabeth Closs. 1 989. On the rise of epistemic meani ngs i n Engl ish: an exa mple of subjectification i n semantic change. Lallguage 65: 3 1 -55. Traugott. Elizabeth Closs. 1 992. Old Engl ish syntax. I n Richard Hogg. ed. , The Cambridge History ofEllglish, vol . I : Old English: the Beginning to 1066. 1 68-289. Cambridge: Cambridge University Press. Traugott, Eli zabeth Cl oss. 1 995. Subjecti fication in grammatica l i zation. In Dieter Stein and Susan Wright, eds. , Subjectivity and Subjectivisation in Language. 3 I -54. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs. 1 996. Grammatical ization and lexical izati on. In Keith B rown and Jim M i l l er, eds Concise Encyclopedia of Syntactic Theories. 1 8 1 -7 . Oxford : Pergamon. Traugott, Elizabeth Closs. 1 999. The role of pragmatics i n a theory of semantic change. i n Jef Verschueren, ed. , Pragmatics in 1 998: Selected Papers from the 6th Illtema tional Pragmatics Conferellce. vol . 2 : 93- 1 02. Antwerp: I nternational Pragmatics Assoc. Traugott. Elizabeth Closs. Forthcoming. Lexicalization and gram matical izati on. In D . A l a n Cruse, Franz Hundsnurscher, Michael Job, a n d Peter Rol f Lutzeier. eds . . Lexikologie-Lexicology, vol . 3. Berl i n : Wal ter de Gruyter. Traugott, Elizabeth Closs. and Richard B. Dasher. 2002. Regularity in Semantic C/wnge. Cambridge: Cambridge University Press. (Cambridge S tudies in Li ngu istics 96) Traugott, Elizabeth Closs, and Bernd Heine. eds. 1 99 1 . Approaches to Grammalicalizatioll . Amsterd a m : Benjamins, 2 vols. (Ty pological Studies in Language 1 9) Traugott, Elizabeth Closs. and Ekkehard Konig. 1 99 1 . The semanti cs-pragmatics of grammatica l i zation revisited. I n Traugott and Heine. eds. , vol . I : 1 89-2 1 8 . Traugott. Elizabeth Closs, Rebecca La Brum, and Susan Shepherd , eds . . 1 980. Papers/rom the 4th Illlemational Conference on Historical Linguistics. Stanford, Marcil 26-30, 1979. Amsterdam: Benjamins. (Current I ssues in Linguistic Theory 1 4) Trubetzkoy. N. 1 929. Zur allgemeinen Theorie der phonologischen Vokalsysteme. TrU l'aux du Cercle Lillgllistique de Prague I : 39-67 . .
.•
.•
.•
.•
References
263
Tu rchetta. Barbara. 1 998. On the appl ication of the notion of gra mmatical ization to Wesl A frican Pidgi n English. In Giacalone Ramat and Hopper. eds 273-88 . U l l munn. Stephen. 1 962. Semantics: all illtroduction to the Sciellce o/ Meaning. Oxford : B l ackwel l . Valdman. Albert . a n d Arnold H ighfield. eds. 1 980. Theoretical Orielltatiolls ill Creole Studies. New York : Academic Press. Val i d u v i. Enric. 1 992. The Ill/ormatiol/al Componelll. New York: Garland . Vennemunn. Theo. 1 975. An explanation of dri ft. I n Li . ed. , 269-305. Vergne. George H . de la. 1 989. Hawaiiall Sketches. San Francisco: H . S. Crocker. Vincent. Nigel . 1 978. Is sound change teleological? In Jacek Fisiak, ed. , Recent Developmellls in Historical Phonology. 409-30. The Hague: M outon. ([rends in Lingui stics. Studies and Monographs 4) Vincent, Nigel . 1 979. Word order and gram matical theory. In Jiirgen M. Meisel and Martin D. Pam . eds Lillear Order and Generative Theory, 1 -22. Amsterdam: Benjamins. (Current Issues i n Li ngu i stic Theory 7) Vi ncent. Nigel . 1 982. The development of the auxiliaries H A B E R E and E S S E i n Romunce. In Nigel Vi ncent and M artin Harris. eds. , Studies in the Romance Verb. 7 1 -96. London: Croom Hel m . Vi ncent. Nigel. 1 995. Exuptation and gra mmaticalizalion. In Henning Andersen. e d Historical Lingllistics 1993. Selected Papers /rom the I l th IlItemational COIl/er ence on Historical Linguistics. 43 3-48. Amsterdum: Benjamins. (Current Issues i n Linguistic Theory 1 24) Visser. F. Th . 1 963-73. An Historical Syntax o/the English Language. Parts 1-/ I I. Leiden : E. J. Brill. Wackernagel. J acob. 1 892. O ber ein Gesetz der i ndogermanischen Wortstellung. Illdoger manische Forsc/ltmgen I : 333-435. Warner. A nthony. 1 982. Complementation ill Middle English and the Methodology 0/ Historical Syntax: a Study o/ the Wyclifite Sermons. University Park: Pennsylvania State University Press. Warner. A nthony . 1 993. English A uxiliaries: Structure and History. Cambridge: Cambridge U n i versity Press. (Cambridge Studies in Linguistics 66) Walkins. Calvert. 1 964. Prel i m i naries to the reconstruction of Indo-European sentence structure. In H. Lunt, ed Proceedings o/the 9th IlItel7lational Congress o/Linguists ( Cambridge. Mass.), 1 035-45. The H ague: Mouton. Wei"n reich. Uriel . 1 953. Lallguages ill COil tact: Findings and Problems. The Hague: Mouton. 8th pri nting. Wei n reich. Uriel. Wi lliam Labov, and Marvi n I. Herzog. 1 968. Empirical foundations for a theory of language change. In Lehmann and Malkiel. eds 95- 1 89. Werner. Heinz. and Bernard Kaplan. 1 963 . Symbol-Formatioll: all Organismic Developmelllal Approach to Language and the Expression o/ Thougllt. New York: Wiley. Whinnom. Kei th. 1 97 1 . Linguistic hybridi zation and the "special case" of pidgins and creoles. In Hy mes. ed. , 9 1 - 1 1 5. Wi nford . Dona ld. 2000. Tense and aspect i n S ranan and the creole prototype. In McWhorter. ed .. 3 83-442 . Wi scher, l i se. 2000. Grammatica l i zation versus lexicalization - "methi n ks" there is some con fusion. In Fi scher, Rosenbach. and Stei n. eds., 355-70. .•
.•
.•
.•
.•
264
References
Wi scher, lise, and Gabriele Diewald, eds. 2002. New Reflections on Grammaticaliza tion: Proceedingsfrom the Intemational Symposium on G rammaticalizatioll, 1 7-19 June 1 999, Potsdam, Germany. Amsterdam: Benjamins. (TYpological Studies in
Language 49) Zwicky, Arnold. 1 985. Cli tics and particles. Language 6 1 : 283-305. Zwicky, Arnold. 1 993. Heads, bases and functions. In Grevi lle G. Corbett, Scott McGlashen, and Norman Fraser, eds., Heads in Grammatical TheOlY, 292-3 1 5 . Cambridge: Cambridge University Press.
Index of names
Abraham, Werner 35
Bogers, Marije xvi
Adams, Karen L. 1 0 1 -2
Bol inger, Dwight 28, 78, 1 8 1
Aijmer, Karin 97
Bossong, Georg 1 66-8
Aikhenvald, Alexandra Y. 37
Boyland, Joyce 1 27
Akatsuka, Noriko 1 86
Braunmiiller, Kurt 206
Allan, Keith 1 02
Breal , Michel 1 28
Al len, Cynthia L. 5 2
Bresnan, Joan 1 33
Andersen, Henning 6, 7, 4 1 , 42-3, 47, 1 3 1 , 1 37,
Brinton, Laurel 37, 49, 88, 1 29, 1 34
1 38, 1 45-7, 1 63-4, 232 Anderson, Lloyd 1 1 5, 1 87 Anderson, Stephen 1 44 Antiila, Raimo 40, 4 1 , 42-3, 58, 64, 88, 1 03, 1 05 Archangeli, Diana 1 33 Atlas, Jay D. 79 Austin, Peter 1 76
Browman, Catherine P. 1 27
Bach, Eminon 63
Brown, Penelope 222, 225-6 Bruyn, Adrienne 37, 22 1 , 228-9 Burridge, Kate 1 02, 1 37 Burrow, T. S. 1 50 Butt, Miriam 1 1 4, 1 7 1 Bybee, Joan L. xvi, 4, 30, 33, 35-6, 44, 73-5,
77-8, 85-6, 89, 97, 1 0 1 , 1 04, 1 1 3, 1 27, 1 34, 1 38, 1 42, 1 50-4, 1 76, 208, 2 1 7, 226, 236 Byrne, Francis 220
Bailey, Charles-James N. 236 Baker, Philip 37, 2 1 7, 22 1 Bakker, P. 2 1 3
Campbel l, Lyle 25, 34, 35, 40, 5 1 , 59, 1 32, 1 37,
1 7 1 , 20 1 , 209, 234
Barcelona, Antonio 88
Cameron, Angus 36 Carrington, Lawrence D. 2 1 9
Baldi, Philip 1 59 Baronian, Luc xvi
Chafe, Wallace 36, 6 1 , 1 4 1 , 1 8 1 , 1 87
Barth-Weingarten, Dagmar 1 29 Bates, E�izabeth 72
Charpentier, J . - M . 228
Bean, Marian C. 68 Benven iste, Emile 26, 62
Chaudenson, Robert 223 Chomsky, Noam 43, 45, 1 53
Berndt, C. H. 1 85 Berndt, R. M. 1 85
Chung, Sandra 1 7 1
Besnier, N i ko 25
Clark, Brady Z. xvi
Chambers, 1. K. 44
Clancy, Patricia 36, 37
Bhattacharya, S . 1 50
Clark, Eve V. 53
B ickerton, Derek 30, 2 1 3, 2 1 7-22, 229 B innick, Robert I . 1 53
Claudi , Ulrike 6, I I , 1 9, 33, 59, 73, 75, 84-5,
Birner, Betty 1. 6 1 B isang, Walter 37
Cole, Peter I I , 8 1 Company, Concepcion 1 30 Comrie, Bernard 30, 1 04, I I I , 1 25-6, 1 4 1 ,
Bjork, Robert E. 235 B lakemore, Diane 72, 80 Boersma, Pau l 1 33
92, 1 1 0, 1 1 1 , 234
1 49-50, 1 52, 1 69, 1 87, 1 88, 202-3, 234 Conklin, Nancy Frances 1 02
265
266
Index a/names
Couper-Kuhlen. Elisabeth 1 29. 1 8 1 Cowie. Claire xvi Craig. Colette 90. 1 1 4 Croft. Wi l l iam xvi. 1 7. 35. 46. 7 1 . 78. 1 07. 1 33.
1 37. 1 66 Cruse. D. A. 1 0 1 Dahl. Osten 75. 82. 85. I I I . 1 38. 1 42. 1 52. 1 53.
234 Dasher. Richard B . 35. 37. 82. 1 29 Deacon. Terence W. 46 DeGraff. Michel 2 1 3. 2 1 9. 22 1 DeLancey. Scott 1 66. 1 68. 204-5 Denison. David 55. 95 Deutscher. Guy 1 90. 1 94-6 Dietrich. W. 1 24 Di ngare. Shipra 1 33 Dirven. Rene 87 Disterheft. Dorothy 1 37-8 Dixon. R. M. W. 1 65 . 1 68. 1 73 Downing. W. H . 226 Dressler. Wolfgang U . 72. 1 09, 1 42
Genett i. Carol 1 88
Giacalone Ramat. Anna 35 Gianto. A. xvi G i l l ieron. Jules Lou i s 1 03 G iv6n. T. x v i . 1 2. 20. 26. 27. 29-30. 32. 33. 46. 73-4. 1 0 I . 1 30. 1 42. 1 48-50. 1 68-7 1 . 1 76. 1 78 Glass. A n t hon y x v i Good l uck. Helen 3 6 Goldstein. Louis 1 27 G reen. Georgia M . 76. 85 G reenbaum. Sidney 1 0. 1 25 Greenberg. Joseph H. x v i . 1 7 . 28-9. 30. 46. 60. 7 1 . 1 35-6. 1 55 . 1 65-6. 1 74 G rice. H. Paul 78-9. 8 1 Groefsema. Malj olein 77 G Unther. S usanne 9 1 . 2 1 0- 1 I Haase. Martin 37 Hagege. Claude xvi. 1 35 . 1 38 Hai man. John 27. 35. 6 1 . 78. 1 27. 1 55 . 1 76.
1 82. 1 84. 1 86. 207 . 236
Dryer. Matthew S. 60
Hale. Kenneth 1 55 . 1 62. 1 97
Du Bois, John W. 35. 36, 72. 1 68-7 1
Halle. Morris 43-4. 64 . 97. 222 Hall iday. M. A. K. 4. 6. 1 85 . 235 Hal pern . Aaron 5 . 1 44 Hammond. M. 4 Han·is. Al ice 35. 40. 5 1 . 59. 1 7 1 . 20 1 . 209. 234 Harris. Mart i n 62. 94. 1 87. 2 1 1 Harri s. Roy 20 Hasan . Ruqaia 1 85 Haspel math. Mart i n 4-5 . 34. 35. 1 32. 1 42. 1 50. 1 89. 20 1 Hawkins. John A. 60 Hayes. B ruce 1 33 Healey. A ntonette d i Paolo 234 Hei ne. Bernd xvi. 6. I I . 1 9. 30. 32-3 . 53. 59. 73. 75. 84-5. 92. 93-4. 1 08. 1 1 0. 1 1 3. 1 1 8. 1 24. 1 27-8. 1 33-4. 1 37. 2 1 2. 234. 236 Held. Warren H J r 1 99-202 Herzog. Marvin 46. 47. 2 1 2. 235 Heusler. Andreas 1 59-60 Highfield. Anthony 2 1 3 H i m melmann. Nikolaus 35 H i nt. Mati 63 Hock. Hans Henrich 25. 40. 65 . 1 62-3 Hockett. Charles F. 1 57 Hodge. Carlton T. 9 Hoenigswald. Henry 40 Holm. John 2 1 4 Hook. Peter E. 37. 1 1 2- 1 4. 1 1 6. 1 30
Ebert. Robert 2 5 Eckert, Penelope 4 4 . 47 El leg ard, Alvar 95 Faltz. Leonard 1 1 5 Ferguson, Charles A. 1 7. 30 Fernandez-Vest, M. 37, 1 45 Fill more, Charles J. 1 0 Finegan. Edward 25 Fischer. Olga x v i , 37, 1 90 Fleischman. S u zanne x v i . 32. 52. 54. 62-3. 65 .
1 0 1 . 235 Foley. William A . 1 52 . 1 76. 2 1 4. 236 Ford. Cecilia 1 8 1 Fox. Barbara 36 Fraj zyngier. Zygmunt 37. 1 32. 1 85 Frances. W. Nelson 36 Fraser. Bruce 37 Fries. Charles C . 67 Gabelentz. Georg von der 20- 1 . 93-4 Gamillscheg. Ernst 1 1 7 . 1 89 Garrett. Andrew x v i . 9 5 . 1 7 1 Geeraells. Dirk 1 02 Geis. Michael L. 8 1 Gelb. Ignace J . 1 95
.•
.
Index ofllames Hopper. Pau l J . 1 2. 26, 34. 35--6, 49, 50, 73, 96. 1 07, 1 1 6, 1 1 8-20. 1 25 . 1 30. 1 73, 1 92, 208 Horn. Laurence R. 79, 8 1 H u mboldt. Wilhelm von 1 9-2 1 , 26. 30 H undl, Marianne 1 30 H iin nemeyer, Friederike 6, I I . 1 9, 33. 59. 73, 75, 84-5 . 92, 1 1 0. 1 1 3 . 234 H u rren. Tony 234 Hyman. Larry M. 1 55 Hymes, Del l 1 4 1 . 2 1 3 I halainen. Ossi 95 Jackendoff, Ray 84 J akobson. Roman 27. 64, 87 Janda. Richard D. 1 32. 1 36--7, 1 73 Je lTers, Robel1 J . 5. 1 62 Johansson. Slig 36 Johnson, Mark 77, 84 Jones. M i chael A. 55 Joseph, B rian D. 34, 1 32, 1 73 J USIUS, Carol 1 99-200 Kaburak i, Etseko 1 66 Kaisse, Ellen M . 1 43 Kaplan. Bernard 73 Kaufman, Terrence 45, 2 1 2. 23 1 -3 Kay. Paul 1 0 Keenan. Edward L . 1 97-8. 200, 202-3 Keesi ng, Roger M. 227-8 Kel ler, Rudi 74 Kemenade. Ans van 34, 35, 37. 60, 68, 1 33, 1 38 Kemmer. Suzanne )lvi, 77. 1 1 5 , 1 23 , 1 6 1 Kienast. B u rkhardl 1 95 K i l ian-Halz. Christa 37 K i parsky. Pau l )l v i , 40, 64, 1 90. 222 Kiss, Katalin E. 6 1 K l ai man, M . H . 1 99. 200 K l amer. M arian 1 87 K l avans, Judith L. 5 Konig, Ekkehard 30, 73, 75. 90, 1 08, 1 87, 21 1 Konmann, Berndt 1 08 KOSIer. Jan 1 94 Kovecses, Zoltan 84. 88 K rach, An thony S. 56--7 Krug, Man fred 1 3 . 35. 36. 5 5 , 58, 73, 1 27 K ru pa. Viklor 1 56 Kucera. Henry 36 Kuno. Susumu 1 66
267
Kuryfowicz, Jerzy 25, 64--6, 1 0 1 Kuteva, Tania 32, 75. 206--7. 2 1 2 Kyto, Merja 1 30 Labov, Will iam 44, 46, 47, 50. 1 38. 2 1 2. 232. 235. 236 Lakoff, George 77, 84, 1 83 Lakoff, Robin 1 00 Lambrecht. Knud 1 5 , 6 1 , 223 Langacker, Ronald W. 35, 5 1 , 72-3, 92, 94, 1 23, 1 77, 1 79, 236 Langendoen, D. Terence 1 33 Lass, Roger )lvi, 34, 7 1 , 1 32, 1 35 Laury, Rilva 36, 1 29 Lausberg, Heinrich 1 40- 1 Leech, Geoffrey 1 0, 36 Lehi ste, lise 1 62 Lehmann, Christian )lvi, [ 9, 25, 30-2, 35, 92 . [ [ 0, [ 32. [ 35 , [ 38-9, [ 77-8, [ 82, 236 Lehman n. Wi n fred P. 60 Le Page, Roben B. 2 [ 9 Lerner, Gene H . 36 Lev i n , Belh 1 0 1 -2 Lev i nson, Stephen C. 30, 75, 76, 79, 82, 84 Li, Charles N. 27-9. 60, 6 [ , 234 Lichtenberk, Frantisek 46. 84 Lightfoot, David 43, 45, 47, 55, 5 7-8, 75 Lightfoot. Doug[as )l v i li ndstrom. Therese x v i Lipka, Leonhard [ 34 Longacre, Roben [ 76--7, 1 80, [ 82 Lord, Carol [ 3, 96-7, [ 87, 234 Los, Bettelou [ 90 Liidtke, He[mut 94 Luraghi, Silvia [ 37 Lyons, John 53, 1 0 [ MacWhinney, B rian 72 Mair, Christian 1 30 Ma[kiel, Yakov 1 00, 1 03 Mann ing, Christopher 1 33 Manin, Janice [ 30 Masica, Col in P. 2 1 3 Mas[ova, Elena 7 [ Matth iessen, Cluistian [ 77, [ 83 Matisoff, James A. 1 57-8 Matsumoto, Yo 2 1 0- 1 1 McCawley. James D. 7 7 McE[hinney, Bonnie 229 McMahon. Apri l 25, 40 McWhoner, John H . )lvi, 2 1 9, 220, 223
268
Index a/ names
Meillet, Antoine 1 9, 2 1 -6, 28, 59, 93--4, 99, 1 33, 1 8 1 M ikola, Tibor 66 Milroy, James 44 Mitchell, Bruce 83, 1 9 1
RicHord, John R. xvi, 2 1 3, 2 1 8, 22 1 , 235 Rissanen, Matti 36 Rittel, Theodora 1 46 Roberts, Ian 35, 52-5, 59, 1 33, 235 Roberts, Sarah 1. xvi, 2 1 5, 2 1 6, 2 1 9, 22 1 -2
M ithun, Marianne 63, 1 42, 1 50, 1 55, 1 8 1 , 1 85 , 207
Romaine, Suzanne 37, 50, 7 1 , 203, 2 1 3, 2 1 4,
Moravcsik, Edith 1 7, 30
Rosenbach, Anette 37, 235
Mufwene, Salikoko S. 2 1 3, 22 1 , 223, 229
Rude, Noel 1 56
226, 230, 236
Muhlhausler, Peter 2 1 5, 2 1 9 Mulac, Anthony 1 30, 208-9
Sampson, Geoffrey 45
Muysken, Peter 2 1 3
Samuels, M. L. 1 03
Newmeyer, Frederick J . xvi, 34, 35, 1 32-3, 1 38
Sandberg, Kristian 2 1 3 Sankoff, G i l lian 30, 22 1 -2, 225-6 Sapi r, Edward 1 00
Nichols, Johanna 1 3 1 , 1 87, 232
Sapir, J. David 84, 87, 1 00
Noonan, M ichael 4, 1 73, 1 76, 1 89 Norde, Muriel 1 32, 1 34, 1 36, 1 37
Saussure, Ferdinand de 2, 26, 3 1 Schachter, Pau l 1 1 9 Schegloff, Emanuel 36
Obermeier, Anita 235 Ochs, Elinor 36 O' Connor, Mary Catherine 1 0
Scheibman, Joanne 35, 73 Schneider, G. D. 2 1 6 Schwegler, Armin 65, 1 23, 1 42
Ohori, Toshio 37 O ' Nei l, Wayne 1 92 , 1 97-8
Schwenter, Scott x vi
Ono, Tsuyoshi 36
Seiler, Hansjakob 30
Searle, John R. 84 Selting, Margret 1 8 1
Pagliuca, Will iam 30, 33, 75, 85-6, 89, 97, 1 0 1 , 1 04, 1 1 3, 1 34, 1 38, 2 1 7, 226 Patrick, Peter 2 1 8 Pau l, Hermann 43-5
Sharma, Devyani xvi Shi, Yuzhi 37 Shibatani , Masayoshi 1 7 1 Shopen, Ti mothy 1 76
Peirce, Charles Sanders 27, 42-3
Singler, John V. 220
Perez, Aveline 89 Perkins, Revere 30, 33, 74, 85, 1 1 3, 1 34, 1 38, 2 1 7, 226
Siobin, Dan I. 30, 44, 73
Phi l pot, Brad xvi Pinkster, Harm 52-4, 62
Silverstein, M ichael 1 65 Smith, Henry 1 57-8 Smith, I. R . 230 Smolen sky, Paul xvi, 1 33
Plag, Ingo 2 1 7- 1 8, 22 1 , 229
Sperber, Dan 76, 79 Stein, Dieter 37, 95, 1 30, 1 37
Plank, Frans 55, 1 6 8 Prince, Alan x v i , 1 33
Stoffel, C. 1 22
Pintzuk, Susan 68
Prince, Ellen 6 1 Pullum, Geoffrey K . 1 42 Quirk, Randolph 1 0, 234 Radden, Giinter x v i Ramat, Paolo 1 34 Rankin, Robert L. 1 32 Ravid, Dorit Diskin 44 Reddy, Michael J. 84 Reh, Mechthi l d 32, 59, 93, 94, 1 1 8, 1 24, 234
Stern, G ustav 88 Sun, Chaofen 37 Svartvik, Jan 1 0 Svourou, Soteria 8 5 Sweetser, Eve E . xvi, 33, 75, 77, 84, 86-7, 94, 21 1 Syea, Anand 37 Tabor, Whitney 35, 68, 1 38-9 Tabouret-Ke l ler, A. 2 1 9 Tal my, Leonard 84, 86, 235 Taylor, Ann 2 1 3
Index a/ names Taylor, Talbot 20
Venezky, Richard L. 234
Thomason, Sarah Grey 45, 2 1 2, 2 1 7, 230
Vennemann, Theo 60, 1 00
Thompson, Sandra A . xvi, 27-9, 36, 6 1 ,
Vergne, George H. de la 2 1 5- 1 6
1 07- 1 1 , 1 27, 1 30, 1 76-7, 1 80, 1 83-4,
Vincent, Nigel xvi, 60, 62, 65, 68, 74, 1 36
208-9, 234, 236
Visser, F. Th. 1 90
Timberlake, Alan 1 04--6, 1 3 1 , 232, 235 Tomasello, Michael 45-6 Trask, R. L. 25
Wackernagel, Jacob 1 44-5
Traugott, Elizabeth Closs 33, 35, 37, 55, 73, 75,
Warner, Anthony 5 5 , 1 90, 1 94
82, 85, 90, 92, 1 1 5, 1 29, 1 34, 1 3 8-9, 1 86, 191, 21 1, 216 Trubetzkoy, N . 1 55 Turchetta, Barbara 37
Ward, Gregory 6 1 Watkins, Calvert 26, 1 4 1 Weinreich, Uriel 46, 47, 2 1 2, 235 Werner, Heinz 73 Wheeler, Max xvi Whinnom, Keith 223
U l l mann, Stephen 1 0 1
Wilson, Deirdre 76, 79 Win ford, Donald 2 1 7
Valdman, Albert 2 1 3, 2 1 9
Wischer, lise 1 35
Valduvf, Enric 6 1 Van Val in, Robert D . 1 52, 1 76, 236
Zwicky, Arnold xvi, 5, 8 1 , 1 42, 235, 236
269
I ndex of l a nguages
A k kadian 1 90. 1 94-6 A leut. Mednyi 2 1 2 A m haric 63. 1 48-9 A rabic 1 03 A rmenian. Classical 1 26. 20 1 Austronesian. Eastern Oceanic 227 B asque 60. 1 68-9 Bengal i 1 1 3 Bislama. Vanuatu 22 7-8 B u l garian 1 63-4. 207 Cape Verde Creole 2 1 4 Cayuga 1 5 3 Chickasaw 1 86 Chi nese 28. 1 80. 1 86 Danish 8. 6 1 . 1 59. 206-7 Dutch 206 Egyptian 1 96 English 1 0- 1 3 . 60. 207-9 Ol d (OE) (c. 600- 1 1 25) 1 2 . 39. 48-9. 5 1 . 56. 79. 88. 89-9 1 . 9 5 . 97. 1 02-3 . 1 1 9. 1 23. 1 36. 1 43 . 1 72. 1 90, 1 9 1 -4. 1 97-8. 202 M iddle ( M E) (c. 1 1 25- 1 500) 40. 49. 5 7-8. 95. 1 03. 1 23. 1 36 . 1 37-8. 1 43. 1 90. 1 94. 1 99 Earl y New ' Engl ish (ENE) (C. 1 500- 1 750) 56-7 . 89. 9 1 Estonian 59. 63. 1 2 5-6. 1 37 Evenk i 1 88 Ewe 1 3- 1 6. 49. 1 26 Fa d ' A mbu 2 1 7 Fi n n ish 1 04-5. 1 29 French 8. 1 5. 22-3 . 26. 3 1 -2. 52-5 . 65-6. 88, 94, 1 02. 1 1 7- 1 8. 1 23-4. 1 34. 1 40-2, 1 48-9. 1 56-7. 1 72, 1 89. 223-4
270
Ga 35. 1 1 5 German 8. 24. 35, 1 1 0. 1 34. 1 37. 1 72 . 2 1 0 Gothic 24 G reek Classical 24. 99. 1 87 Homeric 1 43-4 G ujerat i 1 1 3 Gunwi nggu 1 85 G uyanese Creole 2 1 6- 1 7 . 2 1 8. 235 Haitian Creole 2 1 4 " Hawai ian Creole 2 1 4. 2 1 8. 220 H awaiian Pidgin 2 1 4. 2 1 5 Hebrew 60. 203 Bibl ical 1 1 5 . 1 30 H i ndi 8. 1 1 2- 1 4. 1 1 6 H i ndi-Urdu 1 1 3- 1 4. 1 1 6 H i n ite 1 86. 1 99-202 H ua 1 82. 1 83. 1 86 H u ngarian I I I I ndo-Aryan see H i nd i I ndo-European 9. 1 44 I ndonesian 5. 1 44 . 1 86 I ranian see Persian I rish 1 73 Italian 8. 5 5 Japanese 6 0 . 1 8 1 . 1 86. 209- 1 0 Kashmiri 1 1 3 Kewa 1 52 K waio 227-8 Lahu 1 5 7-8 Lati n 5 . 8. 1 5 . 2 1 . 23. 26. 3 1 . 5 2-5 . 64-5. 88. 94. 96, 1 02. I I I . 1 26. 1 40- 1 . 1 44. 1 48. 1 57. 1 5 8-9. 1 80. 1 83-4. 1 86. 1 89. 1 99 Lhasa 204-6
Index of languages Malay 60. 1 1 9-2 1 . 1 7 1 Maori 1 55-6. 1 62. 1 88-9 Maral hi 1 1 3- 1 4. 1 1 6 Marwari 1 1 3 Masai 60 Maurilian Creole 2 1 7 Melanesian Pidgin 228 M i angkabau 1 86 Mongolian Buryat 1 4 1 Classical 1 49-50 Kal myk 1 49 Newari 1 88 Nez Perce 1 56-7 Old Norse 8. 1 59 Olgolo 1 73 Palenquero 2 1 4 Pal i 1 62-3 Pa piamenlu 2 1 4 Pe ngo 1 50 Persian 1 65-8 Pol ish 1 45-7 Port uguese 229-30
27 1
Romanian. ISlro- 8 Russian 63. 1 3 1 . 1 86 Saami 1 45 SacapuJtec 1 69 Samoyedic 66 Sanskrit 1 62-3. 1 86 Saramaccan Creole 2 1 7. 220- 1 Shina 1 1 3 S iouan 1 32 Sololl1ons Pij i n 227. 228 Spanish 5. 1 03. 1 30. 1 4 1 -2 Sranan Creole 2 1 8. 228-9 S ri Lanka Portuguese Creole 229-30 S ursel van 1 23 Swahili 8. 3 1 . 60. 1 86 Tagalog 1 86 Tamil 200. 229-30 Tel ugu 1 1 5 Ti w i 1 52 Tok Pisin 2 1 4- 1 5 . 222. 225-7 Turkish 1 55 Vetmbao 227-8
Qucchua 60
Welsh 60 West African Pidgin Engl ish 2 1 4. 2 1 5. 2 1 6. 224
Ran1