Studies on Grammaticalization 9783110211764, 9783110205824

Citation preview

Studies on Grammaticalization

≥

Trends in Linguistics Studies and Monographs 205

Editors

Walter Bisang Hans Henrich Hock (main editor for this volume)

Werner Winter

Mouton de Gruyter Berlin · New York

Studies on Grammaticalization

edited by

Elisabeth Verhoeven Stavros Skopeteas Yong-Min Shin Yoko Nishina Johannes Helmbrecht

Mouton de Gruyter Berlin · New York

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

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

Library of Congress Cataloging-in-Publication Data is available at http://catalog.loc.gov.

ISBN 978-3-11-020582-4 ISSN 1861-4302 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. ” Copyright 2008 by Walter de Gruyter GmbH & Co. KG, D-10785 Berlin All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without permission in writing from the publisher. Cover design: Christopher Schneider, Berlin. Printed in Germany.

Preface

The articles in this book are a joint effort to advance research in the field of grammaticalization and grammaticalization theory. Approximately two years ago, the authors of the sixteen contributions in this volume agreed to participate in this project. Most of them are long-standing experts in grammaticalization and all of the contributions are reports of the results of their own original research in this field. The topics which are dealt with cover the grammaticalization of nominal categories, verbal categories and grammaticalization of more complex syntactic constructions in various and typologically quite different languages. In addition, theoretical questions such as the hypothesis of the unidirectionality of grammaticalization processes are discussed. As the subtitle of this volume indicates, the papers of this volume are dedicated to Christian Lehmann who turned sixty in November 2008. Christian Lehmann is one of the leading linguists of our times and was enormously influential in many fields and areas of linguistic research. This holds in particular for the field of grammaticalization. It is no exaggeration to say that Christian Lehmann established grammaticalization as a research paradigm in the first place and prepared the field for systematic research with his famous book “Thoughts on grammaticalization”. It is in this sense that Christian Lehmann’s research paved the way for the contributions in this volume. The present volume relates to another collective volume dealing with questions within the realm of “Functional-Typological Linguistics and Language Theory”. It presents new and original research in typology and language theory and is also dedicated to Christian Lehmann and his scientific work. Today, grammaticalization is a well-established part of the functionaltypological research paradigm in linguistics which led the way to overcome (in the good Hegelian sense) the old Saussureian dichotomy between synchrony and diachrony. From a historical point of view, it was grammaticalization that evoked a new interest in the historical development and the evolution of language and languages among typologists. We would like to thank all contributors of the present volume for their participation and cooperation with us. In addition, we are grateful to Birgit

vi Preface Sievert, Monika Wendland and – of course – Anke Beck from Mouton de Gruyter who supported us in many and various ways to make this project possible. Last but not least, we would like to thank Ingrid Stitz and our student assistants (University of Regensburg) who helped with all kinds of organizational work.

The editors.

Contents

Introduction Elisabeth Verhoeven, Stavros Skopeteas, Yong-Min Shin, Yoko Nishina, and Johannes Helmbrecht

1

I. Directionality of grammaticalization processes Unidirectionality of grammaticalization in an evolutionary perspective Elena Maslova Grammaticalization and sets of form-function pairs: Encoding spatial concepts in Greek Stavros Skopeteas

15

25

II. Nominal categories and adpositions Grammaticalization of Korean numeral classifiers Myung-Chul Koo

59

Grammaticalization of spatial adpositions in Nànáfwê Amani Bohoussou and Stavros Skopeteas

77

On the grammaticalization of the German preposition von as a genitive equivalent John Ole Askedal

105

III. Verbal categories The grammaticalization of agreement in Chibchan J. Diego Quesada Decay and loss of applicatives in Siouan languages: A grammaticalization perspective Johannes Helmbrecht

121

135

viii Contents Reciprocals in the making: multiple grammaticalization in Manambu Alexandra Y. Aikhenvald The Maltese continuative: A grammaticalization borderliner Thomas Stolz and Andreas Ammann What do “do” verbs do? The semantic diversity of generalised action verbs Eva Schultze-Berndt Predicting a future change: Relative clauses of Japanese Tasaku Tsunoda

157

169

185

209

IV. Grammaticalization and constructions The catalytic function of constructional restrictions in grammaticalization Gabriele Diewald Grammaticalization of periphrastic constructions Ilse Wischer Grammaticalization in constructions: Clitic doubling with experiencers in Modern Greek Elisabeth Verhoeven

219

241

251

Grammaticalization of honorific constructions in Japanese Yoko Nishina

283

From cause to contrast: A study in semantic change Anna Giacalone Ramat and Caterina Mauri

303

Index

321

Glosses

1 2 3 A ABL ACC ACT.FOC ADJ ALL AOR AP APPL.BEN

first person second person third person cross-reference set A ablative accusative action focus adjectiviser allative aorist active participle benefactive applicative APPL.INESS inessive applicative APPL.INST instrumental applicative APPL.SUPESS superessive applicative ASS associative AUX auxiliary BEN beneficiary CAUS causative CL classifier CMPR comparative CNJ conjunction COLL collective COM comitative DAT dative DECL declarative DEF definite DEM demonstrative DET determiner DO direct object DU dual EP epenthetic sound

EXCL EXP F FOC FUT GEN H1 H2 H3 HAB HON IMP IMPF INCH INCL INDEF INS INV IO IRR LK LOC M MED N NEG NOM NPST OBL OBV PASS PAST PCL PFV PL

exclusive experiencer feminine focus future genitive honorific-stylistic honorific-respectful honorific-humble habitual honorific imperative imperfective inchoative inclusive indefinite instrumental inverse indirect object irrealis linker locative masculine medial neuter negation nominative nonpast oblique obviative passive past tense particle perfective plural

x Glosses POSS POT PRED PRES PROG PROH PROX PRS REC REFL REL

possessive potential predicative marker present progressive prohibitive proximal present reciprocal reflexive relative

SBJ SEQ SG SPEC SS SUB TERM TM TOP U

subject sequential singular specificity marker same subject subordinate terminative t-marbuta topic marker undergoer

Introduction Elisabeth Verhoeven, Stavros Skopeteas, Yong-Min Shin, Yoko Nishina, and Johannes Helmbrecht

1. Preliminaries Grammaticalization research has played a major role in the developments in language typology and functional linguistics during the last three decades. Though the study of grammaticalization does not imply a particular linguistic framework, this paradigm of research has been a crucial field of contemporary functional linguistics, since it advocates the viewpoint that the evolution of grammatical formatives and structures exhibits universal tendencies which follow from the language-independent foundations of grammar in human cognition and communication. Crucially for our understanding of grammar, grammaticalization phenomena show that language change takes place in a continuous way rather than in form of a categorical choice among clear-cut universally available options. Moreover, the fact that particular grammaticalization paths are attested across languages gives rise to hypotheses concerning their functional explanation. This book contains contributions devoted to Christian Lehmann whose seminal work on grammaticalization has inspired a large number of studies in the last two decades. The aim of the present introduction is to outline the major issues in grammaticalization research in order to embed the contributions of this volume into the contemporary research agenda. Section 2 introduces the parameters of a grammaticalization process and discusses evidence about the co-occurrence of changes at different layers of grammar. Section 3 discusses the major issues of a grammaticalization process, i.e. the beginning of the diachronic change, the grammaticalization path, and the relevance of particular constructional environments for the development of grammatical elements. Section 4 is devoted to the root of many debates with respect to the grammaticalization research, namely the issue of unidirectionality. Finally, section 5 offers an outline of the parts of this volume.

2

Elisabeth Verhoeven et al.

2. The grammaticalization phenomenon Grammaticalization is a diachronic process that refers to the loss of autonomy of a linguistic unit in the paradigmatic and syntagmatic axes (see Lehmann 2005). The loss of autonomy is diagnosed by means of changes that take place with respect to a number of parameters (or a subset of them), which are summarized in Table 1, cited from Lehmann (22002). The parameters in Table 1 are gradient, i.e., integrity, paradigmaticity, paradigmatic variability, structural scope, bondedness, and syntagmatic variability do not determine binary but scalar distinctions (see Lehmann 22002: 143– 145 about the measurement of these parameters). Hence, the processes in this table describe changes from lexical to grammatical units entailing a gradient conception of the distinction between lexical and grammatical categories. Table 1. Grammaticalization parameters (Lehmann 22002: 146) strong grammaticalization

parameter

weak grammaticalization

process

→

integrity

bundle of semantic features; possibly polysyllabic

attrition

few semantic fea→ tures; oligo- or monosegmental

paradigmaticity

item participates loosely in semantic field

paradigmaticization

→

paradigmatic variability

free choice of items according to communicative intentions

obligatorification

choice systematically constrained, → use largely obligatory

structural scope

item relates to constituent of arbitrary complexity

condensation

→

bondedness

item is independently juxtaposed

coalescence

item is affix or → even phonological feature of carrier

syntagmatic variability

item can be shifted around freely

fixation

→

small, tightly integrated paradigm

item modifies word or stem

item occupies fixed slot

Introduction

3

The first three parameters in Table 1 relate to the paradigmatic properties of grammaticalized units. The process of attrition refers to the reduction of phonological and semantic weight of the linguistic unit at issue. At the semantic layer, this process corresponds to the loss of semantic features, i.e. to ‘desemanticization’ or ‘bleaching’. Desemanticization processes are discussed by several articles in this volume: Stolz and Ammann examine the desemanticization of lexical verbs that develop to aspectual auxiliaries in Maltese; Bohoussou and Skopeteas discuss evidence concerning the desemanticization of body part nouns and their development to adpositions in Nànáfwê; Skopeteas presents the reflexes of desemanticization in the text frequency of Greek prepositions; Diewald shows the gradient loss of contextual restrictions in the occurrences of the modal particle ruhig in German; Verhoeven investigates the desemanticization of object clitics in particular constructions in Modern Greek. The contribution by Schultze-Berndt sheds light on the content of desemanticized units. Comparing generalized action verbs (e.g., English do) in several languages, this article shows that there is considerable cross-linguistic diversity with respect to the range of functions that these elements fulfill. Paradigmaticization refers to the reduction of size and the development of functional homogeneity of the paradigm to which the linguistic unit at issue belongs. An illustrative example for the evolution of a small size paradigm out of an open lexical class is the development of numeral classifiers in Korean (see Koo, this volume). While an open set of attributes is used with quantified nouns, a small subset of them is selected and develops to a grammaticalized paradigm of default classifiers encoding a small number of oppositions among several classes of entities. The parameter of obligatorification is related to the decrease of paradigmatic variability, i.e. the loss of the possibility to substitute certain linguistic units in a particular function. This process is exemplified by the preposition von ‘from’ that develops to the sole possibility to encode nounto-noun dependencies in particular contexts in German (see Askedal, this volume). Another example is the obligatorification of object clitics in particular constructions in Modern Greek (see Verhoeven, this volume). The articles by Wischer (on the evolution of ANTERIOR grams in English) and Quesada (on the evolution of agreement markers in Chibchan languages) give to obligatorification a particular status among the processes related to grammaticalization. Both authors argue that obligatorification corresponds to the last stage of a grammaticalization process.

4

Elisabeth Verhoeven et al.

The three latter parameters in Table 1 relate to the syntagmatic properties of linguistic units. Condensation refers to narrowing down the structural scope of a linguistic unit. An example of condensation is the development from lexical verbs bearing a VP complement to auxiliaries whose structural scope is the V (see Lehmann 22002: 128). Coalescence is the process of phonological integration of an element into its host during grammaticalization, hence it reflects the increase of bondedness that ranges from juxtaposed elements (not bound) to affixes (bound). A typical example of this process is the development of agreement markers out of free pronouns (dealt with in Quesada, this volume). Finally, fixation corresponds to the reduction of syntagmatic variability and targets the development of certain fixed slots for grammatical formatives. Related to this parameter, the article by Koo (this volume) describes the development of distributional restrictions that lead to strict adjacency of the numeral classifiers and the head noun. A major issue in current grammaticalization research is to what extent the processes introduced constitute instances of a unique phenomenon or simply diverse developments in language change, which are not necessarily related to each other (see Lehmann 2005 for detailed discussion). The conceptual basis for the interrelation of these processes is the fact that they represent several facets of the loss of autonomy of a linguistic sign. The empirical evidence is the frequent co-occurrence of these processes in a single instance of language change. The contributions to this volume present rich empirical evidence for the interaction of the grammaticalization parameters. Quesada proposes an arrangement of several genetically related languages in a continuum of grammaticalization of agreement, based on the criteria of degree of syntactic cohesion (bondedness), degree of formal distinguishability of functional properties (paradigmaticity) and degree of obligatoriness of person markers (paradigmatic variability). The study of Koo on Korean numeral classifiers shows that desemanticization goes in hand with paradigmaticity and bondedness. On the other hand, Stolz and Ammann demonstrate that the cooccurrence of developments in the semantic and the formal properties of linguistic units is not a necessary condition for a grammaticalization process. The change of lexical verbs to aspectual auxiliaries in Maltese is manifested through the functional properties of these elements, but is not accompanied by changes in their formal properties.

Introduction

5

3. The grammaticalization process The major issue at the beginning of the evolution of a grammatical formative is the selection among a number of lexical alternatives. The contribution by Elena Maslova is devoted to the interaction of mutation and selection in the incipient stage of a grammaticalization process. The situation in which more than one linguistic units are alternative options for the fulfillment of one and the same function is reported in several papers in this volume (e.g., a set of lexical verbs in Maltese may be used to encode continuative aspect, see Stolz and Ammann; a set of possible attributes in Korean may be used as numeral classifiers, see Koo). Furthermore, Skopeteas argues that the selection among the lexical alternatives is determined by cross-linguistic eligibility hierarchies which are based on cognitive asymmetries between the available options. A different outcome is reported for the evolution of reciprocal markers in Manambu in the contribution by Aikhenvald. In this case, paradigmatic variability is not narrowed down by a process of selection; more than one alternatives are grammaticalized in parallel, rendering a set of co-existent grammatical mechanisms that are not fully synonymous but share the same function of reciprocity. This phenomenon, called ‘multiple grammaticalization’ by Aikhenvald is especially challenging for accounts advocating the idea that language change is determined by economy. Grammaticalization processes are not random instances of language change, but they follow cross-linguistically attested paths, which are well described in the existing introductions to grammaticalization (see Heine and Reh 1984, Heine, Claudi, and Hünnemeyer 1991, Lehmann 22002, Hopper and Traugott 22003). Fresh evidence for some well established paths is presented in this volume, for instance the grammaticalization of full verbs to auxiliaries in English and Maltese in the articles by Wischer and Stolz and Ammann, respectively, the grammaticalization of serial verbs and body part nouns to adpositions in Nànáfwê in the contribution by Bohoussou and Skopeteas, the grammaticalization of pronominal elements to agreement markers in Chibchan in the article by Quesada. Furthermore, this volume includes some hitherto non-documented types of change. Giacalone Ramat and Mauri present evidence for the subjectification of the coordinating marker però in the history of Italian and study the development of a speaker-oriented interpretation out of the inference of contrast. Nishina shows that honorific constructions emerge out of valency changing operations in Japanese. The article by Askedal includes a general discus-

6

Elisabeth Verhoeven et al.

sion on the relevance of grammaticalization ‘clines’. It challenges the view that the position in a grammaticalization cline reflects the degree of grammaticalization in showing that analytic and synthetic exponents of genitive relations in German are in complementary distribution. A major issue in recent grammaticalization research is the relation between grammaticalization and constructions (see Hopper 2001, Traugott 2003, Diewald 2006). This research paradigm is based on the idea that grammaticalization is not just a change of isolated linguistic items but a change affecting the properties of the whole construction in which these items are embedded (see Lehmann 1992: 406). It is supported by the current interest on the role of constructions in the language system (see Goldberg 1995, Croft 2001). The role of constructions in grammaticalization is very well represented in our volume. Wischer argues with respect to the grammaticalization of auxiliaries in Modern English that the diachronic process involves the functionalization/specialization of the entire construction. Verhoeven shows that the obligatorification of object clitics in Modern Greek crucially differs in particular constructional contexts. Based on her previous research on the development of modal particles in German (Diewald 2002, 2006), Diewald concludes that grammaticalization arises in particular constructions that provide the critical contexts in which further developments may take place; grammaticalization proceeds in abandoning restrictions to particular construction types.

4. Directionality of grammaticalization The (uni)directionality of grammaticalization is probably the most debated issue, accompanying this research paradigm from its beginning. The basic observation is that the diachronic process which is reverse to grammaticalization, namely ‘degrammaticalization’, does not occur with the same systematicity and frequency as grammaticalization proper. Although a process of degrammaticalization, in which linguistic signs gain in autonomy, may be conceptually constructed, the empirical cases of language change which are discussed under this rubric are scarce (for instance, see Luraghi 1998, Ramat 2001). It should be noted that the concept of degrammaticalization does not include every case of development of lexical units out of grammatical formatives (most of them being cases of lexicalization) but properly refers to the reversion of the grammaticalization parameters, meaning that a grammatical unit gains additional semantic weight and paradigmatic

Introduction

7

variability, that its bondedness and fixation decreases, etc. However, the existence of genuine examples of degrammaticalization does not lead to the rejection of the unidirectionality hypothesis, but proves that unidirectionality is a falsifiable hypothesis. The empirical proof comes from the quantitative dominance of cases of grammaticalization, which is beyond doubt. Based on this asymmetry, grammaticalization theorists conclude that the grammaticalization process is unidirectional, which reflects the fact that the loss of autonomy of linguistic signs is a natural consequence of economy principles in human communication, while degrammaticalization lacks a corresponding functional motivation (see Haspelmath 1999, Lehmann 22002: 14–17, 2005 for further discussion). The directionality of grammaticalization processes is the subject of the paper ‘Unidirectionality of grammaticalization in an evolutionary perspective’ by Maslova. This paper takes an evolutionary perspective on language change in distinguishing between the mutation of particular linguistic features or expressions and the selection among them (see Croft 2000). The main argument of the paper – based on an empirical study in Russian – is that the mutation step of language change is random, but the selection among the range of mutated alternatives is based on communicative principles which are universal in nature. This fact explains the unidirectionality of the cross-linguistically observed changes. The paper by Tsunoda presents an empirical study on an ongoing change in Japanese: relative clauses with a relativized undergoer constituent display a strong tendency to occur in a passive construction. Based on a study on the micro-variation in present-day Japanese and assuming a unidirectional process of language change, the author makes predictions about the generalization of the evolving pattern. The relation between grammaticalization and lexicalization and the understanding of the qualitative differences and the diachronic interactions between these phenomena is crucial for the directionality hypothesis (see Lehmann 2002, 2005). Helmbrecht concentrates on this interaction with reference to the Siouan applicative morphemes. He shows that the grammaticalization of applicative markers which is accomplished in ProtoSiouan is followed by processes of lexicalization in the individual Siouan languages which lead to the loss of productivity and compositionality of applicative markers. This article allows for a direct comparison between lexicalization and grammaticalization, since it examines evidence related to the parameters of Table 1 in the lexicalization process.

8

Elisabeth Verhoeven et al.

In a broader diachronic perspective, it has been argued that grammaticalization is part of a cyclic type of change, also referred to as Jespersen’s cycle. The cyclic appearance of grammaticalization phenomena is based on the idea that the loss of particular functional properties invokes a regeneration process (either reinforcement or renovation) which creates new linguistic units that fulfill the former functions of grammaticalized elements. The new units are subject to grammaticalization, rendering thus diachronic chains of grammaticalization and regeneration cycles. Theoretically, this concept is at odds with the creative force of grammaticalization, which frequently innovates the grammatical system in inserting new grammatical categories (see Lehmann 2005). Grammaticalization results from the loss of semantic and phonological weight, which is a product of the automaticization of human communication and not motivated by the necessity to ‘fill a gap’ in the inventory of grammatical formatives, as argued by Wischer in this volume. Skopeteas shows that cyclicity involves the assumption that there is a set of functions which are cross-linguistically invariant and moreover that there is a language independent functional motivation to retain them in the course of language change. He argues that this assumption contradicts the facts from language variation known from current typological research.

5. Outline of the volume The book is organized as follows. Part I is foundational and contains two contributions which deal with the issue of directionality of change. The contribution by Elena Maslova investigates the grammaticalization process in evolutionary perspective and distinguishes the role of mutation and selection in the diachronic change with evidence from Russian. Stavros Skopeteas discusses the cyclicity of grammaticalization phenomena and presents evidence from spatial relators in Greek which shows that retaining particular functions is not a necessary consequence of the loss of semantic weight. Part II is devoted to grammaticalization processes in noun and prepositional phrases. It includes a chapter on Korean numeral classifiers written by Myung Chul Koo which demonstrates how members of a lexical category are selected as grammaticalized expressions of numeral classification. Two chapters on prepositions follow: Amani Bohoussou and Stavros Skopeteas report on the grammaticalization status of adpositions in Nànáfwê.

Introduction

9

John Ole Askedal investigates the complementary distribution between morphological genitives and prepositional phrases in German. Part III deals with grammaticalization processes in the verbal system. J. Diego Quesada investigates the grammaticalization of agreement in Chibchan languages. Johannes Helmbrecht presents evidence for the lexicalization of (the formerly grammaticalized) applicative markers in Siouan. Alexandra Aikhenvald examines the sources of grammaticalization of reciprocal markers in Manambu with emphasis on the phenomenon of multiple grammaticalization. Stolz and Ammann investigate the development of aspectual auxiliaries out of lexical verbs in Modern Maltese. Eva Schultze-Berndt discusses the range of uses of generalized action verbs in several languages. Finally, Tasaku Tsunoda examines the predictive force of grammaticalization research on the basis of evidence from microvariation in present day Japanese. The last part of the volume, part IV, contains papers about the interaction between grammaticalization and constructions. Gabriele Diewald studies constructional restrictions in the grammaticalization of modal particles in German. Ilse Wischer investigates the grammaticalization of perfect auxiliaries in English. Elisabeth Verhoeven presents evidence for constructional restrictions on the occurrence of object clitics in Modern Greek. Yoko Nishina examines the grammaticalization of honorific constructions in Japanese. Finally, Anna Giacalone Ramat and Caterina Mauri investigate the emergence of a speaker-oriented interpretation in the Italian connective peró.

References Croft, William 2000 Explaining Language Change: An Evolutionary Approach. Harlow, New York: Longman. 2001 Radical construction grammar: Syntactic theory in typological perspective. Oxford: Oxford University Press. Diewald, Gabriele 2002 A model for relevant types of contexts in grammaticalization. In New Reflections on Grammaticalization. International Symposium, Potsdam, 17–19 June, 1999, Ilse Wischer and Gabriele Diewald (eds.), 103–120. (Typological Studies in Language 49.) Amsterdam/Philadelphia: Benjamins.

10 Elisabeth Verhoeven et al. 2006

Konstruktionen in der diachronen Sprachwissenschaft. In Konstruktionsgrammatik. Von der Anwendung zur Theorie, Anatol Stefanowitsch and Kerstin Fischer (eds.), 79−103. Tübingen: Stauffenburg. Goldberg, Adele 1995 Constructions. A construction grammar approach to argument structure. Chicago/London: The University of Chicago Press. Haspelmath, Martin 1999 Why is grammaticalization irreversible? Linguistics 37.6, 1043– 1068. Heine, Bernd, Ulrike Claudi, and Friederike Hünnemeyer 1991 Grammaticalization: A conceptual framework. Chicago/London: The University of Chicago Press. Heine, Bernd, and Mechthild Reh 1984 Grammaticalization and Reanalysis in African languages. Hamburg: Buske. Hopper, Paul 2001 Grammatical constructions and their discourse origins: prototype or family resemblance? In Applied Cognitive Linguistics I: Theory and Language Acquisition, Martin Pütz, Susanne Niemeier, and René Dirven (eds.), 109−129. Berlin: Mouton de Gruyter. Hopper, Paul J., and Elizabeth Closs Traugott 2 2003 Grammaticalization. Cambridge: Cambridge University Press. Lehmann, Christian 1992 Word order change by grammaticalization. In Internal and external factors in syntactic change, Marinel Gerritsen and Dieter Stein (eds.), 395−416. (Trends in linguistics. Studies and monographs 61.) Berlin/New York: Mouton de Gruyter. 2 2002 Thoughts on Grammaticalization. Assidue (Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt), Nr. 9. 2002 New reflections on grammaticalization and lexicalization. In New reflections on grammaticalization, Ilse Wischer and Gabriele Diewald (eds.), 1–18. Amsterdam/Philadelphia: Benjamins. 2005 Theory and method in grammaticalization. Zeitschrift für Germanistische Linguistik 32 (2): 152–187. Luraghi, Silvia 1998 On the directionality of grammaticalization. Sprachtypologie und Universalienforschung 51: 355–365. Ramat, Paolo 2001 Degrammaticalization or recategorization? In Naturally! Linguistic studies in honour of Wolfgang Ulrich Dressler presented on the occasion of his 60th birthday, Chris Schaner-Wolles, John Rennison, and Friedrich Neubarth (eds.), 393–401. Torino: Rosenberg and Sellier.

Introduction 11 Traugott, Elizabeth Closs 2003 Constructions in Grammaticalization. In The Handbook of Historical Linguistics, Brian D. Joseph and Richard Janda (eds.), 624−647. (Blackwell Handbooks in Linguistics.) Oxford: Blackwell.

Part I Directionality of grammaticalization processes

Unidirectionality of grammaticalization in an evolutionary perspective Elena Maslova

1. Introduction The underlying causes of directionality in grammaticalization can be hidden at two different levels, roughly corresponding to mutations and selection (propagation of mutations) in evolutionary frameworks: a mutation happens in an individual language, or mental grammar, and offers itself for further propagation by consequent changes in individual speech, while selection happens in a language community, via communicative interaction, whereby some novel behaviors can be replicated by other speakers and integrated into their individual languages (including both adult speakers and language learners). Whereas grammaticalization research tends to look for causes of directionality in non-randomness of mutations, this paper is intended to demonstrate that selection-level processes may play a more prominent role in determining universal tendencies of language change than commonly assumed. The paper focuses on one aspect of directionality, namely, the tendency towards semantic extension, or pattern spread. In Section 2, I outline the selection pressures that may determine this tendency even if mutations in pattern use are entirely random, followed by examples of mutations of two types in modern Russian: one expands the range of a pattern and has a good chance to turn into a language change (Section 3), whereas the other reduces the range of a pattern and is unlikely to get “selected for”, i.e. adopted by other speakers (Section 4).

2. A selection-based explanation for directionality in language change The concept of (uni)directionality in grammaticalization invokes “language” as a shared system, which retains its identity as generations of speakers go by; Östen Dahl (2004: 66) has recently suggested to call it “Planguage” (where P may stand for ‘Plato’ or ‘Popper’), to distinguish this

16 Elena Maslova concept from the Chomskyan I- and E-languages: P-language is opposed to I-language in that it is “external” with respect to any individual speaker (whereas I-language(s) are the internal mental representations of language knowledge in the minds of its speakers); on the other hand, it is opposed to E-language in that it is thought of as a system, an ordered set of patterns (rather than just the chaotic external observable behavior of language speakers). The search for causes of universal tendencies of language change tends to concentrate on possible reasons of individual mutations in I-languages (which may be functionally motivated), whereas the selection and propagation processes are taken to be heavily influenced (or even fully determined) by social pressures (Labov 1994; Croft 2000: 166-195; inter alia). Indeed, the propagation processes often remains beyond the scope of grammaticalization-related discussions, as though relevant changes might simply happen in all I-languages simultaneously: P-level processes are in effect equated to I-level processes. This contrasts sharply with what happens in prototypical cases of evolution, where mutations are random, and it is selection that is functionally motivated. As I will try to show, selection-level effects may be responsible at least for some aspects of P-level directionality: some types of individual variations in linguistic behavior are more likely to be “selected for” (i.e. adopted by other speakers) than others. A salient aspect of directionality in grammaticalization is the P-level tendency towards expansion of semantic range, or pattern spread: a Planguage pattern is more likely to spread to novel contexts than to retreat from some of its contexts, unless it is being “pushed” from a certain functional niche by another expanding pattern. The P-level tendency for pattern spread is commonly accounted for by universal “I-level” propensity to overuse expressive patterns (see Bybee, Perkins and Pagliuca 1994: 281292; Lehmann 1985, 1995; Heine 1994; Keller 1990/94: 57, 71; Dahl 2004: 121-127; inter alia). In this paper, I argue that there is a simple and universal selection-level cause for this P-level tendency; it is to be expected even if individual speakers are equally likely to expand and to reduce semantic ranges of linguistic patterns, and even if “expanding” mutations are less likely. In a nutshell, I will try to show that a range-expanding mutation has a considerably greater chance to be replicated (and thus selected for) than a range-reducing mutation. Generally speaking, a P-language pattern can be represented in a language community by a variety of slightly different strategies of use, most of which are mutually compatible, so that the variation remains impercep-

Unidirectionality of grammaticalization in an evolutionary perspective 17

tible to the speakers. However, the difference becomes “visible” if some speakers use the pattern in contexts where it is prohibited from the point of view of other speakers. It can be assumed, for the sake of simplicity, that such a variation in the use of a non-obligatory pattern E can be described in terms of two strategies, one of which (A) licenses E in a certain context C, and the other (B) does not. Accordingly, there are A-speakers, who use E in C, and B-speakers, who do not. This situation involves a simple asymmetry with far-reaching repercussions for the selection process: the very existence of inter-speaker variation is more conspicuous for B-speakers than for A-speakers. Whenever an A-speaker uses E in C in a conversation with a B-speaker, the latter can hardly fail to notice the difference in their linguistic behavior, insofar as, from the B-speaker's point of view, a constraint on the use of E is violated. On the other hand, it is much harder, if not impossible, for an A-speaker to recognize this difference: even if a Ccontext occurs in a conversation, a single non-use of E does not indicate a difference in linguistic habits, since E is optional for both speakers. Accordingly, B-speakers have a possibility to adopt the A-strategy (i.e. to replicate the behavior of A-speakers), while A-speakers are unlikely even to notice that there is another strategy to adopt. If the B-strategy is conservative and the A-strategy is a recent “mutation”, the latter can acquire new speakers and, eventually, come to dominate the community (a rangeexpanding change characteristic of grammaticalization processes). Conversely, if the B-strategy is a recent mutation in a language community where the A-strategy is dominant, the novel strategy can hardly be adopted by conservative speakers because they simply would not notice its existence (hence, a range-reducing mutation will never turn into a P-language change).

3. A visible range-expanding mutation Although the idea introduced in the previous section can be illustrated by virtually any token of grammaticalization process, I would like to use this opportunity to introduce an example which, as far as I know, has not been discussed in the grammaticalization literature so far, presumably because the language change in question has begun very recently. The Russian expression ni razu ‘not once, not a (single) time’ belongs to a lexically closed class of emphatic negations sharing the negative item ni (e.g. ni minuty ‘not a (single) minute’ and ni shagu ‘not a (single) step’), based on a regu-

18 Elena Maslova lar pattern which requires a semantically superfluous numeral (odin 'one', in the appropriate gender and case form), e.g. ni odnoi knigi ‘not a single book’ (as in He hasn’t read a single book), ni odnogo posetitelya ‘not a single visitor’, etc. This expression seems to have recently entered what might become the very first stage of “Jespersen’s Cycle” (Dahl 1979): its use in negative sentences can, at the present time, be described in terms of co-existence of two strategies. The conservative strategy is linked to the compositional meaning of ni razu ‘not a (single) time’. In accordance with the semantics of raz ‘(one) time’, this strategy licenses the use of this emphatic negation only in descriptions of situations which might have recurred multiple times; it also disallows its use in present-tense sentences. The function of the conservative ni razu can be illustrated by the following pair of examples: (1)

a.

b.

On ne byl v Londone. He NEG be:PAST in London:OBL ‘He hasn’t been to London.’ or ‘He was not in London.’ On ni razu ne byl v Londone. He NEG once NEG be:PAST in London:OBL ‘He has never been to London.’

The first sentence can mean either that the subject referent has never visited London or that he was not there on a particular occasion (which must be clear from the context); ni razu resolves the ambiguity in favor of the former meaning and adds some emphasis to it. It seems worth emphasizing at this point that Russian is a “double-negation” language, so ni razu does not replace the basic pre-verbal negative marker, but always co-occurs with it, i.e. it cannot serve as the sole marker of negation. The novel strategy in effect breaks the link between ni razu and raz and uses ni razu just for emphatic negation; in particular, it can be used with individual-level predicates in the present tense, e.g.: (2) a.

Vy chto, ni razu ne gramotnyi? you what, NEG once NEG literate ‘Are you totally illiterate?’ (Y. Petrosjan. ‘Army humour’ (2004); cited acc. to (NCR 2003-2006))

Unidirectionality of grammaticalization in an evolutionary perspective 19

b.

[...] tozhe ni razu ne vladelets [...] also NEG once NEG owner ‘[...] by no means the owner (of a factory) either...’ (K. Krylov, an article published in Specnaz Rossii (2003); cited acc. to (NCR 2003-2006)).

What is essential here is that (a) conservative listeners can readily recognize the intended (emphatic) meaning on the basis of their own linguistic knowledge, even if it is the first time such a speaker encounters this usage, but (b) the usage is also immediately perceived as indicating a “mutation”, a violation of a strong semantic constraint. These conditions make it possible for conservative speakers to replicate this usage, i.e. to adopt the novel strategy. Such examples seem to be quite common in informal speech. The National Corpus of Russian (NCR 2003-2006) contains only two unmistakable instances of the novel usage among around one thousand occurrences of ni razu found in its 21st century sub-corpus, both given in (2): one from a collection of “army humour”, which frames it as a sign of uneducated speech and thus demonstrates the sociolinguistic “meaning” attached to this usage, and the other from a nationalist opposition outlet, without any apparent indication of mocking anyone’s speech errors, but probably positioning itself as being “closer to the people” than other magazines, not only in its political standing, but also in its language. The corpus also seems to indicate a conspicuous rise in the overall frequency of ni razu over the last couple of decades (there are about twice less occurrences in a sub-corpus of a comparable size from the last decade of the 20th century). I hypothesize that this dimension of sociolinguistic variation must be much more salient for those who use ni razu conservatively than for those who tend to use it as a general emphatic negation. As described above, a conservative speaker perceives novel uses of ni razu as violations of a strong semantic constraint, which are immediately noticeable. On the other hand, an innovative speaker can at most notice that someone has not used this expression in a certain context where it might have been used for emphasis (according to the novel strategy); however, the absence of emphatic negation on any particular occasion does not imply any general difference in speech patterns, since there are no contexts where ni razu is obligatory: it may be absent simply because the speaker has decided that no emphasis is necessary, or has chosen another emphatic device. Accordingly, those who do use ni razu in novel contexts can remain unaware that some of their interlocutors

20 Elena Maslova use ni razu in a more restrictive way. Thus, the inter-speaker variation is asymmetric, in that the conservative speakers regularly encounter novel usages and thus can replicate the novel strategy, whereas the conservative strategy remains inconspicuous for innovative speakers. In this situation, it is much more likely that more and more conservative speakers will adopt the innovation than that innovative speakers will change their linguistic behavior to conform to the more conservative pattern, and indeed the novel strategy seems to be gaining new speakers very rapidly. Furthermore, it seems that the spread of ni razu to novel contexts has, as a side effect, weakened semantic constraints associated with other negative adverbs as well, as witnessed by the following sentence: (3)

Ja zh ne gumanitarij nigde I PCL NEG scholar.of.humanties nowhere ni razu NEG once ‘I am not a scholar of humanities at all (lit. I am not a scholar of humanties nowhere not once).’ (personal observation)

The speaker has apparently decided that ni razu does not provide enough emphasis to her negation, and added another negative adverb, nigde ‘nowhere’ (equally semantically incompatible with the context from the conservative point of view). This speaker’s level of education (a medical doctor) and age (around forty) seem to indicate that the novel strategy (which was unthinkable twenty years ago) is now used even by relatively conservative speakers, at least in informal conversations.

4. An invisible range-reducing mutation It is much more difficult to find a compelling example of range-reducing mutation, precisely because such a mutation would normally remain invisible to conservative speakers and would never turn into even the incipient stage of a P-language change (for example, if any speakers of Russian use ni razu more restrictively than licensed by the semantically motivated conservative strategy described in Section 3, it would take a large-scale linguistic study with multiple informants to discover this fact). The best example of an apparent range-reducing mutation I have found is related to

Unidirectionality of grammaticalization in an evolutionary perspective 21

another Russian expression, na samom dele ‘actually, the truth is, as a matter of fact’ (literally, ‘on the very deed, on the deed itself’). The meaning of this expression is doubly non-compositional, albeit still transparently related to the component meanings. First, the semantics of the slightly outdated expression na dele (literally, ‘on (the) deed’) cannot be derived from the meanings of delo ‘deed, action’ and the locative preposition na ‘on’; this meaning invokes an implicit semantic opposition with the equally non-compositional expression na slovah (literally, ‘on the words’) and is intended to contrast what one really does to what one claims to do (or preaches). Secondly, the adjective samyj generally expresses superlative meaning (usually with adjectives, as in samyj krasivyj ‘(the) most beautiful’, but also with some nouns, as in samyj verh ‘the very top (of something)’). The combination *samoje delo has no sense whatsoever outside the context of a locative preposition, and the locative preposition introduces an unpredictable meaning: whereas na samom dele indicates a contrast, an unexpected turn of thought, which negates some previous statements, v samom dele (literally ‘in the very deed’) is used to introduce additional arguments or evidence in favour of a previous statement; in a sense, the discourse functions of these two expressions are nearly opposite, and the difference is wholly dissociated from the semantic difference between na ‘on’ vs. v ‘in’. As far as can be judged from a preliminary analysis of the National Corpus of Russian (NCR 2003-2006), the discourse function of na samom dele does not seem to have undergone any significant changes at the level of Planguage over the last couple of centuries: informally, it serves as a part of a rhetoric device, a contrastive discourse marker introducing an unexpected turn of thought or a negation of an apparently feasible assumption. In what follows, I refer to this use as “weak” (for the reasons that will become clear shortly). Here is a typical example from the late 19th century, where na samom dele indicates that a semantic presupposition of the previous sentence (the unexpectedness of a social change) is negated in the sentence introduced by this expression: (4)

Chem zhe objasnit' stol' rezkiy perevorot? Na samom dele, pri vsei grandioznosti etogo perevorota v n'em ne bylo ničego neožidannogo ili slučajnogo. ‘How can one explain such a radical change? As a matter of fact, however grandiose the change, there was nothing unexpected or

22 Elena Maslova random about it.’ (B.Porozovskaya. Martin Luther (1895); cited acc. to (NCR 2003-2006)). Although the expression functions as a kind of discourse particle, it has not been formally reduced. As a result, the remaining transparent link with delo ‘action, deed’ allows for a more literal interpretation of na samom dele, whereby it is intended to contrast some “hard facts” to “opinions”, “appearances”, or “words” (rather than just any two propositions). This possibility seems to have brought about a range-reducing mutation in the modern Russian language, whereby na samom dele is felicitous, for some speakers, only in the context of a clear and unquestionable opposition between “appearance” and “reality”; this meaning will be referred to has “strong”. Under the strong interpretation, the expression is severely overused in examples like (4), where one opinion is simply replaced by another, arguably less trivial, one. Moreover, the weak use of na samom dele sounds, for such speakers, as a pretentious attempt to present one's opinions, beliefs, or theories as “facts” (this interpretation even found its way into some culturological discussions (Rudnev 1997)). As can be judged from an informal survey of about thirty speakers of Russian, there are at least two perceptions of na samom dele, which currently co-exist in the Russian-speaking community. Some speakers carry the range-reducing mutation (i.e. prefer the “strong” interpretation), perceive an inter-speaker variation in how the expression is used, and even attach a very specific sociolinguistic meaning to it: according to them, na samom dele is overused by the former Soviet intelligentsia, especially with scientific background (which was prominent on the public scene from the sixties till the collapse of the Soviet Union), as well as by some of their genetic and cultural descendants. As shown by my preliminary analysis of the National Corpus of Russian, this perception does not capture the real sociolinguistic distribution: na samom dele is commonly used as a discourse marker by people of very different backgrounds and generations. What is essential in the context of this paper is that those who use na samom dele as a discourse marker (i.e. don’t carry the range-reducing mutation) are completely unaware of any variation in usage; this is straightforwardly explained by the fact that it is nearly impossible for such a speaker even to notice the absence of na samom dele in a context where it might have been present for rhetoric purposes, let alone to recognize this absence as a sign of a more general difference in linguistic behavior. Indeed, although I am probably more attentive than most speakers of Russian

Unidirectionality of grammaticalization in an evolutionary perspective 23

to its linguistic peculiarities, I have stumbled upon this phenomenon only accidentally, when its social interpretation cited above was explicitly mentioned in a conversation. An accurate description of this variation will require further work, yet the “invisibility” effect seems quite obvious: those who use na samom dele freely, with relatively weak constraints on appropriate discourse contexts, remain perfectly unaware that not everyone does it. On the other hand, the 19th-century examples like in (4) demonstrate that that the range-reducing strategy (based on the “strong” interpretation) represents, at the present time, an innovation (albeit analyzable as a “reversal” of a former rangeexpanding mutation). However, this range-reducing mutation is unlikely to lead to a language change, precisely because it remains perfectly invisible to the majority of speakers in the course of everyday communication. In other words, some speakers seem to have reversed the initial step of grammaticalization of this expression in their I-languages and linguistic behavior, but this cannot change the behavior of other speakers because they do not notice it. To conclude, the asymmetry in “visibility” of variations in linguistic behavior straightforwardly explains the tendency towards pattern spread (as opposed to reduction of semantic range): even if the inter-speaker variations arise randomly, so that deviations from the dominant strategy in both directions are equiprobable, this asymmetry suffices to ensure a higher probability of propagation for range-expanding innovations, simply because they are more visible to conservative speakers and thus lend themselves to replication. As described here, the phenomenon is limited to initial stages of grammaticalization, or, more specifically, to contexts where the expression is optional. It seems, however, that universal selection-level pressures (as opposed to non-randomness of mutations) may play a broader role in emergence of universal P-level tendencies.

References Bybee, Joan L., Revere D. Perkins, and William Pagliuca 1994 The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press. Croft, William 2000 Explaining Language Change: An Evolutionary Approach. Harlow, New York: Longman.

24 Elena Maslova Dahl, Östen 1979 2004

Typology of sentence negation. Linguistics 17: 79–106. The Growth and Maintenance of Linguistic Complexity. Philadelphia: Benjamins.

Heine, Bernd 1994 Grammaticalization as an explanatory parameter. In Perspectives on Grammaticalization, William Pagliuca (ed.), 225–287. Amsterdam: Benjamins. Keller, Rudi 1990/94 On Language Change: the Invisible Hand in Language. London: Routledge. Labov, William 1994 Principles of Linguistic Change. Oxford (UK)/Cambridge (Mass.): Blackwell. Lehmann, Christian 1985 Grammaticalization: synchronic variation and diachronic change. Lingua e Stile 20, 303–318. 1995 Thoughts on Grammaticalization. München: Lincom Europa. NCR 2003-2006 Национальный корпус русского языка [National Corpus of the Russian Language]: http://www.ruscorpora.ru. Rudnev, V. P. 1997 Slovarj kuljtury XX veka. (Dictionary of the 20th century culture.) Moscow: Agraf.

Grammaticalization and sets of form-function pairs: Encoding spatial concepts in Greek Stavros Skopeteas

1. Preliminaries1 Grammaticalization is a theory about the historical development of single linguistic units. It refers to the evolution of grammatical formatives out of lexical elements in terms of a set of parameters in the paradigmatic and syntagmatic axes (i.e. weight, cohesion, and variability, see Lehmann 1995b: 123). Crucially for the purposes of this article, a grammaticalization process involves by definition a change in the functional properties of the element at issue. Within the framework of Lehmann (1995b: 164), the relevant change relates to the parameter of semantic integrity and is a process of ‘attrition’ which at the semantic layer means loss of semantic features, also known as ‘desemanticization’, ‘semantic depletion’ or ‘bleaching’. The consequences of the loss of functional properties were observed already in the early research on the diachronic phenomena that were later encompassed by the grammaticalization framework. It has been observed that some diachronic changes seem to be concomitant to the grammaticalization processes, since they appear as ‘repair’ strategies invoked by the fact that elements that once fulfilled certain functions lose their functional properties due to the grammaticalization process. This is illustrated in the rationale of Meillet (1912) in the following quotation. Les langues suivent ainsi une sorte de développement en spirale; elles ajoutent de mots accessories pour obtenir une expression intense; ces mots s’affaiblissent, se dégradent et tombent au niveau de simples outils grammaticaux; on ajoute de nouveaux mots ou des mots différents en vue de l’expression; l’affaiblissement recommence, et ainsi sans fin (Meillet [1912] 1921/1936: 140f.).

The general idea is that some diachronic processes, i.e., reinforcement and renovation, are used as ‘repair’ strategies that are invoked by the loss of functional properties due to the grammaticalization of particular linguistic units. Lehmann (1995b: 22) formulates the interaction between these phenomena as follows:

26 Stavros Skopeteas If an element is weakened through grammaticalization, there are, in fact, two possibilities open to linguistic conservatism. The first is to give it up and replace it by a new but similar one. This is renovation […]. The second is to reinforce it, thus compensating for and cheeking the decay (Lehmann 1995b: 22).

The phenomena reported above are intensively discussed in grammaticalization studies, in particular with respect to the cyclic appearance of grammaticalization phenomena. The elements that either renovate or reinforce their grammaticalized predecessors are eligible for new grammaticalization processes, rendering thus a chain of loss and regeneration processes that is cyclically repeated in the course of language change. The conceptual basis of this phenomenon involves a crucial issue concerning the foundations of linguistic functions. It entails the assumption that there is a set of linguistic functions whose fulfillment is indispensable for any diachronic stage of a grammar. Hence, there is a language situation at which particular elements may be “found insufficient” (in terms of Jespersen 1917) in order to serve these linguistic functions, which is the functional motivation for a ‘repair’ strategy of either the reinforcement or the renovation type.2 Though the correlation between grammaticalization and regenerating processes is empirically attested in a vast number of studies, its underlying conceptual assumptions are by no means obvious, in particular in light of the cross-linguistic variation of categories and functions (see Haspelmath 2008). In view of the amount of variation shown in contemporary typological studies, the assumption that the loss of particular functional properties evokes a ‘repair’ strategy is not obviously motivated. The aim of this paper is to exploit the possible outcomes of a grammaticalization process that leads to the loss of functional properties of particular elements. Recent studies in language change suggest that the motivation of diachronic processes may be better understood if we consider the impact of these processes in a model of verbal interaction that contains a speaker, a hearer and particular communicative intentions of the former (see Croft 2000, Keller 1994, Haspelmath 1999). This paper considers the cyclic appearance of grammaticalization processes in view of a model of verbal interaction as proposed in the above studies. Section 2 establishes the necessary assumptions: in order to assess the impact of the phenomenon at issue, we assume that form/function pairs in a language are not isolated lemmata in the mental lexicon, but are members of sets of pairs that are determined by some common denominator that is functional in nature, i.e.,

Grammaticalization and sets of form-function pairs 27

functional domains. A process that affects the functional properties of a particular element necessarily involves a change in the respective functional domain in the language at issue. Changes of this type influence the potential to convey particular propositional contents in a model of human interaction and this is the source of concomitant changes. Section 3 discusses evidence from Greek and exploits the consequences of the grammaticalization of adpositions in this language, presenting quantitative data from a diachronic text corpus. Section 4 concludes.

2. Grammaticalization and functional domains This section sets out a conceptual basis that makes the assumptions involved in grammaticalization cycles explicit. In order to assess the need to realize particular linguistic functions, the assumption of a model of verbal interaction is due. This involves a speaker and a hearer, as well as the intention of the speaker to convey to the hearer certain propositional contents.3 It is only with respect to the potential of a speaker to realize particular communicative intentions that statements about the appropriateness of particular linguistic structures may be tested.

2.1. Structure of functional domains Linguistic units are not separate lemmata in a mental inventory, but are members of functionally interacting sets of units. We assume that functional concepts are organized in functional domains, i.e., sets of such units that share a common denominator that is functional in nature (see Seiler 1988, Lehmann 1989). In the following, we will illustrate the notion of functional domain by means of the domain of localization that encompasses the concepts that are used to localize entities in space. A functional domain has a syntagmatic and a paradigmatic structure. The syntagmatic structure of the domain of localization contains the following components (see Lehmann 1992, 1995a, Wunderlich 1982, Talmy 1983): (a) a ‘localized object’, which is the entity that is localized in space, (b) a ‘place’ which is the spatial sector in which the localized object is located, and (c) a ‘spatial relation’ between the localized object and the place, which may be either static or dynamic depending on whether the localized object changes location in time. Places are usually identified with

28 Stavros Skopeteas respect to further entities whose location is assumed to be known to the hearer. In this case, a place contains the following sub-components: (i) a ‘reference object’, which is the entity that is used to identify a place in space and (ii) a ‘spatial region’ of the reference object which is a spatial sector identified with relation to the reference object, e.g. its inner space, the space above it or the space in front of it. The lexicalization of this type of situation in English is illustrated in (1a-b). These examples illustrate a typical pattern of encoding the components of localization in this language: the subject constituent encodes the localized object, the verb the static vs. dynamic relation between localized object and place, the PP encodes the place, whereby the preposition encodes a conflation of region and relation (see the contrast between in and into) and the complement of the preposition encodes the reference object. (1)

English (sp-relation (xLOC, (region (yREF)))) [ Sbj [VP V [PP P NP ] ] ] a. John is standing in the house. b. John is going into the house.

It is crucial, that languages differ with respect to the association of functional components with syntactic constituents (see Lehmann 1992, Talmy 2000: 49). Example (2) shows that in Nànáfwê (dialect of Baoule, Kwa branch of Niger-Congo languages, Côte d’Ivoire) the (static/dynamic) relation between localized object and place is exclusively encoded through the verb (see further discussion in Bohoussou and Skopeteas, this volume). The lexicalization pattern in (2) may be observed in several languages (see for instance Lehmann 1992, 1995a on Yucatec Maya, Rennison 1997: 173 on Koromfe). (2)

Nànáfwê (sp-relation (xLOC, (region (yREF))))

a.

[ Sbj [VP V [PP NP P ] ] ] kòfí ó swā-n nú. Kofi be.located house-DEF inner.side ‘Kofi is in the house.’

Grammaticalization and sets of form-function pairs 29

b.

kòfí kò‰ swā-n nú. Kofi go house-DEF inner.side ‘Kofi goes into the house.’

The paradigmatic structure contains the sets of alternative concepts that may instantiate a particular component. These sets are hierarchical in nature, since the concepts at issue stand in taxonomic relations (i.e., form pairs of hypernyms and hyponyms). Hence, a spatial relation between a localized object and a place may be ‘static’ (if it does not change in time) or ‘dynamic’ (if it does change in time). In case it is dynamic, the relation is furthermore either ‘allative’ (if the place is the target of motion) or ‘ablative’ (if the place is the source of motion) or ‘perlative’ (if the place is the path of motion). Similarly, spatial regions are parts of taxonomic structures too. The most generic instance of a spatial region is the localization of an entity in the part of the space that is identified by the reference object, but the kind of identification is not further specified, i.e. it may be vicinity or inclusion, depending on the relevance in particular situations, vgl. the region denoted by the Spanish preposition a or French à. If the region is further specified, there are several spatial dimensions that may be alternatively chosen, such as inclusion (English in vs. out), proximity (English near vs. far from), vertical axis (English on/above vs. under), saggital axis (English in the front of vs. behind), and lateral axis (English right of vs. left of). These regions may be further specified in combination with further properties, e.g. in the upper region of the vertical axis English displays a further distinction depending on the property of contact (contact on, noncontact above). The taxonomic relations among the alternative concepts within a functional domain have a crucial implication for our discussion. Concepts that stay in a taxonomic relation may be alternatively used in order to refer to the same situation. Ηence, for the propositional content of ‘a candle that stands in the upper side of a table and in contact with it’, the English expression in (3a) encodes the concept of ‘superior’ (localization in the positive part of the vertical axis) and the concept of ‘contact’ (adjacency to the reference object). A speaker of Nànáfwê uses for the same situation a postposition that only encodes the concept of ‘superior’. I.e., the expression in Nànáfwê is underspecified which means that it holds true for a larger set of situations than the corresponding English expression.

30 Stavros Skopeteas (3)

a.

English The candle is on the table. b. Nànáfwê bùzí-n ó táblí-n be.located table-DEF candle-DEF ‘The candle is on/above the table.’

sú on/above

In this section, we have shown two properties of functional domains that are differentiated across languages, as summarized in (4). The fact that there is cross-linguistic evidence that the realization of a functional domain may differ at least in these two aspects, implies that there is no functional motivation to maintain the typological properties of a given language in language change. (4)

a.

Syntagmatic variation The distribution of semantic components to syntactic constituents may differ across languages. b. Paradigmatic variation The sets of concepts that instantiate a semantic component differ across languages.

2.2. How does grammaticalization affect a functional domain? We assume a form-function pair (fi, si) containing a form fi that encodes a spatial concept si in a particular grammar G. The sign (fi, si) is part of the functional domain of localization in G, i.e. the spatial concept si is member of a set of concepts S ={si … sn} that may be encoded in G. In a subsequent diachronic stage of the same grammar, namely G´, the form fi is grammaticalized, which involves a loss of its function of encoding the spatial concept si. If no further diachronic change takes place, the grammaticalization of fi has the consequence that the set of spatial concepts S´ in the later diachronic stage contains a functional vacuum, i.e. no expression for the spatial concept si. In other words, S´ is a proper subset of S, S´ ⊂ S. If diachrony had to preserve isomorphism among the sets of formfunction pairs that realize a functional domain in two subsequent diachronic stages S and S´, the change sketched so far would be impossible, i.e., the functional vacuum in S´ should be accommodated by a concomitant change. However, this assumption is contradicted by the facts of lan-

Grammaticalization and sets of form-function pairs 31

guage variation within functional domains as stated in (4b). Since languages with different sets of concepts are possible, language change to a target stage that displays a different set of concepts from the source stage should be a possible change too. What does this difference mean in view of a model of verbal interaction? The critical communicative situation is one at which the speaker intends to commit to the hearer a propositional content involving the spatial concept si. If the set S that contains si is available, the speaker is expected to select (fi, si). However, in S´ the critical spatial concept is not available. A first possibility is that the speaker of S´ uses an expression that is underspecified with respect to si. In a natural communicative situation, the conveyed message does not exclusively depend on the encoded meaning, since it is enriched by contextual and situational knowledge as well as pragmatic inferences. However, this means that the speaker of S selects a different propositional content than the speaker of S´. Section 3.2 provides empirical evidence that this is a possible diachronic change. Section 2.1 shows that the encoding of semantic components differs across languages, see (4a). This aspect of language variation adds a further option to the diachronic situation at issue. The loss of the exponent fi of si certainly leads to a ‘local’ change, i.e., the concept si disappears from a particular paradigm, but the relative issue for the interactive model is not the set of functions that can be encoded through particular paradigms, but the set of functions that can be encoded in utterances. We saw that English encodes the component of spatial relation through verbs and prepositions (see example (1)), while Nànáfwê encodes the same component only through the verbs (2). In a hypothetical situation, in which a local preposition in a language of the English-type loses the feature of ‘spatial relation’, this local change would not affect the possibility of encoding ‘spatial relation’ in a clause, but it would simply mean that this language shifts to an encoding pattern of the Nànáfwê-type. Since the latter is a possible human language, there is no reason to assume any functional motivation for a ‘repair’ strategy. Evidence for this type of diachronic change is discussed in section 3.3. Finally, a further possible outcome of the diachronic development at issue is a process that compensates the possible loss, i.e., a diachronic process that leads to a new form-function pair (fi, si) that encodes the critical spatial concept si. The outcome of the regeneration is that the set of concepts S´ of the subsequent diachronic stage is identical to the set of con-

32 Stavros Skopeteas cepts S of the previous stage. Empirical evidence for the diachronic change of this type is presented in section 3.4. In this section, we considered the possible consequences of a grammaticalization process for a functional domain. We hypothesized a formfunction pair (fj, si) such that si belongs to the set S of spatial concepts expressed in a particular diachronic stage of a language. In case fj be grammaticalized, the following outcomes are theoretically possible: either (a) si is not available at all in the set S´ of spatial concepts expressed in the subsequent diachronic stage, or (b) si is mapped to another constituent, or (c) a regeneration process replaces (fi, si) by (fj, si). In the case (a), there is a change in the functional domain such that the set S´ of spatial concepts in the subsequent stage is a proper subset of the set S of spatial concepts in the preceding stage. In cases (b) and (c), the set of spatial concepts that are available in both stages are equal. So far we argued for these possible historical developments on purely conceptual grounds. In the following sections, we provide empirical evidence for these types of diachronic change from the grammaticalization of Greek local adpositions.

3. Types of diachronic change and Greek local adpositions 3.1. Preliminaries In the oldest documented stage of Greek, there is a paradigm of prepositions/adverbs, whose ‘core’ relates to the encoding of spatial concepts. Some of these elements are already grammaticalized to prepositions (e.g. diá ‘through’) while others alternate between an adverbial use and a use with an (optional) nominal dependent. In the course of Greek antiquity, these elements grammaticalize to prepositions that select genitive, accusative, and dative NPs, a stage which is attested in the Classical era (5th and 4th century BC). The grammaticalization process goes on beyond the Classical era, leading to the desemanticization of the prepositional elements and the neutralization of the case oppositions of their NP complements. These developments reach their peak in the Medieval era. In the earliest stages of Modern Greek we find a completely different situation: a few members of the old prepositional paradigm are highly grammaticalized and are preserved for the encoding of abstract thematic relations, while the further members of the paradigm become extinct. In the encoding of spatial concepts, the Ancient Greek prepositions are renovated by adverbs.

Grammaticalization and sets of form-function pairs 33

What we are observing in Greek, is a textbook example of grammaticalization. A former lexical category develops to a grammatical category and loses its functional load in the course of the language history until it is renovated by another category. The main consequences of this change for the functional domain of localization are summarized in the following. In Classical Greek, the distinction between static and allative relations are encoded through the contrast between the cases of the NP complement (dative vs. accusative respectively), and in particular for the region ‘interior’ through the contrast between the prepositions en ‘inessive’ and eis ‘illative’ (see (5a)). In the subsequent stages of Greek, the contrast between these prepositions is desemanticized and the local cases of the NP complement are replaced by a structural accusative that solely encodes the government relation between the preposition and its complement (see Theophanopoulou-Kontou 2000). Hence, in the target stage, the prepositional elements do not encode the distinction between static and allative (see (5b), whereby ‘∅’ means that no element of the categories P or P-Case denote the spatial relations at issue). (5)

a.

Encoding relations in the source stage static | P{en} P-Case{DAT}

allative | P{eis} P-Case{ACC}

b. Encoding relations in the target stage static | P{∅} P-Case{∅}

allative | P{∅} P-Case{∅}

Spatial regions are encoded in Classical Greek through members of two structural paradigms: prepositions and adverbs (see (6a)). The distinction between the two sets of elements is structural, i.e., only the former but not the latter license an NP complement. In the course of language change, these prepositions are desemanticized and at least in their spatial uses are completely replaced by the corresponding adverbs (or their descendants). Hence, in the target stage, only adverbial exponents of spatial region are available (see (6b)).

34 Stavros Skopeteas (6)

a.

Encoding regions in the source stage interior superior | | P{en, eis} P{epí} Adv{éndon, eísō} Adv{ánō}

anterior | P{pró} Adv{prósthen}

exterior |

posterior |

Adv{éksō}

inferior | P{hupó} Adv{kátō}

Adv{opísō}

b. Encoding regions in the target stage interior | Adv{mésa}

superior | Adv{epánō}

anterior | Adv{brostá}

exterior | Adv{éksō}

inferior | Adv{kátō}

posterior | Adv{písō}

The properties of these changes are the research object of the following sections. Assuming an interactive model, we address the question what are the consequences of these changes for the interaction of a speaker and the hearer and in which cases the properties of this interaction may motivate concomitant changes. Our observations are based on a diachronic text sample4 which is presented in the appendix, Table 1.

3.2. Change of propositional content The spatial prepositions en ‘inessive’ and eis ‘illative’ in Classical Greek encode the region ‘interior’ in static and allative relations respectively. Already in Classical Greek, these prepositions display some non-local usages too. The preposition en occurs with abstract nouns in usages that may be accounted for through the Container Metaphor, and the preposition eis occurs in the thematic roles of ‘purpose’ and ‘recipient’ (see Luraghi 2003). In the course of language change, the semantic extension of these prepositions spreads to a wide range of usages (see Bortone 2000: 223– 226), ending up to as the default markers of oblique arguments bearing

Grammaticalization and sets of form-function pairs 35

diverse thematic roles in Modern Greek. Before we examine the alternation between the two prepositions (see section 3.3), we will discuss the consequences of this change for their common functional feature, namely the encoding of the region ‘interior’. Figure 1 presents the frequencies of prepositions in the text sample (proportions are calculated with relation to the total n of words of the examined texts pro diachronic stage; see the exact counts in the appendix, Table 1).5 The relevant finding is the increase in frequency of these prepositions in the two later diachronic stages. These prepositions occur at a frequency of 13.43‰ in Classical Greek (n = 3 420) and 12.73‰ in PostClassical Greek (n = 2 300), but they occur at a higher frequency level in Medieval (23.29‰, n = 7 067) and Early Modern Greek (22.08‰, n = 1 421). This difference implies an increase of the range of contexts in which these prepositions occur, which may be predicted by the fact that these prepositions are desemanticized. A chi-square test on the frequencies of the prepositions en/eis revealed that the difference between the two older and the two later stages is statistically significant, χ2(1)=1132.35, p < .001. 30

‰ of n words

25 20 15 en/eis 10 5 0 Classical

Post-Classical

Medieval

Early Modern

Figure 1. Frequency of prepositions originally encoding the region ‘interior’

The question of this paper is what are the consequences of this development for a model of verbal interaction in which a speaker intends to convey a localizing statement to the hearer. Letting the metaphoric uses aside, the

36 Stavros Skopeteas use of either preposition in Classical Greek denotes inclusion within the boundaries of the spatial configuration of the reference object, as exemplified in (7). (7)

heurōÁn ekeî touÁs find:PART.AOR:NOM.SG.M there DEF:ACC.PL.M presbutérous … oikoûntas en older:ACC.PL.M dwell:PART.PRS:ACC.PL.M in taîs oikíais DEF:DAT.PL.F house:ACC.PL.F ‘since he found there the older people dwelling in the houses…’ (Xenophon, Hellenica 6.5.12)

The concept of inclusion requires a reference object with a spatial configuration that may be conceptualized as delimiting an inner space. Very different types of reference objects occur in this conceptualization, e.g. containers and houses, but also masses, groups of individuals, as well as twodimensional reference objects such as areal districts or islands (see a summary in Skopeteas 2003: 275, 2004: 479). Since all spatial prepositions that occur in static and allative relations denote a particular region, Classical Greek belongs to the languages in which the semantic component of spatial region has to be specified in any expression of localization. In previous stages of Greek, the expression of spatial region was not obligatory, since reference objects could be introduced in an adverbial case (dative for static relations, accusative for allative relations, and genitive for ablative relations), as exemplified for the locative dative in (8). However, in Classical Greek, expression of local relation through oblique case NPs is restricted to typical locations such as ‘ground’, ‘house’, etc. and may not be productively used with any type of reference object.6 (8)

toûto gígnetai sîtos this:NOM.SG.N become:3.SG wheat:NOM.SG.M tēÏi gēÏi DEF:DAT.SG.F ground:DAT.SG.F ‘this becomes wheat in the ground’ (Xenophon, Oeconomicus 17.10.4)

Grammaticalization and sets of form-function pairs 37

In the course of desemanticization, the prepositions en/eis lose the feature ‘interior’ and encode merely a non-specific region. Hence, in later stages of Greek, we increasingly find examples in which the concept of inclusion within the boundaries of the reference object is not applicable. Example (9) exemplifies this fact by means of the preposition en: inclusion of a human being in the boundaries of plank is excluded. (9)

ho odusseús mónos DEF:NOM.SG.M Odysseus:NOM.SG:M alone:NOM.SG.M en sanídi toû ploíou in plank:DAT.SG.F DEF:GEN.SG.N ship:GEN.SG.N en tōÏi pelágei ephéreto in DEF:DAT.SG.N see:DAT.SG.N carry:MEDP.IMPF:3.SG ‘Odysseus was carried away on a plank in the sea alone’ (J. Malalas, Chronographia 121.15–16)

The functional consequence of this change for the expression of local relations is that speakers of different diachronic varieties of Greek conveyed different propositional contents in discourse situations in which they had presumably very similar intentions. A speaker of Classical Greek had to choose a spatial region in order to encode an event of localization (with the exception of some typical locations that may be encoded through oblique case NPs). A speaker of a later variety of Greek did not have to encode spatial region. He/she may choose the non-specific prepositions en/eis with any reference object. This data does not answer to the question how speakers of later varieties of Greek encoded the concept of inclusion, when they intended to do so (see section 3.5). What this data shows, is that for a given concept, the speakers of the later varieties did not encode it in all cases that speakers of previous varieties did so. From a purely functional point of view, the desemanticization of this preposition results in an innovation with respect to the model of interaction. It creates the possibility to express the localization in a non-specific region. The literal translation of (9) in Classical Greek would be an expression with the locative dative, but this expression at this diachronic stage is only selected for typical locations, that do certainly not include planks of wrecked ships. The important conclusion for the typology of diachronic changes is that the desemanticization of these prepositions results into a change that is not accommodated through concomitant changes that would preserve the properties of the functional domain at issue.

38 Stavros Skopeteas 3.3. Change of encoding pattern A further type of change, as predicted in section 2.2, does not involve a change of the set of functional concepts, but just a change in the encoding pattern. The contrast of the prepositions en ‘inessive’ and eis ‘illative’ encodes an opposition of spatial relations. The transparency of this contrast decreases already in Post-Classical Greek, as an effect of the desemanticization process already discussed in 2.2. Several local and non-local usages occur with both prepositions, so that they occur in free variation in a wide range of contexts. This development reaches its peak in the Medieval era; in the Modern variety of Greek the functions of these prepositions are undertaken by eis or its phonological descendants is/se/s, while the preposition en becomes extinct (it only occurs in expressions borrowed from Ancient varieties of Greek). 30

‰ of n words

25 20 15

en eis

10 5 0 Classical

Post-Classical

Medieval

Early Modern

Figure 2. Frequencies of eis (orig. ‘illative’) and en (orig. ‘inessive’)

Figure 2 presents the frequencies of these prepositions in the diachronic text sample. The frequency of eis shows an large-size increase in the last era of the sample texts, in which it appears to have completely replaced the preposition en (the difference between the number of occurrences of eis in the Medieval (8.77‰, n = 2 661) and the Early Modern era (21.95‰, n = 1 412) is statistically significant, χ2(1)=841.58, p < .001). The preposition en shows a significant frequency increase in the Medieval era (the differ-

Grammaticalization and sets of form-function pairs 39

ence to Post-Classical Greek is significant, χ2(1)=991.71, p < .001). We speculate that this increase is due to the fact that in the Medieval texts the preposition en undertakes the functions of the dative case (s. Wolf 1912: 43f.). In the texts of the Early Modern Greek stage, the frequency of this preposition decreases significantly (χ2(1)=925.53, p < .001): it actually occurs only 9 times (0.14‰) in idiomatic expressions. The aim of this section is to consider the impact of the loss of contrast between these prepositions on the model of verbal interaction. We assume a speaker that has the intention to convey to the hearer a propositional content involving the spatial relation between two objects in space. How can this intention be realized in the diachronic stages at issue? The concept of spatial relation is encoded in Ancient Greek through prepositions and cases. In case of the region ‘interior’, the opposition between static and allative relations is encoded through the contrast of the prepositions en and eis, as exemplified in (10a) and (10b).7 (10)

a.

b.

en mèn tōÏi stratopédōi … in LK1 DEF:DAT.SG.N barracks:DAT.SG.N pûr ouk ékae fire:NOM.SG.N NEG burn:IMPF:3.SG ‘no fire burned in the barracks’ (Xenophon, Hellenica 6.2.29) píptei keraunòs eis fall:PRS:3.SG lightning:NOM.SG.M into tò stratópedon DEF:ACC.SG.N barracks:ACC.SG.N ‘a lightning fell into the barracks’ (Xenophon, Hellenica 4.7.7)

With some further prepositions, namely epí ‘on’, pará ‘near’, hupó ‘under’ and prós ‘on the side of’ in Classical Greek, the dative~accusative contrast of their complements encodes different spatial relations, namely static~allative respectively (see Smyth 21956: 378-388, Skopeteas 2003: 89). The preposition and case constellations develop several idiosyncratic meanings that reduce the transparency of case in PPs (see a systematic account in Luraghi 2003), but the contrast between dative and accusative for the encoding of spatial relations is still available in Classical Greek as illustrated in (11a) for static relations and in (11b) for allative ones.

40 Stavros Skopeteas (11)

a.

b.

estratopedeúsanto epì tōÏi apantikrù encamp:AOR:3.PL on DEF:DAT:SG.M opposite lóphōi … ridge:DAT.SG.M ‘they encamped on the opposite ridge’ (Hellenica 6.4.4) pémpei lúkion… epì tòn send:3.SG Lykios:ACC.SG.M on DEF:ACC.SG.M lóphon … ridge:ACC.SG.M ‘he sends Lykion onto the ridge’ (Xenophon, Anabasis 1.10.14)

Spatial relations are not exclusively encoded by a feature of the PP (realized on either the P or the case of its complement), it is also encoded through the verb valency. In particular, verbs of motion license a PP complement with the thematic properties of an allative. In Classical Greek, this is not a selectional restriction, namely verbs of motion may also combine with a static PP. Static PPs in examples with motion verbs, such as (12), have a resultative meaning, i.e. they assert that the movement of the localized object ends up in a static relation with the goal of motion (a construction known as ‘constructio praegnans’ in Ancient Greek grammar, see detailed discussion in Skopeteas 2008). Though there is distributional freedom between the verb and the static/allative property of the PP complement at this stage of Greek, it is unambiguous that the PP complement of a motion verb denotes the goal of motion, which implies that the thematic property of an allative complement is determined by the valency of the verbal head (the choice of a static or an allative complement influences the aspectual interpretation, but in both cases the place denoted by the PP is the goal of motion). (12)

hoi mèn autōÏn euthùs en DEF:NOM.SG.M LNK1 3.GEN.PL.M directly in tōÏi potamōÏi épeson DEF:DAT.SG.M river:DAT.SG.M fall:AOR:3.PL ‘some of them directly fell into the river’ (Xenophon, Hellenica 3.4.24)

The crucial historical development is that the contrast between static and allative PPs (both the contrast of en/eis as well as the dative/accusative

Grammaticalization and sets of form-function pairs 41

contrast) is obscured in later diachronic stages. The loss of the feature of static relation is already documented in the early Post-Classical papyri (see Skopeteas 2008: 63). Hence, examples such as (13a-b) illustrate that the same motion verb (in the same aspectual form) occurs with a PP complement that is either introduced by eis or en in free variation. (13)

a.

eiselthoûsa en tēÏi enter:PART.AOR:NOM.SG.F in DEF:DAT:SG.F tēÏs aigúptou khōÏrai country:DAT.SG.F DEF:GEN.SG.F Egypt:GEN.SG.F ‘after she entered to the country of Egypt’ (J. Malalas, Chronographia 28.19)

b.

eis tò palátion eisēÏlthen enter:AOR:3.SG into DEF:ACC.SG.N palace:ACC.SG.N ‘he/she entered to the palace.’ (J. Malalas, Chronographia 371.16)

The same development takes place with respect to the local uses of case within PPs. In the texts of the Medieval era in our sample, we find many cases of accusative with static relations which would not be possible in the Classical era, see example (14). (14)

hoútōs ēÏn hē oíkēsis thus be:PAST:3.SG DEF:NOM.SG.F dwelling:NOM.SG.F autēÏs epì tò óros 3.SG:GEN.F on DEF:ACC.SG.N hill:ACC.SG.N ‘Her dwelling on the hill was like that.’ (J. Malalas, Chronographia 205.21)

The free variation between en and eis or between dative and accusative is characteristic of the texts of the Medieval era. In the Early Modern Greek documents, the free variation ends up by selecting a single option: the preposition eis (see (15)); the case of PP complements is the accusative case, which is reanalyzed as a structural case marking governed NPs (either by a verbal or by a prepositional head) and does not have any contribution to the encoding of spatial concepts anymore.

42 Stavros Skopeteas (15)

a.

nà meínō brádu eis to stay:SUBJ.PFV:1.SG night:ACC.SG.N at tò paláti sou. DEF:ACC.SG.N palace:ACC.SG.N 2.SG:GEN ‘…to stay night at your palace’ (Anonymous, Historia Alexandri Magni, rec. F, 11.4)

The examples so far provided evidence for the loss of the exponents of spatial relation. Though this property is lost in all exponents within the PP complement (either the lexical opposition en/eis or the case opposition), no ‘repair’ strategy took ever place. Does this phenomenon imply a change of the propositional content that speakers exchange in different diachronic stages, similarly to the loss of the concept of ‘interior’? Similarly to the loss of the concept of ‘interior’, the contrast between static and allative relation within the PP becomes extinct. Differently to the concept of ‘interior’, the opposition between spatial relations is also encoded through the verb. What is relevant for the interactive model, is that the utterances in (13), (14), and (15) are unambiguous with respect to the encoded spatial relation though this is not encoded by the preposition or its case properties. Desemanticization has a local effect on the encoding of spatial relations through prepositions/cases but in the global perspective of the whole sentence, it does not affect the possibility to convey particular propositional contents; hence, there is no functional reason for a ‘repair’ strategy to be activated. Desemanticization affects the encoding pattern: while in the older stages of Greek spatial relation is encoded through a combination of the semantic properties of the verb valency with the semantic properties of the PP, in the latter stages of Greek spatial relation is encoded through the verb only. Based on the illustrative language types in (1)-(2), Greek shifts diachronically from the English-type to the Nànáfwê type with respect to the static/allative distinction.

3.4. Eligibility hierarchy for grammaticalization processes In the last two sections, we observed the diachronic development of the prepositions that encode the region ‘interior’. These prepositions are members of a set of local prepositions that encode further spatial regions, e.g. epí ‘superior’, hupó ‘inferior’, pró ‘anterior’, as shown in (6). The latter prepositions do not conflate spatial region and spatial relation as the prepo-

Grammaticalization and sets of form-function pairs 43

sitions en and eis. They only encode spatial regions and combine with local cases for the encoding of different spatial relations in the manner illustrated in (11a-b). Next to the local usages, these prepositions are attested in Classical Greek with several non-local usages that may be derived by metaphorical extension (see details in Luraghi 2003). In the critical era in which the prepositions en and eis are generalized (in the later stages of our diachronic sample), the further prepositions that encode spatial regions become extinct. The diachronic development may be observed in Figure 3, which repeats the frequencies of the prepositions of ‘interior’ (see Figure 1) together with the frequencies of three further prepositions that encode spatial regions, i.e., epí ‘superior’, hupó ‘inferior’, and pró ‘anterior’. 30

‰ of n words

25 20 en/eis 15

epí hupó

10

pró

5 0 Classical

Post-Classical

Medieval

Early Modern

Figure 3. Frequencies of local prepositions

The data presented in Figure 3 reveals an asymmetry in the frequency of spatial regions with striking stability across diachronic stages, presented in (16). The frequencies of the prepositions en/eis are given in sum, since these prepositions encode the same spatial region and their alternation is equivalent to the dative/accusative alternation with the other prepositions. However, note that the frequency of each of these prepositions alone is higher than the frequencies of the other prepositions, a difference which is statistically significant. Across the three earlier diachronic stages, it holds that neis (7 240) > nepí (3 830), χ2(1)=1057.7, p < .001, and nen (6 968) > nepí (3 830), χ2(1)=918.1, p < .001. Furthermore, the differences among the

44 Stavros Skopeteas further prepositions are significant as well: nepí (3 830) > nhupó (1 905), χ2(1)=648.45, p < .001, and nhupó (1 905) > npró (346), χ2(1)=1081.25, p < .001. From these measurements, we conclude that the hierarchy in (16) holds for Greek across diachronic stages.8 (16)

interior (en/eis) > superior (epí) > inferior (hupó) > anterior (pró)

The hierarchy in (16) is in line with asymmetries already reported in the literature on spatial regions, based on different types of data. A number of earlier psycholinguistic studies (Clark 1973: 33f., Clark & Clark 1978: 240–243 among others) propose that the antonymic pairs of spatial regions display markedness asymmetries that reflect asymmetries in the perceptual accessibility of spatial concepts. Note that the lexical gaps in the encoding of spatial regions through prepositions in Ancient Greek are in line with these asymmetries, hence there are prepositional exponents of ‘interior’ (en/eis), but no corresponding exponent for ‘exterior’; similarly, there is a prepositional exponent of ‘anterior’ (pró), but not corresponding exponent for ‘posterior’ (see section 3.5 for the encoding of these regions through adverbs). (17)

Perceptual asymmetry (Clark 1973: 33f.) interior >accessibility exterior superior > accessibility inferior anterior > accessibility posterior

Johnston and Slobin (1979) investigated language acquisition by English, Italian, Serbo-Croatian, and Turkish children. Their findings are summarized in a cross-linguistic hierarchy, a short version of which is given in (18). This hierarchy corresponds to the order of acquisition of spatial adpositions in all examined languages. (18)

Order in language acquisition (Johnston and Slobin 1979: 529) (interior | superior | inferior) (anterior | posterior | lateral)

The hierarchy of frequencies in the Greek corpus data in (16) is in line with all asymmetries presented above. We may also speculate that the empirical data in Figure 3 and the choice of the prepositions of ‘interior’ to enter the grammaticalization process is not the result of random evolution, but reflects asymmetries inherent to the functional domain at issue attested in different types of data from a wide range of languages.

3.5. Compensatory processes The previous sections provided evidence that the prepositions of interior are grammaticalized and occur in a wide range of contexts in the later diachonic stages of Greek. In section 3.2, we discussed the consequences of the former development for the interactive model: while in the older stage, a speaker of Greek has to choose a spatial region, in the newer stage he/she does not have to do so. This section discusses the question how the speaker may express the concept of region if he/she intends to do so. This question equally applies to all regions that were once encoded through prepositions which were either desemanticized or became extinct in the later stages of Greek. In view of the interactive model, the problem for the speaker is in both cases the same: to select (or to develop) an alternative to the prepositional expression of spatial regions. A paradigmatic alternative to the expression of spatial regions is already available in Ancient Greek grammar, namely the use of spatial adverbs. Prepositions and adverbs are distinguished by their governing properties: the former but not the latter license a NP complement (see Lehmann 1985). The consequences of these syntactic properties in the view of an interactive

46 Stavros Skopeteas model are not straightforward, as will be shown in the following. When a speaker intends to encode the spatial region of an implicit reference object that is uniquely identified in the discourse situation (either the deictic center or a referent that is highly accessible in the context), he/she selects a spatial adverb, as illustrated in (21).9 (21)

polloì éksō apéthnēiskon many:NOM.SG.M outside die:IPFV:3.PL ‘many died outside’ (Xenophon, Hellenica 6.2.15)

When the speaker intends to encode the reference object, then two structural options are available. The first option is to select a local preposition that encodes the respective spatial region and to encode the reference object as its NP complement (see examples (7) and (11) above). Alternatively, he/she selects a syntactic constituent involving an adverb and a reference object NP, which is possible with two syntactic constructions: (a) the reference object NP appears as a genitive modifier of the spatial adverb, as exemplified in (22a) (in Modern Greek a PP introduced by apó ‘ABL’ is used instead of the genitive NP), or (b) the spatial adverb modifies a local PP with a preposition that encodes a non-specific spatial region, as exemplified in (22b) from a text of the Medieval era in which the preposition eis is arguably desemanticized (see section 3.2). The latter example is a typical case of reinforcement. (22)

a.

b.

hoplítas apobibáseien heavy.armed:ACC.PL.M get.off:CAUS:AOR.OPT:3.SG pulō›n eísō kaì éksō tō›n to.inside and outside DEF:GEN.PL.F gate:GEN.PL.F ‘he would have let the heavy armed soldiers get off inside and outside the gates’ (Xenophon, Anabasis 1.4.5) ho rhōmanòs ē›ton be:IPFV:3.SG DEF:NOM.SG.M Romanos:NOM.SG.M ésō eis tò loutròn inside at DEF:ACC.SG.N bath:ACC.SG.N ‘Romanos was in the bathroom’ (Georgios Monachos, Chronicon breve 110.1221)

In sum, a speaker who intends to convey a propositional content involving the spatial region of a reference object has the following encoding possi-

Grammaticalization and sets of form-function pairs 47

bilities: (a) for some spatial regions, there is only an adverbial option available, (b) for some other spatial regions, there are two available options, a prepositional and an adverbial one (see listing in (6)). By hypothesis, in cases that two alternatives are available, the unmarked case is the choice of preposition; the use of an adverb for these regions is the marked option, which may be selected if the speaker intends to attract the attention of the hearer to the spatial region (e.g. in the case of focus). These speculations allow for the following predictions: Spatial adverbs encoding the same spatial region as a corresponding preposition in Ancient Greek are expected to rarely occur in a construction with an NP (of either type illustrated in (22). The use of adverbs in these constructions is expected to increase in the course of language change, parallel to the desemanticization of the corresponding prepositions. Spatial adverbs that do not correspond to a synonymous preposition are expected to occur frequently in a construction with an NP across diachronic stages. Figure 4 presents the proportion of occurrences of spatial adverbs that involve a syntactic construction with a reference object NP, as illustrated in (22). The spatial adverbs are grouped in two subsets following the classification in (6). Adverbs that have a corresponding preposition (~ prep) are the adverbs that encode the regions ‘interior’ (éndon, eísō, ésō, éndothen, ésōthen, apésō, mésa), ‘superior’ (ánō, epánō, apánō, ánōthen, epánōthen), ‘inferior’ (kátō, kátōthen, hupokátō, hupokátōthen), and ‘anterior’ (prósō, prósthe(n), émprosthen, emprós). Adverbs that do not have a corresponding preposition (no ~ prep) are those that encode the regions ‘exterior’ (éksō, éksōthen, apéksō) and ‘posterior’ (opísō, písō, eksopísō, ópisthe(n), eksópisthe(n)). The alternative forms of the adverbs are either different phonological forms occurring in different eras or they display semantic differences that are not expected to have an influence on the syntactic properties at issue.10 Figure 4 largely confirms our predictions. Adverbs with a corresponding preposition (~ prep), such as the adverbs denoting inclusion, e.g., éndon ‘inside’, occur with a reference object NP only rarely in the Ancient stages (n=16 out of 288 occurrences, 5.56%, in Classical Greek, and n=8 out of 187 occurrences, 4.28%, in Post-Classical Greek) (see exact counts per spatial region in the appendix, Table 4). The desemanticization of prepositions in Medieval Greek correlates with a significant increase of the occurrence of these adverbs with NPs (n=132 out of 284 occurrences, 46.48%), which reaches an even higher level in Early Modern documents (n=146 out of 221 occurrences, 66.97%). A chi-square test on the pre-

48 Stavros Skopeteas dicted difference reveals that the frequency increase of these adverbs in the Medieval era is statistically significant (χ2(1)=96.13, p < .001). Presumably, the gradient frequency increase reflects the gradient grammaticalization of the corresponding prepositions. Hence, the data pattern of these adverbs fully corresponds to our expectations.

% with NP of n occurrences

100 80 60

~P no ~ P

40 20 0 Classical

Post-Classical

Medieval

Early Modern

Figure 4. ‘Adverb + NP’ constructions

Adverbs that do not have a corresponding preposition, such as éksō ‘outside’ occur frequently with reference object NPs already in the old stages (33.90% in Classical Greek and 27.78% in Post-Classical Greek, see Table 4) (both frequencies differ significantly from the respective frequencies of ‘~ prep’ adverbs of the same era, χ2(1)=42.84, p < .001 for Classical Greek, and χ2(1)=20.26, p < .001 for Post-Classical Greek). Furthermore, we observe a (non-predicted) slight increase of this construction in Medieval Greek and a (non-predicted) drop of the frequency in the Early Modern Greek texts. These effects are not predicted by our hypotheses and are not interpretable. Further research is required in order to identify whether these findings correspond to some grammatical changes in Greek or are simply artifacts of the small text sample (especially the observations for Early Modern Greek relate to 51 textual occurrences of ‘exterior’/‘posterior’ adverbs). However, note that our hypothesis predicts the low frequency of relational constructions with adverbs that display a more or less synonymous preposition, which is the case for interior and anterior regions in the

Grammaticalization and sets of form-function pairs 49

older stages of our empirical basis. The gradient differences between all other cases is actually irrelevant for this hypothesis. This section provided evidence for the impact of the grammaticalization of particular items to the speaker’s choices within an interactive model of verbal communication. The desemanticization of the prepositions en/eis and the extinction of the further prepositions that were used for the encoding of spatial regions in Ancient Greek resulted in the increase of the use of spatial adverbs in constructions with a reference object NP. If we consider this change outside its functional environment, we may conclude that it is an instance of renovation: the prepositions en ‘inessive’/eis ‘illative’ are renovated by the adverb mésa ‘inside’ in Modern Greek. However, the empirical data presented in this paper do not reveal any change in the relational properties of the adverbs in Greek, since these elements could be used in constructions with reference object NPs already in Ancient Greek. The reference object NP did not became an obligatory complement (which would imply that the adverbs shifted to prepositions), but they occur more frequently in later stages as a result of the loss of competing paradigmatic alternatives for the encoding of the same propositional content.

4. Conclusions This paper discussed the underlying assumptions of grammaticalization cycles. We argued that the observed variation in the functional concepts that are available in the languages of the world suggests that functional concepts should be subject to language change. When a linguistic unit loses its functional properties due to a grammaticalization process, a reinforcement or renovation process that preserves the functional properties of the source grammar is only one among the possible historical options. By means of empirical evidence from the history of Greek adpositions we were able to show that ‘linguistic conservatism’ is not necessary, and that loss of particular functional properties may result in an innovation of the properties of the functional domain. In section 3.2, we have shown that speakers of the recent stages of Greek do not have to encode spatial regions in all contexts in which speakers of the older stages had to. The process of loss may locally affect the set of functional concepts that are encoded by a particular paradigm, but without having crucial consequences for the set of propositional contents that may be encoded in an utterance. An example of this type was the loss of the distinction

50 Stavros Skopeteas static/allative in the Greek prepositions and the cases of the prepositional complements (see section 3.3). Since this distinction is encoded through the verb, there is no functional motivation for a ‘repair’ strategy. In this case, the loss is only local (with respect to the particular paradigm); the effect of the grammaticalization is a change in the encoding pattern. In these cases, the history of Greek adpositions does not provide evidence for a compensatory change that is invoked in order to accommodate the functional consequences of the grammaticalization process. The interaction between the syntax of adverbs and the grammaticalization of the corresponding prepositions (3.5) provided evidence for a change to this direction. The generalization of the empirical study is that spatial adverbs occur more frequently in constructions with a reference object NP, when there is no corresponding preposition for the encoding of the same region. The consequence of the grammaticalization or the extinction of spatial prepositions is that the corresponding adverb increasingly occurs with reference object NPs. The aim of this paper was to exploit the possible outcomes of a grammaticalization process. The empirical data present two alternative diachronic scenarios to the functionally-driven compensation that suggest that functions are not inevitable and their preservation does not always determine language change. A last question is to what extent are the observed phenomena generalizable for all linguistic functions. A linguistic function that is extensively studied with respect to the cycle of grammaticalization and regeneration processes is the expression of negation (see Jespersen 1917). Though the expression of negation seems to be an inevitable communicative function at first sight, evidence from language typology shows that languages differ in the encoding patterns of this function, but also in the availability of a affirmative-negative contrast (in some languages negation is not straightforwardly contrasted to affirmation, but is a subordinate concept of the irrealis moods, see Bhatt 2004: 1211). Hence, the attested language variation predicts that the effects of grammaticalization on the expression of negation should not necessarily be accompanied by a regeneration process, which is subject to further research. However, we may speculate that the likelihood for a particular concept to be persistent in diachronic change is a function of its global relevance for a cross-linguistic model of human interaction which is reflected in cross-linguistic eligibility hierarchies as presented for the spatial concepts in section 3.4.

Grammaticalization and sets of form-function pairs 51

Notes 1.

The main idea of this paper is the product of previous discussions with Christian Lehmann for which I am particularly indebted. Grateful thanks are due to Thanasis Georgakopoulos (University of Athens) and Katerina Stathi (Freie Universität Berlin) for their comments on a previous manuscript. 2. Note that Lehmann (1995c: 22) does not treat renovation/reinforcement as necessary consequences of grammaticalization, but as instances of ‘linguistic conservatism’ which implies that they are optional in nature (see quotation in section 1). 3. See Haspelmath (1999: 1054f.) for the role of maxims of human interaction in the explanation of grammaticalization processes. 4. The text corpus used in this study is necessarily a convenience sample. Aristophanes, Lysias, and Xenophon belong to different varieties of Classic Attic. Post-Classical Greek is represented in the sample through the new Attic Comedy (Menander), and the Hellenistic Romances (Chariton, Xenophon Ephesius, Longus, and Heliodorus). The Medieval and Early Modern Greek works are considered to be close to the colloquial variety. 5. The corpus measurements that are reported in this paper take into account all allomorphs of the prepositions (e.g., eis~es and later se) or adverbs (e.g., páno~epáno) at issue, as well as alternations that are phonologically conditioned (e.g., epí~ep~eph). 6. Jannaris (1897: 347) observes that local cases are replaced by prepositions, whereby the notion of ‘replacement’ should be understood very roughly: Ancient Greek prepositions are not synonymous to the local cases. 7. Furthermore, ablative relations with respect to the region ‘interior’ are encoded through the preposition ek ‘from within’. 8. The frequency hierarchy ‘interior’ > ‘superior’ > ‘inferior’ > ‘anterior’ > other regions also holds for (Present-Day) Modern Greek adverbs, see corpus study in Goutsos (2007: 44). 9. This generalization holds for Classical Greek, while in previous stages of Greek the elements at issue were adverbs (see Jannaris 1897: 366, Smyth 2 1956: 365, Horrocks 1981: 45). See also Chila-Markopoulou (2007) concerning the fluent transition between these categories and the corresponding problems in classification. 10. In particular, the suffixed forms in -then encodes ablative relations in the oldest documents of Greek, but their use with spatial adverbs is by large nontransparent in the diachronic stages at issue.

52 Stavros Skopeteas

References Bhatt, D.N.S. 2004 Negation. In Morphology: An International Handbook on Inflection and Word-Formation, Geert Booij, Christian Lehmann, Joachim Mugdan, and Stavros Skopeteas (eds.), 1207–1212. Berlin, New York: De Gruyter. Bortone, Pietro 2000 Aspects of the history of Greek prepositions. A ‘localistic’ view. Ph.D. dissertation, University of Oxford. Chila-Markopoulou, Despina 2007 Apo tin istoria tu pera(n). In Glossikos Periplous: Meletes afieromenes sti Dimitra Theophanopoulou-Kontou, Tomeas Glossologias Panepistimiou Athinon (ed.), 387–397. Athens: Kardamitsa. Clark, Herbert 1973 Space, time, semantics and the child. In Cognitive development and the acquisition of language, Timothy E. Moore (ed.), 27–63. New York u. a.: Academic Press. Clark, Eve V. & Clark, Herbert H. 1978 Universals, relativity, and language processing. In Universals of human language, vol. 1, Joseph H. Greenberg (ed.), 225–277. Stanford: Stanford University Press. Croft, William 2000 Explaining Language Change. An Evolutionary Approach. Harlow: Longman. Drossard, Werner 1992 Lokale Relationen: Vom Einfacheren (Topologischen) zum Komplexeren (Dimensionalen). Sprachliche Reflexe einer psycholinguistischen Erkenntnis. Arbeiten des Kölner Universalienprojekts 86, 16–56. Goutsos, Dionysis 2007 Vasika topika epirimata se ilektronika somata kimenon: Prokatarktikes paratirisis. In Glossikos Periplous: Meletes afieromenes sti Dimitra Theophanopoulou-Kontou, Tomeas Glossologias Panepistimiou Athinon (eds.), 36–46. Athens: Kardamitsa. Haspelmath, Martin 1999 Why is grammaticalization irreversible? Linguistics 37 (6): 1043– 1068. 2008 Comparative concepts and descriptive categories in cross-linguistic studies. Ms., Max-Planck-Institute for evolutionary anthropology, Leipzig.

Grammaticalization and sets of form-function pairs 53 Horrocks, Geoffrey C. 1981 Space and time in Homer. Prepositional and adverbial particles in the Greek Epic. New York: Arno Press. Jannaris, Antonius N. 1897 An historical Greek grammar chiefly of the Attic dialect. As written and spoken from classical antiquity down to the present time; founded upon the ancient texts, inscriptions, papyri and present popular Greek [Nachdruck: Hildesheim: Olms, 1968]. London. Jespersen, Otto 1917 Negation in English and other Languages. Copenhagen: Høst and Søn. Johnston, Judith R., and Dan I. Slobin 1979 The development of locative expressions in English, Italian, SerboCroatian and Turkish. Journal of child language 6: 529–545. Keller, Rudi 1994 Sprachwandel [2nd ed.; 11990]. Tübingen/Basel: Francke. Lehmann, Christian 1985 On grammatical relationality. Folia Linguistica 19, 67–109. 1989 Language description and general comparative grammar. In Reference grammars and modern linguistic theory, Gottfried Graustein and Gerhard Leitner (eds.), 133–162. Tübingen: Niemeyer. 1992 Yukatekische lokale Relatoren in typologischer Perspektive. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 45: 626–641. 1995a Raumkonstruktion in funktionaler Sicht. Unpublished ms., University of Erfurt. 1995b Thoughts on Grammaticalization. Munich: Lincom-Europa. Luraghi, Silvia 2003 On the semantics of cases and prepositions: A study of the expression of semantic roles in Ancient Greek. Amsterdam: Benjamins. Meillet, Antoine 1921/1935 Linguistique historique et linguistique générale. 2 vols. Paris: Klincksieck. Rennison, Joհn R. 1997 Koromfe. London/New York: Routledge. Seiler, Hansjakob 1988 Die universalen Dimensionen der Sprache: Eine vorläufige Bilanz. Köln: Institut für Sprachwissenschaft der Universität (Arbeiten des Kölner Universalien-Projekts, 75). Skopeteas, Stavros 2003 Lokale Konstruktionen im Griechischen: Sprachwandel in funktionaler Sicht. Ph. D. dissertation, University of Erfurt.

54 Stavros Skopeteas 2004

Prepositions and Adverbs in Classical Greek. In Word Classes and Related Topics in Ancient Greek, Emilio Crespo, Jesus De La Villa, Antonio R. Revuelta (eds.), 471–485. Louvain-La-Neuve: Peeters. 2008 Encoding spatial relations: language typology and diachronic change in Greek. Language Typology and Universals 61 (1): 54–66. Smyth, Herbert Weir 2 1956 Greek Grammar [11920]. Cambridge/Mass.: Harvard University Press. Svorou, Soteria 2002 Semantic constraints in the grammaticalization of locative constructions. In New Reflections on Grammaticalization, Ilse Wisher and Gabriele Diewald (eds.), 121–142. Amsterdam/Philadelphia: Benjamins. Talmy, Leonard 1983 How language structures space. In Spatial orientation: theory, research and application, Herbert L. Pick and Linda P. Acredolo (eds.), 225–282. New York/London: Plenum Press. 2000 Toward a cognitive Semantics, vol. II: Typology and Process in concept structuring. Cambridge, Mas./London: MIT Press. Theophanopoulou-Kontou, Dimitra 2000 Topika epirrimata ke ptosi stin Eliniki: Diaxroniki prosegisi. Glossologia 11/12: 1–40. Wolf, Karl 1912 Studien zur Sprache des Malalas. II. Teil: Syntax. München: Akademische Buchdruckerei von F. Straub. Wunderlich, Dieter 1982 Sprache und Raum. Studium Linguistik 12, 1–19; 13, 37–59.

Grammaticalization and sets of form-function pairs 55

Appendix Table 1. Text sample diachronic stage Classical

Post-Classical

Medieval

Early Modern

texts Aristophanes, 5th–4th c. BC: Opera Omnia Lysias, 5th–4th c. BC: Opera Omnia Xenophon, 4th c. BC: Hellenica Total Menander, 4th–3rd c. BC: Aspis, Dyscolus, Epitrepontes, Misumenus, Periciromene, Samia, Sicyonus Chariton, 2nd c. AD: De Chaerea et Callirhoe Xenophon Ephesius, 2nd c. AD: Ephesiaca Longus, 3rd c. AD: Daphnis et Chloe Heliodorus, 3rd c. AD: Aethiopica Total Ioannes Malalas, 6th c. AD: Chronographia Georgius Monachus, 9th c. AD: Chronicon breve Total Anonymous, 12th c. AD: Digenes Akritas Anonymous, 14th c. AD: Hist. Alexandri Magni (F) Total

n of words 119 562 67 059 67 939 254 560

26 845 35 523 17 197 20 929 80 126 180 620 91 268 212 141 303 409 16 000 48 332 64 332

56 Stavros Skopeteas Table 2. Frequencies of prepositions en, eis, epí, en, en, hupó, pró, (n: occurrences of preposition in the texts of the era; ‰: proportion of n with respect to the total number of words of the text sample of the respective era, see counts in Table 1)

inessive illative superior inferior anterior Total

Classical n ‰ 1685 6.6 1735 6.8 1163 4.5 681 2.6 92 0.3 5356 21.1

Post-Classical n ‰ 868 4.8 1432 7.9 1002 5.5 314 1.7 91 0.5 3707 20.5

Medieval n ‰ 4406 14.5 2661 8.7 1660 5.4 910 3.0 28 0.1 9665 31.8

Early Modern n ‰ 9 0.1 1412 21.9 5 0.1 0 0.0 1 0.0 1427 22.1

Table 3. Frequencies of adverbs (n: occurrences of adverbs in the texts of the era; ‰: proportion of n with respect to the total number of words of the text sample of the respective era, see counts in Table 1)

interior exterior superior inferior anterior posterior Total

Classical n ‰ 120 0.47 59 0.23 56 0.22 36 0.14 76 0.30 32 0.13 379 1.49

Post-Classical n ‰ 103 0.57 36 0.20 33 0.18 24 0.13 27 0.15 8 0.04 231 1.28

Medieval n ‰ 69 0.23 86 0.28 109 0.36 57 0.19 49 0.16 47 0.15 417 1.37

Early Modern n ‰ 75 1.17 30 0.47 77 1.20 39 0.61 30 0.47 21 0.33 272 4.23

Table 4. Occurrence of adverbs with reference object NPs (n: occurrences of adverbs in constructions with reference object NPs; %: proportion with respect to the entire number of occurrences of the corresponding adverb(s) in the respective era, see counts in Table 3)

interior exterior superior inferior anterior posterior Total

Classical n % 5 4.17 20 33.90 5 8.93 0 0.00 6 7.89 8 25.00 44

Post-Classical n % 3 2.91 10 27.78 0 0.00 0 0.00 5 18.52 1 12.50 19

Medieval n % 30 43.48 40 46.51 43 39.45 28 49.12 31 63.27 28 59.57 200

Early Modern n % 49 65.33 7 23.33 57 75.00 13 34.21 27 93.10 1 4.76 154

Part II Nominal categories and adpositions

Grammaticalization of Korean numeral classifiers Myung-Chul Koo

1. Introduction It is well-known that Korean is a language with numeral classifiers. In this sense, it is not surprising that many Korean linguists such as Yang (1995, 1996), Lee (1998), Woo (2001), and Woo et al. (2005) have already discussed this topic. However, none of these works treats the grammaticalization of Korean numeral classifiers as its primary subject. The purpose of this paper therefore is to introduce Korean numeral classifiers and to discuss their grammaticalization. A substantial part of the Korean numeral classifiers has developed from common nouns and can be considered grammaticalized. Diachronic and synchronic evidence supports this fact (cf. section 3.2). This paper primarily discusses some typical examples of Korean numeral classifiers that show characteristics of the kind of grammaticalization described by Lehmann (1995, 2005a). Korean numeral classifiers can be divided into several groups according to semantic criteria such as animacy, shape, and function. Each of these groups contains at least four classifiers, but one among these classifiers tends to gain preeminence and replace the others within the group. Hence, although all members of the paradigm are of common origin, a numeral classifier exhibits a higher degree of grammaticality than the others. The relation of this type of change to grammaticalization will be a central subject of this paper.

2. Numeral classifiers in Korean According to Woo (2001: 105), there are more than 100 numeral classifiers in Korean,1 and all of them are derived from nouns.2 According to the inherent properties of the counted object, numeral classifiers can be divided into numeral classifiers for human beings, for animals, for plants (i.e. vegetables and grains), and for inanimate entities, as in (1).3

60 Myung-Chul Koo (1)

a.

b.

c.

d.

tasPs kyPngchal policeman five ‘five policemen’ so tasPs cow five ‘five cows’ sonamu tasPs k¨lu pine.tree five ‘five pine trees’ ciukae tasPs eraser five ‘five erasers’

myPng CL.name mali CL.animal

CL.stump kae CL.piece

Following Yang (1996: 198), myPng ‘name’ is a prototypical numeral classifier for human beings, but different classifiers can be used “according to speaker’s respect for or relationships to the referent.” In this way, he distinguishes between the various Korean numeral classifiers that refer to human beings:4 (2)

a. b. c.

myPng ‘name’: neutral pun ‘esteemed person’: highly respective, honorific nom ‘guy’: slightly downgrading, originally referred to male, but neutralized

Following Woo (2001: 186), salam ‘person’ is also a numeral classifier that is neutral with regard to speaker’s respect for the referent. The categorization of salam as a numeral classifier is not in doubt, since it is combined with numerals in order to count the number of people (e.g., kyPngchal yPl salam (lit. policeman ten person) ‘ten policemen’). salam as a numeral classifier is distinguished from myPng ‘name’ in the sense that the former is usually restricted to a head noun that conveys personal intimacy, while the latter has not such restrictions (cf. Woo 2001: 203). In the case of animals, the prototypical numeral classifier is mali ‘head’, but tu ‘head’, phil ‘one’, su ‘head’, etc. can also be taken into consideration. tu is primarily used to refer to cows, and occasionally to horses. By contrast, phil is primarily used to refer to horses, and occasionally to cows. su is used for counting chickens or ducks. To count plants, classifiers such as k¨lu ‘stump’, thol ‘grain’, phoki ‘body of vegetables’, ppuli ‘root’, and songi ‘blossom’ representing the

Grammaticalization of Korean numeral classifiers 61

prominent or useful parts of plants are used. Their uses can be categorized as follows: (3)

a. b. c. d. e.

k¨lu ‘stump’: for trees thol ‘grain’: for grains or nuts phoki ‘body of vegetables’: for weeds or Chinese cabbages ppuli ‘root’: for ginsengs or radishes, edible roots songi ‘blossom’: for flowers

Since a part of the referent is used for the classification, the selection of numeral classifier is determined by synecdoche in this case. The items kae ‘piece’, kaepi ‘piece of split wood’, calu ‘grip, helve’, cang ‘sheet’, kwPn ‘volume’, al ‘egg’, chae ‘bulk’, tae ‘unit’, chPk ‘vessel’, etc. belong to the numeral classifiers for inanimate entities. The appropriate classifier is selected primarily according to the physical shape of the objects being counted (cf. Craig 1994: 567 on semantic properties of classifiers): For one-dimensional objects, kaepi or calu is used; for twodimensional ones, cang or kwPn; for three-dimensional ones, kae, al, chae, tae, or chPk. The items kaepi and calu can be further distinguished according to the size of their referents: kaepi is used for small and thin one-dimensional objects such as cigarettes, matches, etc. and calu for relatively big objects, such as daggers and shotguns as well as for writing instruments such as pens and pencils. For two-dimensional objects, cang is used as an unmarked classifier. It is associated with nouns that refer to flat objects such as paper, bank notes, credit cards, diskettes, etc. The item kwPn is functionally motivated and is used for all sorts of books, including magazines and pamphlets (on functionally motivated classifiers, cf. Woo 2001, section 5.4). According to Woo (2005: 313), kae is a numeral classifier for three-dimensional objects. If an object has a spherical form such as pills or bullets, then al is used as its classifier. For some objects such as buildings, cars, buses, airplanes, and ships, special classifiers that are functionally motivated are put to use:5 chae for buildings; tae for cars, buses, airplanes, (musical) instruments, and machinery;6 chPk for ships. The descriptions of Korean numeral classifiers are summarized in Table 1 (cf. Woo 2001: 235).

62 Myung-Chul Koo Table 1. Korean numeral classifiers semantic class

human

animals

plants (vegetables, grains)

one-dimensional

inanimate entities

twodimensional

threedimensional

formally motivated functionally motivated formally motivated functionally motivated

numeral original classifier meaning myPng name esteemed pun person guy nom person salam head mali head tu one phil head su k¨lu stump grain thol body of phoki vegetables root ppuli blossom songi kaepi split wood calu grip

gloss

CL.guy CL.person CL.animal CL.cattle CL.horse CL.poultry CL.stump CL.grain CL.vegetables

cang

sheet

CL.sheet

kwPn

volume

CL.volume

kae al chae tae chPk

piece egg bulk unit vessel

CL.piece CL.egg CL.building CL.unit CL.vessel

CL.name CL.person(HON)

CL.root CL.blossom CL.split_wood CL.grip

3. Arguments for the grammaticalization of numeral classifiers 3.1. General properties The various Korean numeral classifiers have some characteristics of grammaticalization in common. First, they have lost their lexical meaning. The degree of their desemanticization varies according to the degree to

Grammaticalization of Korean numeral classifiers 63

which the classifier is grammaticalized (cf. 3.2). At any rate, they show some loss of paradigmatic weight, namely their integrity (cf. Lehmann 1995: 126f. on integrity and cf. Lehmann 2005a on grammaticalization in general). Second, they permit no plural marking as in (4) (cf. Lehmann 2000 on Stück and Mann as a numeral classifier in German). This means that the Korean numeral classifiers have more or less lost their lexical meaning and have little or no referring function (cf. Allan 1977: 293 on the loss of lexemic status for classifiers). (4)

a.

b.

c.

d.

kyPngchal tasPs policeman five ‘five policemen’ so tasPs cow five ‘five cows’ sonamu tasPs pine.tree five ‘five pine trees’ ciukae tasPs eraser five ‘five erasers’

*myPng-t¨l/*salam-t¨l CL.name-PL/CL.person-PL *mali-t¨l/*tu-t¨l CL.animal-PL/CL.cattle-PL *k¨lu-t¨l CL.stump-PL *kae-t¨l CL.piece-PL

Third, whenever objects are counted, an appropriate classifier must be selected according to its semantic property, as has been shown in (1). This means that the classifiers have a paradigmatic relationship. Fourth, they cannot in principle appear alone without a numeral, although they are free forms (see details in 3.2.1). That is, they lost their autonomy and are regarded as ‘dependent nouns’ in Korean grammar (cf. Nam and Ko 1985: 74). Furthermore, nothing can intervene between the numeral and the classifier, regardless of the degree of its grammaticalization, i.e., the bondedness of Korean numeral classifiers to the numeral is very high (cf. some examples of bondedness in Lehmann 1995: 170). (5)

a.

kyPngchal tasPs policeman five ‘five tall policemen’

(*kh¨n) myPng/salam (big) CL.name/CL.person

64 Myung-Chul Koo b.

c.

d.

so tasPs cow five ‘five big cows’ sonamu tasPs pine.tree five ‘five big pine trees’ ciukae tasPs eraser five ‘five big erasers’

(*kh¨n) mali/tu (big) CL.animal/CL.cattle (*kh¨n) k¨lu (big) CL.stump (*kh¨n) kae (big) CL.piece

Last, because the cohesion of the classifier X with the numeral A is more or less strong, ‘A X and A Y’ cannot be reduced to ‘A [X and Y]’ (cf. Lehmann 1995: 150). Examples in (6) demonstrate that the coordination reduction in the b-constructions is impossible. This fact leads to the conclusion that the bondedness of Korean numeral classifiers is very high. (6)

a.

b.

(chaek) se kwPn (yPnphil) se calu-oa (pencil) three CL.grip-CONJ (book) three CL.volume ‘three pencils and three books’ *se calu-oa kwPn three CL.grip-CONJ CL.volume ‘three grips and volumes’

3.2. Grammaticalization asymmetries within groups of numeral classifiers 3.2.1. Numeral classifiers for human beings As we have seen in section 2, human nouns take myPng, pun, nom, or salam as a numeral classifier. myPng, pun, nom, and possibly salam are dependent nouns in Korean grammar because they cannot occur without any modifier. Of these, myPng seems to be the most grammaticalized classifier. Two lines of evidence support this argument: myPng can only be combined with numerals, not with adjectives and determiners (see (7)), which does not hold for the further classifiers of this group, namely pun, nom, and salam (see (8)).

Grammaticalization of Korean numeral classifiers 65

(7)

a.

b.

c.

(8)

a.

b.

c.

kyPngchal tasPs myPng policeman five CL.name ‘five policemen’ myPng *i/k¨/cP this/the/that CL.name ‘this/the/that person’ myPng *cak¨n small CL.name ‘a small person’ sPnsaengnim/chinku/kyPngchal tasPs pun/ teacher/friend/policeman five CL.person(HON) nom/salam /CL.guy/CL.person ‘five teachers/friends/policemen’ pun/nom/salam i/k¨/cP this/the/that CL.person(HON)/CL.guy/CL.person ‘this/the/that person/guy’ pun/nom/salam cak¨n small CL.person(HON)/CL.guy/CL.person ‘a small (esteemed) person/guy/person’

This means that myPng occupies a fixed syntactic slot (i.e. only after numerals). It may be concluded that myPng is the most grammaticalized classifier because it has no syntagmatic variability. By contrast, pun, nom, and salam have some syntagmatic variability, and therefore they are less grammaticalized than myPng. The item salam, especially, permits a quantifier like mot¨n and myPchmyPch, while the combination of myPng, pun, and nom with this quantifier is impossible or marginal. (9)

a.

b.

c.

*mot¨n/*myPchmyPch myPng all/some CL.name ‘all/some persons’ ? mot¨n/?myPchmyPch pun/nom all/some CL.person(HON)/CL.guy ‘all/some (esteemed) persons/guys’ mot¨n/myPchmyPch salam all/some CL.person ‘all/some persons’

66 Myung-Chul Koo The item salam can even occur as an independent common noun: (10)

salam-¨n sahoicPk tongmul-i-ta person-TOP social animal-be-DECL ‘Human beings are social animals.’

This shows that the numeral classifier salam has evolved from a common noun.7 Its diachronic development to numeral classifier has the properties of a grammaticalization process, since it is losing its lexical meaning and entering the classifier paradigm.8 This grammaticalization process began in the middle Korean period, in which the earliest record of the use of saram as a numeral classifier is attested (Woo 2001: 189). Although the grammaticalization process of salam has lasted for a long time, it is the least grammaticalized of the classifiers for human beings, as has been shown. The fact that salam is combined with a restricted set of nouns is also an argument for its low degree of grammaticalization.9 (11)

a.

b.

c.

myPng/salam kyPngchal tasPs policeman five CL.name/CL.person ‘five policemen’ chinku tasPs myPng/?salam friend five CL.name/CL.person ‘five friends’ Plini tasPs myPng/*salam child five CL.name/CL.person ‘five children’

Therefore, it may be argued that grammaticalization increases from salam over pun/nom to myPng. There is another evidence for this argument. The item myPng can always substitute for salam, pun, and nom. This means that myPng is already desemanticized and does not impose any semantic restrictions.10 (12)

a. b.

sPnsaengnim/haksaeng/Plini/chinku/kangto yPl myPng teacher/student/child/friend/robber ten CL.person sPnsaengnim/*/?haksaeng/*Plini/*chinku/*kangto teacher/student/child/friend/robber yPl pun ten CL.person(HON)

Grammaticalization of Korean numeral classifiers 67

c. d.

*sPnsaengnim/?haksaeng/?Plini/chinku/kangto yPl nom teacher/student/child/friend/robber ten CL.guy ? sPnsaengnim/haksaeng/*Plini/ chinku/kangto teacher/student/child/friend/robber yPl salam ten CL.person ‘ten teachers/students/children/friends/robbers’

Because of the desemanticization of myPng, the numeral classifier phrase used with myPng has little or no referring function, so that myPng generally needs a head noun. In other words, without a head noun, myPng cannot easily be used as an anaphoric device. For other classifiers such as salam, pun, and nom, this is not the case.11 (13)

a.

b.

c.

d.

(sPnsaengnim/chinku/kyPngchal) yPl myPng-i (teacher/friend/policeman) ten CL.name-NOM o-ass-ta. come-PAST-DECL ‘Ten teachers/friends/policemen came.’ (sPnsaengnim) yPl pun-i (teacher) ten CL.person(HON)-NOM o-ssi-Pss-ta. come-HON-PAST-DECL ‘Ten (teachers) came.’ o-ass-ta. (chinku) yPl nom-i (friend) ten CL.guy-NOM come-PAST-DECL ‘Ten (friends) came.’ o-ass-ta. (kyPngchal) yPl salam-i (policeman) ten CL.person-NOM come-PAST-DECL ‘Ten (policemen) came.’ ?

3.2.2. Numeral classifiers for animals The items mali, tu, phil, su, etc. are used as numeral classifiers for animals (cf. section 2). Over time the uses of tu, phil, and su became restricted to special domains (i.e. for trading of farm animals), and they can be replaced with mali (see (14)).

68 Myung-Chul Koo (14)

a. b. c. d.

yPl tu so/?mal/*talk/*oli cow/horse/chicken/duck ten CL.cattle ? so /mal/*talk/*oli yPl phil cow/horse/chicken/duck ten CL.horse *so/*mal/talk/oli yPl su cow/horse/chicken/duck ten CL.poultry so/mal/talk/oli yPl mali cow/horse/chicken/duck ten CL.animal ‘ten cows/horses/chickens/ducks’

Furthermore, mali can be used to refer to all sorts of insects for which there are no other classifiers: (15)

a. b. c. d.

*napi/*camcali/*phali/*moki yPl tu butterfly/dragonfly/fly/mosquito ten CL.cattle *napi/*camcali/*phali/*moki yPl phil butterfly/dragonfly/fly/mosquito ten CL.horse *napi/*camcali/*phali/*moki yPl su butterfly/dragonfly/fly/mosquito ten CL.poultry napi/camcali/phali/moki yPl mali butterfly/dragonfly/fly/mosquito ten CL.animal ‘ten butterflies/dragonflies/flies/mosquitoes’

The fact that mali can be combined with all sorts of animals seems to be related to its desemanticization. Very few Korean speakers know the original meaning of mali, whereas the meanings of tu, su, for instance, are wellknown. “The Great Dictionary of Standard Korean” defines its original meaning as ‘head’. Also “The Dictionary of 17th Century Korean” provides some examples of usage: mali k¨lk-ta (head scratch-DECL) ‘to scratch the head’, mali pis-ta (head comb-DECL) ‘to comb the hair’. 12 Thus, it is apparent that mali is grammaticalized from a common noun with the meaning of ‘head’.13 Moreover, the evolution of mali into a numeral classifier is accomplished, to the effect that mali cannot be used to refer to ‘head’ anymore. In order to refer to ‘head’, a phonological variant with an ablaut (mPri) has evolved.

Grammaticalization of Korean numeral classifiers 69

3.2.3. Numeral classifiers for inanimate entities The classifier kae can, in principle, replace almost all classifiers for inanimate entities regardless of their shape:14 (16)

a.

b.

c.

d.

tampae se kaepi cigarette three CL.split.wood ‘three cigarettes’ → tampae se kae cigarette three CL.piece polphen se calu ball.pen three CL.grip ‘three ball pens’ → polphen se kae ball.pen three CL.piece cang sinyongkhat¨/siti se credit.card/CD three CL.sheet ‘three credit cards/CDs’ se kae → sinyongkhat¨/siti credit.card/CD three CL.piece yak se al pill three CL.egg ‘three pills’ → yak se kae pill three CL.piece

Replacing with kae is not possible, however, in the case of classifiers that are functionally motivated: (17)

a.

b.

c.

chaek se kwPn → book three CL.volume ‘three books’ cip se chae → house three CL.building ‘three houses’ catongcha se tae → car three CL.unit ‘three cars’

*chaek se kae book three CL.piece *cip house

se kae three CL.piece

*catongcha car

se kae three CL.piece

70 Myung-Chul Koo d.

→ *pae se kae pae se chPk ship three CL.vessel ship three CL.piece ‘three ships’

However, if the objects are not genuine members of the class, then the functionally motivated classifiers cannot be used (cf. Chae 1996, Woo 2001: 92). For examples, toy houses, toy cars, and toy ships are not paired with chae, tae, or chPk but with kae: (18)

a.

b.

c.

*cangnankam cip se chae toy house three CL.building ‘three toy houses’ → cangnankam cip se kae toy house three CL.piece *cangnankam catongcha se tae toy car three CL.unit ‘three toy cars’ → cangnankam catongcha se kae toy car three CL.piece *cangnankam pae se chPk toy ship three CL.vessel ‘three toy ships’ → cangnankam pae se kae toy ship three CL.piece

Furthermore, kae can also replace some classifiers for plants such as thol, phoki, ppuli, etc:15 (19)

a.

b.

c.

ssal se thol → ?ssal se kae rice three CL.grain rice three CL.piece ‘three grains of rice’ paechu se phoki → paechu se kae cabbage three CL.root cabbage three CL.piece ‘three cabbages’ mu se ppuli → mu se kae16 radish three CL.root radish three CL.piece ‘three radishes’

Grammaticalization of Korean numeral classifiers 71

This means that although kae was once used only for three-dimensional objects, its use is expanded to almost all objects regardless of their shape (16), even plants (19). The following examples illustrate this expansion in present-day Korean:17 (20)

a.

b.

yPnphil se pencil three ‘three pencils’ ? polphen se ball.pen three ‘three ball pens’

calu → ?/*yPnphil CL.grip pencil calu → polphen CL.grip ball.pen

se kae three CL.piece se kae three CL.piece

Both yPnphil and polphen denote writing instruments, but they differ with respect to their compatibility with the classifiers calu and kae. This difference relates to the fact that polphen is a word of foreign origin. The noun shaph¨ ‘sharp pencil’, which is also a loan word but borrowed in a later time than polphen, also takes kae (and not calu) as its classifier. (21)

*shaph¨ se calu → shaph¨ se kae sharp.pencil three CL.grip sharp.pencil three CL.piece ‘three sharp pencils’ ?/

This means that kae has been supplanting calu since the inflow of ball pens and sharp pencils into Korea. The decrease of semantic restrictions of kae has made this replacement of calu with kae possible. Further evidence for the desemanticization of kae comes from the fact that in the register of young Korean speakers today, yPnphil se kae is equally acceptable as yPnphil se calu (20a).18 This suggests that the grammaticalization of kae has now been nearly accomplished.

4. Conclusion This paper began with the assumption that Korean numeral classifiers are grammaticalized to different degree. Most notably, it was observed that numeral classifiers such as salam and mali emerged out of common nouns. Since they have lost their lexical meaning and now fit into the paradigm of numeral classifiers, they are considered grammaticalized.

72 Myung-Chul Koo Furthermore, distinctions were drawn between four groups of numeral classifiers according to the animacy: numeral classifiers for human beings, animals, plants, and inanimate entities. Each group contains a prototypical classifier that can be used to replace the others: myPng for human beings, mali for animals, and kae for inanimate entities. kae, which is the prototypical classifier for inanimate entities, can even replace the classifiers for plants. Such a replacement is possible because these classifiers have lost all or many of their semantic components and do not apply any selectional restrictions to the head noun. On the basis of this evidence, it may be concluded that the selection of an appropriate numeral classifier is more grammatically motivated than in the previous diachronic stages of the language and that the classifier itself is more grammaticalized.

Notes 1. 2. 3.

4. 5. 6. 7. 8. 9.

Cf. Craig (1994: 566): “numeral classifiers may come in large numbers (dozens to hundreds), […].” It is a well-known fact that numeral classifiers have often developed from nouns (Lehmann 1995: 60, 2000: 250, 2005b: 6, Downing 1996: 3). In Korean, two types of numeral classifier constructions can be distinguished according to the order of the numeral classifier and the head noun: the primary numeral classifier construction and the secondary numeral classifier construction. In the primary numeral classifier construction, the head noun is followed by a numeral classifier as in (1); in the secondary numeral classifier construction, the numeral classifier precedes the head noun, as in tasPs myPng-¨i kyPngchal (five CL.name-GEN policeman) ‘five policemen’ (cf. Sohn 1999: 352f.). According to Lee (2000), the primary construction is more common than the secondary. (2) is a slightly modified version of the examples in Yang (1996: 198). See Adams (1986: 246): “Inanimate objects are typically counted according to shape or function.” To count small (musical) instruments and apparatuses such as triangles, flutes, wristwatches, and electronic shavers, kae is used, not tae (Woo 2001: 273). Woo (2001: 76) regards salam as a common noun that has a classifier function. Like other classifiers such as myPng, nom, etc. (see section 3.1), salam also does not permit plural marking. Following Yang (1995: 215), salam “is used with a specific set of nouns and can often occur without the head noun present”. Additionally, in 10 of 11 cases, salam occurred without a head noun.

Grammaticalization of Korean numeral classifiers 73 10. Because the use of pun and nom is semantically and pragmatically restricted, they can be combined only with semantically compatible nouns. As a result, the less restricted salam can be combined with more nouns than pun and nom. 11. See Yang (1995: 215): “The speaker does not need to mention the head noun if the classifier constructions are used as an anaphoric device.” See also Yang (1995: 215): “[...] the more general a classifier is, the more likely it is to need a head noun.” 12. Modern Korean still exhibits traces of the middle Korean mali. The compound silmali ‘clue’ has the following morphological structure: sil ‘thread’ + mali ‘head’. The phrase means that a clue is like the head, or beginning, of a thread. 13. On the development of mali, see Woo (2001: 113, 2005: 69). 14. In the case of congi ‘paper’, kae cannot be used as a classifier: congi/poksaci han cang (paper/copying paper one CL.sheet) ‘a sheet of paper/copying paper’ → *congi/*poksaci han kae (paper/copying paper one CL.piece). On the other hand, it is very interesting that envelopes that are made of paper can match with kae: pongthu han cang (envelope one CL.sheet) ‘an envelope’ → pongthu han kae (envelope one CL.piece). 15. Note that songi and k¨lu cannot be easily replaced with kae: cangmikkoch han songi (rose one CL.blossom) ‘a rose’ → ?/*cangmikkoch han kae (rose one CL.piece) ‘a rose’; namu han k¨lu (tree one CL.stump) ‘a tree’ → *namu han kae (tree one CL.piece) ‘a tree’. 16. insam, which takes also ppuli as a numeral classifier, cannot be combined with kae: insam han ppuli (ginseng one CL.root) ‘a root of ginseng’ → *insam han kae (ginseng one CL.piece) ‘a piece of ginseng’. 17. The combination of yPnphil and kae is not found in the ‘Sejong corpus’, while two tokens with polphen and kae are found. On the contrary, no combination of polphen and calu is found. 18. The item kae can even be used as a numeral classifier for some abstract nouns such as pPmcu ‘a category’, chauPn ‘a dimension’, somang ‘a hope’, if they are to be counted.

References Adams, Karen 1986 Numeral classifier in Austroasiatic. In Noun Classes and Categorization, Colette Craig (ed.), 241–262. Amsterdam/Philadelphia: John Benjamins Publishing Company. Allan, Keith 1977 Classifiers. Language 53: 285–311.

74 Myung-Chul Koo Chae, Wan 1996

The borrowing and meaning of the numeral classifier ge. Chin-Tan Hakpo (Journal of Chin-Tan Society) 82: 193–215. Craig, Colette A. 1994 Classifier languages. In The Encyclopedia of Language and Linguistics, vol. 2, R.E. Asher, and J.M.Y. Simpson (eds.), 565–569. Edinburgh: Pergamon Press. Downing, Pamela 1996 Numeral Classifier Systems. The Case of Japanese. Amsterdam/Philadelphia: Benjamins. Hong, Yunpo et al. (eds.) 1995 The Dictionary of 17th Century Korean [in Korean]. Seoul: Thaehaksa. Lee, Chungmin 2000 Numeral classifiers, (in-)definites and incremental theme in Korean. In Korean Syntax and Semantics: LSA Institute Workshop, Santa Cruz, '91, C. Lee and J. Whitman (eds.). Thaehaksa: Seoul. Lee, Heechul 1998 A study on some Korean classifiers with respect to cognitive semantics. PnPhak (Journal of the Linguistic Association of Korea) 6 (2): 83–101. Lehmann, Christian 1995 Thoughts on Grammaticalization. Unterschleissheim: Lincom (LINCOM Studies in Theoretical Linguistics, 1). 2000 On the German numeral classifier system. Naturally! Linguistic studies in honour of Wolfgang Ulrich Dressler presented on the occasion of his 60th birthday, Chris Schaner-Wolles, John R. Rennison, Friedrich Neubarth (eds.), 249–253. Torino: Rosenberg and Sellier. 2005a Theory and method in grammaticalization. Zeitschrift für Germanistische Linguistik 32 (2): 152–187. 2005b Wortarten und Grammatikalisierung. Wortarten und Grammatikalisierung. Perspektiven in System und Erwerb, Clemens Knobloch and Burkhard Schaeder (eds.), 1–20. Berlin/New York: De Gruyter. Nam, Ki-Shim and Young-Gun Ko 1985 Standard Korean Grammar [in Korean]. Seoul: Tower. Sohn, Ho-Min 1999 The Korean Language. Cambridge: Cambridge University Press. The National Institute of the Korean Language 1999 The Great Dictionary of Standard Korean [in Korean]. Seoul: Dusandonga.

Grammaticalization of Korean numeral classifiers 75 Woo, Hyungshik 2001 A Study of the Categorizing Functions of Korean Classifiers [in Korean]. Seoul: PakicPng. Woo, Hyungshik et al. 2005 A Contrastive Study of the Noun Classifying Functions of Korean and Japanese Numeral Classifiers [in Korean]. Seoul: J & C. Yang, Byong-seon 1995 Korean numeral classifier and topic continuity in written discourse. PnPhak (Journal of the Linguistic Association of Korea) 3: 207–224. 1996 What categorizes numeral classifier construction: Classifier or head noun? Three types of Korean numeral classifiers. Humanities Research (CPncu University) 1: 191–211.

Grammaticalization of spatial adpositions in Nànáfwê Amani Bohoussou and Stavros Skopeteas

1. Introduction Typical sources for adpositions across languages are relational nouns and transitive verbs, hence elements that already display an argument slot (see Lehmann 1995b:104). The grammaticalization process is similar in both cases. In case of relational nouns, grammaticalization applies to the dependency between the noun and its head in the clause. While a relational noun may serve as an argument, its adpositional counterpart displays the distributional properties of an adverb (with the difference that it is accompanied by a complement NP). In case of transitive verbs, grammaticalization takes place in serial verb constructions: what was once a part of a verb series, is the dependent of a verbal head after grammaticalization. It is well known in West African linguistics that languages in this broad area display adpositions that emerge out of these two sources, namely nouns and verbs. The corresponding adpositional elements undergo distinct grammaticalization paths and form sub-paradigms with characteristically distinct semantic and syntactic properties (see Heine, Claudi, and Hünnemeyer 1991: 140–141): (a) V-adpositions form-small size classes, while Nadpositions form large-size classes; (b) V-adpositions typically denote different spatial relations (i.e., static, allative, ablative, perlative, etc.) between the localized object and the reference object of a locative construction, while N-adpositions typically denote different spatial regions of the reference object (e.g., interior, exterior, superior, inferior, etc.); (c) Vadpositions introduce adjuncts, while N-adpositions may introduce either adjuncts or complements. Furthermore, V-adpositions generally follow the ordering rules within VPs, while N-adpositions follow the ordering rules within complex NPs. Hence, in languages like Ewe (Ghana, Togo: Kwa) or Koromfe (Burkina Faso, Mali: Gur), where VPs are head-initial and NPs head-final, two subclasses of adpositions may be distinguished: prepositions, which emerge out of verbs, and postpositions, which emerge out of

78

Amani Bohoussou and Stavros Skopeteas

nouns (see Heine and Reh 1984: 253–258, Heine, Claudi, and Hünnemeyer 1991: 142ff. for Ewe, see Rennison 1997: 169ff. for Koromfe). In this paper, we present evidence for the emergence of spatial adpositions in Nànáfwê, which is a dialect of Baule, a language of the Kwa branch of the Niger-Congo family. It is spoken in the central part of Côte d’Ivoire in the district of Yamoussoukro by a population of approximately 15 000 speakers. The grammatical properties of Baule have been the subject of several publications including descriptive grammars (Carteron 1966, Creissels and Kouadio 1977, Timyan 1977) and numerous publications on phonology (Gross 1967, Creissels and Kouadio 1979, Ahoua 1996, Leben and Ahoua 1997 among others) and syntax (Larson 2002, 2005, Kouadio 2000, Kouadio and Creissels 2007 among others) (see a bibliographical outline in Bohoussou 2008). The particular variety of Baule that we examine in this paper is virtually unexplored. A few recent works on this dialect shed light on the syntax of the simple clause (Bohoussou 1996), on clause linkage (Bohoussou 2008), and on copulative constructions (Bohoussou and Skopeteas 2005). An outline of the basic grammatical properties of Nánàfwê is given in section 2. The data from Nànáfwê adpositions is very similar to the data reported from other West African languages. It is clear that a particular class of nouns and a small set of verbs occur as heads of NPs forming a constituent that may be used as verb dependent, i.e. in a syntactic function in which we expect to find an adpositional phrase. Similarly to Ewe and Koromfe, adposition-like elements that originate in nouns form head-final constituents as exemplified in (1a), while adposition-like elements that originate in verbs form head-initial constituents as exemplified in (1b). We refer to the two classes of elements as N-relators and V-relators without anticipating their grammaticalization status which is the empirical question of this paper. The term ‘relator’ is used with functional content, indicating the exponent of a component of a localizing function (see Lehmann 1992). (1)

a. N-relator nànnán ó [swā-n sîn]. grandfather be.located house-DEF back.side1 ‘The grandfather is behind the house.’ b. V-relator nànnán wàndí [kò‰ dímbókrô]. grandfather run go Dimbokro ‘The grandfather runs to Dimbokro.’

Grammaticalization of spatial adpositions in Nànáfwê

79

The data presented in (1) is not enough to conclude that the relational elements are grammaticalized. A grammaticalization process is not implied by the occurrence in a particular syntactic function but has to be diagnosticized through a set of heuristics (outlined in Lehmann 1995b, see introduction to this volume) that reflect a change in the distributional properties of the elements at issue. With this conceptual background, the aim of the current contribution is descriptive: we examine two classes of elements in Nànáfwê that are usually involved in grammaticalization processes across languages, and we address the question whether the available evidence from this language suggests such a diachronic process. Sections 3 and 4 present the main body of the empirical evidence providing a detailed account of N-relators and V-relators, respectively. Our account primarily relates to the syntactic phenomena applying to locative constructions; we will refer to the abstract uses of adpositions only when these are relevant for the syntax. Section 5 summarizes the differences between N-relators and V-relators and section 6 draws the conclusions of this empirical study with respect to grammaticalization.

2. Basic grammatical properties of Nànáfwê Most syntactic structures in Nànáfwê are head-initial. Verbs precede complements and adjuncts and are preceded by the subject constituent rendering a rigid SVOAdv order, which is exemplified in (2a) for temporal adverbs and in (2b) for manner adverbs that are VP-internal and display different word order properties in some languages (see, e.g., manner adverbs in Edo, Stewart 2001: 20). Temporal or local adjuncts may be leftdislocated in which case they serve as frame setters, see (2c). (2)

a. nànnán tò‰ntòÊÊn-ní dwó-n nnè‰. grandfather cook-PFV yam-DEF today ‘The grandfather cooked the yam today.’ b. nànnán kàn waÌwlé-n ndèÌdè‰. grandfather speak baule-DEF rapidly ‘The grandfather speaks Baule rapidly.’ c. nnè‰ nànnán tò‰ntòÊÊn-ní dwó-n. today grandfather cook-PFV yam-DEF ‘Today, the grandfather cooked the yam.’

80

Amani Bohoussou and Stavros Skopeteas

Nouns precede adjectival modifiers and quantifiers as illustrated in (3a), while the definite determiner is a suffix, as shown in (3b) and (3c). In contrast, pronominal and nominal dependents precede the head NP, as demonstrated in (3b) and (3c), respectively. (3)

a.

b.

c.

tràlè‰ blê nñòÊÊn shirt blue/black/green four ‘four blue/black/green shirts’ mí tràlè‰-n 1.SG shirt-DEF ‘my shirt’ kòf í tràlè‰-n Kofi shirt-DEF ‘the shirt of Kofi’

The word order facts are summarized in (4). The facts that verbs precede their complements and adjuncts, and that nouns precede their modifiers lead to the generalization that heads precede their dependents in (4a). An apparent exception to this generalization is that possessor NPs precede possessed NPs in (4b). Finally, subjects precede predicates. (4)

a. head b. NPdependent c. subject

p dependent p NPhead p predicate

Syntactic relations generally do not have morphological exponents: Nouns do not inflect for case and verbs do not bear agreement affixes, as may be observed in (2). Argument functions are unambiguously encoded by linear order. Subject constituents have to be overtly realized (see Creissels and Kouadio 2007: 3), while object constituents have to be dropped under particular circumstances (object drop occurs with a lexically conditioned subset of verbs; with these verbs, inanimate singular referents in clause final position that may be contextually retrieved cannot be overtly realized, see Larson 2002: 90). Verbal inflection in Nànáfwê comprises a number of affixes that encode aspectual/temporal/modal distinctions. Some of them are suffixes as illustrated in (5a) by means of the perfective, while others are independent morphemes that are anteposed to the lexical verb, as shown in (5b) by means of

Grammaticalization of spatial adpositions in Nànáfwê

81

the progressive. The latter elements are diachronically derived from serial verb constructions (see Larson 2002). (5)

a.

b.

kòfí nántì-lí. Kofi walk-PFV ‘Kofi walked.’ kòfí sú nàntí. Kofi PROG walk ‘Kofi is walking.’

Nouns are not inflected. The only morpheme that occurs with nouns is an enclitic that conflates definiteness and number (singular vs. plural) that is attached at the right edge of the NP constituent (see Bohoussou 2008: 21). (6)

wákā-n (tree-DEF) ‘the tree’ wákā-mú (tree-DEF.PL) ‘the trees’ wákā blê-n (tree blue/black/green-DEF) ‘the blue/black/green tree’

Personal pronouns display a contrast between independent and cliticized forms which are segmentally different in the singular2: mí ‘1.SG’ vs. n ‘1.SG.SBJ.CL’, wò‰ ‘2.SG’ vs. a ‘2.SG.SBJ.CL’, and í ‘3.SG’ vs. ò ‘3.SG.SBJ.CL’ (see Bohoussou 2008: 23). Cliticized forms do not bear lexical tone: their tonal properties are determined by the tonal structure of the verb (see Creissels and Kouadio 2007: 3). Independent pronouns occur in object function or as possessors of NPs, as illustrated in (7a-b). (7)

a.

b.

kòfí wùn-ní í Kofi see-PFV 3.SG ‘Kofi saw him today.’ í swā-n 3.SG house-DEF ‘his house’

nnè‰. today

When the possessor is realized in a lexical NP, the possessive pronoun may occur also. The pronominal expression is optional in singular (see (8a-b))3. (8)

a.

nànnán í swā-n grandfather 3.SG house-DEF ‘the house of the grandfather’

82

Amani Bohoussou and Stavros Skopeteas

b.

nànnán swā-n grandfather house-DEF

Cliticized pronouns occur in subject function, when the subject constituent is contextually retrievable and non-emphatic. The reference of the 3rd person clitic pronoun in (9a) may not identify a referent in the discourse situation (as a demonstrative), but is co-indexed with a contextual antecedent. The subject clitics are obligatory, when the lexical subject is left dislocated rendering a clitic doubling construction, as illustrated in (9b). (9)

a.

b.

òÊ wùn-ní kòfí 3.SG.SBJ.CL see-PFV Kofi ‘He saw Kofi.’ nànnán òÊ wùn-ní grandfather 3.SG.SBJ.CL see-PFV ‘The grandfatheri, hei saw Kofi.’

kòfí Kofi

In sum, the major grammatical properties of Nànáfwê, that are presented in this section, are: (a) word order is generally head-initial; (b) possessor NPs precede possessed NPs and may be optionally cross-referenced by a co-indexed pronoun; (c) inflection is poor and does not involve encoding of syntactic relations; (d) pronouns display a contrast between independent and clitic forms, whereby the distribution of the latter is restricted to the subject function. With this grammatical background, we discuss evidence concerning locative relational morphemes in the next sections.

3. N-relators 3.1. Preliminaries The diachronic change from nouns to adpositions implies an initial grammatical stage which involves two noun phrases, NP1 and NP2 as illustrated in (10), stage S1. NP1 is the head of this constituent, NP2 is an optional dependent of NP1, and both form a complex NP, whereby the order of constituents in (10) does not relate to the realized word order (both NP1pNP2 and NP2pNP1 are possible and depend on language-specific word order rules). The target grammatical stage S2 involves two changes. First, the entire constituent is an AdpP and not an NP anymore; this change implies

Grammaticalization of spatial adpositions in Nànáfwê

83

some differences in the distribution of this constituent in syntactic constructions. Second, NP1 – or the N contained within NP1 – has turned to an adposition, i.e. NP2 is not an optional dependent anymore, but is the complement of the adposition. (10) a. b. c.

stage S1: stage S2: S 1 > S2

[NP NP1 NP2] [AdpP Adp NP]

It is important to bear in mind that the sketched grammaticalization process involves two syntactic changes, which may well be independent from one another. The first change relates to the distributional properties of the entire constituent, i.e., to the syntactic relation between the head of this constituent and its head in the clause, while the second relates to the syntactic relation between head and dependent within the constituent at issue. Evidence for this development in Nànáfwê is found in a (semantically defined) set of nouns that encode parts of spatial configurations (henceforth, locative nouns). In line with the ‘NPdependent p NPhead’ ordering principle, the locative noun follows the noun that encodes the reference object. As already observed for possessive constructions in (8), the reference object is optionally cross-referenced by a co-indexed element of the class of independent pronouns. (11) a.

b.

ánúmān-n ó swā bird-DEF be.located house ‘The bird is inside the house.’ ánúmān-n ó swā bird-DEF be.located house ‘The bird is above the house.’

(í) nú. 3.SG inner.side (í) sú. 3.SG topside

As already implied by the contrast between (11a) and (11b), there is a set of locative nouns that may be used in this construction. The exponents of the basic oppositions are given in Table 1.

84

Amani Bohoussou and Stavros Skopeteas

Table 1. N-relators in Nànáfwê spatial region interior4 superior inferior anterior posterior lateral dextral sinistral

noun nú sú bô Ôrún sîn nwán fámānú bèÌnú

There is ample evidence from grammaticalization studies that locative nouns encoding parts of spatial configurations such as those exemplified in Table 1 develop into adpositions in several languages (see among others Heine 1989: 88, Lehmann 1990: 172, Heine, Claudi, and Hünnemeyer 1991: 129, Svorou 1993: 70, Rubba 1994: 86, Lehmann 1995b: 77). A frequently reported semantic change is concomitant to the syntactic changes in (10). While the nouns denote parts of spatial configurations, typically body parts, the derived adpositional elements denote spatial regions, i.e., fragments of space that are determined either by their contiguity to particular parts of the reference object (for instance the spatial region denoted by the English preposition on is contiguous to the top part of the reference object) or by axes projected by these parts (for instance the spatial region denoted by the English preposition above contains possible locations on the axis projected to the top part of the reference object). In the following sections, we discuss the categorical status of the locative nouns in Nànáfwê. We first discuss the syntactic properties of the entire constituent in section 3.2. Section 3.3 outlines the syntactic relation between the parts of this constituent.

3.2. Syntactic properties of the locative constituent The evidence that the elements listed in Table 1 are nouns is syntactic. All these elements also occur in argument functions as illustrated by means of the noun nú ‘inside’ in (12a). From a semantic viewpoint, example (12a) shows that the exponents of spatial region are referential in Nànáfwê, i.e. they may be used to denote a particular part of the spatial configuration of

Grammaticalization of spatial adpositions in Nànáfwê

85

the dependent NP’s referent. Furthermore, as a noun it may be modified by adjectives as illustrated in (12b). (12) a.

b.

swā-n sú tí blê . house-DEF topside be blue/black/green ‘The topside of the house is blue/black/green.’ swā-n sú klánmān tí blê . house-DEF topside beautiful-DEF be blue/black/green ‘The beautiful topside of the house is blue/black/green.’

The categorical status of these nouns when they occur in the locative construction in (11) is at issue. The thematic properties of these elements, i.e., their role as locative constituents, are not encoded by the noun itself or its position, but by the verb. Note that the verb ó ‘be.located’ in (11) is a locative verb, while property assignment is either expressed through a zero copula or through the linking element tí ‘be’. Hence, the thematic role of location is carried by the verbal valency and does not imply a change in the categorical status of these elements in the first sight. Moreover, we observe that further nouns are eligible arguments for the locative copula ó ‘be.located’, as shown in (13a). The same phenomenon is illustrated in (13b) by means of a verb of motion and common nouns. (13) a. ánúmān-n ó dímbókrô. bird-DEF be.located Dimbokro ‘The bird is in Dimbokro.’ b. n‰ kò‰ sùkluÌ-n/ klòÌ-n. 1.SG.SBJ.CL go school-DEF village-DEF ‘I will go to the school/the village.’ Further evidence that the thematic properties of these elements are determined by the verb is the fact that they are not specified for spatial relation. Hence, the distinction between static and different subtypes of dynamic (i.e., allative, ablative, perlative) spatial relations is specified completely through the verbal valency. Example (14a) illustrates the thematic underspecification by means of a motion verb and an N-relator, (14b) by means of a motion verb and a proper noun, and (14c) by means of a verb of transport and an N-relator. This phenomenon is very widespread in languages that denote spatial regions through locative nouns, and by no means re-

86

Amani Bohoussou and Stavros Skopeteas

stricted to this language type (see facts from Koromfe in Rennison 1997: 173, and from Yucatec Maya in Lehmann 1992: 636, 1995a). (14) a.

b.

c.

kòfí kò‰/ fìn/ sìn swā-n nú. Kofi go leave pass house-DEF inner.side (literally) ‘Kofi goes into the house/leaves from inside the house/passes through the house.’ kòfí kò‰/ fìn/ sìn dímbókrô nú. Kofi go leave pass Dimbokro inner.side (literally) ‘Kofi goes into Dimbokro/leaves from inside Dimbokro /passes through Dimbokro.’ Ôìn bútèlí-n bjá-n sú. kòfí Kofi put bottle-DEF chair-DEF topside ‘Kofi puts the bottle on the chair.’

The occurrence of nouns in this construction and their thematic underspecification suggest that locative verbs do not put categorical restrictions on their complements. However, this is the wrong conclusion: first, not every noun is eligible in this construction but only nouns that denote places. Second, locative verbs differ from transitive verbs in that they may take an adverbial complement. The deictic adverbs ‘here’ and ‘there’ may fill the argument slot of a locative verb as exemplified in (15), but not the argument slot of a transitive verb. (15)

kòfí kò‰/ fìn/ sìn l‰. Kofi go leave pass there (literally) ‘Kofi goes there/leaves from there/passes through there.’

Assuming that locative verbs license an adverbial complement with the thematic properties of place (unspecified for the exact spatial relation), the restriction to nouns denoting places is expected. Only those nouns that may be reinterpreted as adverbial expressions of place may occur in this construction. Locative nouns may occur as complements of locative verbs, but they display a distributional peculiarity: In contrast to their referential use, locative nouns may not be accompanied by adjectival modifiers in their use as adpositions. Hence, the ungrammaticality of (16) contrasts to the grammaticality of (12b). Note that the availability of a pronominal possessor does not interact with the grammaticality of this structure.

Grammaticalization of spatial adpositions in Nànáfwê

(16) a.

b.

87

*kòfí ó swā-n (í) sú klánmān. Kofi be.located house-DEF 3.SG topside beautiful (intended) ‘Kofi is at the beautiful topside of the house.’ *ánúmān-n ó swā (í) nú blê. bird-DEF be.located house 3.SG blue/black/green (intended) ‘The bird is at the blue/black/green inner side of the house.’

In concluding, distributional criteria suggest that locative nouns in their use as complements of locative verbs are in the incipient stage of a grammaticalization process. The critical restriction is their capability to occur with adjectival modifiers, which is a possible syntactic configuration outside the locative constructions (compare similar evidence concerning postpositions in Ewe in Heine and Reh 1984: 257).

3.3. Head-dependent relation This section examines the syntactic relation between the head and the governed noun in the locative construction. Svorou (1986) points out that a critical point in the development of adpositions out of nouns is the loss of relational morphology (e.g., genitive affixes). The relevant issue in Nànáfwê is the behavior of the pronominal markers in the context of nounto-noun dependencies. We have already mentioned that pronominal possessors are optional with most nouns (see (8c-d)), and so do they with locative nouns too (see (11a) and (11b)). The question is whether this alternation is functionally vacuous (hence, free variation) or reflects a semantic opposition. In a compositional view, the occurrence of a possessive pronoun is expected to indicate a possessive relation between the locative noun and the dependent noun. The construction without a locative pronoun is expected to be underspecified in this respect. Keeping in mind the compositional predictions, we observe some critical examples in the following. The sentences in (17a) and (17b) are only partly synonymous: The version in (17a), without a co-indexed pronominal, denotes that the localized object is located in the spatial region that is projected by the backside of the reference object, i.e. behind it. However, the version in (17b), with a pronominal possessor, has an additional reading: it may denote that the localized object is located within the part of the spatial configuration of the reference

88

Amani Bohoussou and Stavros Skopeteas

object that is referred to as ‘back part’. The latter meaning is exactly the expected compositional reading of the construction with a pronominal possessor. The assertion of the possessive relation through the pronoun licenses a reading in which a meronomic (part-whole) relation holds between the two nouns. Note that the unspecified version is (17b), i.e., the version with the pronominal possessor. (17) a.

b.

kòfí ó swā-n sîn. Kofi be.located house-DEF back.side ‘Kofi is behind the house.’ *‘Kofi is at the back part of the house.’ kòfí ó swā-n í sîn. Kofi be.located house-DEF 3.SG back.side ‘Kofi is behind the house.’/ ‘Kofi is at the back part of the house.’

Similar minimal pairs occur with further adpositions as well. The noun sú ‘on/above’ denotes superposition in the vertical axis. In the version (18a) without pronominal possessor, the localized object is located in this axis; due to inferences that relate to the posture verb Ôín ‘stand’ and world knowledge, this example is interpreted as involving contact to the upper side of the reference object, but this inference is defeasible (see Skopeteas 2007). The version (18b) with pronominal possession is semantically different: the possessive morphology licenses the interpretation of a partwhole relation between the denoted region ‘on/above’ and the reference object. This is not necessarily the same location, depending on the spatial configuration of the chair. (18) a.

b.

Ôín bjá-n sú. bútèlí-n bottle-DEF stand chair-DEF topside ‘The bottle is standing on the chair.’ *‘The bottle is standing on the top of the chair.’ Ôín bjá-n í sú. bútèlí-n bottle-DEF stand chair-DEF 3.SG topside ‘The bottle is standing on the chair.’/ ‘The bottle is standing on the top of the chair.’

These examples show that the alternation between the versions of the adpositional phrase with and without a pronominal possessor is not semanti-

Grammaticalization of spatial adpositions in Nànáfwê

89

cally vacuous. Recall that the compositional view predicts that the uses of the version with a pronominal possessor are a subset of the uses of the version without a pronominal possessor, since the latter version is underspecified. However, our data suggest the opposite pattern (we come back below to this apparent mismatch). The semantic contrast evinced in (17) and (18) is well attested in languages that develop adpositions out of nouns encoding body parts. For instance, similar facts are reported for Mixtec (see Macauley 1993: 172ff.). In order to evaluate the relevance of these contrasts concerning grammaticalization, we need comparative evidence from the referential use of the same elements. A similar semantic effect occurs when the locative nouns are used as arguments. (19a) is the version without a pronominal possessor which is now the unspecified version. (19b) illustrates the version with a pronominal possessor: the only reading is one in which the predicate holds for an inherent part of the spatial configuration of the referent encoded by the noun. (19) a.

b.

táblí-n bô tí blê. table-DEF bottom.side be blue/black/green ‘The bottom of the table is blue/black/green.’/ ‘The place under the table is blue/black/green.’ táblí-n í bô tí blê. table-DEF 3.SG bottom.side be blue/black/green ‘The bottom of the table is blue/black/green.’ *‘The place under the table is blue/black/green.’

These examples provide evidence for a semantic difference between the nominal and the adpositional use of the locative nouns. In the nominal use, the version with a pronominal possession denotes a meronomic relation between the referent of the head noun and the referent of the dependent noun as exemplified in (19b), while the construction without a pronominal possessor is unspecified (see (19a)), exactly as predicted by the compositional account. In the adpositional use, exemplified in (17) and (18), the occurrence of pronominal possessors is desemanticized, i.e., it looses its capability to denote a meronomic relation. The version without pronominal possession is in the adpositional use restricted to the abstract meaning, i.e. the meaning that we except to find for adpositional elements.

90

Amani Bohoussou and Stavros Skopeteas

So far we examined compositional uses of locative nouns and we found a systematic alternation between a version with possessive morphology and a version without. Moreover, we were able to identify minimal pairs in which these two versions are not synonymous and we claimed that the version with possessive morphology displays evidence for desemanticization in the adpositional uses of locative nouns. Next to the compositional uses of the locative nouns in adpositional function, these elements display a large number of uses that arise through metaphorical extension. When a spatial noun relates to a temporal or further abstract concept, the partwhole interpretation encoded through the pronominal possessor cannot apply literally. The noun sîn ‘back.side’ denotes the spatial region of ‘posterior’. However, it also occurs as a comitative, which is illustrated by (20a). This abstract meaning of the noun is only available for the version without possessive morphology. Evidence for this restriction is given in (20b): the version with possessive morphology only allows the concrete spatial interpretation of the noun sîn ‘back.side’. (20) a.

b.

kwàsí tò kòfí sîn. Kouassi play Kofi back.side ‘Kouassi plays (together) with Kofi.’ kwàsí tò kòfí í sîn. Kouassi play Kofi 3.SG back.side ‘Kouassi plays behind Kofi.’ *‘Kouassi plays (together) with Kofi.’

The noun bô ‘under’ denotes the spatial concept of inferior. However, in the context of motion verbs it may be used to introduce complex targets of motion. This use is exemplified in (21). The use of the possessive pronoun in (21) would denote the literal region of the bottom part of the dance, which would not make sense in this context. (21) a.

sùklú mmá-mú bé hòÊ-lí àblé-n school kid-DEF.PL 3.PL.SBJ go-PERF dance-DEF (*i) bô. 3.SG bottom.side ‘The school kids (=pupils) go to the dance.’

Grammaticalization of spatial adpositions in Nànáfwê

91

The choice of adposition is idiomaticized for some nouns denoting typical places. Hence, the noun fjeÌ ‘field’ typically combines with the adposition sú ‘topside’ and the noun gwá ‘public.place’ with the adposition bô ‘bottom.side’. The choice of adposition is not transparent in these cases. Crucially, the idiomatic meaning is only available for the versions without a pronominal possessor, as illustrated in (22a) and (23a). (22) a.

b.

(23) a.

b.

n‰ kò‰ mān fjeÌ 1.SG.SBJ.CL go.IMP NEG field ‘I do not go to the field.’ n‰ kò‰ mān fjeÌ 1.SG.SBJ.CL go.IMP NEG field ‘I do not go to the top of the field.’ *‘I do not go to the field.’ n‰ kò‰ mān 1.SG.SBJ.CL go.IMP NEG ‘I do not go to the market.’ n‰ kò‰ mān 1.SG.SBJ.CL go.IMP NEG ‘I do not go below the market.’ *‘I do not go to the market.’

sú. topside í sú. 3.SG topside

gwá public.place gwá market

bô. bottom.side

í bô. 3.SG bottom.side

The data from non-compositional uses and from idiomatic uses complicate the facts. Since we found evidence for the desemanticization of the construction with pronominal possessors, we could hypothesize that this construction would be eligible in all contexts. However, the incompatibility of this construction with non-compositional and idiomatic uses shows that this is not the case. How do these generalizations fit together? We assume that the complex data pattern from the use of adpositions in Nànáfwê is an effect of grammaticalization applying to certain constructions. Pronominal possessors are excluded in the uses of adpositional phrases that do not involve a meronomic relation between the head noun and the dependent noun. Pronominal possessors are desemanticized in the adpositional use of locative nouns, i.e. in the context of a particular syntactic construction, but not in other contexts. If these properties are the result of diachronic developments, we may speculate that the development of the idiomatic and non-compositional uses took place at a diachronic stage at which pronominal possessors were not yet desemanticized.

92

Amani Bohoussou and Stavros Skopeteas

4. V-relators 4.1. Preliminaries Verbs are the second source of adpositions in West African languages (see Heine, Claudi, and Hünnemeyer 1991: 140–141). The grammaticalization path originates in a stage S1 at which two verbs are in a coordination construction, usually without a coordinative conjunction (‘asyndeton’ or ‘covert coordination’). In many of the languages of this linguistic area, serial verbs develop from this construction as illustrated in stage S2 below. This stage involves the development of a complex head containing two verbs that denotes a single event. Within the serial verb construction of the type V1 V2 NP, the second verb V2 may develop to an adposition which reflects the final stage of this grammaticalization path. (24) a. b. c. d.

NP] (coordination) stage S1: [VP1 V1 ] [VP2 V2 V2 ] NP] (serial verb) stage S2: [VP [V V1 V [AdpP Adp NP]] (verb and adposition) stage S3: [VP S 1 > S2 > S3

Assuming that the verbs that are eligible for this process have an argument slot, there is no change in the governing properties of the head of the embedded constituent: the complement of the verb in the initial stage S1 is the complement of the adposition at the final stage S2. The change affects the categorical status of the head. Evidence for the change V > Adp is the fact that the V is not (part of) the head of the predicate anymore; a concomitant change may be the loss of verbal inflection (see Lehmann 1995b: 104). The development from verb to adposition is already attested in several Benue-Kwa languages (see Lord 1993: 29). For instance, the verb gye ‘take’ in Twi is used for the concept of ‘except’ in which case it does not take tense/aspect, negation and agreement. In Ewe, some verbs developed to prepositions and lost their capability of conjugation. However, loss of inflection is not a necessary condition: the verb gyaw ‘leave’ in Twi is used with the meaning ‘without’ retaining its inflectional properties (see Lord 1993: 141–147). Aim of this section is to examine the locative verbs of Nànáfwê and to seek for evidence for the grammaticalization of V-relators. Typical V-relators that enter this grammaticalization path are verbs that encode spatial relations (static, allative, ablative, perlative). Table 2 presents the verbs that are used for the encoding of such relations in Nànáfwê.

Grammaticalization of spatial adpositions in Nànáfwê

93

Their occurrence in syntactic constructions has already been introduced in example (14). Table 2. V-relators in Nànáfwê spatial relation static allative 1 allative 2 allative 3 perlative ablative

verb o ‘be.located’ kò ‘go’ u ‘arrive’ ba ‘arrive, come’ sin ‘pass’ fin ‘leave’

These verbs occur in combination with other motion verbs as is exemplified in (25). In the following, we examine two constructions that may be instantiated through several verbs of the corresponding paradigms. The first is a construction of ‘manner + relation’, encoded both through separate verbs as illustrated in (25a). The second is a construction of ‘relation + relation’, encoded through the combination of two verbs of the list in Table 2, as exemplified in (25b). (25) a.

b.

kòfí nàntí kò‰ dímbókrô. Kofi walk go Dimbokro ‘Kofi walks to Dimbokro.’ kòfí fìn dímbókrô kò‰ ábìÔān. Kofi leave Dimbokro go Abidjan ‘Kofi goes from Dimbokro to Abidjan.’

Only the verbs of motion are involved in serial verb constructions such as those exemplified in (25). The static locative verb does not form corresponding constructions, since static relations are expressed through the absence of an exponent of spatial relation (the use of the locative verb in static adjuncts is attested in Ewe, see Heine, Claudi, and Hünnemeyer 1991). This is illustrated in (26a) by means of a static locative adjunct that modifies a non-spatial event and in (26b) by a static locative adjunct that modifies a verb of manner of motion (this example correspond to the construction of motion verbs with dative PPs in German).

94

Amani Bohoussou and Stavros Skopeteas

(26) a.

b.

nànnán tò‰ntòÊÊn-ní dwó-n (*ó) grandfather cook-PFV yam-DEF be.located swā sîn. house back.side ‘The grandfather cooked the yam behind the house.’ kòfí nàntí (*ó) dímbókrô nú. Kofi walk be.located Dimbokro inside ‘Kofi walks inside Dimbokro.’

4.2. Verb serialization The first question is whether the examples in (25) constitute serial verbs or instances of covert coordination. ‘Covert coordination’ is the concatenation of either clausal or embedded constituents without any conjunction (asyndeton). The main criterion for the distinction between covert coordination and verb serialization is argument sharing (see Stewart 2001:6-11 for an outline of the different approaches in recent literature). Serial verbs are assumed to be concatenated parts of the same head and to share the same arguments. Furthermore, covert coordination is expected to be fully compositional and to be applicable to any elements of a particular constituent type, while verb serialization may involve several restrictions as to the exact types of verbs that may form part of a series or to their linear order. The property of argument sharing applies to the verb constructions in (25) only with respect to the subject constituents, which is not unambiguous evidence since covert coordination may also apply to predicate constituents. The verb nàntí ‘walk’ cannot take a noun as a complement, as shown in (27), hence argument sharing is excluded for the locative complement of (24a). In (24b), the two motion verbs have clearly different complements. (27) *kòfí nàntí dímbókrô. Kofi walk Dimbokro (intended) ‘Kofi walks to Dimbokro.’ However, these constructions display some non-compositional properties that are not expected to occur in covert coordination constructions. First, the order of the involved verbs is irreversible. The verb encoding manner of motion has to precede the verb encoding the spatial relation in (25a)

Grammaticalization of spatial adpositions in Nànáfwê

95

(compare (28a), and the verb encoding ablative relation has to precede the verb encoding allative relation in (25b) (compare (28b). (28) a.

b.

*kòfí kò‰ dímbókrô nàntí. Kofi go Dimbokro walk (intended) ‘Kofi walks to Dimbokro.’ dímbókrô . *kòfí kò‰ ábìÔān fìn Kofi go Abidjan leave Dimbokro (intended) ‘Kofi goes from Dimbokro to Abidjan.’

The ungrammaticality of (28a-b) is not the result of general rhetorical constraints on the linearization of particular event types, but it is associated with the particular type of construction in which the two verbs are involved. The ungrammatical linearizations in (28) are possible in (29), which illustrates clear cases of covert coordination (since subject sharing does not apply). The contrast in the grammaticality of (29) versus (28) implies that the constructions displaying subject sharing involve some conventionalized linearization properties. (29) a.

b.

kòfí kò‰ dímbókrô òÊ nàntí. Kofi go Dimbokro 3.SG.SBJ.CL walk ‘Kofi goes to Dimbokro (and) he walks.’ fìn dímbókrô kòfí kò‰ ábìÔān òÊ Kofi go Abidjan 3.SG.SBJ.CL leave Dimbokro ‘Kofi goes to Abidjan (and) he leaves Dimbokro.’

.

Further evidence comes from the inflectional properties of the verbs in the constructions with subject sharing. The tense properties of the verbs involved in this construction have to be uniform (see (30a)), the corresponding inflectional markers may be affixed to either verb without any semantic difference (see (30b-e)), and marking of these inflectional properties on both verbs invokes the intuition of redundancy (see (30f-g)). Taken together, this data suggests that both verbs form a complex verb head in a construction of subject sharing which is accompanied by inflectional affixes once. (30) a.

*kòfí sú nàntí àá kò‰ dímbókrô. Kofi PROG walk FUT go Dimbokro (intended) ‘Kofi is walking and will go to Dimbokro.’

96

Amani Bohoussou and Stavros Skopeteas

b.

c.

d.

e.

f.

g.

kòfí sú nàntí kò‰ dímbókrô. Kofi PROG walk go Dimbokro ‘Kofi is walking to Dimbokro.’ kòfí àá fìn dímbókrô kò‰ ábìÔān. Kofi FUT leave Dimbokro go Abidjan ‘Kofi will go from Dimbokro to Abidjan.’ kòfí nàntí sú kò‰ dímbókrô. Kofi walk PROG go Dimbokro ‘Kofi is walking to Dimbokro.’ kòfí fìn dímbókrô àá kò‰ ábìÔān. Kofi leave Dimbokro FUT go Abidjan ‘Kofi will go from Dimbokro to Abidjan.’ ? kòfí sú nàntí sú kò‰ dímbókrô. Kofi PROG walk PROG go Dimbokro (redudant) ‘Kofi is walking to Dimbokro.’ ? kòfí àá fìn dímbókrô àá kò‰ ábìÔān. Kofi FUT leave Dimbokro FUT go Abidjan ‘Kofi will go from Dimbokro to Abidjan.’

Further evidence comes from negation. The negation particle mān is placed immediately after the verbal head in Nànáfwê, as exemplified in (31a). In constructions of ‘manner + relation’, the negative particle has to follow the last verb, as shown through the contrast between (31b) and (31c). The scope of negation in (31b) is not restricted to the preceding verb, but to the complex event, i.e., this expression may be followed either by the continuation ‘… but he walks to Abidjan’ or by the continuation ‘… but he runs to Dimbokro’, i.e. both verbs may be negated separately. This evidence supports the view that the manner verb and the relation verb form a complex event head in this construction. (31) a.

b.

c.

kòfí kò‰ mān dímbókrô. Kofi go NEG Dimbokro ‘Kofi does not go to Dimbokro.’ kòfí nàntí kò‰ mān dímbókrô. Kofi walk go NEG Dimbokro ‘Kofi does not walk to Dimbokro.’ *kòfí nàntí mān kò‰ dímbókrô. Kofi walk NEG go Dimbokro (intended) ‘Kofi does not walk to Dimbokro.’

Grammaticalization of spatial adpositions in Nànáfwê

97

Exactly the same data pattern appears in the ‘relation + relation’ construction. The negative particle has to appear right adjacent to the second verb and its scope does not only relate to the destination of the movement, but to the entire complex event. (32) a.

b.

kòfí fìn dímbókrô kò‰ mān ábìÔān. Kofi leave Dimbokro go NEG Abidjan ‘Kofi goes from Dimbokro to Abidjan.’ *kòfí fìn mān dímbókrô kò‰ ábìÔān. Kofi leave NEG Dimbokro go Abidjan (intended) ‘Kofi goes from Dimbokro to Abidjan.’

Adverbs display properties similar to the negation particle. As mentioned in section 2, adverbs are placed after the verbal complements in Nànáfwê. Examples (33a-b) show that the manner adverb ndèÌdè‰ ‘rapidly’ may be placed either after the manner verb or after the complement of the relation verb. This evidence shows that the dependent noun is not reanalyzed as complement of the first verb in which case the adverb could not occur in the position exemplified in (33a). The crucial evidence is however the interpretation of these examples. There is no discrete semantic difference between (33a) and (33b), i.e. it is not the case that the adverb in the former construction modifies the manner of motion and in the latter construction the event of reaching the destination. Both verbs constitute a complex event head that may be modified as a unit. (33) a.

b.

kòfí nàntí ndèÌdè‰ kò‰ dímbókrô. Kofi walks rapidly go Dimbokro ‘Kofi walks rapidly to Dimbokro.’ kòfí nàntí kò‰ dímbókrô ndèÌdè‰. Kofi walks go Dimbokro rapidly ‘Kofi walks rapidly to Dimbokro.’

This fact is exemplified through the contrast between (34a) and (34b) for the ‘relation + relation’ construction. An interpretation in which the temporal adverb either relates only to the origin of motion or only to the target of motion is not available. The adverb obligatorily relates to the complex event.

98

Amani Bohoussou and Stavros Skopeteas

(34) a.

b.

kòfí fìn dímbókrô ndè‰ kò‰ ábìÔān. Kofi leave Dimbokro early go Abidjan ‘Kofi goes early from Dimbokro to Abidjan.’ kòfí fìn dímbókrô kò‰ ábìÔān ndè‰. Kofi leave Dimbokro go Abidjan early ‘Kofi goes early from Dimbokro to Abidjan.’

Ιn conclusion, this section provided rich evidence that the two constructions involving V-relators, namely the ‘manner + relation’ construction and the ‘relation + relation’ construction, are serial verb constructions. The linearization of verbs in these constructions is conventionalized and cannot be altered under the condition of subject sharing. The encoding of inflectional categories and of negation as well as the interpretational properties of adverbs show that the two verbs involved in either construction form a complex head.

4.3. Evidence for grammaticalization to adpositions The next question is whether there is evidence for grammaticalization of the V-relators in these constructions to adpositions. The evidence we presented in the previous section clearly supports the view that this is not the case. The loss of inflection that is reported for verbs turning to adpositions in other languages of the area (see for instance Ewe or Akan in Lord 1993: 141–147) does not apply for Nànáfwê, see in particular examples (30e-g). Example (35) shows that the second verb of motion may be accompanied by the perfective suffix. (35)

kòfí nàntí hòÊ-lí dímbókrô. Kofi walk go-PFV Dimbokro ‘Kofi walked to Dimbokro.’

Furthermore, the distributional properties of the verbs in the previous section provide clear evidence that a reanalysis V>Adp did not take place in Nànáfwê. Adverbs are placed after verb complements, hence if the second verb was reanalyzed as an adposition, the placement of the adverb after the first verb would be excluded, which is not the case as demonstrated in (33a) and (34a). In the same vein, if the ‘V2 NP’ string were an adpositional phrase, then it should be possible for the negative particle to

Grammaticalization of spatial adpositions in Nànáfwê

99

precede this constituent, which also renders an ungrammatical serialization as (31c) shows. On the basis of this evidence, we conclude that V-relators are verbs in Nànáfwê that also occur in serial verb constructions.

5. Further differences between N-relators and V-relators The previous sections lead to the conclusion that N-relators and V-relators do not have the same categorical status. N-relators are nouns that display the properties of an incipient stage of grammaticalization to adpositions. V-relators are still at the first stage of the cross-linguistically attested grammaticalization path: they are serial verbs. That the two types of relators belong to distinct grammatical categories is indicated by their word order properties: V-relators precede the dependent noun, while N-relators follow it. If our estimation of the different categorical status is on the right track, then we should be able to identify further distributional differences between the two categories. A first difference relates to the properties of the governing slot. V-relators are essentially verbal heads, hence they may govern adverbs, as illustrated in (36a). This is not expected to apply to N-relators (see data from further languages of the same area in Heine, Claudi, and Hünnemeyer 1991). Example (36b) shows indeed that placement of the deictic adverb in the position of dependents of the N-relator results in ungrammaticality. (36) a.

b.

kòfí kò‰/ fìn/ sìn lè‰. Kofi go leave pass there ‘Kofi goes/leaves from/passes through there.’ *kòfí là lè‰ nú. Kofi lay there inner.side (intended) ‘Kofi lays there inside.’

Evidence from cleft sentence formation shows that the string ‘Vhead NP’ and the string ‘NP NPhead’ are not constituents of the same type. Example (37a) illustrates a canonical sentence with a positional verb, an N-relator and its nominal dependent. (37b-c) illustrate two different possibilities of forming cleft sentences: either through extraction of the dependent noun or through extraction of the whole verb complement including both the N-relator and its dependent.

100 Amani Bohoussou and Stavros Skopeteas (37) a.

b.

c.

bútèlí-n Ôín bjá-n sú. bottle-DEF stand chair-DEF topside ‘The bottle is standing on the chair.’ Ôín bjá-n sú jè‰ bútèlí-n chair-DEF topside REL bottle-DEF stand ‘It is on the chair that the bottle is standing.’ Ôín sú bjá-n jè‰ bútèlí-n chair-DEF REL bottle-DEF stand topside ‘It is the chair that the bottle is standing on.’

òÊ. PRES òÊ. PRES

These possibilities of extraction are not available for ‘Vhead NP’ strings. The examples in (38) are the clefted counterparts of (25a) and show that it is not possible to separate the two verbal heads of a serial verb construction, while it is possible to extract the nominal complement. (38) a.

b.

*kò‰ dímbókrô jè‰ kòfí nàntí go Dimbokro REL Kofi walk ‘It is to Dimbokro that Kofi is walking.’ dímbókrô jè‰ kòfí nàntí kò‰ Dimborko REL Kofi walk go ‘It it Dimborko that Kofi is walking to.’

òÊ. PRES òÊ. PRES

The same phenomenon is illustrated in (39) for the clefted counterparts of (25b). It is not possible to extract the second verb while its complement may be extracted. (39) a.

b.

kòfí fìn dímbókrô òÊ. *kò‰ ábìÔān jè‰ go Abidjan REL Kofi leave Dimbokro PRES (intended) ‘It is to Abidjan that Kofi goes from Dimbokro.’ kòfí fìn dímbókrô kò‰ òÊ. ábìÔān jè‰ Abidjan REL Kofi leave Dimbokro go PRES ‘It it to Abidjan that Kofi goes from Dimbokro.’

6. Conclusions This investigation began with the question whether there is evidence for grammaticalization to adpositions in Nànáfwê. Knowing from further West African languages that adpositions develop out of nouns and verbs, we

Grammaticalization of spatial adpositions in Nànáfwê 101

undertook a systematic examination of the distributional and interpretational properties of two classes of spatial relators in our object language: N-relators and V-relators. N-relators are nouns that denote parts of spatial configurations (typically termed ‘body part’ nouns). These nouns are at the incipient stage of a grammaticalization process. When used for the introduction of verbal complements, their use with pronominal possessors looses its semantic transparency and their denotation indicates a process of desemanticization. Furthermore, in particular constructions such as some conventionalized constructions with particular nominals or some metonymic uses they may not combine anymore with a possessive pronoun as in their nominal usage. On the semantic layer, the grammaticalized uses of these elements do not denote parts of spatial configurations but spatial regions. V-relators are verbs that also occur in verb series. We provided rich evidence for the distributional properties of these elements that unambiguously shows that in particular constructions they form a complex verb head together with other verbs. Both verbs encode a unique event that may be modified once as we illustrated by means of restrictions on the use of aspectual morphology, on the placement and scope of negation, on the placement and interpretation of adverbs. However, we argued that there is no evidence for a further development on this grammaticalization path, which would lead from (parts of) serial verbs to local adpositions. This argument was based on the fact that these elements are fully inflected and on the positional properties of adverbs and negation that indicate that these verbs (together with their complement NPs) are not reanalyzed as verbal complements. Furthermore, the extraction possibilities suggest that the two verbs in these constructions form a complex head that cannot be separated and that the second verb together with its complement does not form a unique constituent. The empirical relevance of our investigation is that it figures out the syntactic properties of Nànáfwê that provide us with evidence how the incipient stages of grammaticalization processes of nouns and verbs to adpositions look like. The theoretical significance of our findings is that they provide evidence for the independence between function and form in grammaticalization processes. In functional viewpoint, both N-relators as well as V-relators in Nànáfwê do not essentially differ from prepositions in familiar European languages. The corresponding constructions in several languages of the area where Nànáfwê is spoken underwent a grammaticalization process that led to the development of genuine adpositions in these

102 Amani Bohoussou and Stavros Skopeteas languages (see, e.g. Ewe in Heine, Claudi, and Hünnemeyer 1991). The corresponding elements in Nànáfwê are either in a very incipient stage of this process (N-relators) or have not entered this process yet (V-relators).

Notes 1.

2. 3. 4.

In order to maintain consistency in the glosses, we use the morphemic translation that applies to the lexical element (noun or verb) throughout its uses (also in the adpositional ones). In plural, independent and cliticized forms only differ in their tonal properties. The use of the pronominal possessor is obligatory with plural possessors (see Creissels and Kouadio 1977, Bohoussou 2008: 21). The region ‘exterior’, which is the antonym to ‘interior’ is not encoded by a member of this paradigm. This concept may be rendered either by the locative noun for ‘posterior’ in metonymical use or by more complex expressions.

References Ahoua, Firmin 1996 Prosodic Aspects of Baule: with special reference to the German of Baule speakers. Köln: Köppe. Bohoussou, Amani 1996 La structure interne de l’énoncé verbal du Nanafwe. Mémoire de maîtrise, Université de Cocody (Abidjan). 2008 L’énoncé complexe du Nanafwe. Unterschleissheim: Lincom. Bohoussou, Amani, and Stavros Skopeteas 2005 En nanafwe. In Typologie de la syntaxe connective, Christos Clairis, Claudine Chamoreau, Denis Costaouec, and Françoise Guérin (eds.), 155–169. Rennes: Presses Universitaires de Rennes. Carteron, Michel 1966 Essai de grammaire de la langue "baoulé". Bocanda. Creissels, Denis, and N´guessan Kouadio 1977 Description phonologique et grammaticale d’un parler baoulé. Abidjan: Institut de Linguistique appliqué (Linguistique africaine, 59). 1979 Les tons du baoulé: parler de la région de Toumodi. Abidjan: Université d’Abidjan, Institut de linguistique appliquée.

Grammaticalization of spatial adpositions in Nànáfwê 103 Creissels, Denis, and Jérémie Kouadio 2007 Resumptive pronouns in ditransitive constructions: The case of Baule. Handout presented in the Conference on Ditransitive Constructions, Max-Planck Institute for Evolutionary Anthropology, Leipzig 23-25 November 2007. Gross, Marcel 1967 Essai pour une phonologie du baoulé. Paris: CEDEV/CNRS (Bulletin de la SELAF, 2). Heine, Bernd 1989 Adpositions in African languages. Linguistique Africaine 2, 77–127. Heine, Bernd, Ulrike Claudi, and Friederike Hünnemeyer 1991 Grammaticalization: A Conceptual Framework. Chicago/London: The University of Chicago Press. Heine, Bernd, and Mechthild Reh 1984 Grammaticalization and Reanalysis in African languages. Hamburg: Buske. Kouadio, Jérémie 2000 Les séries verbales en baoulé: questions de morphosyntaxe et de sémantique. Studies in African Linguistics 29 (1): 75–90. Larson, Martha 2002 Baule SVCs: Two distinct varieties of missing objects’. In Proceedings of the Legon/Trondheim Linguistics Project Annual Colloquium, M.E. Dakubu and E. Osam (eds.), 87–109. The Empty Object Construction and related phenomena. Ph.D. dis2005 sertation, Cornell University. Leben, William R., and Firmin Ahoua 1997 Prosodic domains in Baule. Phonology 14: 113–132. Lehmann, Christian 1990 Towards lexical typology. In Studies in Typology and Diachrony, William Croft, Denning Keith, and Suzanne Kemmer (eds.), 161– 185. Amsterdam: Benjamins. 1992 Yukatekische lokale Relatoren in typologischer Perspektive. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 45: 626–641. 1995a Raumkonstruktion in funktionaler Sicht. Unpublished ms., University of Erfurt. 1995b Thoughts on Grammaticalization. Unterschleissheim: Lincom. Lord, Carol 1993 Historical Change in Serial Verb Constructions. Amsterdam/Philadelphia: Benjamins.

104 Amani Bohoussou and Stavros Skopeteas Macauley, Monica 1993 A Grammar of Chalcatongo Mixtec. Berkeley: University of California Press. Rennison, Joհn R. 1997 Koromfe. London/New York: Routledge. Rubba, Jo 1994 Grammaticalization as semantic change. A case study of preposition development. In Perspectives on Grammaticalization, William Pagliuca (ed.), 81–101. Amsterdam/Philadelphia: Benjamins. Skopeteas, Stavros 2007 Semantic categorizations and reasoning in the domain of superposition. In Ontolinguistics: Correlations between Ontological Status and Linguistic Coding, Dietmar Zaefferer and Andrea Schalley (eds.), 331–356. Berlin/New York: Mouton De Gruyter. Stewart, Osamuyimen Thompson 2001 The Serial Verb Construction Parameter. New York/London: Garland. Svorou, Soteria 1986 On the evolutionary paths of locative expressions. Papers from the Annual Meeting of the Berkeley Linguistics Society 12: 515–527. 1993 The grammar of space. Amsterdam/Philadelphia: Benjamins. Timyan, Judith 1977 A discourse-based grammar of Baule: The Kode dialect. Ph.D. dissertation, City University of New York. 1978 N wan yo: Cours de baoulé. Université d’Abidjan.

On the grammaticalization of the German preposition von as a genitive equivalent John Ole Askedal

The present contribution deals with the grammaticalization of the German preposition von as a genitive equivalent in possessive constructions. One strong argument in favour of assuming grammaticalization of von is the obligatory use of a von-PP in cases where no distinctive marking of genitive case is available for morphological reasons. Further supporting evidence is provided by coordinative and appositive possessor chains consisting of both genitive-marked and von-marked possessors. Such morphosyntactically mixed possessor sequences are typologically interesting in that they show coexistence and functional equivalence of synthetic and analytic means of expression within one syntactic frame.

1. Introduction In his seminal article on Grammaticalization and Related Changes in Contemporary German, Christian Lehmann (1991: 498–499) observes that the preposition von, long recognized as a functional equivalent of the adnominal genitive (cf. for instance Wilmanns 1906: 609–610; Behaghel 1923: 533–539), is in certain instances obligatory. This is the case when a noun to be used as a post-nominal modifier neither allows for morphological genitive marking nor is accompanied by a determiner or adjective enabling genitive marking of the NP. This state of affairs is illustrated by the examples in (1a), (1c) (from Lehmann 1991: 499), to which (1b) is added here for contrast: (1)

a. b. c.

die Erzielung stilistischer Effekte ‘the attainment of stylistic effects’ *die Erzielung Effekte die Erzielung von Effekten

Zifonun et al. (1997) make a note of the functional similarity of von and genitive morphology in German NPs:

106 John Ole Askedal In ähnlicher Weise wie das morphologische Merkmal gen [genitive, JOA] wirkt auch die Präposition von in attributiven Nominalphrasen, wie etwa der Schatten von Hans/von einer Buche. Auch sie kann explizit als Operator verstanden werden, der aus Flexionstermen im Sinne von Charakteristiken Nomenmodifikatoren macht. [The preposition von ‘of’ functions in a way similar to the morphological feature gen [genitive], compare der Schatten von Hans/von einer Buche ‘the shadow of Hans/of a beech tree’. The preposition can likewise be understood as an operator transforming characterizing inflected arguments into noun modifiers.] (Zifonun et al. 1997: 999)

According to Diewald (1997: 68), the use of the preposition von is in such cases strongly grammaticalized (“stark grammatikalisiert [...]”). To illustrate this, Diewald gives the example in (2) (= Diewald’s (9)), where the von-PP is an optional alternative to the genitive: (2)

Das Buch von Peter (Peters Buch) muß noch bei uns herumliegen. (Diewald 1997: 68) ‘Peter’s book must be lying around somewhere at our place.’

The descriptions of von as a modifier-adjoining element given by Zifonun et al. (1997) and Diewald (1997) seem to be typical in that they do not deal explicitly with obligatory von. Nor do they address the question of morphosyntactically heterogeneous post-nominal possessor sequences consisting of both morphologically marked genitive NP(s) and von-PP(s) and the linearization of different possessor expressions within such sequences. In what follows I shall present and discuss a modest corpus of altogether 25 examples of morphosyntactically heterogeneous, paratactic or appositional,1 possessor chains that I have been able to cull from a variety of sources in the course of the past couple of years. In my view, these examples provide even stronger empirical evidence for assuming a high degree of grammaticalization of von as a genitive equivalent than the examples given in the standard descriptions quoted above.

On the grammaticalization of the German preposition von 107

2. Presentation and discussion of morphosyntactically heterogeneous possessor sequences involving genitive and von-PP The paratactic or appositive structure of the possessor sequences in question is evidenced by the use of paratactic conjunctions (und, oder, und/oder, aber; sowie, cf. Zifonun et al. 1997: 2396–2398); a conjunctionlike element (bzw.); an apposition-introducing particle (als) or other words with a similar function (insbesondere, genauer, z. B.); or a simple comma. In sequences containing more than two possessor elements, internal hierarchic structure in the form of ‘layering’ of paratactic or appositional chains on different levels is a possibility to be taken into consideration (cf. (9)– (10) below). Sequences consisting of one genitive and one von-PP show both theoretically possible linear orderings of the possessor expressions, cf. (3) and (4), respectively:2 (3)

Genitive – von-PP (7 examples): a. Die Verweigerung der Zustimmung oder zumindest von Verständnis für den Sprecher ... (Zifonun et al. 1997: 939) ‘the refusal of consent or at least of an understanding of the speaker …’ b. durch Angabe eines Resultats und/oder von Handlungsfolgen (Zifonun et al. 1997: 126) ‘by indicating a result and/or a sequence of actions’ c. unter Einbezug der Responsive bzw. von ParaVerwendungen (Zifonun et al. 1997: 406) ‘by including responsive elements or para-applications’ d. Partizipien bestimmter Verben, z. B. von Verben mit effiziertem Objekt, ... (Zifonun et al. 1997: 1820) ‘the participles of certain verbs, for instance of verbs with an object ensuing from the action denoted by the verb’ e. die Unterstützung der Päpste, insbesondere von Innozenz III., ... (Schaeffner, ed., 1981: 56) ‘the support of the popes, in particular of Pope Innocent III…’ f. Die Genese des modalen Passivs im Deutschen ist in den größeren Zusammenhang der Entstehung analytischer Verbformen, insbesondere von Passivkonstruktionen gestellt worden. (Demske-Neumann 1994: 275)

108 John Ole Askedal

g.

(4)

‘The genesis of the German modal passive has been placed in the wider context of the development of analytic verb forms, in particular of passive constructions.’ Als ihr älterer Bruder William Aigret als Kind starb, wurde sie zur Erbin Aquitaniens und von sieben weiteren Ländern ... (Pohanka 2006: 75) ‘When her older brother William Aigret died as a child, she became heiress to Aquitaine and seven more countries ...’

von-PP – genitive (10 examples) a. durch Einsparung von Verknüpfungselementen und des Subjektes ... (Zifonun et al. 1997: 2229) ‘by omission of combining elements and of the subject’ b. aus der nicht einzudämmenden Flut von Sinneslust und des Vergnügens. (Jung, ed., 2002: 135) ‘from an uninhibited flood of sensual desire and enjoyment’ c. Die Organisation versammelt die Parlamente der Welt zur Erörterung von Fragen von internationalem Interesse und weltweiter Anliegen, ... (Faszination 2003: 9) ‘The organization gathers the world’s parliaments for the purpose of discussing questions of international interest and worldwide concerns.’ d. aufgrund von Ortsnamen und der in Kap. 2 erwähnten Wörter mit anlautendem „p“ (von Polenz 1978: 28) ‘because of place-names and the words with initial p mentioned in Ch. 2’ e. Hier sind erhebliche Versäumnisse von Schulen und in Sonderheit der Medien festzustellen. (Informationen für Neugierige 2003: 9) ‘Here considerable neglect on the part of schools and in particular the media has to be noted.’ f. ohne Anwendung von „sich lassen“ oder eines anderen Modalitätsoperators (Zifonun et al. 1997: 1856) ‘without the use of sich lassen or some other modal operator’ g. Die Kenntnis von Illokutionstypen, genauer der Bündel von konstituierenden Bedingungen für Illokutionen eines bestimmten Typs ... (Brandt et al. 1983: 114) ‘The knowledge of illocution types, more specifically of the constitutive conditions of a specific kind of illocutions …’

On the grammaticalization of the German preposition von 109

h.

i.

j.

nach der symbolischen Bedeutung von Godot oder anderer Charaktere (Schwanitz 2002: 347) ‘according to the symbolic value of Godot and other characters’ das Bild von der Sprache als eines Behälters für zu transportierende Inhalte (Albrecht 1998: 68) ‘the image of language as a container of content to be transported’ ein einigermaßen eingehendes Studium von Thomasius’ deutschen Schriften und anderer Werke (Eggers 1986: 289) ‘a relatively thorough study of the German writings and other works of Thomasius’

Sequences consisting of more than two elements appear to be rare but do occur. Consider first (5)–(6), where the possessor sequences comprise three elements: (5)

Genitive – genitive – von-PP (2 examples) a. eine bunte Sammlung profaner literarischer Werke, z. B. mancher alter Sagen, aber auch von Erzeugnissen frühmittelalterlicher „filid“. (Gaechter 2003: 42; author’s italics) ‘a varied collection of profane literary works, for instance a number of old sagas, but also of the products of early medieval filid’ b. Entwürfe der CDU/FDP, der SPD und von BÜNDNIS 90/DIE GRÜNEN (Reiher and Baumann, eds., 2000: 101) ‘proposals of the CDU/FDP, the SPD and BÜNDNIS 90/DIE GRÜNEN’

(6)

Genitive – von – genitive (1 example)3 Er kann zudem nicht im Skopus einer Frage, von Gradpartikeln oder der Negation stehen, ... (Pittner 1999: 266)

In five further examples, elliptical omission of von or a genitive-marked determiner can be assumed, yielding sequences of three (7)–(8), four (9) and six (10) possessor elements respectively:4

110 John Ole Askedal (7)

Genitive – von-PP – (von) NP Die Formulierung ihrer Ziele schließt den Wettstreit der Ideen der Mitglieder, von Plattformen und (von) innerparteilichen Strömungen ein. (Programm der PDS 1990: 90) ‘The statement of its [the party’s] goals includes the competing ideas of its members, the various platforms and internal trends within the party.’

(8)

von-PP – (von-) N – genitive a. Für ihn, und von ihm vermittelt, für Goethe und die Stürmer und Dränger war Geschichte mehr das Wirken von Individuen als die dürre Abfolge von Schlachten, (von) Krönungsdaten und der Todesfälle. (von Borries and von Borries 1991: 237) ‘For him, and through his mediation, for Goethe and the Sturm und Drang poets, history was to a greater extent the actions of individuals than a dry sequence of battles, coronation dates and deaths.’ b. die Gegnerschaften der Erzbischöfe von Köln, (von) Mainz und auch Salzburgs, ... (Schneidmüller and Weinfurter, eds., 2003: 195) ‘the antagonisms of the archbishops of Cologne, Mainz and, in addition, Salzburg, …’

(9)

Genitive – genitive – von-PP – (von) (definite article der) N Identität/Nicht-Identität der jeweiligen Adressaten sowie der Sprechereignisorte bzw. von der Konstanz oder ((von) der) NichtKonstanz der Relation zu einem zeitdeiktischen Bezugspunkt (Zifonun et al. 1997: 1760) ‘identity/non-identity of the addressee in question and the location of the speech event or of the constancy or non-constancy of the relationship to a temporal-deictic point of reference’

(10)

von-PP – (von) N – genitive – (genitive des of the definite article) – von-PP – (von) N die von der Bodenforschung erhärtete Entfaltung von Ackerbau und (von) Viehzucht, des Siedlungs- und (des) Wohnwesens, von Technik und (von) Schiffahrt. (Adolf Bach; quoted from Vennemann 1984: 107)

On the grammaticalization of the German preposition von 111

‘the development, confirmed by archaeological research, of agriculture and cattle breeding, settlements and house building, technology and sea-faring activities’ There is no obvious reason to analyze (5) and (6) in terms of a hierarchic layering of paratactic structures or to associate the omission of possessor marking in (7) and (8) with such layering. However, the examples in (9)–(10), where certain possessor markers are left out, naturally lend themselves to a layering analysis. Cf. (11)–(12): (11)

Identität/Nicht-Identität [[der jeweiligen Adressaten sowie der Sprechereignisorte] bzw. [von der Konstanz oder Nicht-Konstanz der Relation zu einem zeitdeiktischen Bezugspunkt]]

(12)

die von der Bodenforschung erhärtete Entfaltung [[von Ackerbau und von Viehzucht], [des Siedlungs- und Wohnwesens], [von Technik und Schiffahrt]]

In the majority of the cases in (3)–(10), the use of a von-PP as a substitute for the genitive is forced by the morphological deficit described by Lehmann (1991: 498 f.), i.e. the inability of a large number (probably the majority) of German nouns to show morphologically distinctive genitive marking. In a few instances, however, morphological genitive marking is possible or its replacement by von seems to require another reason than lack of morphological distinctiveness. In (3g) the von-PP can no doubt be replaced by genitive marking (cf. zur Erbin Aquitaniens und sieben weiterer Länder). In (9), the use of von instead of the morphologically possible genitive in von der Konstanz oder Nicht-Konstanz der Relation... seems to serve the semantic purpose of distinguishing the two conceptual groups of der jeweiligen Adressaten sowie der Sprechereignisorte on the one hand, and von der Konstanz oder NichtKonstanz der Relation zu einem zeitdeiktischen Bezugspunkt on the other. In other cases other factors seem to be at work. In (4h), the genitive Godots can be substituted for von Godot,5 but the von-PP may still have been chosen because of the foreign, French character of the name Godot. In (4i) the use of a von-PP in das Bild von der Sprache als eines Behälters probably has lexical motivation (cf. sich ein Bild von etwas machen). In (4j), replacing von in ein einigermaßen eingehendes Studium von Thomasius’ deutschen Schriften und anderer Werke by genitive marking of deutsch als

112 John Ole Askedal deutscher would have yielded possessive genitive marking on two subordinate syntactic levels within the NP (cf. ein einigermaßen eingehendes Studium Thomasius’ deutscher Schriften und anderer Werke), which is at the very least stylistically highly infelicitous. Finally, using theoretically possible case agreement in Entwürfe der CDU/FDP, der SPD und von BÜNDNIS 90/DIE GRÜNEN would have resulted in Entwürfe der CDU/FDP, der SPD und des BÜNDNISSES 90/DER GRÜNEN, which appears highly cumbersome and might even detract from the recognizability of BÜNDNIS 90/DIE GRÜNEN as the proper name of a political party. The general conclusion thus has to be that the majority of the von-PPs occurring in possessor sequences in the admittedly small number of examples I have been able to assemble are morphologically conditioned obligatory substitutes for genitive marking. The examples in (4), (6), (8) and (10), with one or two von-PPs preceding the genitive, provide evidence that, on the stylistic level of the texts dealt with here, the genitive is the unmarked, primary option, to which the author reverts once the initial morphological default problem has been dealt with by the use of a von-PP. The flexibility of the linearization of the post-nominal genitives and von-PPs in (3)–(10) contrasts with the fixed linearization of categorically different complements to the noun, which are distributed in accordance with the topological schema for complex NPs in (13a) (cf. Duden 2005: 816)6: (13)

a. b.

# Det. | Adj. | Noun | Genitive | PP(s) # # die | große | Freude | der Sport- | über ihren Sieg # lerin

Paratactic conjunctions etc. are never used to combine elements belonging in the genitive and the PP slots respectively in the linearization schema in (13a). This goes to prove that a possessor chain consisting of morphological genitive(s) and von-PP(s) constitutes one syntactic unit belonging in the Genitive slot of (13a). It should be noted that paratactic and appositional possessor sequences also differ syntactically from the ‘distance genitives’ occasionally found in cases like (14a–c), where a genitive is added secondarily to an NP(1) that already conforms to the canonical topological pattern in (13a), yielding an NP(2) conforming to the extended hierarchical pattern illustrated in (15):

On the grammaticalization of the German preposition von 113

(14)

a.

b.

c.

(15)

Institut für Germanistik der Universität Wien (Offizieller Briefkopf des Instituts für Germanistik der Universität Wien.) ‘Department of Germanic Studies at the University of Vienna’ die zweisprachige Neuausgabe der Bibel auf griechisch und latein des Erasmus von Rotterdam (Schwanitz 2002: 138) ‘the new bilingual edition of the Bible in Greek and Latin by Erasmus of Rotterdam’ In einem Lied der Neuen Hohen Minne Walthers ... (Weddige 2003: 133) ‘in a song of New Courtly Love by Walter’

[[Institut [für Germanistik] PP] NP1 der Universität Wien] NP2

3. Concluding remarks on coordination and grammaticalization The paratactic and appositional possessor sequences in (3)–(10) are theoretically interesting in connection with general coordination theory, in particular common assumptions concerning “structural homogeneousness” (Lang 1984: 19) or “functional similarity” (cf. Zifonun et al. 1997: 2260 f.) and the like as a basic coordination requirement. As the genitive NPs and the von-PPs in (3)–(10) are clearly morphosyntactically different, yet functionally equivalent, one has to assume that the German preposition von has been grammaticalized as a possession marker to the extent that the categorical disparity between morphological genitive and von-PP is overridden by other syntactic, functional demands. Moreover, given the obligatory character of von in syntagmatic environments like Lehmann’s example in (1) and the majority of the examples presented in section 2, we are in the case of von as a partly optional partly syntagmatically conditioned obligatory substitute for morphological genitive marking also dealing with a process of “increasing obligatoriness” or “obligatorification” (cf. Lehmann 1995: 139–143; Heine 2003: 588), testifying to a high degree of grammaticalization. Of particular interest is the fact that the default function of von is not restricted to a position following a morphological genitive; a von-PP may just as well precede a morphological genitive. In fact, the von-PP precedes a genitive in a small majority – altogether 14 out of 25 – of the possessor

114 John Ole Askedal chains in the corpus (cf. (4), (6), (8), (10)). These examples are all the more remarkable because in each and every case the possessor chain question could have been continued by the use of a dative governed by the preceding preposition von (cf. e.g. (4a) and durch Einsparung von Verknüpfungselementen und dem Subjekt). This indicates that the genitive is the unmarked means for marking possessors. However, given the fact that the use of the genitive as a possessor morpheme is morphologically restricted, whereas von is universally possible, one might surmise that a markedness reversal between the two ways of marking possessors is under way. Still, stylistic preferences obviously work strongly in favour of the genitive in present-day literary German. In a typological perspective, the syntagmatic interplay of the morphological genitive and von-PP in possessor constructions is a case of functional equivalence of a synthetic and an analytic means of expression within the same syntactic domain. This tallies well both with August Wilhelm Schlegel’s ([1818] 1971: 14) original conception of analytic languages as later developments of synthetic languages and with the common observation that German has in the course of its known history developed a number of analytic constructions, while at the same time retaining basic synthetic characteristics in its overall morphosyntactic structure.7 From this typological perspective, the functional and syntagmatic equivalence of synthetic morphological genitive and analytic von-marking casts doubt on the usefulness of “clines” like, e.g., the one presented by Hopper and Traugott (2003: 7) as a general measure for degree or level of grammaticalization (cf. the critical comments in Joseph 2004: 28). The functional equivalences described in this article rather indicate that grammaticalization assessments not only need to take into consideration position on a morphosyntactic or constructional “cline” but also the – possibly varied (or “inconsistent”) – typological character of specific language systems.

On the grammaticalization of the German preposition von 115

Notes 1.

2. 3. 4. 5.

6. 7.

In the following, the distinction between paratactic and appositional status is not descriptively relevant for the present purpose and is therefore left out of consideration. For ease of exposition, alternating morphological genitive and von-markings are set in boldface in the examples in (3)–(10). This example is adduced with the proviso that der Negation might also be taken to be governed by the preceding preposition von. In (7)–(10) the omitted but retrievable genitive or von-markings are supplied in parentheses. Cf. for instance: Als man die [sic] Insassen nach Ende der Aufführung die unvermeidliche Frage nach der Identität Godots stellte, kamen Antworten wie der Sinn des Lebens,... [When after the performance the inmates were asked the inevitable question of the identity of Godot, there were answers like the meaning of life…] (http://www.google.com/search?q=Godots&hl=en&lr =lang_de&as_qdr=all&start=10&sa=N; 30.01.2008). For the present purpose, the slots for appositions indicated in Duden (2005) are not relevant and therefore omitted. Cf. for instance Stedje (2006: 21–22, 109–111, 227) and Askedal (1996: in particular 377–379).

Sources Albrecht, Jörn 1998 Literarische Übersetzung: Geschichte – Theorie – Kulturelle Wirkung. Darmstadt: Wissenschaftliche Buchgesellschaft. Borries, Ernst von, and Erika von Borries 1991 Deutsche Literaturgeschichte. Vol. 2: Aufklärung und Empfindsamkeit. Sturm und Drang. (dtv 3342.) München: Deutscher Taschenbuch Verlag. Brandt, Margareta, Wolfgang Koch, Wolfgang Motsch, Inger Rosengren, and Dieter Viehweger 1983 Der Einfluß der kommunikativen Strategie auf die Textstruktur – dargestellt am Beispiel des Geschäftsbriefes. In Lunder Symposium 1982, Inger Rosengren (ed.), 105–135. Stockholm: Almqvist & Wiksell International. Demske-Neumann, Ulrike 1994 Modales Passiv und Tough Movement. Zur strukturellen Kausalität eines syntaktischen Wandels im Deutschen und Englischen. (Linguistische Arbeiten 326.) Tübingen: Max Niemeyer.

116 John Ole Askedal Eggers, Hans 1986 Deutsche Sprachgeschichte. Band 2: Das Frühneuhochdeutsche und das Neuhochdeutsche. (Rowohlts Enzyklopädie 426.) Reinbek bei Hamburg: Rowohlt Taschenbuch Verlag. Faszination 2003 Faszination. Das Philatelie Journal für Sammler Nr. 296. Genf: Postverwaltung der Vereinten Nationen. Informationen für Neugierige 2003 Informationen für Neugierige. Verein Deutsche Sprache. Bautzen: Druckerei Schleppers. Gaechter, Paul 2003 Die Gedächtniskultur in Irland. 2. Auflage, bearbeitet von Wolfgang Meid. (Innsbrucker Beiträge zur Kulturwissenschaft, Sonderheft 114.) Innsbruck: Institut für Sprachen und Literaturen der Universität Innsbruck. Jung, Thomas (ed.) 2002 Alles nur Pop? Anmerkungen zur deutschen und niederländischen Literatur des zurückliegenden Jahrzehnts. (Osloer Beiträge zur Germanistik 32.) Frankfurt am Main: Peter Lang Europäischer Verlag der Wissenschaften. Pittner, Karin 1999 Adverbiale im Deutschen. Untersuchungen zu ihrer Stellung und Interpretation. (Studien zur deutschen Grammatik 60.) Tübingen: Stauffenburg. Pohanka, Reinhard 2006 Die Herrscher und Gestalten des Mittelalters. Wiesbaden: Marix Verlag. Polenz, Peter von 1978 Geschichte der deutschen Sprache. (Sammlung Göschen 2206.) Berlin/New York: Walter de Gruyter. Programm der PDS 1990 Programm der Partei des demokratischen Sozialismus von 1990. In Wahlparteitag der Partei des demokratischen Sozialismus, 24./25. Februar. Berlin, 88–108. Reiher, Ruth, and Antje Baumann (eds.) 2000 Mit gespaltener Zunge. Die deutsche Sprache nach dem Fall der Mauer. Berlin: Aufbau Taschenbuch Verlag. Schaeffner, Claude (ed.) 1981 Stadt und Land im Mittelalter. Kirche und Reich. Die italienischen Städte. (Weltgeschichte in Bildern [8].) Bayreuth: Gondrom. Schwanitz, Dietrich 2002 Bildung: Alles, was man wissen muß. München: Goldmann.

On the grammaticalization of the German preposition von 117 Schneidmüller, Bernd, and Stefan Weinfurter (eds.) 2003 Die deutschen Herrscher des Mittelalters. Historische Portraits von Heinrich I. bis Maximilian I. (919–1519). Darmstadt: Wissenschaftliche Buchgesellschaft. Vennemann, Theo 1984 Bemerkung zum frühgermanischen Wortschatz. In Studia Linguistica et Philologica: Festschrift für Klaus Matzel zum sechzigsten Geburtstag überreicht von Schülern, Freunden und Kollegen, HansWerner Eroms, Herbert Kolb, and Bernhard Gajek (eds.), 105–119. Heidelberg: Carl Winter Universitätsverlag. Weddige, Hilkert 2003 Mittelhochdeutsch: Eine Einführung. 5. Auflage. München: C. H. Beck. Zifonun, Gisela, Ludger Hoffmann, Bruno Strecker et al. 1997 Grammatik der deutschen Sprache. 3 Vols. Berlin/New York: Walter de Gruyter.

References Askedal, John Ole 1996 Deutsch als sprachtypologischer “Mischtyp”. In Deutsch – typologisch, Ewald Lang and Gisela Zifonun (eds.), 369–383. (Jahrbuch des Instituts fıür deutsche Sprache 1995.) Berlin: Walter de Gruyter. Behaghel, Otto 1923 Deutsche Syntax. Eine geschichtliche Darstellung. Vol. 1: Die Wortklassen und Wortformen. A. Nomen. B. Pronomen. Heidelberg: Carl Winter. Diewald, Gabriele 1997 Grammatikalisierung. Eine Einführung in Sein und Werden grammatischer Formen. (Germanistische Arbeitshefte 36.) Tübingen: Max Niemeyer. Duden 2005 Duden. Die Grammatik. 7., völlig neu erarbeitete und erweiterte Auflage, Dudenredaktion (eds.). (Der Duden in zwölf Bänden 4.) Mannheim/Leipzig/Wien/Zürich: Dudenverlag. Heine, Bernd 2003 Grammaticalization. In The Handbook of Historical Linguistics, Brian D. Joseph and Richard D. Janda (eds.), 575–601. Oxford etc.: Blackwell Publishing.

118 John Ole Askedal Hopper, Paul J., and Elizabeth Closs Traugott 2003 Grammaticalization. Second Edition. Cambridge: Cambridge University Press. Joseph, Brian D. 2004 Rescuing traditional (historical) linguistics from grammaticalization theory. In Up and Down the Cline: The Nature of Grammaticalization, Olga Fischer, Muriel Norde and Harry Perridon (eds.), 45–71. (Typological Studies in Language 59). Amsterdam/Philadelphia PA: John Benjamins. Lang, Ewald 1984 The Semantics of Coordination. Amsterdam: John Benjamins. Lehmann, Christian 1991 Grammaticalization and Related Changes in Contemporary German. In Approaches to Grammaticalization. Vol. II: Focus on Types of Grammatical Markers, Elizabeth Closs Traugott and Bernd Heine (eds.), 493–535. (Typological Studies in Language 19:2.) Amsterdam/Philadelphia PA: John Benjamins. 1995 Thoughts on Grammaticalization. Revised and expanded version. First published edition. (LINCOM Studies in Theoretical Linguistics 01). München/Newcastle: LINCOM Europa. Schlegel, August Wilhelm [1818] 1971 Observations sur la Langue et la Litterature Provençales. (Tübinger Beiträge zur Linguistik 7.) Tübingen: Tübinger Beiträge zur Linguistik. Stedje, Astrid (with Heinz-Peter Prell) 2007 Deutsche Sprache gestern und heute. 6. Auflage. (UTB 1499.) Paderborn: Wilhelm Fink. Wilmanns, Wilhelm 1906 Deutsche Grammatik: Gotisch, Alt-, Mittel- und Neuhochdeutsch. Dritte Abteilung: Flexion. 1. Hälfte: Verbum. Strassburg: Karl J. Trübner. Zifonun, Gisela, Ludger Hoffmann, Bruno Strecker et al. 1997 Grammatik der deutschen Sprache. 3 Vols. Berlin/New York: Walter de Gruyter.

Part III Verbal categories

The grammaticalization of agreement in Chibchan J. Diego Quesada

1. Introduction The Chibchan languages extend from Northeastern Honduras, through the Atlantic Coast of Nicaragua, most of Costa Rica, Panama, Colombia, and West of Venezuela. Areally, the Chibchan languages are a part of the Intermediate Area, which borders with Mesoamerica to the North and with the Peruvian and Amazonian areas to the South. On the basis of important differences that recent archaeological, anthropological and linguistic research has established between Central America and Colombia, the Chibchan world has been divided into these two geographic zones (cf. Quesada 2007). Some of those differences have to do with the category of auxiliaries, which seems to be more relevant and more grammaticalized in Colombia. Another difference concerns modal systems; the Colombian languages have the category of data source, and appear thus close to those of the Peruvian Area. The third main difference concerns clause connectors as part of verbal morphology, which are absent in Central America (with exception of Rama), but present in Colombia. One more determining factor of this division concerns agreement, which appears by far more syntacticized in Colombia than in Central America.1 All these differences point to a more elaborate verbal complex in Colombia than in Central America. Rather than venturing an explanation about the origin of this division, this paper intends to show that some Chibchan languages of Central America find themselves in the process of grammaticalizing agreement patterns; the various sources of this process will be discussed in detail in the light of Lehmann’s theory of grammaticalization (Lehmann 1995).

2. Agreement When talking about the grammaticalization of agreement, two things have to be kept in mind. First, grammaticalization proceeds from discourse and semantics to syntax and morphology. Second, agreement, as categorial covariation which expresses syntactic reference-relations morphologically,

122 J. Diego Quesada originates in anaphoric relations (cf. Lehmann 1982); the categories of a noun phrase are reproduced by another element. The grammaticalization of (person) agreement thus suggests a discourse-driven process, motivated by the need to identify the participants of situations. If the agreement patterns of a language are highly grammaticalized, their use is mechanical (e.g. English third person singular), if not, these will be heavily dependent on discourse (e.g. the Spanish clitic pronouns ceasing to be pronominal and becoming object agreement markers, cf. Quesada (1995)). In addition, as in every gradual process (and grammaticalization is one such process), the existence of identifiable intermediate levels is expected; in the case of agreement, these will be characterized, language internally, by variability (between semantic and syntactic agreement) and, cross-linguistically, by the arrangement of the languages compared into at least three stages/groups: syntactic agreement, variable agreement, and zero or incipient agreement. According to Lehmann, agreement markers originate from pronouns, usually external to the clause; gradually, these become integrated into it, until they cliticize to the verb; “since its referent is ultimately in the same clause, its function ceases to be anaphora and becomes agreement” (Lehmann 1995: 114). The remainder of this paper explores the paths followed by four Chibchan languages of Central America in the rise of agreement.

3. Chibchan: to agree or not to agree 3.1. Teribe As mentioned in footnote 1, with the exception of Paya (Honduras) and Guatuso (Costa Rica) most Central American Chibchan languages lack agreement. The languages that seem to be undergoing the process are Rama (Nicaragua), Boruca (Costa Rica), Teribe and Cuna (Panama). Let us begin with Teribe. There are three basic orders in Teribe transitive clauses: SOV order is used discourse-initially, for grounding participants, and to reinforce their identity in some discourse passages (1); the more frequent OV-s order, where ‘-s’ stands for a person-indexing suffix, is used for running discourse (2); and the inverse construction OVSdë, where S is a full noun phrase marked as obviative (by dë) in postverbal position (3):

The grammaticalization of agreement in Chibchan 123

(1)

Ta Jacinto shpo-no. 1.SG Jacinto hit-PFV ‘I hit Jacinto.’

(2)

Jacinto shpo-ro-r. Jacinto hit-PFV-1.SG ‘I hit Jacinto.’

(3)

Jacinto shpo-ra Carlos dë. Jacinto hit-PFV.INV Carlos OBV ‘Carlos hit Jacinto.’

The Teribe personal pronouns fall into two paradigms, one that I call “nominal” and another that I term “oblique”. The former is used to code referents in subject and object relations, while the latter codes objects and objects of postpositions. Members of the oblique paradigm are also used as possessive determiners, in prenominal position. The Teribe pronoun system is illustrated in Table 1; what could be called “agreement markers” in Teribe (see Table 2) are grammaticalized instances of the nominal pronouns: Table 1. The pronoun system in Teribe

SINGULAR

PLURAL

1. 2. 3. 1. EXCLUSIVE 1. INCLUSIVE 2. 3. SAME 3. DIFFERENT

NOMINAL

OBLIQUE

ta pa ∅ tawa shi pay ∅ ebga

bor bop ba borwa bi bomi ba ba

124 J. Diego Quesada Table 2. Teribe person markers SINGULAR

PLURAL

1. 2. 3. 1. EXCLUSIVE 1. INCLUSIVE 2. 3. SAME 3. DIFFERENT

-r -p -a ~ -∅ (depending on verb class) -rwa -y -mi -a ~ - ∅ (depending on verb class) ... lok -ba

To the extent that the subject is indexed in the verb in the OV-s order, it can be said that agreement is highly syntacticized; on the other hand, the existence of the discourse-determined alternative word order SOV shows that it is not totally grammaticalized. In principle, the SOV and OV-s orders are mutually exclusive; however, there are already instances in which both “agreement” and word order are used (notice the SVO order in (4)): (4)

Kone kone kro-ro-rwa some find-PFV-1.PL.EXCL ‘Some [of us] found a husband.’

borwa lanma. 1.PL.EXCL.POSS husband

This process is a clear instance of semantic agreement, which usually appears at the onset of the grammaticalization of agreement. That this is the case in (4) is evidenced by the fact that the exceptionally infrequent SVO order is not an instance of the OV-s order, which is the order that exhibits regular agreement in the language; equally rare are instances of agreement on the SOV order: (5)

Oba junikong om woyoje-r-a lok p’ir. finish people this.side FOC forget-PFV-3 PL ‘The people from here have have forgotten totally about it.’

(6)

Tawa borwa llëbo thing 1.PL.EXCL 1.PL.EXCL.POSS ‘We respect those things of ours.’

e DEM

bankrë-rwa. fear-1.PL.EXCL

The admittedly rare cases of agreement in the SVO and SOV orders thus hint at an extension of the agreement pattern. It becomes clear then that Teribe is somewhere between syntactic and semantic agreement. In other

The grammaticalization of agreement in Chibchan 125

words, agreement occurs in some particular constructions whose syntactic structure is more complicated than a canonical SOV clause.

3.2. Rama Let us now turn to Rama. This language is SOV (29), with an alternative Os-V word order, where ‘s-’ stands for a subject-agreement prefix (7): (7)

naas glaas aark-u 1.SG glass break-PAST ‘I broke the glass’

(8)

chiicha i-ngw-i chicha 3.SG-drink-PRES ‘he drinks chicha’ (CIDCA 1990: 51-2).

As in Teribe, the two orders are discourse-determined; the former is used discourse-initially, to introduce and to reactivate participants, while the latter is used in running discourse (CIDCA 1990: 72). As shown in (8), Rama makes use of prefixes to express subject-agreement; free subject pronouns are used for emphasis and/or discourse-opening. While the distribution of the two word orders in Rama mirrors that of Teribe, in that person markers are used in the absence of the free personal pronouns (or full NP subjects), the Rama system of participant identification differs in two important respects from that of Teribe, thus revealing a higher degree of grammaticalization of person-agreement. First, there are two paradigms of personal pronouns, one for subject and one for object. The former are of two kinds, free and bound. The free forms are: (8a) 1. 2. 3.

singular naas ~ nah maa yaing

plural nsut mulut anut

These forms are used in sentence-initial position in the SOV order. The bound pronouns are short (grammaticalized) forms of the free ones, and are used in the (O)s-V order:

126 J. Diego Quesada (8b)

singular 1. n-, ni2. m-, mi 3. y-, i(CIDCA 1990: 72)

plural2 ns-, s- nsu-, sum-, mul-, mlan-

As for the object pronouns, these all end in a and are all free (3.SG is usually ∅): (8c)

singular 1. n-a 2. m-a 3. y-a ~ ∅ (CIDCA 1990: 74)

plural nsul-a mulul-a anul-a

Second, the subject-prefixes are used regularly with intransitive verbs: (9)

y-almalng-u 3.SG-die-PAST ‘he died’ (CIDCA 1990: 52)

(9) shows that, as opposed to Teribe, where the person suffixes are to a great extent still determined by discursive needs, the use of the Rama person agreement markers is more syntacticized. In fact, the absence of the subject prefix in an intransitive construction with no free subject NP is ungrammatical (e.g. in (9), *almalng-u).

3.3. Boruca The next language in this survey, Boruca, exhibits a pattern of participant identification which is completely the opposite of the Rama one. Boruca is SOV; participants are not cross-referenced on the verb (but see below), their role being made explicit by word order and by a set of contrast and focus markers. In addition, the Boruca pronominal system does not distinguish the core relations. The Boruca pronominal system is represented in (10); these forms code both subjects and objects; in addition, they function as obliques and possessive pronouns:

The grammaticalization of agreement in Chibchan 127

(10) 1. 2. 3.

singular a bá i

plural di? (róhk) bi? (róhk) i? (róhk)

The plural marker róhk is used only when the plural pronouns appear in subject-function. When functioning as subjects, these forms rarely if ever stand alone; in most cases they are accompanied by the informationstructure markers ang or ki, or else by abí, as in (11a), (11b), and (11c).3 Actually, their function as either subject or object is marked by the presence or absence, respectively, of the markers – that is, none of these markers (with certain exceptions in the case of abí) is used with non-subject pronominal NPs (this is why ki is glossed as subject in (11b)): (11a)

ba-ng daba-krá arrive-PFV 2.SG-FOC ‘you, you came’

(11b)

ba ki 2.SG SBJ ‘you came’

(11c)

abi-ng daba-krá a?r4 arrive-PFV 1.SG EMPH-FOC ‘it was me who arrived’

daba-krá arrive-PFV

Of the three markers, however, ang is restricted to subject noun phrases (full or pronominal); on the other hand, ki and abí can accompany nonpronominal object noun phrases and postpositional phrases. A ki-marked pronoun in object function is ungrammatical: (12a)

*At ki 1.SG(OBJ) ‘He hits me.’

bahd-r-ing. hit-PRES-3.SG

(12b)

*At ki kwik-krá i ki mang. 1.SG(SBJ) dance-PFV 3.SG (OBL) with ‘I danced with her.’

128 J. Diego Quesada It is the role of ang which is of interest here. A characteristic of this marker is that it fuses with the pronouns, producing portmanteau (bound) morphemes: (13) 1. 2. 3.

singular át + ang → ang bá + ang → bang i + ang → ing

plural → ding di? + ang ? → bing bi + ang i róhk + ang → ing róhk

These new forms, in turn, are fused and affixed to the conjugated verb forms in cases in which the basic SOV order is disrupted by reordering of constituents, as in (14), where S is postposed, or by reordering, as in (15). Fusion takes place with the forms beginning in a vowel (ang, first singular; and ing third singular and plural). The three forms with an initial voiced stop (bang, second singular; ding, first plural; and bing, second plural) also cliticize to the last verbal suffix, which in Boruca always ends in a vowel; stops in intervocalic position become fricative. Although total fusion does not occur with these forms, the lenis articulation of the fricative segment may eventually cause the total attrition of this element: (14)

ai?-kr-ang e?tse auh one dog kill-PFV-1.SG ‘(I) killed a dog’

(15)

we?

kuchi uge? at tsan-kr-ing róhk DEM pig because.of 1.SG put.in-PFV-3.SG PL ta ku? jail to ‘because of this pig (they) put me in jail’

Ang thus has two forms, a free or ‘heavy’ one used to focus subjects, and a bound or ‘light’ (more grammaticalized) one, reminiscent of the relation between free subject pronouns and prefixes and suffixes of Rama and Teribe, respectively. The light form of ang can be affixed to free personal pronouns, still as a focus marker, as in (11a) – a free portmanteau morpheme; in addition, the portmanteau morpheme can lose its free-form status, being affixed to the verbal complex, in which case it functions as a topic continuity marker and ultimately as an agreement – marker; the result is an incipient alternative OV-s word order, much like that in Teribe. In

The grammaticalization of agreement in Chibchan 129

fact, the most common use of ang is that of highly topical (‘active’) subject; thus ang marks same subjects, while ki in its function as subject marker5 tends to appear when subjects are new in discourse. Thus Boruca has two basic word orders for transitive clauses: A(ki)OV and OVs(-ang), whose distribution parallels that of the word orders of Rama and Teribe. Compare (16), where ki marks the subject of the presentative sentence (which although is not transitive does illustrate the discourse-function of ki as marker of new subjects), but ang performs the anaphoric function: (16)

ta kwíng kawí?-kra e?tse sí?kwa ki. Brúnkahk kí6 SPEC in many live-PFV one foreigner SPEC Boruca Wá ki ya? ka?yéng-íra Johnson. DEM DEF REFL call-IMP Johnson. kwíng kákba? róhk ki ta, Tunxuá krí?-ír-íng PL SPEC in, policeman big-IMP-3.SG many day e?tse brúnkahk mang i-ng wá? róhk ba-kr-í-ng have-PFV-3-SG one Boruca then 3-SG child PL e?de, bú?k kóngat e?tse ramát. with two males one female. ‘There lived right in Boruca for a long time a [certain, specific] foreigner [that was] called Johnson. He was a police officer for many years and had three children with a Boruca [woman], two boys and a girl.’ (Adapted from Constenla and Maroto 1986: 88)

While the distribution of the two word orders and the two markers functions as explained above, their use is not yet obligatory. Zero anaphora is still a very productive mechanism in Boruca, whereby the identification of participants is still heavily dependent on discourse (cf. the texts included in Constenla and Maroto 1986).

3.4. Cuna In the last language under study here, Cuna, the situation is less complicated, as the process is at a very early stage. Personal pronouns in Cuna are formed by the possessive forms plus the marker t(t)i, in singular forms only; however, it seems that the longer form is used for emphasis. Holmer (1946: 190) explains that there is a distinction between an soke and anti soke, where the former means ‘I say’, and the latter ‘It is I who says’. The

130 J. Diego Quesada personal pronouns, as in Rama, have a reduced form which is prefixed to the verb, yielding an (O)s-V order. Thus pe-takke, 2.SG-see, can mean ‘you see him’ or ‘he sees you’, because third person tends to be expressed by ∅. As for plural, the suffix -mala is added to the verb, as in (17). However, with transitive verbs -mala is used only with objects (18); a plural subject has to be expressed by the full plural pronoun (19). (17)

an-na-mala 1-go-PL ‘we go’

(18)

an-tai-mala 1-see-PL ‘he sees us’

(19)

pemal takke see 2.PL ‘you (pl.) see him’ (Holmer 1946: 192).

According to Holmer (1947), the person prefixes find themselves in a process of grammaticalization, concretely cohesion; as a result of which their “prefix character is not always apparent” (Holmer 1947: 127). In fact, the discontinuous nature of the plural marker hints at some degree of cohesion between the person markers and the verbal stem, which can be taken as a symptom of, at least, cliticization; the person markers also tend to cliticize to postpositions (an-ka 1.SG-for, ‘for me’). In Llerena (2000: 66) there are examples providing phonological evidence of cliticization by a rule of vowel fusion: toa pe amie [to˘abami˘e], who 2.SG look for, ‘who are you looking for?’, or pe ittoe-sa [bitto˘sa], 2.SG hear-PAST ‘you heard’.7 Elsewhere in the grammar of Cuna, however, the free form status of the person markers is unquestionable, as in (20) and (21). Thus, the person markers in Cuna are still intermediate in their status as free or bound. A sentence like (22) can therefore be regarded as an instance of SOV or s-o-V. (20)

Ti pe an-kala water 2.SG 1.SG-to ‘Bring me water!’

se-take! bring-?

The grammaticalization of agreement in Chibchan 131

(21)

An-kine okop attursae-sa-mala. coconuts steal-PAST-PL 1.SG-LOC ‘They stole my coconuts.’ (Llerena 2000: 64)

(22)

an-pe-sape 1.SG-2.SG-love ‘I love you.’ (Holmer 1947: 126)

3.5. Comparative remarks The four languages just discussed share one typological feature: all four have two basic word order patterns whose use is determined by the discourse: SOV for discourse onset and participant activation, and OVS (Teribe and Boruca) and OSV (Rama) for running discourse (Cuna keeps SOV); both the OSV and the OVS orders have in common that the subject is removed from sentence-initial position because of its information structure status (given information, active participants). However, they differ in terms of the degree of grammaticalization of the identification strategies. These languages can be arranged in a continuum of grammaticalization of agreement, where each of them represents a particular stage in the process. The three stages can be characterized in terms of three parameters discussed above for each language, and advanced by Lehmann (1995): (a)

degree of syntactic cohesion (bondedness) of person markers;

(b) degree of formal distinguishability of core relations (paradigmaticity); (c)

degree of obligatoriness of person markers (syntacticization).

The behavior of the three languages in terms of these parameters is summarized in Table 3:

132 J. Diego Quesada Table 3. Agreement and participant identification in four Chibchan languages Rama

Teribe

Boruca

Cuna

(a)

highly head-marking

partially head-marking

incipient head-marking

incipient head-marking

(b)

full distinguishability

partial distinguishability

∅ distinguishability

∅ distinguishability

(c)

identification in verb

identification in verb/text

identification in verb/text

still relatively free word order

more grammaticalized

less grammaticalized

According to Table 3, Rama constitutes a case of higher and Boruca one of lower grammaticalization of agreement, while Teribe falls half-way between the two, with Cuna only starting. In terms of the main point of this paper, the lesson to be drawn is that the Chibchan languages provide living evidence of an agreement pattern characterized by partial syntacticization of participant identification and still somewhat dependent on discourse needs. The developments in Rama, Boruca, Teribe, and Cuna are more than just the acquisition of person morphology. They are to be regarded as the enhancement of grammatical relations, especially that of subject. Thus, while Cuna is developing subject and object agreement prefixes in keeping with both the basic order s-o-V and the agglutinating nature of the language, and Boruca is developing subject prefixes,8 Teribe has developed subject suffixes, and a pronominal paradigm for non-subjects. By developing headmarking patterns, grammatical relations, especially that of subject, are reinforced. The grammaticalization of agreement of these four Chibchan languages evidences a variety of ways to accomplish the same goal despite structural heterogeneity. The current on-going process (with the exception of Boruca) certainly provides evidence in support of Lehmann’s parameters of grammaticalization (bondedness, paradigmaticity and syntacticization); what is interesting is the fact that aspects such as word order seem to line up with the process; whether this can be taken as teleology in language change is beyond this presentation.

The grammaticalization of agreement in Chibchan 133

4. Conclusion From the perspective of both Chibchan comparative grammar and linguistic typology the paths that languages from the same family can follow to accomplish the same goal are rather significant. In addition, the admittedly quick look at how agreement is being developed in the Chibchan languages of Central America has provided evidence in favor of Lehmann’s well-established framework. The only discrepancy between Lehmann’s framework and the Chibchan case perhaps concerns the source of the agreement marker ang in Boruca; this erstwhile focus marker in Boruca moved along the topicalityfocality continuum through a complex interaction with word order and ended up practically as a subject agreement marker. This is certainly not predicted by the framework; however, insofar as the focus marker underwent a stage of anaphora, this intermediate stage does tend to confirm the path.

Notes 1.

2. 3. 4. 5.

6.

7.

8.

Two notable exceptions are worth mentioning here, namely Tunebo in Colombia (with no agreement) and Paya and Guatuso in Central America (with agreement). There is no indication in the source about the conditioning factor in the use of all the allomorphs listed in (8b). A detailed description of the distribution and function of these three markers goes beyond the scope of this paper. See Quesada (2001). a?r is a phonetically conditioned variant of at. ki has other functions in the language that include co-presence with demonstratives, relative clauses, proper names, and pronouns. Going into details of these functions exceeds the scope of this presentation; see Quesada (2001). The accent mark in ki represents high tone (Boruca has two tones, high and low, unmarked). There are rules of tone assimilation and placement; one of them is at work in (16). See Quesada-Pacheco (1995) for details on this phenomenon of Boruca phonology. This situation is reminiscent of the phonology of some Romance clitics; e.g. Sp. le digo [»li.Fo] ‘I tell him’, la agarró [la.Fa.»ro] ‘he grabbed her’; or French je le dit [ZOl.»di] ‘I told him’. Actually, Boruca will go no further than this, as the language has been declared extinct in 2003.

134 J. Diego Quesada

References CIDCA 1990

Rama Kuup. Gramática rama. Managua: Centro de Información y Documentación de la Costa Atlántica. Constenla, Adolfo, and Espíritu Santo Maroto 1986 Leyendas y tradiciones borucas. San José: Editorial de la Universidad de Costa Rica. Holmer, Nils 1946 Outline of Cuna Grammar. International Journal of American Linguistics 12 (4): 185–197. 1947 Critical and Comparative Grammar of the Cuna Language. Göteborg: Elandes Boktryckeri Aktiebolag. Lehmann, Christian 1982 Universal and typological aspects of agreement. In Apprehension: das sprachliche Erfassen von Gegenständen, Hansjakob Seiler and Franz J. Stachowiak (eds.), Vol. 2, 201–267. Tübingen: Narr. 1995 Thoughts on Grammaticalization. Munich: Lincom-Europa. Llerena, Rito 2000 Elementos de gramática y fonología de la lengua cuna. In Lenguas Indígenas de Colombia. Una visión descriptiva, González and Rodríguez (eds.), 59–72. Bogotá: Instituto Caro y Cuervo Quesada, J. Diego 1995 Case, agreement and topicality in Spanish. Orbis 38: 2–15. 2001 En route to subject-coding: Evidence from Boruca. Theoretical Linguistics 27: 55–86. 2007 The Chibchan Languages. Cartago: Editorial Tecnológica. Quesada-Pacheco, Miguel A. 1995 Hablemos boruca. San José: Ministerio de Educación Pública.

Decay and loss of applicatives in Siouan languages: A grammaticalization perspective Johannes Helmbrecht

Grammaticalization theory investigates the various ways in which lexical morphemes develop into grammatical morphemes. These processes are considered to be unidirectional. They involve the increasing loss of semantic and structural autonomy of the affected lexical unit. Lexicalization, on the other hand, is a reduction process too, but has been received much less attention than the study of the emergence of grammatical forms. Morphological and syntactic constructions loose their internal compositionality and become lexical units no matter, whether they can be segmented formally into smaller sub-units or not. Lexicalization may lead to the disappearance of a formerly productive morphological operation. The present paper will show that grammaticalization and lexicalization are connected to each other historically at least with regard to the fate and destiny of the applicative markers in Siouan languages. After the emergence of applicative prefixes in pre-Proto-Siouan time, they have begun to undergo lexicalization with the verbal hosts, which lead to their entire disappearance in at least some of the Siouan languages. This process and the intermediate stages will be dealt with in this paper.

1. Introduction Application is defined as a morphological operation on the verb that increases or changes the argument structure of the verbal predicate. By means of a verbal affix, a peripheral participant (that may otherwise be expressed as indirect object or as clausal adjunct) is promoted to direct object function. The operation of application may apply to intransitive, transitive, and ditransitive verb bases with different results. The application of an intransitive verb base increases the number of arguments of the verb resulting in a bivalent verb with a transitive subject and a direct object; compare the example from Hočank, a Siouan language of the Mississippi Valley group still spoken in Wisconsin1, in (1a-c).

136 Johannes Helmbrecht (1) HOČ

a. b. c.

kąąné ‘to fall over’, ‘to topple over’ ha-kąné ‘to fall on sth.’ hanį́kąne ‘I fall on you’ /ha-nįį-kąąne/ APPL.SUPESS-1&2-fall.over

Kąąné ‘to fall over’ is an intransitive inactive verb, which becomes a bivalent verb after the addition of a prefix ha-. This applicative marker traditionally termed preverb raises a GOAL participant in direct object function. Hočank is an active/inactive language with regard to the coding of the main syntactic relations, hence the notions A (=actor corresponding to transitive subject) and U (=undergoer corresponding to direct object) are employed. The application of a transitive verb base may either increase the number of arguments to the effect that the resulting verb is a trivalent verb, or the argument structure of the verb base is rearranged to the effect that a peripheral participant is promoted to direct object (DO), and the former direct object is demoted to an oblique object (OBL) or dropped entirely from the clause. The latter case can be illustrated with the German be- derivation, cf. (2a-c) (cf. also Comrie 1985; Lehmann and Verhoeven 2005). (2) GER

a. b. c.

Hans pflanzt [Bäume] DO [im Garten] OBL ‘Hans plants trees in the garden.’ Hans be-pflanzt [den Garten] DO [mit Bäumen] OBL ‘Hans plants the garden with trees.’ Hans be-pflanzt [den Garten] DO. ‘Hans plants the garden.’

The oblique object (OBL) with the semantic role GOAL in (2a) is promoted to direct object function (DO) in (2b). The former direct object Bäume ‘trees’ is demoted to oblique object, which is optional, i.e. it can be dropped from the clause as can be seen in (2c). Application with transitive verb bases can be found also in Hočank, cf. (3a-c). The transitive verb gúuč ‘to shoot sth./so.’ can take an instrumental applicative prefix which adds a third (instrumental) argument to the argument frame of the verb base. The former direct object hų́ųč is still pronominally marked on the verb and not demoted or dropped. The 3SG.U appears as zero in example (3c). The resulting derived verb is a trivalent verb.

Decay and loss of applicatives in Siouan languages 137

(3) HOČ

a. b. c.

gúuč ‘to shoot so./sth.’ hi-gúč ‘to shoot so./sth. with sth.’ hųųčrá wiižúk wawigúčire /hųųč-rá wiižúk wa-hi-Ø-gúuč-ire/ bear-DEF rifle 3PL.U-APPL.INST-3SG.U-shoot-3PL.SBJ ‘They shot the bears with a rifle’

The application of a ditransitive verb base usually does not lead to an increase in arguments, but rather to a rearrangement of the argument structure. In German, for instance, the indirect object, i.e. the RECIPIENT of a gift, is promoted to direct object function, and the former direct object, i.e. the gift, is demoted to oblique object; cf. (4a-b). (4) GER

a. b.

Peter schenkt [seinem Vater] IO [Bücher] DO. ‘Peter gives books (as gift) to his father.’ Peter be-schenkt [seinen Vater] DO [mit Büchern] OBL ‘Peter gives his father books (as gift).’

The general semantic effect of application is that the peripheral participant brought into direct object/undergoer function is presented as much more or completely affected by the action or event designated by the verb. Application may also have discourse-pragmatic functions such as foregrounding of the applied participant. Despite the similarities between verbal application in German and Hočank, there are also differences that are significant from a typological perspective. The German applicative prefix be- is not sensitive to the semantic role of the applied participant. There is just one marker for participants playing different semantic roles such as GOAL, LOC, REC and perhaps others. Another instance of such a one-type applicative marker (cf. Peterson 1999) can be found in Swahili. The one-type applicative marker contrasts with multi-type applicative markers, i.e. applicative markers that distinguish the semantic role of the applied participant. Hočank applicatives belong to this type, since there are four different applicative markers which all indicate different semantic roles of the undergoer: ha-/a- = ‘on’, ‘upon’, ‘over’ (cf. (1b-c)), hi-/i- = ‘with’, ‘by means of’, ‘about’ (cf. (3b-c)), ho-/o- = ‘in’, ‘into’, and gi- = REC, BEN, GOAL, SOURCE. The present paper will focus mainly on the three applicative markers ha-/a- = ‘on’, ‘upon’, ‘over’ (glossed APPL.SUPESS), hi-/i- = ‘with’, ‘by

138 Johannes Helmbrecht means of’, ‘about’ (glossed APPL.INST)), and ho-/o- = ‘in’, ‘into’ (glossed APPL.INESS), which can be found historically in all Siouan languages and which are considered by Siouanists as reflexes of the Proto-Siouan applicative markers *aa-, *oo-, and *ii-, respectively (cf. Carter, Jones, and Rankin in prep.). Rankin, Carter, and Jones (n.d.) summarize their findings as follows. There is fairly good evidence that the three so-called locative prefixes, a‘on, at’, o- ‘in’, i- ‘toward’, as well as the prefix of instrument, í-, were in fact accented long vowels in Proto-Siouan or pre-Proto-Siouan and so probably distinct roots or proclitics, not prefixes, but there is little else that can be said about them at the moment.

Note that the fourth Proto-Siouan applicative prefix *ii- ‘to, toward’, which was mentioned by Rankin, Carter, and Jones in the citation, has been discerned in only a few Siouan languages (cf. Helmbrecht 2006). Its status as a Proto-Siouan form is therefore not firm. Of course, it may be the case that the directional *ii- and the instrumental *ii- merged in the majority of the descendants of Proto-Siouan because of the homophony. Since the three applicative markers are so deeply entrenched and established in the grammar of Siouan, and since they can be traced back to Proto- or even pre-Proto-Siouan, not much can be said about the origins of these forms. However, the process of grammaticalization comprises not only the development of grammatical markers from lexical items within certain constructions, but also the gradual decay of them in later phases of the development. The present paper will focus on the latter part of this process thus contributing to a largely neglected area in the research of grammaticalization. Decay and loss of grammatical forms usually do not figure as interesting topics as such in research on grammaticalization. In particular, it will be shown that the process of the disappearance of applicative markers can be described as a process of lexicalization. Section 2 presents a brief theoretical outline of the grammaticalization and eventually the lexicalization of applicative markers in general. Section 3 summarizes the synchronic properties of the Proto-Siouan applicative markers in the contemporary descendent languages on the syntagmatic and the paradigmatic axis of analysis. The paper ends with some conclusions in section 4.

Decay and loss of applicatives in Siouan languages 139

2. The grammaticalization and lexicalization of applicatives

Grammaticalization

In order to set the background for the investigation of the decay of applicatives in Siouan languages in section 3, a brief outline of the grammaticalization and finally lexicalization of applicatives seems to be useful. Applicatives/preverbs derive historically from relational nouns or from transitive verbs by passing different stages and constructions. The respective grammaticalization chains are summarized in a very rough manner in Figure 1. NOUNREL

coverbs/converbs

adpositions

oblique case markers Lexicalization

VERBTRANSITIVE

applicatives/preverbs

verbs

Figure 1. From relational noun/transitive verb to applicative and finally verb again

Transitive verbs of certain semantic classes may develop into coverbs in serial verb constructions, or converbs, and may finally become adpositions or applicatives, or both. The possible verbal sources for adpositions and/or applicatives are transitive verbs with meanings such as ‘give’ (BEN), ‘be together’ (COM), ‘use’ (INSTR), ‘arrive at’ (GOAL), and so on (cf. Paul 1982; Lehmann 1995: 104–107; Heine and Kuteva (eds.) 2002 and the corresponding entries therein). These source verbs may become adpositions via a stage of being co(n)verbs. As co(n)verbs, they share an argument with the main clause, the subject, and they are restricted with respect to the independent choice

140 Johannes Helmbrecht of tense and mood categories of the verb. Usually, they are not personally inflected. Adpositions loose the ability to take any verbal inflection. However, many languages have personally inflected adpositions crossreferencing its NP complement (e.g. Arabic). If the co(n)verbs become adpositions, the former DO argument of the co(n)verb becomes the complement of the adposition. The former SBJ argument disappears – it changes to an argument slot for the core of the clause it is modifying. Co(n)verbs may also become applicatives/preverbs. This process presupposes that the co(n)verb with its DO argument NP is placed in the direct vicinity of the main verb, i.e. either directly before or directly after the main verb so that they can form a verbal compound. The original co(n)verb more and more merges with the main verb and eventually becomes fused with it. During this process, the co(n)verb looses its subject argument slot and the ability to take any verbal inflection (except of person marking, if this is an inflectional category of verbs in language X at all). The DO argument slot is integrated into the argument frame of the main verb. This may result either in an increase of the argument structure of the main verb (sometimes leading to double-ACC constructions), or it leads to a demotion of the original DO slot in favor of the new one. The principal stages are roughly summarized in Figure 2. Word order does, of course, play an important role in this grammaticalization process. In Figure 2, it is assumed that the language is verb final, i.e. a SOV language in terms of a constituent order typology. [ SBJ-NPi (DO-NP) [SBJ-Øi DO-NP + CON-V2 ] V1]Clause [ SBJ-NP OBL-NP DO-NP APPL+V1]Clause or [ SBJ-NP DO-NP APPL+V1]Clause Figure 2. Stages of grammaticalization from verb to preverb/applicative

The process of the development of adpositions and applicatives/preverbs from a common source may be simultaneous. Relational nouns are another potential lexical source for applicatives/preverbs. They designate e.g. body parts, kinship relations, and spatial relations. They usually have two argument slots in their semantic representation. These semantic relations may be grammatically expressed by means of a, for instance, genitive construction with a possessor NP ([NP2] in (5)) and a NP ([NP1] in (5)) with the referent to be localized.

Decay and loss of applicatives in Siouan languages 141

(5)

[NP1] is [on top] [of NP2]

The relational noun top in (5) indicates the local region with regard to which the referent of the first noun phrase ([NP1] in (5)) is localized. The spatial relation between NP1 and NP2 is expressed by the relational noun top. Often, the genitive construction with a spatial relational noun appears with other elements additionally providing information on the relation between NP1 and NP2. These elements may be local adpositions, local adverbs, or coverbs. It may also be the case that the relational noun – as head of the construction – receives an oblique case marker that indicates the relation of the entire genitive construction to the subject or the core of the clause (outer case marker). In the course of the development from relational noun to adposition, the relational noun looses its nominal properties. It increasingly looses the ability to take nominal modifiers and determiners. Concomitant to this process is a fusion of the original relational noun with the elements that indicate an additional local or other relation such as preposition, adverb or coverb. Once, the ex-relational noun has become an adposition, this free form may gradually undergo a univerbation with the main verb of the clause resulting in a preverb/applicative marker of the base verb. The other possibility is that the adposition may develop further to an oblique case marker (cf. Figure 1), but this path won’t be considered here further. It can be expected that newly grammaticalized applicative markers, no matter whether they stem from adpositions of verbal or nominal origin, still show the semantics of the former adposition. It may even be the case that adpositions and the corresponding applicative markers are both in use in language X with perhaps only slight phonological differences. It can also be expected that the verbs which take the applicative markers still occur in a derived (with applicative marker) and in an underived (without applicative marker) form in the lexicon of the language, i.e. it can be hypothesized that this morphological operation tends to be quite productive within the set of verbs which take the applicative markers. In addition, the semantics of the application in such a phase of the grammaticalization chain can be expected to be quite compositional and transparent. Evidence for these semantic and distributional properties of newly grammaticalized applicatives can be found, for instance, in Rama, a Chibchan language of Meso-America. Rama has a large set of simple and complex postpositions that show slight phonological differences, if they appear preverbally or postverbally (cf. Craig and Hale 1988; Grinevald n.d.).

142 Johannes Helmbrecht Their origin is not entirely clear. It seems that they are grammaticalized from different sources, because some of them take the subject pronouns as complements, others take object or possessive pronouns as complements. Some of these postpositions are further grammaticalized to applicative markers with only minor phonological and semantic modifications. Interestingly, the speakers of Rama have a choice between the applicative marker, i.e. the applied verb construction, and the postpositional strategy, i.e. the usage of the corresponding postposition. Both syntactic participant binding strategies can be chosen in principal freely, but in fact, text counts reveal that the application strategy is mainly used, if the complement (DO) NP is not overtly marked, i.e. under the condition of zero anaphor. Pronouns and zero anaphora are used, if the anaphorically referred participant is topical, i.e. available from the previous discourse context. Based on this observation in Rama, it can be hypothesized that application of verbs and the concomitant raising of participants in DO position has the pragmatic function to increase the topicality of this participant. What are the next steps in the grammaticalization of applicative markers? It can be hypothesized that the ongoing grammaticalization process leads to a lexicalization of the applied verbs. Lexicalization here means that the original compositional or analytic structure of the applied verb looses its internal morphological and semantic compositional structure in favor of a rather holistic lexical item (cf. Lehmann 2002: 15). This includes the following: a) the applicative markers become fossilized as an integral parts of the verb stem. b) The number of opposite pairs between derived and plain verbs decreases constantly in the lexicon (in favor of the derived form). c) The applicative markers more and more undergo morphonological interactions with other morphological material (stem and other affixes) adjacent to the applicative marker. Finally d) the semantics of the applicative marker disappears more and more to the effect that the former applied verb retains the resulting argument structure, but that traces of the original meaning of the applicative marker are lost or are at least no longer recognizable. These developments belong to the lower part of Figure 1. So, with regard to applicative markers, grammaticalization gives way to lexicalization. Developments which resemble the hypothesized ones in a remarkable and quite specific Siouan way can be observed with regard to the three applicative markers (*aa- ‘on’, *oo- ‘in’, *ii- ‘with’) in Siouan languages mentioned above. These observations will be summarized in the following section.

Decay and loss of applicatives in Siouan languages 143

3. The gradual lexicalization of Siouan applicatives 3.1. The Siouan languages The Siouan language family comprises approximately 17 languages, which are grouped as shown in Table 1. The individual languages or dialect continua are given in small letters, capital letters represent sub-groups. There are three main groups of Siouan languages: the so called Missouri River group, the Northwestern branch of the Siouan languages including Mandan, the Mississippi Valley group, which itself is subdivided into three different groups, and the Ohio Valley group, the Southeastern branch of the Siouan languages. Table 1. Siouan languages and their sub-groups CATAWBA

Catawba Waccon MISSOURI RIVER Mandan

Crow Hidatsa

DAKOTAN

SIOUAN

MISSISSIPPI VALLEY

CHIWEREWINNEBAGO

DHEGIHA

OHIO VALLEY

Sioux Assiniboine Stoney Chiwere (Otoe, Missouri, Iowa) Hočank/Winnebago Omaha-Ponca Osage Kansa Quapaw

Ofo Biloxi Tutelo

The geographic names of the three main groups should not be taken too literally, but there is roughly a northwest to southeast cline observable in this linguistic grouping. Catawba/Waccon are extinct languages considered to be the languages which split off first from Common Siouan.2

144 Johannes Helmbrecht As was indicated above, all three applicative markers can be identified in all Siouan languages in Table 1 except Catawba/Waccon by means of the historical-comparative method. However, the semantic and formal properties of these forms vary significantly in these languages both in paradigmatic and syntagmatic perspective.

3.2. The syntagmatic properties Siouan languages have a quite complex verbal morphology. There are at least six morphological positions preceding the verbal root, which can be discerned more or less in most of the Siouan languages, although there is much variation in detail. These prefix positions include - moving from right to left in Table 2 - instrumental prefixes (-1) in the first slot, reflexive, reciprocal, reflexive possessive and benefactive applicative in slot (-2), actor prefixes and undergoer prefixes in slot (-3) and (-4), the three applicatives in slot (-5) and the indefinite direct object marker in slot (-6). Table 2. Relative order of verbal prefixes in Siouan languages -6

-5

-4

-3

INDEF OBJ

APPL.INST APPL.SUPESS APPL.INESS

undergoer prefixes

actor prefixes

-2 APPL.BEN reflexive reciprocal reflexive possessive

-1

0

instrum. prefixes

root

As can be seen in slot -2 in Table 2, there is a fourth applicative marker in many of the Siouan languages. This marker has very different semantic and structural properties. This morpheme is hypothesized to stem from a verb of motion adding a benefactive or recipient to the argument frame of the verb. However, the APPL.BEN markers won’t be considered further in this paper. What is of concern here is that the applicative markers in slot -5 generally precede the two series of pronominal prefixes, the undergoer prefixes and the actor prefixes which cross-reference the main arguments of the clause. From a typological point of view, this order of morphemes is rather rare and dispreferred, since derivational affixes such as applicative affixes are usually positioned closer to the verb stem than inflectional affixes.

Decay and loss of applicatives in Siouan languages 145

In the process of the increasing lexicalization of the applied verbs and the concomitant fossilization of the applicative markers, the pronominal prefixes get trapped within the stem becoming infixes, a rather rare and dispreferred morphological structure as well3. Hočank and other Mississippi Valley Siouan languages display this structure. Some other Siouan languages mainly of the Northwestern branch and of the Southeastern branch, however, show attempts to get rid of this situation. In Hidatsa, one of the Northwestern Siouan languages, some instances of a metathesis between the applicative markers and the pronominal affixes are attested (John Boyle p.c. and in prep.) in the sense that the applicative markers changed their morphological position to get closer to the verb stem. In particular, the pronominal prefixes precede the a- applicative (APPL.SUPESS) in some verbs now. In adjacent Mandan, however, all personal prefixes precede the a- applicative (APPL.SUPESS). The Mississippi Valley Siouan languages preserved the Common or Proto-Siouan order of applicatives and pronominal prefixes. In Hočank, for instance, there are no exceptions found in the lexicon of the language. It should be mentioned, that many Siouan languages innovated some kind of a 1PL.A and 1PL.U form. In Hočank, it is a 1I.DU actor (hį-) and undergoer form (wąąga-). Both pronominal affixes precede the applicative markers in Hočank. In Osage, the instrumental applicative i-(APPL.INST) precedes the 1PL actor form ąk-, the other locative applicatives a(APPL.SUPESS), and o-(APPL.INESS) follow. The order is as follows: [INDEF.OBJ-INSTR.APPL-1PL.A-APPL.SUPESS/INESS-U-AG]. The languages of the Southeastern branch are much more advanced in changing the order of applicatives and pronominal affixes than the Northwestern languages. In Ofo, the instrumental applicative i- (APPL.INST) and the inessive applicative u-/o- (APPL.INESS) follow the actor and the undergoer pronominal prefixes. The position of the superessive applicative a- (APPL.SUPESS) varies. It can be found before or after the U/A pronominals [(a-)-U/A-(a-)/o-/i-]. In Biloxi, all locatives follow the U/A pronominals, i.e. here, a complete metathesis of the entire paradigm has occurred with regard to the morphological position. Interestingly, this process corresponds to a complete fossilization of the applicative markers on the paradigmatic axis; cf. section 3.3 below. Newly grammaticalized applicative markers can be expected to show no or only a few formal modifications with regard to their phonetic or phonological shape. This is indeed the case with the applicative markers in Rama mentioned above. Applicative markers which are deeply entrenched in the

146 Johannes Helmbrecht grammar of a language are expected to undergo more morphonological processes with respect to the adjacent morphological material. This can be observed with applicative markers in Siouan. All Siouan languages show heavy morphophonemic contractions and interactions between the applicative markers and the adjacent prefixes to the right and to the left in the morphological structure. This will be briefly exemplified with a few examples from Hočank; cf. (6a-c). (6) HOČ

a.

b.

c.

hapé ‘to wait for so.’ hįįpéwi ‘we (INCL) wait for someone’ /hį-a-pé-wi/ 1D.I.A-APPL.SUPESS-wait.for-PL hogirák ‘to tell so. sth.’ waagítak ‘I tell him/someone’ /ho-a-gí-tak/ APPL.INESS-1SG.A-APPL.BEN-1SG.A.tell hogirák ‘to tell so. sth.’ hųųgírak ‘he/someone tells me’ /ho-į-gí-rak/ APPL.INESS-1SG.U-APPL.BEN-3SG.A.tell

In (6a), the vowel is entirely assimilated to the preceding pronominal affix resulting in a lengthening. In (6b), the vowel of the applicative marker is assimilated to the vowel of the pronominal prefix to the right. The /h/ in the APPL.INESS ho- is a phonologically conditioned prothetic consonant and an innovation in Hočank. It appears only in word initial position not only with the applicative markers, but also with pronominal affixes, if they happen to appear in word initial position. In (6c), the combination of the APPL.INESS ho- and the 1SG undergoer prefix į- results in a completely different vowel huu-. It should be mentioned that all three example words in (6a-c) are lexicalized, i.e. the respective applicative marker is synchronically a part of the verb stem. There is no verb or even verb stem *-pe in Hočank ((6a). In addition, there is no verb stem *-girak or *-rak in the lexicon of Hočank ((6b-c). The only other form which exists in the lexicon is horák ‘to tell sth.’ without the benefactive applicative gi- (APPL.BEN).

Decay and loss of applicatives in Siouan languages 147

3.3. The paradigmatic properties Some paradigmatic correlates of the continuing grammaticalization of the applicative markers are the gradual loss of their productivity, the reduction of the paradigm itself, and the increasing desemanticization of the forms. Loss of productivity means that the applicative markers are no longer able to derive new verb stems in the lexicon of the language. To the best of my knowledge, this morphological parameter hasn’t been investigated empirically yet, even in the better describes Siouan languages. And of course, for the extinct ones such investigations are even no longer possible. Thus, only impressionistic statements can be made for a few languages. In Hidatsa, the a- and o- applicatives are no longer productive, this holds also for Crow. However, the instrumental applicative i- is productive in both languages to some degree. In Hočank, the applicatives a- and o- are productive to a quite moderate degree, i.e. they may be productively used with verbs of bodily position or rest (but not with the positionals), with motion verbs (except the group of deictic motion verbs), and with verbs of movement of objects to some place. The instrumental applicative i- is comparatively more productive, since it may be used to derive new stems with all activity verbs that allow semantically and pragmatically an instrument to perform the action, which is designated. It can be assumed that we would find similar results in Dakotan and Dhegiha languages. In Ofo (Rankin 2004), none of the three applicatives seems to be productive and the same is probably true for Biloxi (Einaudi 1976). Unfortunately, these claims can no longer be tested since both languages are extinct. There is of course a close correlation between productivity and the degree to which these derivational prefixes are lexicalized or fossilized in the individual Siouan languages. One would expect that the lower the degree of productivity the higher the number of lexicalized and/or fossilized applicative markers. And, indeed this correlation seems to hold. The vast majority of occurrences of the three applicative prefixes – the forms are easily identifiable in most cases – are more or less lexicalized. This means, for instance, that we do not have paradigmatic pairs of derived versus underived verbal stems like the ones in (6a-c) in Hočank. Mostly, there does not exist an underived verbal stem as counterpart to the applied stem. The underived stems/roots can be identified only with regard to other derivations that occur in the lexicon of the language.

148 Johannes Helmbrecht Concomitant to the loss of paradigmatic variability and lexicalization of these markers is the increasing desemanticization of their meaning. In the majority of cases the proposed semantics of the applicatives is hard to identify. In all these cases, the applicative markers are not morphemes, but submorphemic parts of the stem (ISC ‘initial stem components’ in the terminology of Helmbrecht and Lehmann 2005). Again, it is difficult to make empirically firm statements about the varying degree of lexicalization of these applicatives in Siouan languages. To my knowledge, no one has ever empirically checked this parameter with a list of cognate verbs in the different branches of Siouan. However, there exist grammatical descriptions and sometimes dictionaries for many Siouan languages and they have been consulted for the following claims. In Hidatsa, the applicatives a- and o- are semantically less transparent and more bleached than the instrumental applicative i-. Obviously, the high degree of lexicalization of these two forms was the reason that they weren’t even mentioned as such in older descriptive sources such as Robinett (1955b: 160) and Matthews (1965: 58–60). The same seems to hold for Crow, too. Older sources such as Lowie (1941) and Kaschube (1967) did not even mention the o- applicative, but both have the instrumental applicative i- in their grammatical descriptions. The situation in Hočank is similar. The cases in which there exist pairs of derived and underived stems are the exceptions (such as the verbs of bodily position or rest, motion verbs, and with verbs of movement of objects mentioned above) in the Hočank lexicon. Furthermore, the number of semantically opaque occurrence of the three applicative markers in the lexicon outnumbers significantly the number of semantically transparent (compositional) occurrences; cf. the figures in Table 3.

Decay and loss of applicatives in Siouan languages 149 Table 3. Degree of semantic transparency of applicative verbs in the Hočank lexicon Verbs with ha500 Verbs with hi614 Verbs with ho566

Verbs with ‘on’ in their English glosses 42 (8,4%) Verbs with ‘with’ in their English glosses 34 (5,5%) Verbs with ‘in’/‘into’ in their English glosses4 48 (8,4%) ‘in’/ 31 (5,4%) ‘into’

Table 3 presents the results of a corpus-linguistic search in the lexical database produced by the “DOBES project Documentation of the Hočank language” (cf. Helmbrecht and Lehmann (eds.) in prep.). The database contains approximately 7000 entries. Of these, there are 500 verbs with a word initial ha- which is the Hočank reflex of the Proto-Siouan applicative marker *aa-. However, only 42 i.e. 8,4% of these verbs have an English gloss/translation, which reflects the original semantics of the applicative marker ‘on’. Very similar results can be found for the other applicative markers hi- (APPL.INST) and ho- (APPL.INESS). These figures show that the words presented above in (6a-c) indeed represent rather the norm than the exception. The ha- in hapé ‘to wait for so.’ has no discernable meaning and the same holds for ho- in hogirák ‘to tell so. sth.’ and horák ‘to tell sth.’. A similar lexical investigation with regard to Kansa reveals that the ratio between the number of verbal derivations with the applicatives and semantically transparent derivations is slightly higher than in Hočank. The same holds for Quapaw5. Both languages belong to the Mississippi Valley group as well. However, in the Southeastern branch of Siouan, the process of the lexicalization of the applicative constructions (i.e. the fossilization of the applicative markers) has almost ended. In Ofo, there are no semantically transparent derivations with the three applicative markers, and in Biloxi, verbal stems (historically derived from an applicative construction) with the expected meanings occur only sporadically6.

150 Johannes Helmbrecht 4. Conclusions Research in the grammaticalization of applicatives so far has concentrated on the ways in which applicative markers (verbal affixes) arise out of certain transitive verbs and certain relational nouns via a stage of adpositions. However, the further fate of applicatives, i.e. their decay and loss, has been largely neglected. The present study, therefore, concentrates on this process and describes the changing grammatical and semantic properties of applicatives, which are on the way to get lost, in terms of the parameters of grammaticalization theory (cf. Lehmann 1995). The loss of productivity and compositionality of applicative markers in Siouan was described as a process of lexicalization. The internal structure of applied verbs has been disappeared in the Siouan languages to various degrees. The Proto-Siouan grammaticalization of applicative prefixes (presumably out of certain verbs) gives way to the lexicalization of these constructions. Here, lexicalization is the continuation of grammaticalization. The three Siouan applicatives with the adpositional meanings ‘on’, ‘in’, and ‘with’ are Common Siouan, but show significant differences among the contemporary Siouan languages with regard to their morphological properties. The general tendency is that the applicative prefixes in Siouan become more and more a part of the verb stem thus loosing entirely their formal and semantic integrity, paradigmaticity, and paradigmatic variability. Reflexes of this process are the following: (a) There are less and less morphological opposites between derived and underived/plain verb stems (in favor of the derived stems). In the Southeastern branch of Siouan, these applicatives lost their status as grammatical prefixes entirely. (b) The productivity of these applicatives is reduced very much and goes toward zero in the Southeastern Siouan languages. There, they are not productive any more and can be identified only by historical comparison. (c) Because of the high degree of lexicalization, the majority of derived verbs do not show any traces of the original adpositional meaning of the applicative markers. Thus, a high degree of lexicalization goes hand in hand with a strong degree of desemanticization. (d) The emergence of the applicatives in Siouan produced a typologically highly dispreferred morphological structure with inflectional categories (pronominal inflection of actor and undergoer) closer to the verb stem than the derivational categories (applicative prefixes), a structure, which got even worse in the process of the lexicalization of the applicative markers resulting in infixes. The peripheral branches of Siouan, the Northwestern and the Southeastern Siouan languages show

Decay and loss of applicatives in Siouan languages 151

attempts to overcome this situation by means of shifting the inflectional paradigms to the left periphery of the morphological chain of prefixes of the verb. (e) All Siouan languages show heavy morphonological interactions between the applicative markers and the adjacent morphological material.

Notes 1.

2.

3. 4. 5.

6.

The Hočank data are taken from either fieldnotes of the author during numerous fieldtrips and from the Hočank corpus produced by the DOBES project funded by the Volkswagen Foundation for the documentation of the Hočank language. This documentation project was led by Christian Lehmann and Johannes Helmbrecht during the years 2003 through the end of 2007. Hočank is one of so many highly endangered languages of North America; cf. the website of the documentation project: http://www.uni-erfurt.de/sprachwissenschaft /Vgl_SW/ Hocank/index_frames.html. Descriptive sources for the subsequent investigation of applicatives in Siouan languages include the following: Boas (1907), Boas and Deloria (1941), Boyle in prep., Buechel (1939), Einaudi (1976), Graczyk in prep., Helmbrecht in prep., Kaschube (1967), Kennard (1936), Koontz (1984), Lowie (1941), Matthews (1965), Mixco (1997), Oliverio (1996), Quintero (2004), Rankin (2002, 2004, 2005a, 2005b), Robinett (1955a, 1955b), Rood and Taylor (1996), Van Valin (1977), Whitman (1947), Wolff (1952a, 1952b). For a more detailed discussion of infixes in Hočank and its implication for morphological theory, see Helmbrecht & Lehmann (2005). There is a partial overlap in these figures, because some lexical entries have two glosses, one with 'in' and one with 'into'. They were counted then twice. Both results are not based on counts, but are rather impressionistic statements based on a rough examination of the respective dictionaries (cf. Rankin 2005a, 2002, 2005b). Since there is no dictionary for Biloxi, the data for this particular search came from the grammatical description by Einaudi (1976).

152 Johannes Helmbrecht

References Boas, Franz 1907 Notes on the Ponka grammar. In: Proceedings of the 15th International Congress of Americanists in Québec 1906, part 2: 317–337. Québec: Dussault/Proulx. Boas, Franz, and Ella Deloria 1941 Dakota Grammar. Memoirs of the National Academy of Sciences 23.2. Washington, DC. Boyle, John in prep. Hidatsa Grammar. Buechel, Eugene 1939 A Grammar of Lakota: The Language of the Teton Sioux Indians. Rosebud, SD: Rosebud Educational Society. Carter, Richard T., A. Wesley Jones, and Robert L. Rankin (eds.) in prep. Comparative Dictionary of the Siouan Languages. Comrie, Bernard 1985 Causative verb formation and other verb-deriving morphology. In Language Typology and Syntactic Description. Vol. III, Grammatical Categories and the Lexicon, Timothy Shopen (ed.), 309–348. Cambridge: Cambridge University Press. Craig, Colette G., and Kenneth L. Hale 1988 Relational preverbs in some languages of the Americas: typological and historical perspectives. Language 64: 312–344. Einaudi, Paula Ferris 1976 A Grammar of Biloxi. New York/London: Garland. Graczyk, Randy in prep. Crow Grammar. Grinevald, Colette n.d. A Grammar of Rama. Report to National Science Foundation BNS 8511156. Heine, Bernd, and Tanja Kuteva (eds.) 2002 World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Helmbrecht, Johannes 2006 Applicatives in Siouan languages - a study in comparative Siouan grammar. (Paper presented on the annual conference on Siouan and Caddoan Linguistics in Billings, MT June 2006). in prep. A Grammar of Hočank (Winnebago). University of Regensburg. Helmbrecht, Johannes, and Christian Lehmann 2005 Hočank’s challenge to morphological theory. To appear in: A World of Many Voices: Lessons from Documented Endangered Languages.

Decay and loss of applicatives in Siouan languages 153 David Rood, David Harrison, and Arienne Dwyer (eds.). Amsterdam: Benjamins. Helmbrecht, Johannes, and Christian Lehmann (eds.) 2006 Hočank-English/English-Hočank Learners’ Dictionary (second edition). In: Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt (ASSidUE) Nr. 21. Erfurt: University of Erfurt. in prep. The Hočank-English/English-Hočank Dictionary. University of Erfurt. Kaschube, Dorothea V. 1967 Structural Elements of the Language of the Crow Indians of Montana. (University of Colorado Studies – Series in Anthropology No. 14). Boulder, CO: University of Colorado Press. Kennard, Edward 1936 Mandan Grammar. International Journal of American Linguistics 9: 1–43. Koontz, John E. 1984 Preliminary sketch of the Omaha-Ponka language. Ph. D. diss., University of Colorado in Boulder, CO. Lehmann, Christian 1995 Thoughts on Grammaticalization. (LINCOM Studies in Theoretical Linguistics, 1), Unterschleissheim/Newcastle: Lincom. 2002 New reflections on grammaticalization and lexicalization. In New Reflections on Grammaticalization, Wischer, Ilse and Diewald, Gabriele (eds.), 1–18. Amsterdam/Philadelphia: Benjamins. Lehmann, Christian, and Elisabeth Verhoeven 2005 Extraversive transitivization in Yucatec Maya and the nature of the applicative. In Case, Valency and Transitivity, Leonid Kulikov, Andrej Malchukov, and Peter de Swart (eds.), 465–493. Amsterdam/Philadelphia: Benjamins. Lowie, Robert H. 1941 The Crow Language. Grammatical Sketch and Analyzed Text. University of California Publications in American Archeology and Ethnology, Vol. 39, No.1, 1–142. Los Angeles/Berkeley: University of California Press. Matthews, G. H. 1965 Hidatsa Syntax. The Hague etc.: Mouton. Mixco, Mauricio 1997 Mandan. (Languages of the World/Materials 159). München, Newcastle: Lincom Europa. Oliverio, Giulia R. M. 1996 Grammar and Dictionary of Tutelo. Lawrence, KS: University of Kansas.

154 Johannes Helmbrecht Paul, Waltraud 1982 Die Koverben im Chinesischen (with an English summary). (Arbeitspapier Nr. 40). Köln: Institut für Sprachwissenschaft der Universität. Peterson, David A. 1999 Discourse-functional, historical, and typological aspects of applicative constructions. Ph. D. diss., University of California at Berkely. Polinsky, Maria 2005 Applicative constructions. In The World Atlas of Language Structures, Martin Haspelmath, Matthew S. Dryer, David Gil, and Bernard Comrie (eds.), 442–446, Oxford: Oxford University Press. Quintero, Carolyn 2004 Osage Grammar. (Studies in the Anthropology of North American Indians.) Lincoln, NE: University of Nebraska Press. Rankin, Robert L. 2002 Quapaw Dictionary. Manuscript, University of Kansas. 2004 An Ofo Grammar Sketch – Based on Materials Collected by John R. Swanton and John R. Swanton’s Ofo-English Dictionary. Manuscript, University of Kansas. 2005a Kansa-English Lexical File. Manuscript, University of Kansas, Department of Linguistics. 2005b Quapaw. In: Native Languages of the Southeastern United States,. Heather Hardy and Janine Scancarelli (eds.), 454–498. Lincoln, NE: University of Nebraska Press. Rankin, Robert L., Richard T. Carter, and A. Wesley Jones n.d. Proto-Siouan Phonology and Grammar. Robinett, F. M. 1955a Hidatsa I: morphophonemics. International Journal of American Linguistics 21: 1–7. 1955b Hidatsa II: affixes. International Journal of American Linguistics 21: 160–177. Rood, David, and Allan R. Taylor 1996 Sketch of Lakhota, a Siouan Language. In Languages, Ives Goddard (ed.), 440–482, Vol. 17 of Sturtevant, William C. (ed.) Handbook of North American Indians. Washington, DC: Smithsonian Institution. Van Valin, Robert D. Jr. 1977 Aspects of Lakhota syntax: A study of Lakhota (Teton Dakota) syntax and its implications for universal grammar. Ph. D. diss., University of California, Berkeley. Whitman, William 1947 Descriptive grammar of Ioway-Otoe. International Journal of American Linguistics 13: 233–248.

Decay and loss of applicatives in Siouan languages 155 Wolff, Hans 1952a Osage I. International Journal of American Linguistics 18: 63–68. 1952b Osage II. International Journal of American Linguistics 18: 231– 237.

Reciprocals in the making: multiple grammaticalization in Manambu Alexandra Y. Aikhenvald

The aim of this paper is to illustrate the pathways of multiple grammaticalization of reciprocal meanings in Manambu, a Ndu language from the Sepik area of New Guinea. The reciprocal-associative marker ‘each other; together’ comes from a grammaticalized noun meaning ‘side’. Reciprocal (but not associative) meanings can also be expressed with verbal directional markers, some of which come from grammaticalized verbs. The directionals with a reciprocal reading refer to strictly defined directions and always involve movement ‘away’ from the speaker or the reference point (which is in itself typologically unusual). Multiple grammaticalization of essentially the same, reciprocal meaning results in the development of additional polysemous structures and multiple partial synonymy.

1. The problem: multiple grammaticalization Grammaticalization presupposes the development of lexical items into grammatical forms, with concomitant phonological, semantic and other changes (see Lehmann 2002: 112–159). This may involve grammaticalization chains, and polygrammaticalization, whereby one lexeme is the source of more than one grammatical item. For instance, the auxiliary ‘be at’ in Ewe got grammaticalized as a preposition and as a present progressive marker (see Heine 1992: 354–355). The reverse can also happen: multiple exponents of essentially the same grammatical category can come from different sources. We propose to call this, hitherto undocumented, phenomenon, ‘multiple grammaticalization’. The aim of this paper is to demonstrate the pathways of multiple grammaticalization of reciprocal meanings in Manambu, a Ndu language from the Sepik area of New Guinea. The inherently polysemous reciprocal-associative marker ‘each other; together’ comes from a grammaticalized noun meaning ‘side’ (section 2). In addition, Manambu is in the process of reinterpreting combinations of verbal directional markers, themselves the product of grammaticalization

158 Alexandra Aikhenvald of compounded verbs, as reciprocals without any associative connotations. The functional motivation behind this clearly involves the necessity for disambiguation (section 3). The directionals with a reciprocal reading are typologically unusual, in that they refer to strictly defined directions and always involve movement ‘away’ from the speaker or the reference point (section 4). As a result of multiple grammaticalizations involving the reciprocal meaning, we are faced with a set of nearly synonymous structures each of which has additional polysemous patterns of its own.

2. Reciprocal and associative in Manambu Manambu, spoken by about 2000 people in the Sepik area of New Guinea (see Aikhenvald 2008), is a highly synthetic, predominantly suffixing and agglutinating language, with a strong tendency towards verb-final constituent order. Its morphological make-up is quite complex. Nouns distinguish two genders, three numbers and nine case forms. Verbs have an array of grammatical categories, including several modalities, aspects and tenses fused with person marking, a complex system of negation, and of clausechaining. The major means of expressing a reciprocal meaning is by using a reciprocal-associative marker awar-wa ‘each other; all together’. This comes from a comitative case-form of the inherently locational noun awar ‘side, sideways direction’. The comitative case in Manambu has numerous meanings, including associative ‘with’ (e.g. wun-a-wa (I-LK-COM) ‘with me’), and perlative ‘along; through’ (e.g. yabə-wa (road+LK-COM) ‘along the road’). The comitative -wa is also used with a few nouns, with the meaning of ‘all together’, e.g. du-a-wa (man-LK-COM) ‘with man; all the men’, təp-a-wa (village-LK-COM) ‘the whole village’. That is, one can hypothesize that the reciprocal meaning of awar-wa is an extension of its more general collective meaning ‘all together’ (this is comparable to a reciprocal extension of the collective marker in Boumaa Fijian (Dixon 1988: 177– 178) and other Oceanic languages). Reciprocals, their functions, polysemy and origins, are among the major foci of typological studies today (recent surveys include König and Gast (eds.) forthcoming; Nedjalkov (ed.) 2007; and Frajzyngier and Curl (eds.) 1999; also see Heine and Kuteva 2002). None of these has so far reported a reciprocal originating from a comitative form of a locational noun.

Reciprocals in the making 159

The reciprocal-associative marker is typically placed immediately before the verb (where one expects an adverb to go). The marker awarwa has lost its erstwhile nominal properties: it cannot take modifiers or trigger gender or number agreement. In other words, it has become reanalyzed as a marker of a purely grammatical category. This is akin to what Lehmann (2002: 114–118) termed ‘desemanticization’ and ‘morphological degeneration’, processes known as indicative of grammaticalization. The marker awarwa forms a word class of its own (see Aikhenvald 2008: Table 4.4). It shares a few syntactic properties with adverbs: it can modify both verbs and nouns. However, it lacks a number of crucial features most adverbs have: for instance, it cannot undergo reduplication or repetition, it cannot occur with the suffix meaning ‘like’, and cannot be used as a copula complement or be modified by the adverb məy meaning ‘real’. Unlike a noun marked with the comitative case, awarwa can modify another noun: awarwa ma:j (REC/ASS story) means ‘story (told) to or about each other; a joint story’. This highlights its inherent polysemy: (1) can have two readings, equally plausible: (1)

awarwa rək kur-na-bran REC/ASS joke make-ACT.FOC-1DU.SBJ.NPST (a) ‘We two joke about each other; we are in a joking relationship’ (b) ‘We two joke together (about someone else)’1

The form awarwa can be used in both meanings with any verb, noun or adjective. Unlike reciprocals expressed with verbal affixes (e.g. Aikhenvald 2007) it has no effect on the verb’s transitivity. The context typically determines whether the reciprocal or the associative reading of awarwa is appropriate. Consider (2): a man was attacked by a crocodile, and saved from imminent death by a group of friends who got together and pulled him out of the crocodile’s mouth. In this context, awarwa lagu- can only mean ‘pull together’ (reading (a)). The reading (b) makes no sense in the given context.

160 Alexandra Aikhenvald (2)

[awarwa lagu-ma:r-kə-da] REC/ASS pull-NEG.SUB-FUT-3PL.SBJ.NPST [akəs war-k-na-d] NEG.IRR come.up-IRR-ACT.FOC-3M.SG.SBJ.NPST (a) ‘If they had not pulled (him) all together, he would have never come up (ashore)’ (b) *‘If they had not pulled each other, he would have never come up (ashore)’

In another context – as in (3), a command to children who were pulling at each other’s clothes – only the reciprocal meaning is appropriate. (3)

awarwa lagu-tukwa REC/ASS pull-PROH (a) ‘Don’t pull at each other!’ (b) *‘Don’t pull together (at something)!’

In its associative meaning, awarwa can be replaced with the adverb nakamib ‘all together’, itself a lexicalization of nak-a-məy (one-LK-real) marked with the terminative case -b with the meaning of ‘exactly, up to, until’. (The terminative case is often used in adverb formation: Aikhenvald 2008: §4.4). In its reciprocal meaning, awarwa has no alternative synonym. But there are other ways out.

3. Directionals, and reciprocal meanings Manambu has an articulate system of directional markers which may be suffixed to a verb root, e.g. yakə-su- ‘throw-upward’, yakə-saki- ‘throw across away from speaker’. The directional markers -su ‘upwards’, -sada‘downwards’, -saki- ‘across away from speaker or reference point’, -sapra‘across toward speaker or reference point’, -səwəla- ‘towards inside or away from the reference point’, -saku- ‘outwards’ result from the grammaticalization of the verb sə- ‘move’ accompanied by directional suffixes da ‘down’, -u ‘up’, -aki ‘across away’, -apra ‘across towards’, -(ə)wəla ‘inside’ and -aku ‘outwards’. These directional suffixes also occur on demonstratives, e.g. a-l-aku (that-F.SG-ACROSS.OUTWARD) ‘that one across in the outward direction’ and on a few archaic verbs, e.g. kr-aku- (take-ACROSS.OUTWARD)

Reciprocals in the making 161

‘take (something) outwards’. The verb sə- can take some directional suffixes, e.g. sə-da- ‘put down’, the other two are sə-wəla- ‘push inside’, and saku- ‘push outside’ (from sə-aku). That is, these directional forms can be viewed as grammaticalized verbal compounds. The origin of two further directionals, -tay- ‘sideways away from speaker or reference point’ and -tæy- ‘sideways toward the speaker or reference point’, is as yet unknown. The directionals -saki- ‘across away’ and -saku- ‘outwards’ can combine with the directional -səwəla- ‘towards inside or away from the reference point’. Examples are yakə-saki-səwəla- ‘throw to and from’ (across away from speaker-inside or towards speaker within limited space) and yakə-saku-səwəla- ‘throw to and from’ (involving outward movement; no defined space). The directional -tay ‘sideways away from speaker or reference point’ can be repeated, as in yakə-tay-tay- ‘throw back and forth (no defined space)’. A combination of two directionals -saki- ‘across away’ and -səwəla‘towards inside or away from the reference point’ can be used to mark reciprocal activities, with an additional overtone of multiple action involving multiple participants, e.g. vya-saki-səwəla- (hit-ACROSS.AWAYINSIDE) ‘hit each other back and forth’, wa-saki-səwəla- (talkACROSS.AWAY-INSIDE) ‘talk to each other back and forth between’. So can the reduplicated directional -tay, e.g. kui-tay-tay- (giveSIDEWAYS.AWAY-SIDEWAYS.AWAY) ‘give to each other; give back and forth’. The directional -saku- followed by -səwəla- refers to a reciprocal activity in just one idiomatic expression, bla-saku-sala- ‘debate among each other; perform ceremonial talk in men’s house’. Unlike awarwa ‘reciprocal/associative’, directionals in their reciprocal usage have an implication of multiple activity. A verb with a repeated directional in (4a) indicates that the action of ‘giving’ takes place over and over again. In (4b), a verb without a directional, accompanied by awarwa, refers to ‘giving’ happening just once: (4)

a.

kamna:gw food

kui-tay-tay-an give-SIDEWAYS.AWAYSIDEWAYS.AWAY-SEQ

tə-kwa-dian stay-HAB-1PL.SBJ.NPST ‘We keep giving food to each other back and forth’

162 Alexandra Aikhenvald b.

kamna:gw awarwa kui-n tə-kwa-dian food REC/ASS give-SEQ stay-HAB-1PL.SBJ.NPST ‘We give food to each other (once)’

The reciprocal overtone of directionals is just one of their possible readings, and can be regarded as an extension. Yet the frequency of the reciprocal readings of directional combinations in the language (as it is used nowadays) points towards their conventionalization. New reciprocals with no associative overtones resulting from a reinterpretation of directionals are on the rise – this was exemplified in (4a). Examples in (4a-b) constitute a minimal pair where different reciprocal strategies produce different meanings. These markers have the same effect with any verb (including stative verbs). Directional-reciprocal markers appear on verbs of all classes (with the exception of copula verbs).

4. Unusual reciprocals We mentioned above that grammaticalization of a reciprocal/associative marker from a locational noun marked with comitative in Manambu appears to be unique. The polysemy involving a reciprocal, ‘each other’, and an associative or collective marker ‘together’ is not. A similar polysemy has been described for Boumaa Fijian (Dixon 1988: 177–178), and various other Oceanic languages (also see Aikhenvald 2007, on a similar phenomenon in Tariana, from north-west Amazonia, and its neighbours). Polysemy between directionals and reciprocal meanings is widespread in Oceanic languages. However, this typically involves marking of dispersive movement (happening all over the place), and/or occurring in several directions with overtones of repetition (as in Futunan: Moyse-Faurie 2007, and Mekeo: Jones 1993); or happening here and there, without specific direction or aim, as in Nêlêmwâ (Bril 2007) and To’aba’ita (Lichtenberk 2007). And AnejomÕ (Oceanic: Lynch 2000: 75–76; 85–86) has a reflexivereciprocal verb ispõa- and also a set of suffixes which are said to encode ‘random motion and reciprocality’.2 A similar example comes from Tongan (Churchward 1953: 256); the verb hiki means ‘to move from one place to another’; and a derived form fe-hiki-taki can either have a purely reciprocal meaning ‘to change places with one another’, or refer to random motion, meaning ‘to move hither and thither’.

Reciprocals in the making 163

Along similar lines, Mandarin Chinese has a construction VERB-laiVERB-qu (VERB-come-VERB-go) meaning ‘VERB all over again, back and forth, always, here and there; reciprocal’ (Liu 1999; especially p.129 and example (16)). Here, ‘the reciprocal sense (of lai-qu construction) is not part of the core meaning, but an implicature from the particular situation’. Manambu is somewhat like Mandarin Chinese in that the reciprocal meaning of the directionals can still be seen as an implicature and not (yet) as their core meaning. However, the unusual feature of the Manambu directionals used in reciprocal meaning is that, unlike all the examples above, the directionals with reciprocal overtones refer to strictly defined directions and NOT to random direction and motion. Manambu does have a construction referring to random motion. This construction, VERB-yi-VERB-ya (VERB-go-VERBcome) means ‘VERB back and forth, randomly, all over the place’ and is structurally similar to that in Mandarin Chinese. But in Manambu it does not warrant a reciprocal interpretation. The directionals which develop reciprocal extensions (-saki-sala‘across away-inward’ and -tay-tay reduplicated directional ‘to side away’) involve movement away or outwards and never up or down.3 That the directional with the meaning ‘to side away’ has developed overtones of reciprocal action is congruent with the fact that the major reciprocal marker awarwa is associated with the noun meaning ‘side’. The reason for this, cross-linguistically unusual, development is as yet unknown. Different means of expressing reciprocal meanings interact in a number of ways. While one verb cannot be marked for different directionalreciprocals simultaneously, it is not impossible for awarwa ‘REC/ASS’ and a directional-reciprocal to occur in one clause. An example is in (5): the marker awarwa emphasizes the reciprocity of the joint action, and the directional serves to express the idea of ‘going back and forth’ in a dialogue: (5)

awarwa REC/ASS

wa-tay-tay-an say/speak-SIDEWAYS.AWAYSIDEWAYS.AWAY-SEQ

tə-di stay-3PL.SBJ.NPST ‘They talk to each other back and forth’

164 Alexandra Aikhenvald There are no other ways of expressing reciprocal meanings. In contrast to many other languages, none of the reciprocals in Manambu has a reflexive meaning. A reflexive-emphatic meaning can be expressed with the adverb ka:p ‘by itself, by oneself’ or with a reduplicated pronominal stem, e.g. də də-kə (he he-OBL) ‘he himself’. Just occasionally an emphatic pronoun has a reflexive meaning, as illustrated in (6). (6)

amæyik wukə-ku də də-kə miss-AFTER.SS he he-OBL mother+LK+DAT vya-səpa-ku, ata kiyad hit-hit.body-AFTER.SS then die+3M.SG.SBJ.PAST ‘Then he having missed (his) mother, he having hit himself (lit. he himself having hit) died’ The existing reflexive strategies never have any reciprocal overtones.

5. Multiple grammaticalization, and its outcomes The grammatical expression of reciprocals in Manambu can be described as multiple grammaticalization. A locational noun ‘side’ has grammaticalized into a polysemous reciprocal/associative marker. There is a special associative adverb ‘all together’, but no specialized reciprocal. Combinations of directional markers involving movement ‘away’ and ‘outwards’ are developing an unequivocally reciprocal meaning (without losing their directional implications). The polysemous patterns associated with various ways of expressing reciprocals and related meanings are summarized in Figure 1. The top line indicates the origins of each form.

Reciprocals in the making 165 lexicalized from nak-a-məy-b ‘one-LK-real-TERM’ ↓

grammaticalized from awar-wa ‘side-COM’ ↓

nakamib ‘all together’

(a) awarwa ‘associative (all together)’ (b) awarwa ‘reciprocal (each other)’

grammaticalized from verb sə- ‘move’ + dir. suffix ↓

directionals -saki-səwəla ‘across away-inside’; -saku-səwəla ‘across outward-inside’; -tay-tay ‘sideways awaysideways away’

Figure 1. Forms with reciprocal meanings in Manambu, and their polysemous patterns

Multiple grammaticalization in Manambu has resulted in partial disambiguation of polysemous patterns. Further polysemous structures are on the rise, each following a somewhat idiosyncratic path. The existence of several items with similar meanings in a language is not unusual per se. What is unusual in the Manambu case is the co-existence of several grammatical mechanisms, with different origins, expressing various facets of one grammatical meaning: the reciprocal. These grammatical mechanisms are not fully synonymous (see especially examples (4a) and (4b)). Each of them has an additional overtone, and their interaction helps resolve existing polysemous patterns. In terms of relative chronology of the varied expressions of reciprocity in Manambu, the grammaticalization of awarwa appears to be the oldest extant means of reciprocal marking. A similar form is attested with the same meaning in some languages of the family (e.g. Iatmul: Gerd Jendraschek, p.c.). In contrast, reciprocal meanings for the directionals appear to be a Manambu-specific phenomenon which is frequently attested in the spontaneous speech of younger and innovative speakers. This suggests a relatively recent origin of the – cross-linguistically unusual – polysemy of directional and reciprocal meanings. How widespread multiple grammaticalization is cross-linguistically remains an issue for further study.

166 Alexandra Aikhenvald

Notes 1.

2.

3.

I am grateful to my Manambu friends who taught me their language, to R. M. W. Dixon for comments and suggestions, and to Jessica Cleary-Kemp for editorial assistance. Lichtenberk (1999: 55f) convincingly argues that the polysemy of the marker of ‘plurality of action’ as distributive, repetitive, dispersive and reciprocal goes back to Proto-Oceanic. The order of directionals deserves a mention. In Manambu, if two directional specifications follow each other within one clause or one word, the order is typically ‘far’-‘close’, similarly to the ‘go come’ construction and to the directional sequences discussed here, and not the other way round. It appears that languages of the world have different preferences as to the relative ordering of ‘far’ and ‘close’. For instance, the preferred order in English this and that, here and there, come and go is opposite to that in Manambu. But the order in Manambu is reminiscent of Tok Pisin go kam, and in other languages in New Guinea (such as Barupu, from Sko family: Mim Corris, p.c.). This question requires an in-depth typological investigation (also see Liu 1999 on the order ‘come-go’ in Mandarin Chinese, and el ir y el venir in Spanish; cf. similar patterns in French and Portuguese).

References Aikhenvald, Alexandra Y. 2007 Reciprocal and associative in Tariana: their genetic and areal properties. In Nedjalkov (ed.), 1–16. 2008. The Manambu Language of East Sepik, Papua New Guinea. Oxford: Oxford University Press. Bril, Isabelle 2007 Polysemy of the reciprocal marker in Nêlêmwâ. In Nedjalkov (ed.), 1479–1509. Churchward, C. Maxwell 1953 Tongan Grammar. Nuku’alofa, Tonga: Vava'u Press. Dixon, R. M. W. 1988 A Grammar of Boumaa Fijian. Chicago: University of Chicago Press. Frajzyngier, Zygmunt, and Traci S. Curl (eds.) 1999 Reciprocals. Forms and Functions. Amsterdam: John Benjamins. Heine, Bernd 1992 Grammaticalization chains. Studies in Language 16: 335–368.

Reciprocals in the making 167 Heine, Bernd, and Tania Kuteva 2002 World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Jones, Alan A. 1993 Towards a Lexicogrammar of Mekeo. Ph. D. diss., ANU. König, Ekkehard, and Volker Gast (eds.) forth. Reciprocals and Reflexives: Cross-linguistic and Theoretical Explanations. Berlin: Mouton de Gruyter. Lehmann, Christian 2002 Thoughts on Grammaticalization. 2nd revised edition. Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt. Erfurt: Seminar für Sprachwissenschaft der Universität. Lichtenberk, Frantisek 1999 Reciprocals without reflexives. In Frajzyngier and Curl (eds.), 31– 62. 2007 Reciprocals and related meanings in To’aba’ita. In Nedjalkov (ed.). Liu, Meichun 1999 Reciprocal marking with deictic verbs “come” and “go”. In Frajzyngier and Curl (eds.), 123–132. Lynch, John 2000 A Grammar of Anejomõõ. Canberra: Pacific Linguistics. Moyse-Faurie, Claire 2007 Reciprocal, sociative, reflexive, and iterative constructions in East Futunan. In Nedjalkov (ed.). Nedjalkov, Vladimir P. (ed.) (with the assistance of Emma Geniusiene and Zlatka Guentchéva). 2007 Reciprocal Constructions. Amsterdam: John Benjamins.

The Maltese continuative: A grammaticalization borderliner Thomas Stolz and Andreas Ammann

1. Introducing the problem The verbal categories of Modern Maltese are the subject of a variety of relatively recent in-depth studies. Borg (1981), Fabri (1995) and Ebert (2000) and Spagnolo (2007) investigate aspect (and tense) in this AfroAsiatic language spoken on the southern fringes of Europe. Vanhove (1993: 100–330) provides a richly documented inventory of what she calls “auxiliaires verbales” - a terminological solution which we adopt in this study for purely practical reasons. In Borg and Azzopardi-Alexander (1997: 226–36), the various so-called aspects of Modern Maltese are surveyed. Superficially, these many studies dedicated to aspect and related categories of Maltese suggest that there is no urgent need for another look at the system. On closer inspection, however, the extant descriptions of verbal categories of Maltese do not necessarily give answers to all possible questions, in a manner of speaking. Some interesting problems remain unsolved – among them for instance the degree of grammaticalization of the constructions which express aspect categories. For two periphrastically expressed aspects of Modern Maltese, the situation is such that several verbs compete with each other as to the status of aspect “auxiliary”. The list of candidates for the ingressive/inchoative is especially long (although Vanhove (1993: 264) argues that for some of the supposed auxiliaries the evidence is rather dubious). According to Borg and Azzopardi-Alexander (1997: 233), there are six verbs expressing the ingressive aspect, two of which are focused upon in Stolz and Ammann (2007). These ingressive auxiliaries do not easily replace one another and their text frequency is clearly skewed with beda ‘to begin’ ranking highest. Moreover, some of these verbs occur only in combinations with a restricted set of verbs. Very often, these collocation-like combinations are typical of the literary style. Certainly, the grammaticalization of beda has advanced furthest. Still, we do not know exactly how beda interacts with its competitors and whether these express nuances neutralized by beda. Furthermore, all ingressive auxiliaries can still be used as lexical verbs.

170 Thomas Stolz and Andreas Ammann Similarly, the continuative is a category for which Maltese holds a number of competing expressions in store. Three verbs are mentioned as auxiliaries of the continuative in Vanhove (1993: 265–274) and Borg and Azzopardi-Alexander (1997: 231–232): baqa’ ‘to remain’, kompla ‘to continue’ and issokta ‘to continue’ whereas a fourth candidate, żied ‘to add to’, is characterized as marginal, literary and also dialectal (Vanhove 1993: 273–274). For the continuative, the problems are identical to the ones mentioned above for the ingressive. Baqa’ is the most frequent of the competitors but the exact relation of this verb to the other members of the paradigm is still largely unclear. Kompla, issokta and also the peripheral żied occur in combination with a verb of speaking – a preference which is most pronounced with żied. In addition, all these verbs retain many features of full lexical verbs and thus a closer look at their distribution patterns and other properties is called for to help us determine the degree of grammaticalization. In this contribution, we intend to shed light upon the issue of grammaticalization of the Maltese continuative(s) in the sense that we identify the major properties of the various candidates for auxiliary status. This identification is achieved on the basis of a small selection of modern prose. Except otherwise stated, the empirical data stem from three texts: the first volume of Agħsafar qalb l-għolliq (the Maltese translation of Colleen McCullough’s The Thorn Birds), Iċ-Ċkejken Prinċep (the Maltese translation of Antoine de Saint-Exupéry’s Le Petit Prince) and Tużżana (a collection of stories by twelve authors) with altogether 480 pages of running text.1 The study is synchronic and thus cannot give a full account of the chronology of events in the grammaticalization processes that have lead to the present state of affairs in Maltese. The diachrony of our subject matter will be looked into more closely in a follow-up study. The same holds for the inclusion of spoken language data in our empirical basis.

2. The continuative 2.1. Towards a definition In their large-scale treatment of verbal categories and their evolution in the languages of the world, Bybee, Perkins, and Pagliuca (1994: 127) define2 the continuative as a category which

The Maltese continuative 171 includes the progressive meaning – that a dynamic situation is ongoing – and additionally specifies that the agent of the action is deliberately keeping the action going. Continuative is the meaning of ‘keep on doing’ and ‘continue doing’.

The authors emphasize that there is a “close semantic connection between iteratives, continuatives, and frequentatives” (Bybee, Perkins, and Pagliuca 1994: 164). Crosslinguistically, properly distinct continuatives are relatively rare as the continuative meaning is often syncretistically encoded with other categories (Bybee, Perkins, and Pagliuca 1994: 164–165). In addition, the continuative is absent from Heine and Kuteva (2002) although Maltese is among their sample languages, whereas the language is not taken account of by Bybee and associates. The Maltese continuative thus not only constitutes an additional realization of this rare category but also helps us understand better what the essential properties of the continuative are. In the extant literature on Maltese verbal categories, the continuative is seldom mentioned explicitly as Borg (1981), Fabri (1995) and Ebert (2000) focus on the basic distinction of perfective and imperfective to the exclusion of those categories which require a periphrastic expression involving auxiliaries other than kien ‘to be’. Vanhove (1993: 265) introduces the continuative as a “temps accessoire” corresponding largely to French construction of the type continuer à V ‘to go on V-ing; to continue to V’.3 She specifies that La notion de “continuer à” implique nécessairement que le process décrit par le verbe auxilié a déjà commencé avant le moment où est enoncée la périphrase, ce qui n’est pas le cas pour l’accompli et l’inaccompli duratifs (Vanhove 1993: 265).

Thus, the continuative is not identical with the progressive, durative, continuous or sundry notions associated with the imperfective, for instance, although it may be conceptually related to them (cf. above). Nevertheless, Borg and Azzopardi-Alexander (1997: 322) mention the auxiliaries baqa’, kompla and issokta in their section devoted to the continuous aspect(s) where they claim that these verbs “can also reinforce continuous aspect by occurring before the Imperfect form of non-punctual verbs”. This means that there is no separate continuative for these authors as they do not consider the constructions involving the above auxiliaries expressions of a distinct category. Furthermore, the continuative differs from the perfect because in contrast to the latter, the continuative does not highlight the

172 Thomas Stolz and Andreas Ammann result of an event in the past which has some relevance for the reference time (Vanhoeve 1993: 49). In point of fact, the continuative implies that a result state has not been reached yet because the event/action is depicted as being continued at the reference time (and possibly also beyond).

2.2. First examples The above definitions have to stand the empirical test. Sentences (1)-(4) illustrate typical examples of the uses to which the four auxiliaries are put in modern Maltese prose. (1)

baqa’ [Tużżana 65] l-aħwa baqgħu jitlewmu DEF-brother:PL remain.PFV:3PL 3:RECP:struggle.IMPF:PL bejniethom għal ħin twil among:PL:3PL for time long ‘The brothers kept arguing among them for a long time.’

(2)

kompla [Tużżana 30] Se jkompli jkissirli FUT 3:continue.IMPF 3:CAUS:break.IMPF ‘He will continue destroying my house!’

d-dar DEF-house

(3)

issokta [Agħsafar 151] Waqt li naqqas xi ftit il-viġilanza tiegħu while that reduce.PFV some little DEF-vigilance of:3SG.M minn fuq Ralph from on Ralph issokta jħares lejn is-segretarju tiegħu continue.PFV 3:look.IMPF towards DEF-secretary of:3SG.M ‘While he reduced his watch of Ralph, he continued to observe his secretary.’

(4)

żied [Prinċep 68] Id-deżert sabiħ żied jgħid DEF-desert beautiful add.PFV 3:say.IMPF ‘The desert is beautiful, he continued to say.’

The Maltese continuative 173

What the examples (1)-(4) have in common – apart from the syntactic position of the auxiliary to the left of the lexical verb and the absence of conjunctions or connectors in the construction (termed “asyndétique” by Vanhove (1993: 103)) – is the internal make-up of the situation they describe or refer to. The event, action or state represented by the lexical verb is presented such that its beginning is situated to the left of the reference point on the time-arrow and its continuation is located to the right of the same reference point. Moreover, there is an important component which sets the continuative apart from other aspectual categories which highlight extension over time, namely the fact that, at the reference point, the process is briefly suspended and then resumed once more without making the continuation a separate instantiation of a new process of the same ontological class, i.e., both phases of the process before and after the reference point belong together but there is a potentially very short gap between them. This gap does not necessarily translate into a full-blown interruption of the process, it may also come in the guise of a temporary backgrounding of a process which was more prominent before and becomes prominent again after the reference point4. The continuative differs from the habitual, iterative and frequentative in the following way: the continuative transforms several potential (sub-)events into one whereas the other categories treat the (sub-)events as spatio-temporally distinct units. This is especially true of examples (2)-(4). The outcry of the character Midella in (2) is taken from a context in which the antagonist Mollamolla inadvertently acts in such a way that Midella’s house is damaged. However, Mollamolla is not in control of her movements and thus temporarily jumps elsewhere before she returns to the endangered building. The destruction of Midella’s house is not a properly continuous process but rather comes in fits. Nevertheless, the intervals between one destructive act and the next are too short to qualify those acts as two separate events, and thus the acts are depicted as one overarching process which has an internal contour. Furthermore, sentence (3) describes a situation in which one process is discontinued (the close vigilance to which the papal ambassador had put the archbishop-to-be) whereas the other (parallel process) is continued (namely the equally close watch of the secretary). The use of issokta serves to emphasize the contrast and thus also has a pragmatic function. As to sentence (4), the context conditions resemble those mentioned in connection to sentence (2). The Little Prince has already been speaking for a while when the narrator enters the scene only to fail to utter anything.

174 Thomas Stolz and Andreas Ammann However, there is a short interval during which nobody is speaking. Then, the Little Prince speaks again – and this new speech-act is depicted as a continuation of his previous monologue. Sentence (1) does not lend itself easily to a like-minded interpretation. The brothers have been quarrelling – and the initial phase of their dispute is described in some detail. After this description, the arguments they exchange are known to the reader and thus it is no longer necessary to repeat them. What is more interesting for the story-line is how the quarrel comes to an end: the bewitched parents wake up and everybody joins in their lament about having been deceived by the sorcerer. Baqa’ simply shows us that the brothers go on fighting verbally among each other. There is no discernable break or gap between their earlier and their later struggle. In contrast to kompla, issokta and żied, baqa’ associates with a continuous i.e. uninterrupted process whereas the other three auxiliaries invite the reading that there is some element of discontinuity involved. It remains to be seen whether this is always the case.

2.3. Statistics In point of fact, baqa’ displays a much wider distribution over contexts than its potential competitors. In Table 1, we give the frequency of each of the verbs for the individual sample texts. The frequencies are given separately for grammatical (G) and lexical (L) uses, respectively. Table 1. Frequency count Text L Prinċep 8 Agħsafar 76 Tużżana 54 Total 138

baqa’ G 10 101 112 223

Σ 18 177 166 361

kompla L G 1 10 9 14 19 20 29 44

Σ 11 23 39 73

issokta L G 0 0 1 1 0 1 1 2

Σ 0 2 1 3

L 1 0 0 1

żied G 11 0 0 11

Σ 12 0 0 12

In all texts, baqa’ is the most frequent of the four verbs5. It accounts for 80% of the 449 tokens of continuative auxiliaries in the three texts. Furthermore, baqa’ and the much less frequently attested kompla occur in all of the texts whereas the two remaining verbs have a much more restricted distribution over our empirical basis. Since baqa’ occurs six times as often

The Maltese continuative 175

as kompla – the second best in the above statistics, we feel entitled to claim that baqa’ has a more general meaning than each of the other verbs. This is especially evident for żied whose twelve occurrences comprise exclusively combinations with the verb qal / igħied ‘to say’.6 Żied + verb of speaking is a pattern of two closely associated elements which comes very close to collocations and may alternatively be a candidate for lexicalization (Lehmann 2002: 5). For baqa’, the ratio of lexical and grammatical usages is 0.61-to-1, kompla has a ratio of 0.65-to-one. For the two other verbs, the number of attestations is either to small (issokta) or clearly biased in favour of one particular combination (żied) that it does not make much sense to calculate the exact ratio. Nevertheless, in all four cases, grammatical usages outnumber lexical ones. The dichotomy of grammatical vs lexical usages is explained in section 2.5.1 below.

2.4. Further contexts of baqa’ The idea of a more general meaning range of baqa’ is further corroborated by the employment of this verb in contexts where a gap of sorts applies. Sentences (5)-(7) are meant to prove that baqa’ is also used in contexts in which we would expect kompla or issokta. (5)

baqa’ [Tużżana 99] Għandek bżonn tikber ftit qallu at:2SG necessity 2SG:grow.IMPF little say.PFV:IO.3SG.M missieru u wissieh father:POR.3SG.M and admonish.PFV:DO.3SGM biex ma jibqax jagħmel ħmerijiet in_order_to NEG 3:remain.IMPF:NEG 3:do.IMPF idiocy:PL ‘ “You have to grow a little older!” his father said to him and admonished him not to go on acting foolishly.’

(6)

baqa’ [Tużżana 164] Baqgħet tmur hemm minħabba Aaron remain.PFV:3F 3F:go.IMPF there because_of Aaron ‘She continued to go there because of Aaron.’

176 Thomas Stolz and Andreas Ammann (7)

baqa’ [Tużżana 210] Is-Sinjorina tat bewsa lil Romina u DEF-lady_teacher give.PFV:3F kiss to Romina and wara li sellmu lil xulxin after that greet.PFV to one_another Romina baqgħet sejra tiġri għand Romina remain.PFV:3F go.PTC:F 3F:run.IMPF at nannuha biex tgħidlu grand_father:POR.3SG.F in_order_to 3F:say.IMPF:IO.3SG.M bl-aħbar with:DEF-news ‘The lady-teacher gave Romina a kiss and after they had said farewell Romina ran on to her grand-father to give him the news.’

In these cases, baqa’ is used although no continuous i.e. uninterrupted process applies. For example (5), the context is as follows: the main character of the story, a young boy who is afraid of ghosts, has disturbed his parents because of a danger he had imagined. He alerted the whole family for no reason. The danger turned out to be false alarm. After this is settled, the father lectures his son. This parental lecture also marks a gap as during the lecture the poor boy has no chance to continue to act foolishly. He is admonished not to resume his former habit of being afraid of everything. Example (6) is a little different: a young girl is pestered by her enemy and thus considers quitting the youth club a strategy to avoid being harassed. However, Aaron – the boy she fancies – frequents the club. This impels her to keep going there too. In other words, she never abstained from frequenting the youth club, meaning: the gap consists in a moment of doubt in which the girl makes up her mind. In (7), baqa’ combines with the active participle sejra (feminine of sejjer ‘going’) which in turn is specified by the finite lexical verb tiġri ‘she runs’. This is a very common strategy in Maltese to specify in which way a movement is realized. The actual function of baqa’ is more difficult to determine. First of all, the agent Romina has not started the actual running before she said good-bye to her teacher. Thus, one cannot claim that she continued to run. However, the context from which the example is drawn is the description of an emotionally loaded and rather lengthy scene in which Romina, who is about to go home after her visit to the teacher‘s house, learns that she will be given music lessons for free. Since this means that her greatest wish has come true, she wants to hurry home to bring the good

The Maltese continuative 177

news to her family. However, she still has to show her gratefulness to the teacher first – an act of courtesy which delays her departure. In a way, then, baqa’ suggests that the leaving had begun already before they protagonists said good-bye and the process is continued immediately after the farewell. These and sundry examples show that baqa’ fulfils functions similar to adverbs such as Italian ancora ‘still, again’ which allow for more than one reading. Baqa’ indeed covers a wide range of possible meanings. It neutralises the distinctions associated with the other continuative auxiliaries. Does this also imply that baqa’ is the most grammaticalized of the competitors? Lehmann (2006: 68) ascribes special importance to desemanticization in the process of grammaticalization. Baqa’ has undergone desemanticization – but to what extent?

2.5. Degrees of grammaticalization To determine whether or not the above auxiliaries qualify as instances of grammaticalization and, if they do, to what degree their grammaticalization has advanced, we presuppose the parameters of grammaticalization as established by Lehmann (1995: 121–178). However, we start with a review of the lexical uses to which the above verbs are put.

2.5.1. Lexical uses For all of the verbs, our sample texts also contain examples of their employment as full verbs or lexical verbs, cf. (8)-(11). (8)

baqa’ a. [Tużżana 57] żiffa ħelwa sajfija kienet breeze sweet:F summerly:F be.PFV:3SG.F tħajrek tibqa’ barra 3F:compel.IMPF:DO2SG 2:remain.IMPF outside ‘A mild breeze of summer made you stay outdoors.’ b. [Tużżana 226] Kitty baqgħet b’għajnejha Kitty remain.PFV:3SG.F with’eye:PL:POR.3SG.F

178 Thomas Stolz and Andreas Ammann

c.

(9)

mberrqin b’ħarsitha msammra PTC:open_wide:PL with’look:TM:POR.3SG.F PTC:nail:F fuq dawk il-gozz ħbub on those DEF-pile coin:PL ‘Kitty kept her eyes open wide with her look fixed on the pile of coins.’ [Agħsafar 74] Għax Fee baqgħet kwieta because Fee remain.PFV:3SG.F quiet:F ma turi xejn NEG 3F:show.IMPF nothing ‘Since Fee remained silent, she did not give away anything.’

kompla a. [Agħsafar 105] Ma setax ikompli l-kampanja NEG can.PFV:NEG 3:continue.IMPF DEF-campaign tiegħu b’demm biered of:3SG.M with’blood PTC:freeze ‘He could not continue his campaign in cold blood.’ b. [Agħsafar 21] Frank kompla bix-xogħol Frank continue with:DEF-work li kien qed jagħmel that be.PFV PROG 3:do.IMPF ‘Frank went on with the work he was doing.’ c. [Prinċep 44] Iċ-ċkejken prinċep kompla DEF-little prince continue.PFV il-pjaneta tiegħek tant hi hija żgħira DEF-planet of:2SG so she she little:F li ddur dawramejt maggħa that 2SG:turn.IMPF full_circle with:3SG.F fi tliet passi in three step:PL ‘The Little Prince continued: “Your planet is so small that you travel around it in three steps.”’

The Maltese continuative 179

(10)

issokta [Agħsafar 127] Huwa ssokta b’aktar ġentilezza he continue.PFV with’more gentleness iqis kliemu li jkunu 3:measure.IMPF word.COLL:POR.3SG.M that 3:be. FUT:PL ċari u krudili għalkemm clear:PL and cruel:PL although ma kellu ebda intenzjoni NEG have.PFV:3SG.M no intention li jkun krudil that 3:be. FUT cruel ‘He continued more gently, making his words clear and cruel, although he had no intention to be cruel.’

(11)

żied [Prinċep 24] Naħseb li sar il-ħin għall-fatra 1:think.IMPF that become.PFV DEF-time for:DEF-lunch żiedet ftit wara add.PFV:3SG.F little after ‘ “I think that it is time for lunch”, she added a moment later.’

Baqa’ is frequently used in copula-like fashion. It may be used to express location in space (see (8a)) and temporal states with the predicated state being expressed either by NPs/PPs (as in (8b)) or predicative adjectives (see (8c)). What these examples have in common is the idea that allocation and states remain unchanged for a span of time whose extension is not specified any further. Kompla is used transitively in (9a) where it takes the direct object l-kampanja ‘the campaign’. In (9b), the same verb has a prepositional complement (bix-xogħol ‘with the work’). In (9c), kompla seems to have the characteristics of an intransitive verb. However, it can only be used in this way when direct speech is quoted, i.e., the quote itself is a kind of complement. In these contexts, kompla semantically incorporates the general meaning components of verbs of speaking. Whether this usage of kompla constitutes an elliptical shortening of a construction with a sub-ordinated lexical verb of speaking can only be determined on the basis of a diachronic study. The only attestation of the generally rather infrequent issokta as a lexical verb in our sample texts is (10). It is used intransitively like kompla in (9c) because the sentence in which it occurs specifies the way in which the speaker modifies his diction in order to have a

180 Thomas Stolz and Andreas Ammann certain effect on his listener. In a way, issokta incorporates the meaning of a verb of speaking. This is exactly what applies to żied in (11)7. Like kompla in (9b), żied combines with a quote of direct speech and thus is associated with the meaning of a verb of speaking. Kompla, issokta and żied share the ability to associate with the meaning of verbs of speaking whereas baqa’ is not normally used in this way. Example (10) is also indicative of the full verb status of issokta because it can be modified adverbially (by the modal passive participle b’aktar ġentilezza ‘more gently’). However, this accessibility of the verbs to modification is retained also in their non-lexical uses, cf. (12). (12)

baqa’ [Tuzzana 1819 U Adam u Eva ma baqgħux and Adam and Eve NEG remain.PFV:3PL:NEG biss iħobbu u jbusu only 3:love.IMPF:PL and 3:kiss.IMPF:PL lil xulxin weħidhom to one_another one:3PL ‘And Adam and Eve did not continue to love and kiss one another alone.’

In this sentence, the auxiliary baqa’ is negated and the adverb biss ‘only’ is intercalated to separate the auxiliary from the coordinated lexical verbs. The separation of the auxiliary from the lexical verb by intervening adverbs is very common – and not only for the continuative. Adverb intercalation is also attested for the compound tenses whose initial verb form (= the auxiliary kien and other members of the paradigm) can be followed by adverbials of at times considerable length.8

2.5.2. Absence of evidence of grammaticalization All potential continuative auxiliaries may still be used as lexical verbs. For none of the verbs under scrutiny is the share of purely lexical attestations negligible. Has a split occurred such that the lexical and grammatical versions of baqa’, kompla, issokta and żied can be formally distinguished? Has a new word class arisen? Lehmann (2005: 18) argues that the genesis of a new word class via split requires additional restrictions over the distribution of the members of the new word class. The evidence for more re-

The Maltese continuative 181

strictions is strongest with żied because it hardly ever occurs outside combinations with verbs of speaking. Moreover, there is not a single instance of żied being negated. Likewise, żied is attested exclusively in the perfective and only with a subject in the 3rd person singular or plural. These restrictions might turn out to be induced by the size of our empirical basis, though (a possibility which also extends to issokta). Kompla and especially baqa’ display the full range of paradigmatic distinctions. In point of fact, these auxiliaries behave morphologically like the average full verb in Maltese. Thus, they have not (yet) developed into aspect particles as some of the cognates of baqa’ in Maghrebin varieties of Neo-Arabic (Vanhove 1993: 265–266).9 There is neither morphological nor phonological reduction. Univerbation or coalescence with neighbouring items does not apply. A certain degree of syntactic autonomy remains. Moreover, some kind of paradigmatic variability can also be observed as there are contexts in which several of the continuative auxiliaries may replace each other. Apart from the fact that, in combination with these would-be auxiliaries, the subordinated lexical verb is always in the imperfective (or an active participle) whereas the auxiliary can be either in the imperfective or the perfective (Borg and Alexander-Azzopardi 1997: 228), baqa’ and kompla fail to fulfil most of the formal criteria of grammaticalization (Lehmann 1995: 123). The only parameter on which there is evidence of grammaticalization is the semantic one if we accept that the meaning of the above verbs is generalized in those contexts which we consider instances of auxiliation. However, the meaning change is perhaps only minimal. Do the continuative auxiliaries qualify as quasi-auxiliaries only (Heine 1993: 14–16)? Are they grammaticalized at all? We assume that baqa’, kompla and perhaps only to some extent also issokta and żied have indeed advanced on the verb-to-TMA chain (Heine 1993: 53–65) albeit only a few steps. This fact reminds us of Van Pottelberge’s (2005: 186–187) hypothesis that not every instance of grammaticalization must necessarily result from a succession of many discreet steps which together give the impression of a continuous process. The high number of attestations of baqa’ in our sample texts is suggestive of a higher degree of grammaticalization although there is practically nothing else which corroborates this interpretation. Baqa’ and the other verbs have not fallen victim to the classic epiphenomena of grammaticalization. Where there are additional restrictions (as with żied) they do not fully support an analysis in terms of grammaticalization theory. Nevertheless, baqa’, kompla and with some reservation also issokta and żied have clearly

182 Thomas Stolz and Andreas Ammann contoured functions which qualify them as not purely lexical verbs. They are paradigm cases of borderliners which cannot decide on which side of the boundary between grammar and lexicon they are located best. In this way, they illustrate nicely the fuzzy boundaries between the components as assumed in Lehmann (1989).

3. Conclusions Maltese has a lot on offer for those who take an interest in grammaticalization phenomena. This is so not because of some crosslinguistically rare categories. In Maltese, we find a number of grammaticalizations which do not behave prototypically. The above continuative auxiliaries constitute only one class of those categories whose markers seem to escape the usual fate of morphemes or constructions in grammaticalization processes as they often remain immune to reduction on the phonological and on the morphological level. Similarly, their syntagmatic freedom is not always subject to severe restrictions. For them, grammaticalization starts on the semantic level and in many cases this is all that happens for a long time. Therefore, grammaticalization researchers should cast an eye on Maltese data if they want to refine their models. We will continue in this vein and study other so-called auxiliaries of Modern Maltese and their diachrony in a series of follow-ups to this contribution.

Notes 1.

2.

The full bibliographic details are: Colleen McCullough, Agħsafar qalb lgħolliq. L-ewwel parti [translated by N.N.]. Ħamrun: PEG, 1984; Antoine de Saint-Exupéry, Iċ-Ċkejken Prinċep [translated by Toni Aquilina]. Msida: Mireva, 2000; Ġorġ Mallia & Trevor Żahra (eds.), Tużżana. Stejjer għat-tfal minn 12-il kittieb. Blata l-Bajda: Merlin Library, 1998. These texts are referred to by the abbreviated titles Agħsafar, Prinċep and Tużżana, respectively, with additional indication of the page(s) on which the example can be found. In Bussmann’s (2002: 376) terminological dictionary, the entry Kontinuativ treats the continuative as a synonym of either the aktionsart durative or the aspect progressive. This is certainly a simplification of the intricate relationship of continuative and other verbal categories.

The Maltese continuative 183 3. 4. 5.

6.

7.

8.

9.

For the definition of the continuative, Vanhove (1993) draws heavily on Cohen (1924: 265–269). In this way, the continuative belongs to the functional sphere of Lehmann’s (1991) “Situationsperspektion”. Cf. Vanhove (1993: 272) who observes that baqa’ ‘to remain’ is more widely distributed than the other continuative auxiliaries although without going into the statistical details. Vanhove (1993: 272–274) makes the same observation and adds that combinations of żied ‘to add’ with other lexical verbs are no longer acceptable in “standard” Maltese but can still be encountered in the speech of the eldest speakers of the dialect of Mġarr (Malta). The use of verbs meaning ‘to add’ or the like with the function of indicating that a speech-act is continued etc. is of course a commonality also in SAE languages like English, German and French. Cf. [Tużżana 220] Imma Marju kien kull darba jberraq għajnejn ‘But every time Mario opened his eyes wide’ where the nuclear periphrastic construction kien iberraq ‘he stared’ (verb-initial and alternate according to phonological context) is split by the intercalated adverbial kull darba ‘each time’. Interestingly, the four continuative auxiliaries stem from two different historical sources. Baqa’ and żied have a Semitic background whereas issokta and kompla stem from Romance, more precisely from Sicilian, namely from Sicilian cumpliri ‘to continue, to finish’ and siquitari ‘to follow, to continue’, respectively (Vanhove 1993: 270–272). Whether this etymological divergence has any bearing on the grammaticalization processes is an issue that has to be addressed in a diachronic follow-up study.

References Borg, Albert 1981 A Study of Aspect in Maltese. Ann Arbor: Karoma. Borg, Albert, and Marie Azzopardi-Alexander 1997 Maltese. London: Routledge. Bussmann, Hadumod 2002 Lexikon der Sprachwissenschaft. Stuttgart: Kröner. Bybee, Joan, Revere Perkins, and William Pagliuca 1994 The Evolution of Grammar. Tense, Aspect, and Modality in the Languages of the World. Chicago: University Press. Cohen, Marcel 1924 Le système verbale sémitique et l’expression du temps. Paris: Leroux.

184 Thomas Stolz and Andreas Ammann Ebert, Karen H. 2000 Aspect in Maltese. In Tense and Aspect in the Languages of Europe, Östen Dahl (ed.), 753–785. Berlin: Mouton de Gruyter. Fabri, Ray 1995 The tense and aspect system of Maltese. In Tense Systems in European Languages, Rolf Thieroff and Joachim Ballweg (eds.), Vol. II, 327–343. Tübingen: Niemeyer. Heine, Bernd 1993 Auxiliaries. Cognitive Forces and Grammaticalization. Oxford: Oxford University Press. Heine, Bernd, and Tania Kuteva 2002 World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Lehmann, Christian 1989 Grammatikalisierung und Lexikalisierung. Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 42 (1): 11–19. 1991 Strategien der Situationsperspektion. Sprachwissenschaft 16 (1): 1–26. 1995 Thoughts on Grammaticalization. München: LINCOM Europa. 2002 New reflections on grammaticalization and lexicalization. In New Reflections on Grammaticalization, Ilse Wischer and Gabriele Diewald (eds.), 1–18. Amsterdam: John Benjamins. 2005 Wortarten und Grammatikalisierung. In Wortarten und Grammatikalisierung. Perspektiven in System und Erwerb, Clemens Knobloch and Burkhard Schaeder (eds.), 1–20. Berlin: de Gruyter. 2006 Les role sémantiques comme prédicats. Bulletin de la Société de Linguistique de Paris 101 (1), 67–88. Spagnolo, Michael 2007 L-aspett lessikali fil-verb Malti. Dissertazzjoni mressqa lill-Istitut talLingwistika, l-Università ta’ Malta, ghall-grad ta’ Masters fl-Arti – M.A. – fil-Lingwistika. Stolz, Thomas, and Andreas Ammann forthc. Beda u Qabad: The Maltese Inchoative/Ingressive. Romano-Arabica. Special Issue on Peripheral Arabic Dialects. University of Bucharest: Center for Arab Studies. Vanhove, Martine 1991 La langue maltaise. Études syntaxiques d’un dialecte arabe «périphérique». Wiesbaden: Harrassowitz. Van Pottelberge, Jeroen 2005 Ist jedes grammatische Verfahren Ergebnis eines Grammatikalisierungsprozesses? Fragen zur Entwicklung des am-Progressivs. In Grammatikalisierung im Deutschen, Thorsten Leuschner, Tanja Mortelmans, and Sarah De Groodt (eds.), 169–192. Berlin: de Gruyter.

What do “do” verbs do? The semantic diversity of generalised action verbs Eva Schultze-Berndt

1. Introduction All languages seem to have one or more verbs which, like do in the English translations of (1) to (4) below, are used as ‘pro-verbs’ in contexts where the nature of an event is unknown or left unspecified, and represented by a pronominal or interrogative complement – most frequently a noun phrase, but possibly, as in the Jaminjung example, an interrogative “preverb”.1 Following Van Valin and LaPolla (1997), verbs of this nature will be referred to as ‘generalised action verbs’ (or GAV for short) in this paper, interchangeably with ‘do-verbs’. (1)

Samoan ‘o aa lua mea e fai taeao? PRES what 2DU thing.PL FUT do tomorrow ‘what are you two going to do tomorrow?’ (lit. ‘what are your things to be done tomorrow?’) (Mosel and Hovdhaugen 1992: 520)

(2)

Jaminjung warndug=biya yurru-wu-yu, do.what=now 1PL.INCL>3SG-POT-do ‘what are we going to do now (… are we going to camp out?)’ (Schultze-Berndt, own fieldwork)

(3)

Kalam tap tmey g-pay thing bad do-3PL.PST ‘they have done bad things.’ (Pawley 1994: 408)

186 Eva Schultze-Berndt (4)

Ewe ao, nye-mé-wɔ-e o no 1SG-NEG-do-3SG NEG ‘no, I didn’t do it’ (Ameka 1994: 72)

It has been claimed in the literature that the concept ‘DO’ is universal, and moreover, that it is universally linked to the notion of agency (see e.g. Goddard and Wierzbicka 1994: 42–43, Dowty 1979: 110–125). This equation of ‘DO’ with agency is at the heart of proposals to use DO in the semantic decomposition of verbs, that is, in defining the meaning of verbs by means of supposedly semantically primitive concepts. According to a related line of argument (Foley and Van Valin 1984: 47–53, Van Valin and LaPolla 1997: 102–129), an underlying DO operator is said to define Vendlerian activity predicates, i.e. those depicting continuous, atelic events. Activity predicates may be agentive, such as walk, sing or dance in English, or non-agentive roll, rain or cry (in the reading of involuntary crying). However, if one examines the meaning and functions of generalised action verbs cross-linguistically, it becomes clear that they are not necessarily restricted to expressing either agency or activity. Thus, the verbs that translate as ‘do’ in (1) to (4) above actually cover a wide semantic area ranging from causation to manifestation of a quality, inchoativity, verbaliser with sound-symbolic elements, and quotation marker. In this paper, I will argue that these different uses are related, and that the semantic domains potentially covered by ‘do’ verbs can be arranged in the form of a semantic map. While some of the uses are linked by the semantic components of agentivity, causation or activity, others may better be accounted for by a semantic component of internal causation or “manifestation”. Diagnostics for the different uses of ‘do’ verbs are their morphosyntactic valency, different selectional restrictions on their “subject” participant, and the semantics of the predicative complement expression accompanying them. This semantic diversity of ‘do’ verbs also has implications for their place in grammaticalisation theory. While the grammaticalisation of ‘do’ verbs in some well-described languages (particularly English do) has been the subject of considerable study and controversy (see van der Auwera and Genee 2002 for a recent concise overview), more general accounts of the development of ‘do’ verbs are noticeably absent from the literature on grammaticalisation, even from works devoted to the development of auxiliaries such as Heine (1993), Kuteva (2001), and Anderson (2006). Where

The semantic diversity of generalised action verbs 187

the grammaticalisation of ‘do’ verbs is discussed, the focus is on the result, i.e. the grammatical functions fulfilled, rather than on possible paths of development. For example, Lehmann (1995: 31–32, 115) mentions the use of ‘do’ verbs in constructions expressing habituality, predicate focus, predicate topicalisation, and emphasis. Heine and Kuteva (2002: 117–120) list five possible developments of a verb meaning ‘DO’ (which incidentally they equate with ‘MAKE’): causative, progressive marker, emphasis marker, obligation marker, and pro-verb (or “semantically empty predicate marker”). Anderson (2006: 357–358) gives examples of ‘do’ verbs occurring in periphrastic constructions for future as well as past tense, and habitual aspect. Jäger (2006: 268–294), the most extensive survey of grammatical functions of ‘do’ periphrasis, provides a seemingly heterogeneous list of functions which include intensive/emphatic marking, past and future tense, desiderative and conative modality, progressive and habitual aspect, but also completive aspect. He also mentions inchoative marking as a “rare type of periphrastic aspect” marked by a ‘do’ verb (2006: 276). All of these authors treat ‘do’ verbs as a semantically uniform category, e.g. of “schematic action” (Jäger 2006: 29); similarly, ‘do’ as a source of grammaticalisation simply gets subsumed under the “Action Scheme” by Heine (1993: 28, 34f.) and Anderson (2006: 358). While it is beyond the scope of this paper to explore the evidence for the development of generalised action verbs in grammaticalisation, the findings in this paper suggest that it may well be worth looking more closely at the semantic range of the lexical ‘do’ verbs, given that these differ considerably cross-linguistically. This might go some way towards explaining the heterogeneity of grammatical functions expressed by these verbs, since the lexical semantics of the verb may actually provide starting points for different grammaticalisation paths. For example, the causative function seems to depend on the presence of a “creation” sense of the lexical verb (Section 2.1). The use of some ‘do’ verbs in constructions ascribing a physical or emotional condition to an experiencer (i.e. in the sense of ‘feel’, Section 2.4.2) makes a development to a desiderative marker more plausible. Even an inchoative function can be found for non-periphrastic uses of ‘do’ verbs in some languages (Section 2.4.4). The findings presented here are tentative in the sense that they are based on a small, but genetically and areally diverse, convenience sample of languages where ‘do’ verbs exhibit a particularly interesting range of uses, and for which fairly explicit descriptions of the uses of the verb(s) in

188 Eva Schultze-Berndt question and/or texts are available. The languages investigated here (in addition to English and German) and the sources used are listed in Table 1. Table 1. Language sample on which this study is based Language

Affiliation

Sources

Samoan

Polynesian, Oceanic, Austronesian

Mosel and Hovdhaugen 1992; Mosel 1994

Kalam

Kalam, Trans-New Guinea (Papuan)

Pawley 1993, 1994, n.d.

Yimas

Sepik-Ramu (Papuan)

Foley 1991

Jaminjung

Western Mindi, Non-PamaNyungan (Australia)

own fieldwork; Schultze-Berndt 2000

Ewe

Gbe, Kwa, Niger-Congo

Ameka 1994; Essegbey 1999

Hausa

Chadic, Afro-Asiatic

Jaggar 2001; Newman 2000

Kham

Khamish, Tibeto-Burman, Sino-Tibetan

Watters 2002

Chantyal

Tibetic, Tibeto-Burman, SinoTibetan

Noonan 1999, 2001; Noonan and Grunow-Harsta 2002

All the ‘do’ verbs in the sample have the function of generalised action verb illustrated in (1) to (4). Their other functions will be surveyed in Section 2 of this paper. In Section 3, I summarise the range of functions of generalised action verbs in the form of a semantic map and will return to the question of implications for grammaticalisation.

2. Functions of ‘do’ verbs in a cross-linguistic perspective This section summarizes the main, recurring uses of generalised action verbs found in the languages surveyed. For convenience of exposition, the uses are divided into four major categories: causative verbs (section 2.1), activity verbs (section 2.2), mimetic verbs (section 2.3) and verbs of internal causation (2.4). It should be kept in mind though that it is the rule rather than the exception that a ‘do’ verb in a given language has more than one of these uses, and that the boundaries between uses are not always clear-cut (see also Section 3).

The semantic diversity of generalised action verbs 189

2.1. Causative verb 2.1.1. Verb of manufacturing or creation (CREATE) Verbs used as generalised action verbs often have a – presumably more basic – meaning of “creation” or “manufacturing”, i.e. a meaning close to English make. They have two participants, one of which is a true Agent (in the sense that the creation semantics presupposes agentivity and control). The other participant is an entity brought into existence (‘Effected Theme’). Languages with verbs of this nature are not hard to find. They include German (machen) and French (faire) as well as the majority of languages in our sample (all in fact with the exception of Jaminjung). Examples from Samoan and Kalam are included here for illustration. (5)

Samoan …

fai se kakou pe’epe’e do NSPEC.SG 1PL.INCL cream ‘make some cream for us’ (lit. ‘make some of our cream’) (Mosel and Hovdhaugen 1992: 771)

(6)

Kalam yad wog g-ng, mol g … I garden do-SS.FUT ditch do ‘when I’m about to make a garden, I make a ditch…’ (Pawley 1993: 119)

2.1.2. Productive causative marker (CAUSE) It is well known that generalised action verbs are also used in many languages in a more grammaticalised function, i.e. as productive markers in periphrastic causative expressions (see e.g. van der Auwera 1999). Again, French faire is a well-known example; languages in the sample displaying this use of ‘do’ verbs are Ewe and Chantyal.

190 Eva Schultze-Berndt (7)

Ewe tsidzadza lá wɔ-e bé me-tsí megbé rain DEF DO-3SG.DO COMP 1SG-remain back ‘The rain made it that I was late’, ‘The rain caused me to be late’ (Ameka 1994: 71)

(8)

Chantyal Ram-sə nāni-ra yep-nə la-i Ram-ERG baby-DAT stand-RES do-PF ‘Ram stood the baby up’ (lit. ‘Ram caused the baby to stand’) (Noonan and Grunow-Harsta 2002: 83)

Semantically, the causative function differs from the creation function and the GAV function in allowing for the expression of a caused event with a second participant, the Causee.

2.2. Activity verb (ACTIVITY) According to the literature, a typical function of generalised action verbs is their use as verbaliser in phrasal predicates denoting an activity or accomplishment involving a controlling agent, with a non-finite verb, a deverbal noun or any other nonverbal event expression as a complement. This use, in so far as it is regarded as overtly reflecting the decompositional structure of the predicate in question, forms one of the bases for claims about the agentive nature of ‘do’. Indo-Aryan and Iranian languages and Japanese are well-known examples of languages possessing a large number of complex predicates of this type. They are also found in almost all languages of the sample (though data are too scarce for Yimas and Kham), but Hausa is a language where they abound. The complement of the Hausa verb yi ‘do’ in these expressions is described as a ‘dynamic noun’ – presumably meaning a nominal describing an event – by both Jaggar (2001: 430) and Newman (2000: 434).

The semantic diversity of generalised action verbs 191

(9)

Hausa zân yi màganāÁ dà shī FUT:1SG do speaking with 3SG.M ‘I’ll speak with him’ (Jaggar 2001: 430)

(10)

Hausa tā 3SG.F.PFV

yi do

wà to

mijìntà husband:3SG.F.POSS

wāÁtsī throwing

dà kāyansà with thing:3SG.M.POSS ‘she threw out her husband’s things’ (lit. ‘she did to her husband throwing with his things’) (Jaggar 2001: 435) “Agentive” activities of the type discussed here need to be distinguished (as Van Valin and LaPolla 1997 do by the use of two different operators, do´ and DO) from non-agentive expressions which may qualify as activities in the Vendlerian sense. Examples of the latter are expressions of involuntary internal motion (‘tremble’), sound emission (‘crack’) or light emission (‘glitter’). These are subsumed here under the ‘exhibit property, manifest’ use (see Section 2.4.3), and also the ‘ideophone’ use (see section 2.3.2), since in many languages, both non-agentive and agentive activities (such as manner of motion) are expressed by sound-symbolic elements, often in combination with generalised action verbs.

2.3. Mimetic verb The term “mimetic verb” is used here, following Güldemann (2001), to cover two important uses of generalised action verbs, the use as quotative verb and the use as verbaliser with sound-symbolic elements.

2.3.1. Quotative verb (QUOT) Generalised action verbs that are also used to introduce quotations are quite common cross-linguistically. Examples can be found in Northern Australian languages (cf. Rumsey 1990, 1994; McGregor 1994), in a num-

192 Eva Schultze-Berndt ber of Papuan languages (Foley 1986: 119), and in African languages (Güldemann 2001: 237–245). In this use, the second participant of the ‘do’ verb, corresponding to the event brought about or to the effected theme, is a quotation. This use of generalised action verbs has been linked to the absence of both a linguistic and a cultural distinction between use of language and other types of behaviour (Rumsey 1990). In other words, speaking can be regarded as just another form of bringing about or ‘doing’ something. Languages may differ in whether only verbal quotations, i.e. direct speech, can fill the “quotation” slot, or whether reported thought or representation of (non-linguistic) sound can also be treated as a quotation (see section 2.3.2). In some languages, the verb may also take a nominal complement meaning e.g. ‘speech’, ‘word’, or ‘language’; this is the case e.g. for Samoan fai ‘do’ (Mosel and Hovdhaugen 1992: 672; Mosel 1994: 335– 336). For simplicity’s sake, only the reported speech function is considered here; this is illustrated with examples from Jaminjung and Samoan. (11)

Jaminjung “ba-rum” gani-yu=nu IMP-come 3SG>3SG-do=3SG.OBL ‘“come” she said (lit. she did) to him’

(12)

Samoan fai do

atu loa le tamaloa DIR then DEF man

‘o Saetane: PRES S.

“Saefafine, ‘ole’a ‘e fa’atali atu…” S. FUT 2.SG wait DIR ‘Then the man Saetane said: “Saefafine, wait…”’ (Mosel and Hovdhaugen 1992: 672) In many language descriptions, a generalised action verb in this use is assigned a separate sense of ‘say’; this also forms the basis for many claims about the grammaticalisation of a verb ‘say’ as a quotation marker. However, as Güldemann (2001, 2000) has argued in convincing detail, lexical speech verbs are only one possible source of quotative markers. Manner deictics (‘thus’), similative markers (‘like’) and, as demonstrated here, generalised action verbs may all function as quotative markers and may undergo further grammaticalisation to general complementisers not re-

The semantic diversity of generalised action verbs 193

stricted to reported speech and to markers of hearsay evidentiality (see in particular Güldemann 2002: 283).

2.3.2. Verbaliser with sound-symbolic expressions (IDEO) Another function of ‘do’ verbs, which is at least as common as their quotative use, is their use as a verbaliser with sound-symbolic elements (variously termed ‘onomatopoeia’, ‘ideophones’ or ‘sound-symbolic adverbs’). Although these two uses are not necessarily correlated (i.e. do not always co-exist in the same language), it has been proposed that they may be linked in that “an ideophone might be considered a word in quotation marks” (Samarin 1971: 150, attributed to A. Meeussen, cf. also Cohen et al. 2002: 247). Sound-symbolic elements are often, but not necessarily, iconic representations of noises. Cross-linguistically, they also frequently denote manner of motion (see examples (13) and (14)) and other activities, but also shape, texture or colour, which links them not only to the domain of activity verbs discussed in section 2.2, but also to the domain of verbs of internal causation which is the topic of section 2.4 below. In terms of their syntactic behaviour, sound-symbolic elements are not always exclusively linked to ‘do’ verbs, but can often be used adverbially with other semantically appropriate verbs (or nouns), e.g. verbs of motion. Chantyal and Hausa are examples of languages where this option exists (see the interchange in (13)), but there is a tendency for sound-symbolic elements to combine with ‘do’ verbs in a conjoined structure with the (semantically) main predicate. (13)

Chantyal A: khərəkhərə bɦi-ro! – B: cə khərəkhərə la-i steadily say-IMP that steadily do-PFV A: “continue talking!” – B: “I’ll keep on (talking)” (Noonan 1999: 588)

194 Eva Schultze-Berndt (14)

Hausa tā yi wup tā kaamāÁ shi 3SG.F do (IDEOPHONE) 3SG.F catch 3SG.M ‘she grabbed him in a flash’ (lit. ‘she did wup she caught him’) (Newman 2000: 474)

A generalised action verb may also be used to incorporate non-verbal mimetic expressions – i.e. gestures – into speech (also included in the domain of mimesis by Güldemann 2001). This use has been left out of consideration here because information in the literature is regrettably scarce. A further frequent function of generalised action verbs, is that of verbaliser with loanwords; this has been recognised as one of the important functions of ‘do’ verbs in the literature, related to the observation that the borrowing of verbs as verbs is cross-linguistically marked (Moravcsik 1975, van der Auwera 1999, Jäger 2004, Wichman and Wohlgemuth 2005, Muysken 2000: 184–220). This is linked to the functional domain of quotations and ideophonic expressions by Cohen et al. (2002: 247) because of the “expressive nature of … a loanword” and because “[q]uoting is equivalent to posing an item as existing in the language”. For reasons of space, this function will be left out of consideration here.

2.4. Verb of internal causation The functions of generalised action verbs reviewed so far – creation verb or causative verb, verbaliser with activity predicates, and mimetic verb – already manifest considerable diversity but are still compatible with the presumably prototypical association of ‘do’ verbs with agentivity. However, as already indicated in the introduction, generalised action verbs are not always agentive in nature. Rather, they may be used to predicate events that occur without the involvement of an agent, and also emotional conditions, qualities, conditions or states, and even state changes. It will be argued here that these uses can be subsumed under a general notion of ‘internal causation’, which however has to be distinguished from agentivity (or activity).

The semantic diversity of generalised action verbs 195

2.4.1. HAPPEN Generalised action verbs in various languages have uses in which they are best paraphrased as ‘happen’. In this case, there is no Agent participant; rather, the first participant (“subject”) of the verb is an event, which is depicted as coming about without any apparent cause. The resulting expression may be intransitive as in (15), or transitive, if a participant affected by the event can be added as a complement of the verb as in (16). (15)

Kalam mñab nb ak ned wagn ak g g-ek … country such this first origin DEF do do-3SG.PST ‘the place where this originally happened…’ (Pawley 1994: 408)

(16)

Ewe Nú véví wɔ Kofí (…) thing serious do Kofi ‘something serious happened to Kofi’, ‘Kofi got injured’ (lit. ‘something serious did Kofi’) (Essegbey 1999: 90)

I subsume here under the ‘happen’ use occurrences of a generalised action verb in impersonal constructions. In languages like Kalam, this type of construction is the main means of conveying a physical or emotional condition. English translations often involve a verb like ‘feel’, as illustrated in (17). This is however misleading, in that unlike with English feel and the cases discussed in Section 2.4.2, the experiencer is the object, not the subject in this construction. (17)

Kalam yp tap g-p 1SG.OBJ sickness do-PRS ‘I feel/am sick’ (lit. ‘sickness works/happens on me’) (Pawley 1994: 397)

196 Eva Schultze-Berndt 2.4.2. FEEL In a number of languages, ‘do’ verbs can function as true equivalents of ‘feel’. For example, the generalised action verb tì in the Papuan language Yimas not only has an impersonal use similar to that illustrated in (17) for Kalam to convey a physical or emotional condition (see Foley 1991: 301 for examples), but at least in some complex expressions denoting an experience or desire, the experiencer is the subject. (18)

Yimas ŋay, ama tpuk am-t-wampuŋ ama-na-tì-kia-k Mo 1SG sago.dish eat-NFIN-desire 1SG.SBJ-DEF-dodo.at.night-IRR ‘Mama, I feel like eating sago’ (Foley 1991: 472, l. 79).

Another example is Jaminjung, where the verb -yunggu forms complex predicates with non-inflecting predicates (“preverbs”) expressing emotional or physical condition of animates, as shown in (19). (19)

Jaminjung yarl / butharl nga-yunggu-m itchy / sad 1SG>3SG-do-PRS ‘I am/feel itchy / sad’ (lit. ‘I do itchy / sad’)

In Hausa, too, complex predicates formed with yi ‘do’ not only express activities (see (9) and (10)), but also emotions such as ‘being angry’, ‘being jealous’, ‘being sad’, ‘being happy’, ‘feeling regret’, or ‘feeling desire’, with the experiencer as subject. In many of these cases however yi is interchangeable with a more specific verb ji ‘feel’ (Jaggar 2001: 430–432).

2.4.3. ‘exhibit PROPERTY’ Whereas the ‘feel’ use described in the previous subsection at least requires an animate participant, some languages even possess ‘do’ verbs which may appear with a complement referring to a substance or a property predicated of an inanimate participant. In this case, the verb may have to be

The semantic diversity of generalised action verbs 197

translated with a copular verb ‘be’ in languages like English and German. An example of a verb with this range of uses is Ewe wɔ ‘do’, illustrated in (20). (20)

Ewe é-wɔ ké / ba / nogoo / sue 3SG-do sand mud round small ‘it is sandy/muddy/round/small’ (lit. ‘it does sand etc.’) (Ameka 1994: 71)

A very similar use of the ‘do’ verb is also found in Hausa, e.g. in expressions of temperature or weight (Newman 2000: 474, Jaggar 2001: 430).

2.4.4. ‘turn into, become’ (INCHO) In a number of languages, including in this sample Samoan (Mosel and Hovdhaugen 1992: 113), Yimas (Foley 1991: 293–300) and Jaminjung, the generalised action verb can also translate as ‘become’, that is, function as an inchoative verb which encodes the transition, of an animate or inanimate, into a state or the transformation into a different entity, with a stative predicate or nominal as complement. From the point of view of theories associating ‘do’ with agentivity and control, this is of course a rather undesirable phenomenon. Two examples from Jaminjung are (21) and (22). (21)

Jaminjung wurrguru nganthu-wu-yu devil 2SG>3SG-POT-do ‘you will turn into a devil’

(22)

Jaminjung manyirri gan-unggu-m cool 3SG>3SG-do-PRS ‘it gets cool’ (i.e. mild; of yam species after rinsing it and burying it for a few days)

198 Eva Schultze-Berndt I will now argue that the notion of ‘internal causation’ may help to account for all of the uses of ‘do’ verbs discussed in this section, i.e. the ‘happen’ use, the ‘feel’ use, the ‘exhibit property’, and the inchoative use.

2.4.5. The notion of internal causation The term “internal causation” is introduced by Levin and Rappaport Hovav (1994, 1995) to account for the argument structure properties of a subclass of English verbs. They note that the notion of internal causation subsumes agency but goes beyond it, in that verbs like tremble or glitter, with nonagentive or inanimate arguments, can nevertheless describe internally caused eventualities in the sense that these eventualities are conceptualised as arising from inherent properties of their arguments (Levin and Rappaport Hovav 1995: 91)

Thus, verbs of internal causation are neutral with respect to agentivity or control. Rather, they attribute to a participant the manifestation of an event, a state change, a quality, or a condition which is not construed as externally caused, but as corresponding to an inherent property of this participant. For example, the verb –yunggu in Jaminjung is indeed restricted to internally caused state changes, e.g. ‘become big’ = ‘grow’2, ‘become mild’, ‘become night’, or ‘turn into a devil’. State changes like ‘break’ or ‘open’ – corresponding to what Levin and Rappaport Hovav (1995) term externally caused state changes – are encoded in Jaminjung by complex verbs which are not formed with -yu(nggu), but with the verb -ijga ‘go’ in a secondary sense of change of state (for details see Schultze-Berndt 2000: Ch. 5). With this restriction of the term ‘inchoative’ in mind, the difference between the ‘inchoative’ and the ‘manifestation of a condition’ reading of -yu(nggu) ‘do’, discussed in subsection 2.4.3, may be attributed to differences between the predicative elements that the verb combines with. The two functions are thus closely related. The ‘inchoative’ reading arises with nominal predicates and stative predicates: here -yu(nggu) contrasts with the verb -yu ‘be’ which is used to form stative rather than inchoative expressions with the same predicates. Predicates of bodily and emotional condition, such as yarl ‘itchy’ or butharl ‘sad’, on the other hand, generally only combine with -yu(nggu) and not with -yu ‘be’, and therefore have to be regarded not as stative, but as dynamic predicates. Consequently, there is

The semantic diversity of generalised action verbs 199

no way to express a difference e.g. between “inchoative” ‘I am getting itchy’ and “stative” ‘I am itchy’. Although the evidence is currently not conclusive, I would like to suggest that ‘do’ verbs may be used to predicate a condition or property of a participant precisely in the case that it is an inherent property, i.e. not a state resulting from manipulation, but nevertheless externally manifested and relevant. This holds for the examples discussed in subsection 2.4.3. The analysis in terms of internal causation can probably also apply to most cases where a generalised action verb is claimed to have a reading of ‘feel’, as in languages mentioned in subsection 2.4.2 above. Possibly, English feel is not an appropriate translation equivalent in these cases either because this verb ascribes an experience to the participant, while the generalised action verb focuses on the manifestation of the condition in the participant. At least for Jaminjung, there is evidence (summarised in Schultze-Berndt 2000: 461–462) that physical or emotional conditions are not ascribed to a non-speaker unless the speaker has some behaviourally manifested evidence for this, such as lively behaviour for happiness. Finally, the ‘happen’ use (subsection 2.4.1) is also covered by an analysis in terms of internal causation, since here, an event spontaneously occurs without the involvement of an external Effector or Causer. The analysis proposed here is supported by those authors who have not limited themselves to listing translation equivalents such as ‘do’, ‘say’, ‘feel’, or ‘happen’ for a given ‘do’ verb. A very poignant characterisation is offered by Langdon (1977) for corresponding verbs in Yuman languages of California. These verbs, glossed as ‘say’, occur in construction with sound-symbolic elements from the semantic domains of noise, motion, emotion, pattern, consistency and shape – i.e. they have a semantic range very similar to that described here for generalised action verbs. Langdon suggests that the relevant construction can be characterised as descriptive of a characteristic configuration of an individual, object, or situation by focusing on the overt immediately verifiable evidence, while disclaiming any active, deliberate, or volitional intent. (Langdon 1977: 7)

The Yuman examples, as well as the range of uses of ‘do’ verbs in some languages in our sample, also point to a close link between the uses subsumed here under internal causation, on the one hand, and quotative and ideophone constructions, on the other hand.

200 Eva Schultze-Berndt 3. A tentative semantic map In the case of verbs with such general meanings and wide ranges of functions as generalised action verbs, a semantic approach based on paraphrases (such as ‘manifestation of an event’) is clearly limited. Even more problematic is an approach in terms of decomposition since, as we have seen above, the verbs in question may themselves have as translation equivalents more than one of the supposed semantic primitives CAUSE, BECOME, DO, or HAPPEN which are currently used in decompositional approaches. Rather, the meaning of a semantically general verb can be approached, first, by carefully listing the syntactic environments that it can occur in, and second, by exploring its collocational potential, i.e. its potential to combine with other lexical items, e.g. in complex predicate formation. Its range of uses also depends on its degree of grammaticalisation and on the nature of the verb system in the language (in a language with a closed class of verbs, such as Jaminjung or Kalam, each of them, including the verb used as generalised action verb, is likely to have a wider range of functions than in a language with an open class of verbs). By comparing the range of functions of corresponding verbs in a number of languages, it is possible to come up with a network of functions, some of which are more closely related than others. Thus, if we find that a single verb never has two specific functions A and C unless it also has an additional function B, we can state that A and C are more distantly related, and B is an intermediate function in the network. By this method, it is possible to develop a “semantic map” with predictive value (cf. van der Auwera and Plungian 1998, Haspelmath 2003, van der Auwera and Temürcü 2006). A first – and tentative – outline of what such a semantic map might look like for generalised action verbs is provided in Figure 1. For reasons of space and readability, only six languages are represented in the map. The full information on the uses of ‘do’ verbs in all nine languages (which is compatible with the map as proposed here) is provided in Table 2. Since it is based on a small sample, the map represents proximities in “semantic space” rather than definite links between individual uses. The “generalised action verb” use is placed in the centre of the map not because it is claimed here that this is the central or basic function of ‘do’ verbs, but because it was used as a defining criterion to select the verbs.

The semantic diversity of generalised action verbs 201

CAUSE ACT IDEO CREATE DO (GAV)

QUOT HAPPEN FEEL INCHO PROPERTY

Samoan fai Kalam gChantyal la Ewe wɔ Jaminjung –yu(nggu) German machen Figure 1. A tentative semantic map of functions of generalised action verbs, representing information on six languages

202 Eva Schultze-Berndt

CREATE

ACT

IDEO

QUOT

HAPPEN

FEEL

PROPERTY

Samoan

fai

-

3

3

??

3

-

-

-

3

Kalam

g-

3

3

3

3*

-

3

-

-

-

Yimas

tì-

-

3

??

??

-

3

(3)

-

3

Jaminjung

-yunggu

-

-

3

3*

3

-

3

(3)

3

Ewe

wɔ

3

3

3

3*

-

3

3

3

-

-

3

3

3*

-

??

3

3

-

jəi-

-

3

??

(3)

-

-

-

-

-

la-

3

3

3

3

[3]

-

-

-

-

-

3

(3)

3

[3]

-

-

-

Hausa Kham Chantyal German *

yi

mach-

INCHO

Verb

CAUSE

Table 2. Functions of generalised action verbs in nine languages*

-: function not attested; 3: function attested; (3): function attested but appears to be less frequent (in terms of types of collocations); [3]: specific to QUOT: attested only for nonlinguistic quotation (e.g. “ah”), not for direct speech; 3*: specific to IDEO: sound-symbolic elements may combine with ‘do’ verb but also with other verbs; ??: not sufficient information.

The uses of ‘do’ verbs discussed here can be divided into two major categories, agentive (top half of the map) and non-agentive (bottom half of the map). Since the use as GAV (‘do what/it’) implies agentivity, it is not surprising that all verbs included in the survey have agentive uses, but not all have non-agentive uses (the latter holds for German, Chantyal, and Kham). The different non-agentive uses – ‘HAPPEN’, ‘FEEL’, ‘exhibit PROPERTY’ and INCHO(ative) – are not very strongly correlated, except for ‘FEEL’ and ‘PROPERTY’, represented by closer proximity on the map, since all languages in the sample that have the former also have the latter (except for Yimas, where however the ‘FEEL’ use seems to be marginal). As expected on the basis of other cross-linguistic studies, ‘do’ verbs have a strong affinity with creation verbs (GAVs of eight languages in the sample have both uses, that is all except for Jaminjung) and a somewhat weaker affinity with causation (only three of the nine languages); CAUSE is therefore placed at the margin of the map and next to CREATE to represent the strong semantic link between the two. At least in the sample of lan-

The semantic diversity of generalised action verbs 203

guages considered here, all generalised action verbs that are used as causative markers also have the creation sense, and it has been shown elsewhere that causative verbs are frequently derived from creation verbs (Moreno 1993). The link between quotative and ideophonic uses is not strongly confirmed by the sample – almost all of them use the ‘do’ verb with soundsymbolic elements (data are insufficient for Samoan and Yimas), but only two of them with quotations (and only one for both); for this reason, QUOT is placed at the margin of the semantic map. In the languages that do not use the ‘do’ verb, a specific verb of speaking (‘say’) takes over the quotative function. On the other hand, the study reveals a strong link between ideophonic uses and uses with agentive activity expressions (co-occurring in at least five or the nine languages). This is not too surprising, because the two often cannot be deliminated, e.g. in languages like Chantyal with a large repertoire of sound-symbolic expressions of manner of motion. However, the IDEO function is not ideally placed on the map as depicted in Figure 1, because sound-symbolic expressions are in fact also frequently employed to represent non-agentive activities such as involuntary internal motion and sound emission; this is the case e.g. in Jaminjung, Chantyal, Kalam, and German. In any case, the function of verbaliser with sound-symbolic elements turns out to be one of the most frequent functions of ‘do’ verbs. It may also be a central function in the sense that ‘do’ verbs may or may not entail agentivity, but that they always entail the perceivable manifestation of an event.

Notes 1.

2.

I will leave out of consideration here the case of true interrogative verbs, which appear to be rather rare crosslinguistically (around 10% of the languages in the sample considered by Idiatov and van der Auwera 2004, who also propose a functional explanation for this observation). The translation equivalent of ‘grow, become mature’ employs the ‘do’ verb, not one of the two change-of-state verbs, even in Kalam (Pawley n.d., entry wos g- ‘become mature’).

204 Eva Schultze-Berndt

References Ameka, Felix K. 1994 Ewe. In Semantic and Lexical Universals, Cliff Goddard and Anna Wierzbicka (eds.), 57–86. Amsterdam: Benjamins. Anderson, Gregory D.S. 2006 Auxiliary Verb Constructions. Oxford: Oxford University Press. Cohen, David, Marie-Claude Simeone-Senelle, and Martine Vanhove 2001 The grammaticalization of ‘say’ and ‘do’. In Reported discourse: a meeting ground for different linguistic domains, Tom Güldemann and Manfred von Roncador (eds.), 247–251. (Typological Studies in Language 52.) Amsterdam: Benjamins. Dowty, D. R. 1979 Word meaning and Montague grammar. The semantics of verbs and times in Generative Semantics and in Montague's PTQ. Dordrecht: Reidel Publishing Company. Essegbey, James K. 1999 Inherent complement verbs revisited: towards an understanding of argument structure in Ewe. Ph. D. diss., University of Leiden. Foley, William A. 1986 The Papuan Languages of New Guinea. Cambridge: Cambridge University Press. 1991 The Yimas Language of New Guinea. Stanford: Stanford University Press. Foley, Willam A., and Robert D. Van Valin Jr. 1984 Functional Syntax and Universal Grammar. Cambridge: Cambridge University Press. Güldemann, Tom 2001 Quotative Constructions in African Languages: A Synchronic and Diachronic Survey. Leipzig: Habilitationsschrift, Fakultät für Geschichte, Kunst- und Orientwissenschaften der Univ. Leipzig. 2002 When ‘say’ is not say: the functional versatility of the Bantu quotative marker ti with special reference to Shona. In Reported Discourse: A Meeting Ground for Different Linguistic Domains, Tom Güldemann and Manfred von Roncador (eds.), 253–287. (Typological Studies in Language 52.) Amsterdam: Benjamins. Haspelmath, Martin 2003 The geometry of grammatical meaning: semantic maps and crosslinguistic comparison. In The New Psychology of Language, vol. 2, Michael Tomasello (ed.), 211–242. Mahwah/NJ: Erlbaum.

The semantic diversity of generalised action verbs 205 Heine, Bernd 1993 Auxiliaries. Cognitive Forces and Grammaticalisation. Oxford: Oxford University Press. Heine, Bernd, and Tania Kuteva 2002 World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Idiatov, Dmitri, and Johan van der Auwera 2004 On interrogative pro-verbs. In Proceedings of the Workshop on the Syntax, Semantics and Pragmatics of Questions, I. Comorovski and M. Krifka (eds.), 17–23. 16th European Summer School in Logic, Language, and Information (ESSLLI). Jäger, Andreas 2004 The cross-linguistic function of obligatory ‘do’-periphrasis. In Proceedings of the 2004 Conference of the Australian Linguistic Society, Ilana Mushin (ed.). http://hdl.handle.net/2123/111 2006 Typology of Periphrastic ‘do’-Constructions. Bochum: Brockmeyer. Jaggar, Philip J. 2001 Hausa. Amsterdam: Benjamins. Kuteva, Tania 2001 Auxiliation: An Enquiry into the Nature of Grammaticalization. New York and Oxford: Oxford University Press. Langdon, Margaret 1977 Semantics and syntax of expressive “say” constructions in Yuman. Berkeley Linguistic Society 3: 1–11. Lehmann, Christian 1995 Thoughts on Grammaticalization. München/Newcastle: Lincom Europa. Levin, Beth, and Malka Rappaport Hovav 1994 A preliminary analysis of causative verbs in English. Lingua 94: 35– 77. 1995 Unaccusativity: At the Syntax-Lexical Semantics Interface. Cambridge, MA: MIT press. McGregor, William 1994 The grammar of reported speech and thought in Gooniyandi. Australian Journal of Linguistics 14 (1): 63–92. Moravscik, Edith 1975 Borrowed verbs. Wiener Linguistische Gazette 8: 3–30. Moreno, Juan Carlos 1993 ‘Make’ and the semantic origins of causativity: a typological study. In Causatives and Transitivity, Bernard Comrie and Maria Polinsky (eds.), 155–164. Amsterdam: Benjamins.

206 Eva Schultze-Berndt Mosel, Ulrike 1994 Samoan. In Semantic and Lexical Universals, Cliff Goddard and Anna Wierzbicka (eds.), 331–360. Amsterdam: Benjamins. Mosel, Ulrike, and Even Hovdhaugen 1992 Samoan Reference Grammar. Oslo: Scandinavian University Press. Muysken, Peter 2002 Bilingual Speech: A Typology of Code-mixing. Cambridge: Cambridge University Press. Newman, Paul 2000 The Hausa Language: An Encyclopedic Reference Grammar. New Haven: Yale University Press. Noonan, Michael 1999 Chantyal Dictionary and Texts. Berlin: Mouton de Gruyter. 2001 The Chantyal language. In The Sino-Tibetan Languages, Graham Thurgood and Randy LaPolla (eds.), 315–335. London: Curzon Press. Noonan, Michael, and Karen Grunow-Harsta 2002 Posture verbs in two Tibeto-Burman languages of Nepal. In The Linguistics of Sitting, Standing, and Lying, John Newman (ed.), 79– 101. Amsterdam: Benjamins. Pawley, Andrew 1993 A language that defies description by ordinary means. In The role of Theory in Language Description, William A. Foley (ed.), 87–129 (Trends in Linguistics. Studies and Monographs 69). Berlin: Mouton. 1994 Kalam Exponents of lexical and semantic primitives. In Semantic and Lexical Universals, Cliff Goddard and Anna Wierzbicka (eds.), 387–422. Amsterdam: Benjamins. n.d. Kalam Dictionary. Manuscript, Australian National University, Canberra. Rumsey, Alan 1990 Wording, Meaning, and Linguistic Ideology. American Anthropologist 92: 346–361. 1994 On the Transitivity of ‘Say’ Constructions in Bunuba. Australian Journal of Linguistics 14: 137–153. Samarin, William 1971 Survey of Bantu ideophones. African Language Studies 12: 130– 168. Schultze-Berndt, Eva 2000 Simple and complex verbs in Jaminjung. A study of event categorisation in an Australian language. Nijmegen: Ph. D. diss., Katholieke Universiteit Nijmegen.

The semantic diversity of generalised action verbs 207 Van der Auwera, Johan 1999 Periphrastic ‘Do’: Typological Prolegomena. In Thinking English Grammar, Guy A. J. Tops, Betty Devriendt, and Steven Geukens (eds.). Louvain: Peeters. Van der Auwera, Johan, and Inge Genee 2002 English do: On the convergence of languages and linguists. English Language and Linguistics 6 (2): 283–307. Van der Auwera, Johan, and Ceyhan Temürcü 2006 “Semantic maps in Typology”, In Encyclopedia of Language and Linguistics, vol. 11, Keith Brown (ed.), 131–134. Oxford: Elsevier Van der Auwera, Johan, and Vladimir Plungian 1998 Modality’s semantic map. Linguistic Typology 2: 79–124. Van Valin, Robert D., and Randy J. LaPolla 1997 Syntax: Structure, Meaning and Function. Cambridge: Cambridge University Press. Watters, David E. 2002 A Grammar of Kham. Cambridge: Cambridge University Press. Wichmann, Søren, and Jan Wohlgemuth 2005 Loan verbs in a typological perspective. To appear in Aspects of Language Contact: New Theoretical, Methodological and Empirical Findings with Special Focus on Romanisation Processes, Thomas Stolz, Dik Bakker, and Rosa Palomo (eds.). Berlin: Mouton de Gruyter.

Predicting a future change: Relative clauses of Japanese Tasaku Tsunoda

1. Introduction Most of the works in historical linguistics examine changes that occurred in the past. There are also works that look at changes in progress, e.g. Labov (1978). However, there seem to be very few works that make a prediction about a change that will or may occur in the future. For example, Hock (1991) considers changes in the past and those in progress, but it does not seem to consider future changes. The present work is an attempt to forecast a future change, dealing with relative clauses (‘RCs’) of Japanese.

2. Keenan and Comrie’s accessibility hierarchy Keenan and Comrie (1977) proposed the accessibility hierarchy and showed that cross-linguistically a position higher on the hierarchy is more likely to be relativized on than one lower on the hierarchy. See Figure 1. subject Malagasy Welsh Basque Catalan Japanese English

direct object

indirect object

oblique object

genitive

object of comparison

——> ——————> ———————————> ———————————————> ————————————————————> —————————————————————————>

Figure 1. Keenan and Comrie’s accessibility hierarchy

As shown above, in Japanese, relativization is possible from the subject down to the genitive. Examples are given below, cited from Tsunoda (2004).

210 Tasaku Tsunoda (1) The subject is relativized on. sensee ni repooto o oku-t-ta teacher DAT essay ACC send-LK-PAST ‘the student who sent an essay to the teacher’

gakusee student

(2) The direct object is relativized on. gakusee ga sensee ni oku-t-ta student NOM teacher DAT send-LK-PAST ‘the essay that the student sent to the teacher’

repooto essay

(3) The indirect object is relativized on. gakusee ga repooto o oku-t-ta student NOM essay ACC send-LK-PAST ‘the teacher to whom the student sent an essay’

sensee teacher

(4) The oblique object is relativized on. Taroo ga tegami o ka-i-ta Taroo NOM letter ACC write-LK-PAST ‘the pen with which Taroo wrote a letter’

pen pen

(5) The genitive is relativized on. Taroo ga pen o kari-ta gakusee Taroo NOM pen ACC borrow-PAST student ‘the student whose pen Taroo borrowed’ In terms of Lehmann’s (1986: 665–666) typology of relative clauses, Japanese RCs are external-head RCs, embedded, and pronominal.

3. Passives in Japanese There are two types of passives in Japanese: (i) V-rare- passives, e.g. (7), and (ii) V-te ar- passives (-te ‘infinitive’, ar- ‘be, exist’), e.g. (9). Consider the following pairs: (6) active, and (7) passive; and (8) active and (9) passive. (6) Hanako ga Taroo Hanako NOM Taroo ‘Hanako scolded Taroo.’

o ACC

shika-t-a. scold-LK-PAST

Predicting a future change 211

(7) Taroo ga Hanako ni Taroo NOM Hanako DAT ‘Taroo was scolded by Hanako.’

shikar-are-ta. scold-PASS-PAST

(8) Taroo ga kabe ni e Taroo NOM wall DAT picture ‘Taroo hung a picture on the wall.’

o ACC

(9) Kabe ni e ga kake-te wall DAT picture NOM hang-INF ‘A picture is (or, has been) hung on the wall.’

kake-ta. hang-PAST

ar-u. be-NONPAST

It has been noted in a number of works, e.g. Tsunoda (2003), that the Vrare- passives have been increasing, as against their active counterparts. (In contrast, as Tsunoda (2003) shows, V-te ar- passives are decreasing.) The present paper is concerned with V-rare- passives. Both actives and passives can be used in RCs. Compare (10), in which the RC involves an active, and (11), in which the RC contains a passive. (Examples cited from Takahashi 1985: 17.) (10) tenjoo kara tsurush-i-ta dentoo ceiling ABL hang-LK-PAST lamp ‘a lamp that [someone] hung from the ceiling’ (11) tenjoo kara tsurus-are-ta dentoo ceiling ABL hang-PASS-PAST lamp ‘a lamp that was hung from the ceiling’ In (10) the direct object is relativized on, while in (11) the subject of a passive is relativized on. In (11), the underlying clause is passivized first, and subsequently the passive subject (i.e. the erstwhile direct object) is relativized on.

4. Increase of V-rare passives and its effect on RCs As mentioned in Section 3, V-rare passives are increasing, and this is happening in RCs, too. That is, RCs such as (11) are increasing. Consequently,

212 Tasaku Tsunoda RCs such as (10) are decreasing. This change is causing an interesting change in RCs. I consulted the preprints for the 122nd meeting of the Linguistic Society of Japan (held on the 23rd, 24th June 2001), and selected papers that contain a fair number of RC and also whose author informed me of his/her age (as of the 23rd June 2001). They are eight in all. See Table 1. Table 1 shows the following tendencies. Tendency 1. With each of the authors, virtually all the RCs involve the subject of an active (i.e. active subject), the subject of a passive (i.e. passive subject) or the direct object. In contrast, there is no example or there are very few examples involving the indirect object, the oblique object or the genitive. Examples follow, cited from the author H. Table 1. Relative clauses of Japanese author

A B C D E F G H

age 52 49 45 41 39 35 39 27

active subject n % 46 54 18 62 13 57 10 53 31 56 15 50 17 55 12 32

passive subject n % 10 12 3 10 4 17 5 26 16 29 7 23 6 19 19 51

direct indirect oblique genitive object object object n % n % n % n % 29 34 0 0 0 0 0 0 8 28 0 0 0 0 0 0 5 22 1 4 0 0 0 0 3 16 0 0 0 0 1 5 7 13 0 0 1 2 0 0 4 13 0 0 1 3 3 10 8 26 0 0 0 0 0 0 6 16 0 0 0 0 0 0

(12) The active subject is relativized on. inga-kankee o arawas-u causal-relation ACC express-NONPAST ‘the word that expresses a causal relation’

total n 85 29 23 19 55 30 31 37

% 100 100 100 100 100 99 100 99

go word

(13) The passive subject is relativized on. zenken de arawas-are-ru previous:case INST express-PASS-NONPAST zentee presupposition ‘a presupposition that is expressed by means of the previous case’

Predicting a future change 213

(14) The direct object is relativized on. koko made mi-te k-i-ta here ALL consider-INF come-LK-PAST seeteki-riyuu-bun static-cause-sentence ‘the static causal sentences that [I] have been considering up to now’ In Section 2, I stated, on the basis of Keenan and Comrie’s work, that relativization in Japanese is possible from the subject down to the genitive. However, Keenan and Comrie were concerned with the possibilities of relativization. When we examine the frequency of RCs in actual usage, a very different picture emerges. Table 1 is merely the result of a crosssectional study that deals with only eight authors. (There are hundred of millions of Japanese speakers.) Nonetheless, Table 1 strongly suggests that RCs in Japanese are virtually confined to the subject (active subject and passive subject) and the indirect object. Tendency 2. The younger the author is, the smaller the number is of RCs modifying the direct object; cf. (10) and (14). For example, the percentage of such RCs is 34% with the author A (52 years old), but 16% with the author H (27 years old) — less than a third. In contrast, the younger the author is, the larger the number is of RCS modifying the passive subject; cf. (11) and (13). For example, the percentage of such RCs is 12% with the author A, but it is 51% with the author H — more than four times. With the author H, as many as 51%, i.e. half of RCs modify the passive subject. As additional piece of evidence, in July 2007, I asked about 100 students (around 20 years old) concerning the acceptability of (10) and (11). Virtually all of them regarded (11) acceptable and (10) not acceptable.

5. Predicting a future change in Japanese RCs Obviously the data shown in Table 1 do not reflect the entirety of the grammar of all speakers of Japanese. It looks at only eight speakers. They hail from different parts of Japan. Their linguistic biographies are diverse. Table 1 is concerned with technical writing, that is, a very formal style, and it does not take an informal style into account. (The style of technical writing is no doubt among the most formal styles in the languages that possess it.) Nor is there any guarantee that the data in Table 1 accurately reflect a

214 Tasaku Tsunoda change that may be happening in Japanese. The data are based on a crosssectional study, and not a longitudal study. Nonetheless, there is no doubt that V-rare- passives are increasing, as has been noted by a number of linguists, including Tsunoda (2003). In my observation of Japanese, V-rare- passives are far more frequent in a formal style than in an informal style, and they are far more common in technical writing than in other formal styles, such as newspaper articles. In view of the above, Table 1 seems to present the very front of a change that is happening in the RCs of Japanese. Chappell (1986: 1025–1027, 1033–1034) notes regarding Mandarin Chinese that, in one of their uses, the bei passives have begun to be used like the passives of European languages. They have lost (i) the adversity reading, (ii) the requirement of an overt agent, and (iii) the requirement of a perfective predicate. This use of bei passives occurs in the written language only, such as newspapers. Chappell refers to such passives as ‘translatese’ and ‘journalese’, and she attributes this use to ‘the influence of European languages in translation’. It seems likely that the same cause is responsible for the increase of the V-rare- passives of Japanese. Now, what will happen to the RCs of Japanese if this change spreads to other speakers and other styles, and eventually to the entirety of Japanese? As seen in Section 4, if we look at their frequency in actual usage, and not their possibilities, the RCs are virtually confined to the subject (active subject and passive subject) and the direct object. Furthermore, the RCs modifying the direct object, e.g. (10) and (14), are decreasing and those modifying the passive subject (which is the erstwhile direct object), e.g. (11) and (13), are increasing. This strongly indicates that, if this change progresses, in the future Japanese, the RCs will modify the subject only, and not any other position on Keenan and Comrie’s accessibility hierarchy. As shown by Keenan and Comrie (1977: 69–70), in many Western Malayo-Polynesian languages, such as Malagasy and Tagalog, only the subject can be relativized on. That is, if, in Japanese, the change described above progresses, the RCs will be confined to the subject, and, concerning the formation of RCs, Japanese will undergo a typological change and become a language like Tagalog and Malagasy.

Predicting a future change 215

6. Summary In Japanese, relativization is possible from the subject to the genitive, in terms of Keenan and Comrie’s accessibility hierarchy. However, if we look at their frequency in actual usage, and not their possibilities, they seem to be confined to the subject and the direct object. V-rare- passives are increasing, as against their active counterparts, and, as the data obtained show, this is happening in RCs, as well. That is, those modifying the direct object are decreasing, and those modifying the passive subject (which is the erstwhile direct object) are increasing. If this change progresses, the Japanese RCs will be restricted to the subject (active subject and passive subject), and, regarding the formation of RCs, Japanese will undergo a typological change and become a language like Tagalog and Malagasy, in which only the subject can be relativized on. The present study suggests a new method for historical linguistics: on the basis of a cross-sectional study it is possible to make a prediction that may occur in the future.

Notes 1.

It is my pleasure to dedicate this paper to Prof. Dr. Christian Lehmann to honour his enormous contribution to the science and also to express my appreciation of his friendship, which started when we met at the 13th International Congress of Linguists (Tokyo, August 1982). The present paper is a revised and updated version of Tsunoda (2004). I am grateful to (i) the following people who provided invaluable comments on the draft of Tsunoda (2004): Bernard Comrie, Nobuko Kibe, Satoshi Kinsui, Naonori Nagaya, Motoyasu Nojima, Hiroki Nomoto, Ayako Sakamoto, Kan Sasaki, Takuzoo Satoo, the late Taroo Takahashi, Mie Tsunoda, (ii) the eight authors who informed me of their age, and (iii) Kurosio Publishers for allowing me to publish a revised version of Tsunoda (2004).

216 Tasaku Tsunoda

References Chappell, Hilary 1986 Formal and colloquial adversity passives in standard Chinese. Linguistics 24 (4): 1025–1052. Hock, Hans Heinrich 1991 Principles of Historical Linguistics. Second revised and updated edition. Berlin/New York: Mouton de Gruyter. Keenan, Edward L., and Bernard Comrie 1977 Noun phrase accessibility hierarchy and universal grammar. Linguistic Inquiry 8 (1): 63–99. Labov, William 1978 Sociolinguistic Patterns. Oxford: Basil Blackwell. Lehmann, Christian 1986 On the typology of relative clauses. Linguistics 24 (4): 663–680. Takahashi, Taroo 1985 Gendainihongo no boisu ni tsuite [On voice in Modern Japanese]. Nihongogaku April issue: 4–23. Tsunoda, Tasaku 2003 Henna judoobun: ‘Kuma ga yamu o ezu shasatsusaremashita’ [Strange passives: ‘The bear was unavoidably shot dead]. In Nagoya Kotoba no Tsudoi[:] Gengo Kagaku Ronshuu [Nagoya Linguistic Circle: Papers in Language Science], Ikudoo Tajima and Kazuya Niwa (eds.), 35–42. Nagoya: Graduate School of Letters, University of Nagoya. 2004 Nihongo no rentaishuushokusetsu: Firipin o toorisugite madagasukaru ni tassuru? [Relative clauses of Japanese: Will they go past the Philippines and reach Madagascar?] Nihongo no bunseki to gengoreikeeron [Analysis of Japanese and language typology], Taroo Kageyama and Hideki Kishimoto (eds.), 559–571. Tokyo: Kurosio Publishers.

Part IV Grammaticalization and constructions

The catalytic function of constructional restrictions in grammaticalization1 Gabriele Diewald

1. Introduction This paper investigates the catalytic function of constructional restrictions in grammaticalization. Drawing on results of grammaticalization studies and constructional approaches to language, it will be shown that in the diachronic development of grammatical items, constructions of very specific types may arise in particular diachronic stages and disappear again as grammaticalization proceeds. This is illustrated by a case study on the development of the modal particle ruhig in German. The development of an intermediate constructional stage is shown to be a necessary step for encoding the new pragmatic and contextual meaning; whereas in more advanced stages, when the particle has absorbed its new pragmatic meaning, the constructional restrictions become redundant and are given up again. These findings call for a revision of the long-standing assumption that the function of an item as modal particle in German is essentially and by definition associated with particular contexts or speech act types. Though this is true for individual stages in the grammaticalization paths of modal particles, it can no longer be upheld to be true for the fully grammaticalized stages of those items.

2. Constructions, contexts and grammaticalization As the relevance of constructions in grammaticalization has long been acknowledged, there is no need for an extended treatment of this issue (cf. Lehmann 1992: 406, 1995 [1982] Bybee, Perkins, and Pagliuca 1994: 11, Bisang 1998: 20, Traugott 2003, [to appear], Diewald 2006a, 2006b, 2006c). For a definition of the notion of “construction”, this paper follows the widely held view reflected in Goldberg (2006):2 “Any linguistic pattern is recognized as a construction as long as some aspect of its form or function is not strictly predictable from its component parts or from other constructions recognized to exist. In addition, patterns

220 Gabriele Diewald are stored as constructions even if they are fully predictable as long as they occur with sufficient frequency.” (Goldberg 2006: 5)

In diachronic grammaticalization studies, it has proved useful to distinguish between more general, abstract and transparent constructions on one hand, and more specific, idiosyncratic and non-transparent constructions on the other. Although these distinctions are not discrete, and the totality of constructions in a synchronic stage of a language may be seen as forming a continuum between the most general and the most idiosyncratic constructions, it is necessary to single out different types of more idiosyncratic constructions. Fillmore, Kay, and O’Connor (1988) distinguish between different degrees of idiomaticity or idiosyncrasy, and postulate different types of idiomatic constructions. The first type is called “formal or lexically open” idioms, and defined as “syntactic patterns dedicated to semantic and pragmatic purposes not knowable from their form alone.” (Fillmore, Kay, and O’Connor 1988: 505) Formal idioms are constructions whose compositionality is reduced, i.e. at least some part of their form-meaning correspondence has to be treated holistically and cannot be derived in its totality from other constructions or from a combination of other constructions. However, they are fully productive, as their syntactic positions are not filled with lexically fixed items. Another type of constructions is called “extragrammatical idioms”, and described as follows: “Such expressions [i.e. extragrammatical idioms] have grammatical structure, to be sure, but the structures they have are not made intelligible by knowledge of the familiar rules of the grammar and how those rules are most generally applied” (Fillmore, Kay, and O’Connor 1988: 505). In order to characterize the diachronic stages in grammaticalization in a more detailed way, this paper takes up the suggestion in Diewald (2006b, 2006c) to correlate the notions of formal idioms and extragrammatical idioms by Fillmore, Kay, and O’Connor (1988) with two particular context types that are specific to grammaticalization, namely the critical context and the isolating context, which constitute the second and third stages in the relative chronology of grammaticalization processes (cf. Diewald 2002). Critical contexts are new and peripheral structures restricted to a small group of lexical items, they are not fully analyzable by the rules of the relevant linguistic system, and display a kind of multiple structural and semantic opacity that cannot be reduced to a combination of known constructions without losing information. A critical context invites several alternative interpretations – among them the new grammatical meaning –

The catalytic function of constructional restrictions in grammaticalization 221

and functions as a kind of catalyst, which usually disappears in the later development of the grammaticalizing item. Thus, the critical context may be equated to the notion of extragrammatical idiom. Isolating contexts bring about the consolidation of the grammaticalization process, i.e. the re-organization and differentiation of the grammatical formatives and the paradigmatic structure that is the target category of the ongoing grammaticalization process. This is achieved by isolating the new grammaticalized meaning as well as the older, more lexical meaning by the development of specific linguistic constructions that favor one reading to the exclusion of the other. Thereby, the new grammatical meaning becomes independent of conversational implicature and the linguistic element under grammaticalization is truly polysemous. Isolating contexts are a type of formal idioms because they combine the feature of partial similarities (and partial inheritance relations) with other constructions with the feature of a unique form-meaning correspondence, which cannot be reduced to combinations of other constructions. They qualify as partially productive, idiomatic constructions in the constructional sense of the term. The stages and their corresponding construction types are given in Table 1 (adapted from Diewald 2006c). Stage I has been put into the table for the sake of completeness; it will not be discussed any further. Table 1. Context types in grammaticalization (adapted from Diewald 2006c) Meaning / Function

Construction types (Fillmore, Kay and O’Connor 1988)

I) preconditions of untypical grammaticalization contexts

conversational implicature

no particular construction type; compositional

II) triggering of critical grammaticalization context

multiple opacity

“extragrammatical idioms”3

III) reorganization & differentiation

polysemous / heterosemous items

“formal or lexically open idioms”

Stage

Context

isolating contexts

222 Gabriele Diewald 3. Modal particles – definition and constructional restrictions Modal particles (MPs) in Present Day German (PDG) are acknowledged to form a relatively closed class of items sharing the following properties: non-propositional, pragmatic meaning or function (see below), propositional or utterance scope (in contrast to focus particles and scalar particles, which have constituent scope), no constituent value (in contrast to adverbials), restricted to the middle field, i.e. the topological section right of the finite verb in declarative sentences, associated with particular “Satzmodi”. Furthermore, it has been shown that MPs truly are a grammatical class which developed via a grammaticalization process (Diewald 2006a, Autenrieth 2002, Diewald and Ferraresi [in print]). Table 2. Distribution of MPs in different sentence types (adapted from Thurmair 1989: 49)

aber auch bloß denn doch eben eigentlich einfach halt ja mal nur ruhig schon vielleicht wohl

Declarative

whinterrogative

yes-no interrogative

Imperative

+

+ + + +

+

+ +

+ +

+

+ +

+

+

+

+

Exclama tive

whexclamative

Optative

+

+ +

+ + + + +

+

+ +

+ +

+ +

+

+

+

+ +

+ + + + + + +

+

+

Their grammatical function consists of relating the utterance to a proposition which the speaker regards as relevant and given, and to which s/he relates the actual utterance, i.e. to a specific type of pragmatic presupposition. By referring back to something communicatively given, the MP marks the utterance containing it as non-initial. As the interchange of

The catalytic function of constructional restrictions in grammaticalization 223

initial and responsive turns is the decisive structural device of spoken interaction, this relational function of the MPs clearly is an indispensable grammatical device for structuring discourse. With the help of MPs, the speaker marks the turn as non-initial and responsive, and thus is able to manipulate and modify the ongoing interchange. This paper is concerned with a property of the modal particles that is well-known but not very well understood, namely the restrictions of MPs to certain sentence types and/or speech act types (see Table 2). However, these restrictions cannot be described by just referring to structural and/or illocutionary notions (e.g. sentence types like verb first, verb second, verb final, or speech act types like assertives, directives etc.) as they usually imply specific and complex feature constellations including particular lexical and morphological restrictions (see examples below). Therefore, lacking a better term, these complex multi-feature restrictions are referred to by the German term satzmodus. Furthermore, it is claimed here that they qualify as idiomatic constructions, more precisely as formal idioms in the sense of Fillmore, Kay, and O’Connor 1988, i.e. as constructions whose compositionality is reduced in specific ways, though they display a compositional nature on the surface. The following examples illustrate these satzmodus restrictions. The MPs bloß and nur are only found in exclamative and imperative sentences, i.e. in utterances with an expressive or a directive illocutionary force, and in wh-questions with an exclamative function. (1)

Wenn er bloß/nur hier wäre! ‘If he bloß/nur were here!’

(2)

Hättest du damals bloß/nur deinen Mund gehalten! ‘If you had bloß/nur kept your mouth shut!’ (Thurmair 1989: 182)

(3)

Bleib mir bloß/nur vom Leib! ‘Do not bloß/nur bother me!’

(4)

Wo hast du das bloß gelernt? ‘Where did you bloß learn it?’

The MP denn, on the other hand, is generally classified as a question particle, which, indeed, is its most frequent use:

224 Gabriele Diewald (5)

Was machst du denn hier? ‘What are you denn doing here?’

(6)

Kommst du denn? ‘Are you denn coming?’

The MP eben is restricted to assertive speech acts (statements, declarative sentences with an explanatory and expository force) and to directive speech acts: (7)

Deutsch ist eben schwer. ‘German is eben difficult.’

(8)

Mach es eben noch mal! ‘Do it eben once more!’

The MP schon is found in directive speech acts and assertive speech acts (Helbig 1988: 201–206, Thurmair 1989: 146–154). It is not normally allowed in questions with one exception. It does appear in wh-questions, if they have a rhetorical function (Meibauer 1994). (9)

Was ist schon dabei? ‘What does it schon matter?’

(10)

Wer kann das schon wissen? ‘Who should schon know this?’

Old MPs like ja and doch show the fewest restrictions. The MP doch is possible in any sentence type except true questions (see (11) to (13)), the MP particle ja appears everywhere except in questions (see (14) to (16)): (11)

Martin ist doch gestern gekommen. ‘Martin came doch yesterday.’

(12)

Komm doch her! ‘Come doch here!’

(13)

Wenn er doch käme! ‘If he doch came!’

The catalytic function of constructional restrictions in grammaticalization 225

(14)

Martin ist ja gestern gekommen. ‘Martin came ja yesterday.’

(15)

Lies ja den Brief! ‘Read ja the letter!’

(16)

Das ist ja ein Ding! ‘That’s ja a bit thick!’

All these restrictions are well-known standard-issue examples in grammars, textbooks and linguistic studies. Thus, for the moment being, it may be concluded that, in PDG, many MPs are restricted to particular satzmodi. However, it is suggested here that we should not contend ourselves with this observation, as this would confirm a misconception of the function of the grammaticalized class of modal particle. The misconception consists of treating these restrictions as evidence for the grammatical functionality of the MPs. In other word, the satzmodus restriction usually has been wrongly treated as the target or functional endpoint of the grammaticalization of modal particles. There are, however, reasons to challenge this assumption. The first counter-argument concerns one of the basic tenets of grammaticalization theory, namely that fully grammaticalized categories are maximally free of contextual restrictions, display a high degree of paradigmaticity and are largely obligatory (cf. Lehmann’s parameter of grammaticalization). Therefore, the existence of restrictions indicates that the process of grammaticalization is not yet completed. The development of satzmodus restrictions obviously is the opposite of the loss of contextual restrictions and thus can hardly be the endpoint of a grammaticalization process. Secondly, empirical findings on the particle ruhig, which serves here as an exemplar case, indicate that the alleged satzmodus restrictions do not hold to the degree grammar books and linguistic literature suggest. In other words, the particle seems to be losing its satzmodus restrictions in PDG. From these considerations, the following hypotheses are derived: (a)

More grammaticalized MPs tend to show less satzmodus restrictions.

(b)

Younger, less grammaticalized MPs tend to be restricted to particular satzmodi.

226 Gabriele Diewald (c)

The grammaticalization of MPs includes the development of satzmodus restriction as a necessary and catalytic, but transient step on the way to a fully grammaticalized modal particle.

(d)

The grammaticalization process of the MPs includes a phase of building up satzmodus restrictions, which strictly speaking is a precondition to the grammaticalization process itself, and then, if the process continues towards complete grammaticalization, a phase of loss of satzmodus restrictions.

The following section substantiates these hypotheses by an investigation of the development of the MP ruhig.

4. The diachronic development of the modal particle ruhig: the rise and decline of satzmodus restrictions Ruhig is used as a modal particle and as an adjective. In (17), ruhig is an attributive adjective and can be substituted by adjectives with similar meanings like ‘quiet, peaceful, silent’. The predicative use of the adjective ruhig is also possible and frequent. (17)

Er ist ein ruhiger Mensch. ‘He is a quiet person.’

In its use as a modal particle, ruhig is subject to clear-cut satzmodus restrictions. In all grammars, reference books and linguistic studies on the particle use of ruhig, e.g. in Helbig (1988), Thurmair (1989: 188f.), Meibauer (1994: 234), we find two types of possible contexts. In the first context, the MP ruhig appears in sentences with the morphological imperative or its substitutive forms – i.e. the subjunctive I: (18)

Komm/Kommen Sie ruhig herein. ‘Come / (subj. I, 3 pl.) ruhig in.’

In the second context, the MP ruhig appears in declarative sentences (i.e. V2), which contain the modal verb können, dürfen or – less frequently – sollen, and which are used as indirect directive speech acts:

The catalytic function of constructional restrictions in grammaticalization 227

(19)

Du kannst/darfst/sollst ruhig rein kommen. ‘You can/may/should ruhig come in.’

(20)

Da darf es ruhig ein bißchen später, so zwischen 4 und 5 Uhr, sein. ‘It may be ruhig a bit later, about between 4 and 5 o’clock,.’ (Keil 1990: 45)

(19) exemplifies an ambiguous usage, where ruhig can be interpreted as an adjective used predicatively or adverbially, on one hand, or as a modal particle on the other. This ambiguity arises in all instances with ruhig in a declarative sentence in the middle field, accompanied with a modal verb, an infinitive of an action verb, and an animate subject. Example (20), on the other hand, illustrates one type of isolating contexts for the MP use of ruhig. The combination of the formal subject es and the stative predicate excludes the interpretation as an adjective, i.e. neither the predicative nor the adverbial reading are possible here. The only sensible interpretation of ruhig is the one as a modal particle. In short, the MP ruhig is restricted to the broad class of directive speech acts, i.e. to speech acts whose illocutionary point – according to Searle (1979: 2) – is “an attempt to get the hearer to do something”. However, utterances with ruhig do not cover the whole range of directives: ruhig never appears in orders or commands, but is restricted to different types of permissions, advice, general suggestions. An act of permission can be defined as “a directive speech act the recipient has asked for”. Thus, ruhig is restricted to directive speech acts that are recipient-oriented or reactive. These reactive speech acts are not initiated solely by the speaker, but also by the hearer, i.e. in some way or other, they react to an intention or need of the hearer. This is one part of the meaning the modal particle ruhig adds to an utterance. In addition to this, ruhig expresses a particular type of contrast. This contrast concerns the supposed versus the actual attitude of the speaker towards the proposition. Ruhig means that the speaker does not have objections despite of the hearer’s suspicion that he (the speaker) might have objections. It indicates ‘a contrast between the expected attitude of the speaker and the actual attitude of the speaker concerning the imminent action’. By using ruhig the speaker says: ‘in contrast to your/somebody’s expectation (irrelevant reservations), I do not have objections’. This meaning can be represented as an underlying dialog, in which the initial turn is not uttered but is taken to be implicit.

228 Gabriele Diewald A (addressee of the ruhig-utterance): B (speaker):

May (I do) proposition, although you do not want proposition? RUHIG proposition!

Thus, in an example like (20), the particle meaning can be paraphrased as (20´): (20)

Da darf es ruhig ein bißchen später, so zwischen 4 und 5 Uhr, sein. ‘It may be ruhig a bit later, about between 4 and 5 o’clock.’ (Keil 1990: 45)

(20´)

‘In contrast to your presupposition that I might object to it being later, I say that it may be later, about between 4 and 5 o’clock.’

To sum up, ruhig as an MP is restricted to directive speech acts that are recipient oriented/reactive and expressed through highly specific linguistic constructions. However, this is only one part of the story. It is suggested here that ruhig today is expanding to contexts it has not been used in as an MP before. The chronology of the building up and loosening of satzmodus restrictions can be verified by a closer look at the diachronic development. As there is no comprehensive work on the diachronic development of ruhig to build upon, the diachronic study presented here is the first step in a thorough investigation into the history of this item.4 Therefore, the following paragraphs present but intermediate results, which, nevertheless, are highly suggestive and support the hypotheses set out above. The adjective or adverb ruhig first means ‘free of work’, then ‘free of external movement’, and finally ‘free of internal movement’; ‘calm’, ‘quiet’, ‘silent’ (DWB Bd. 14, s. v. ruhig). Like other adjective-derived modal particles (e.g. bloß) ruhig starts out with a privative relational meaning which develops along the known paths of semantic change. According to the handbooks, the modal particle ruhig first appears in the nineteenth century in imperative sentences or in declaratives reporting on imperative speech acts. These statements are confirmed by the corpus data of ENHG and NHG investigated so far. In addition, they provide some new facts which suggest a relative chronology of relevant types of contexts. In the whole EHNG corpus (fourteenth to seventeenth century), there were only four instances of ruhig, all of them used as an adjective or adverb. The corpus for the eighteenth century has 21 instances of ruhig, all of them uses as adjective or adverb. However, beyond the regular corpus, individual instances

The catalytic function of constructional restrictions in grammaticalization 229

of ambiguous cases could be discovered for the eighteenth century. One of them is example (21), taken from a poem by Gellert: (21)

„Wenn kannst du dir ein solches Ansehn geben? Und wenn bewundert dich die Welt?“ “Schweig“, rief der Gaul, „und laß mich ruhig pflügen; …“ ‘“When are you able to present yourself in such a way? And when does the world admire you?” “Be quiet”, the mare exclaimed, “and let me ruhig plough;…’ (Das Kutschpferd, Gellert Werke, Bd. 1, S. 113–114)

The passage represents a dialog between two horses: The first horse tries to distract the other from its work, i.e. ploughing a field; the second horse replies in lines three and four, using a morphological imperative and the item ruhig. This usage is ambiguous and can be interpreted in at least three ways. In the first interpretation ruhig is an adjective, used as a secondary predication on the object mich, i.e. ‘me, who I am in a calm state of mind’. The second interpretation treats ruhig as a verb phrase adverbial. It modifies the infinitive ruhig pflügen ‘plough calmly’. The third interpretation is the PDG modal particle reading of ruhig, which may be paraphrased as follows: (21´)

‘let me plough – in contrast to what you think I do not object to it.’

Examples like these – with ruhig in direct speech in an imperative sentence – may be taken as an indication that the use of ruhig as MP was present in the spoken language in the middle of the eighteenth century. But as this register is only rarely rendered in the written language of that time, this use is not documented in the historical dictionaries. Still, from examples like these, it is possible to derive the hypothesis that the morphological imperative triggers the reading of ruhig as an MP, because the imperative does not have an overt subject, while the implicit subject by necessity is the addressee. In other words: Because of the lack of a subject nominal, the predicative function of ruhig is easily transferred to the hearer, whereby a textual relation is changed into a pragmatic relation, i.e. a relation to an element of the communicative situation. In short, the grammaticalization of the adjective ruhig had its visible beginning in a very specific satzmodus construction, which made up the critical context.

230 Gabriele Diewald The data from the nineteenth century support this observation. All clear cases of ruhig as a modal particle appear in dialogs and in directive speech acts with imperative morphology. Examples are given in (22) and (23): (22)

„Stine! Du sollst nicht brüsk mit ihr brechen, im Gegenteil, besuche sie, solange dich’s dazu treibt; habe deine Plauderstunde mit ihr ruhig weiter; aber es muß der Augenblick kommen, wo sich’s ausgeplaudert hat und wo du deinen Irrtum empfindest.“ ‘Stine! There is no need for you to break with her abruptly, on the contrary, keep calling on her as long as you feel like it; go ruhig on having your chat with her, but it must lead to the moment when all chatting will have stopped and when you will feel your error.’ (Fontane, Stine, number 89)

(23)

„Bleiben Sie ruhig liegen und duseln sich gemütlich aus!“ ‘Keep ruhig lying and have a good nap!’ (Keller-SW. Bd. 3, S: 812)5

Example (23) proves the existence of the use of subjunctive I in imperative sentences (polite/distancing imperatives) in the nineteenth century. It can be hypothesized that the use of ruhig expanded from the morphological imperative to the polite imperative with its explicitly expressed pronominal subject. In spite of the small quantity of data, it is reasonable to assume that the function of ruhig as a modal particle developed in a satzmodus construction with the morphological imperative in the course of the eighteenth century and, in the nineteenth century, expanded to a second construction, i.e. the polite subjunctive I in directive speech acts. Its overall usage remained restricted to permissions and concessions usually uttered in a face-to-face situation. Table 3 summarizes this development: Table 3. Constructions with ruhig as an MP in the 18th and 19th century Stage 1 (18th c.)

Morphological imperative in face-to-face dialog

Stage 2 (19th c.)

Morphological imperative in face-to- face dialog Polite subjunctive I in directive speech acts in face-toface dialog

While we do not find MP uses of ruhig in declarative sentences with modal verbs in the eighteenth and nineteenth century, this construction appears in the first decades of the twentieth century. In a novel by Kafka (Amerika, written

The catalytic function of constructional restrictions in grammaticalization 231

1911–1914, first published 1927), the modal particle ruhig appears in morphological imperatives as in (24), in polite imperatives as in (25), and in the new construction, i.e. in declarative sentences with a modal verb, as in (26): (24)

„Berufe dich nur ruhig darauf, daß du mich um Vertretung gebeten hast.“ ‘Refer nur ruhig to the fact that you asked me for substitution.’ (Kafka-GW.; Bd. 6., S. 191)

(25)

„Erzählen Sie also nur ruhig. Es interessiert mich auch.“ ‘Well then, talk nur ruhig about it. I am interested in it as well.’ (Kafka-GW.; Bd. 6., S. 298)

(26)

„Du kannst ruhig staunen“, fuhr Robinson fort, „selbst ich habe gestaunt, wie mir das der Diener damals erzählt hat.“ ‘“You may ruhig wonder at it”, Robinson continued, “I wondered at it myself, when the servant told me at that time.”’ (Kafka-GW.; Bd. 6, S. 264)

Through the expansion of the MP reading of ruhig to declarative sentences with modal verb constructions, for the first time, a new construction with a different satzmodus got eligible for that reading of ruhig. However, the relevant characteristics for all MP usages of ruhig still remain the same: The MP reading is restricted to directive speech acts of the subtype of permission or concession in direct dialogic face-to-face situations, or reports on such situations. The persistence of the restriction to directive speech acts is the reason why only second and third person subjects are found with these constructions during that period. Table 4 summarizes the development up to that period:

232 Gabriele Diewald Table 4. Constructions with ruhig as an MP in the first decades of the 20th century Old constructions (from 18th and 19th c.)

Morphological imperative in face-toface dialog Polite subjunctive I in directive speech acts in face-to-face dialog

New construction (first attested at the beginning of the 20th century)

Declarative sentence with modal verb & infinitive and 2/3P. subject; used as an indirect directive speech act in face-toface dialog

According to reference books and grammars, the distribution rendered in Table 4 holds up to the present day for the distribution of the modal particle use of ruhig (see section 3). However, a check on the use of ruhig in spoken German in the second half of the twentieth century paints a different picture here. The corpora of spoken German provided by the IDS contain a total of 445 instances of ruhig, the first 100 examples of which were thoroughly analyzed.6 Most of the 62 uses of ruhig as a modal particle (including ambiguous cases) fall into the satzmodus categories that have been set up so far. There are morphological imperatives, as in (27), polite imperatives as in (28), and declaratives with modal verbs and 2nd/3rd person subjects as in (29): (27)

wo man singt da laß dich ruhig nieder ‘where they sing settle ruhig down’ (IDS_DSAV FR049_32)

(28)

Erzählen Sie ruhig ein bisschen! ‘Talk ruhig a bit about it! (IDS_DSAV OS141_98)

(29)

[das ist gut, das ist gut, machen Sie’s ruhig richtig, wie Sie’s denken,] das kann ruhig grob werden! ‘[that's good, that's good, do it ruhig right, as you are thinking] it may ruhig get rough!’ (IDS_DSAV OS101_87)

In addition to these constructions, which were maximally expectable, there are also new constructions, the most prominent ones being a hortative construction and the declarative construction with a first person subject. The hortative uses are illustrated in (30) and (31):

The catalytic function of constructional restrictions in grammaticalization 233

(30)

sagen wir ruhig die Reaktionäre ‘Let’s say (subj., 1.pl.) ruhig say the reactionaries’ (IDS-DSAV FR200_54)

(31)

gehn wir ruhig mal kriminalistisch vor go-SUBJI-1PL we ruhig mal criminologically ahead ‘Let’s proceed (subj., 1.pl.) ruhig mal in a criminological way.’ (IDS-DSAV FR212_60)

As can be seen, the hortative constructions display a V1 pattern with the verb in the subjunctive I, followed by the first person plural subjects (wir) and the item ruhig. Their meaning is no longer permission, but some kind of suggestion, invitation or encouragement including the speaker. The dialogic and reactive component seems to be reduced here. A paraphrase of ruhig in this use might take the following wording: ‘I suggest (we do) proposition although we have refrained from (doing) proposition before’. Thus, the use of the MP ruhig in this new construction also marks a new function, an extension toward another type of directive speech act. The second remarkable type of innovation is the use of first person subjects in the declarative construction, as in (32): (32)

und ich darf das ruhig einmal sagen ohne als sentimental... ‘and I may ruhig einmal say that’ (IDS-DSAV, FR 182_50)

This usage is also found parenthetically and seems to be formulaic to some extend. The meaning of the MP in this construction seems to develop in the direction of a discourse marker, possibly into a speaker signal with the function of attempting to keep the floor. However, as these examples were not frequent in the corpus, the development into a discourse marker will not be discussed here any further. What can be said with certainty is that ruhig has expanded its range of constructions in the spoken language in the second half of the twentieth century. As Table 5 shows, it acquired at least two additional satzmodus constructions:

234 Gabriele Diewald Table 5. Constructions with ruhig as an MP in the spoken language of the second half of the 20th century Old constructions (18th to beginning of 20th c.)

Morphological imperative in face-to-face dialog Polite subjunctive I in directive speech acts in face-to-face dialog Declarative sentence with modal verb & infinitive and 2/3P. subject; used as an indirect directive speech act in face-toface dialog

New construction (first attested in the second half of the 20th century)

Hortative construction Declarative sentence with modal verb & infinitive and 1sg. subject

5. Summary and conclusion The development of the modal particle ruhig starts out in a subclass of directive speech acts, namely the morphological imperative. As has been shown, evidence for this construction is found from the 18th onward. The available data indicate that before that time we can assume a stage zero, where ruhig is a lexical item carrying a privative meaning (‘free from work, movement, emotional disturbance’ etc.) and displaying a correspondingly low frequency in the corpus, where it is found in attributive, predicative and adverbial functions. This construction type (i.e. directive speech act with morphological imperative) forms the critical context where the new, more grammatical meaning (i.e. permission and indication of irrelevant counter-expectation) of the lexical item ruhig develops as one possible reading. The critical context is an idiomatic construction. In the further development the new function expands, leading to the gradual loss of the catalytic satzmodus restrictions. Today, we witness an expansion to other speech act types, which seems to be accompanied by a change of meaning of ruhig toward more general pragmatic functions. In sum, the original restriction of the MP ruhig to a particular satzmodus got lost step by step: The grammaticalization of ruhig as an MP is accompanied by the loss of constructional restrictions. The chronological development of these new constructions, as it could be reconstructed from the data investigated so far, is given in Table 6.

The catalytic function of constructional restrictions in grammaticalization 235 Table 6. Successive expansion to constructions/context types stage

I

II

III

IV

th

18 c.

th

19 c.

beginning of 20th c.

second half of 20th c.

+

+

+

+

+

+

+

+

+

construction imperative subj. I (2/3) decl. & MV (2/3) hortative

+

decl. & MV (1sg.)

+

Generalizing these results, it may be concluded that the grammaticalization of MPs proceeds through a bottleneck of idiomatic constructions towards more general constructions. While constructional restrictions are necessary to create a critical context and to start the grammaticalization process, they are gradually given up as the process of grammaticalization advances. From this angle, grammaticalization can be seen as giving up restrictions to certain construction types.

Notes 1.

2.

3. 4. 5.

6.

Beyond my best congratulations for the anniversary, I would like to express many thanks to Christian Lehmann, whose comments on my first studies on the grammaticalization of modal particles encouraged me to go on with this topic. For compatible definitions confer e.g. Goldberg (1995: 4), Kay and Fillmore (1999: 2-3), Croft (2001: 18-19, 261), Michaelis (2004: 8), Fried and Östman (2004). Instead of talking about extragrammatical idioms, the term extragrammatical constructions is preferred here. For information on the corpora used here, see reference section. Though this example looks like the use as an adjective or adverb (ruhig bleiben/ruhig liegen), the further context clearly favours the MP reading here: The addressee has been sleeping and is waking up, when the visitor utters (23). These 100 instances display the following distribution: ruhig as an adjective: 30; ruhig in unclear contexts (incomplete utterances etc.): 8; ruhig as an MP (including ambiguous cases): 62. This total number figure for ruhig as an MP contains all instances of the third context , i.e. ruhig in declaratives with a modal verb.

236 Gabriele Diewald

References A. Bibliographical references Autenrieth, Tanja 2002 Heterosemie und Grammatikalisierung bei Modalpartikeln. Eine synchrone und diachrone Studie anhand von eben, halt, e(cher)t, einfach, schlicht und glatt. Tübingen: Niemeyer. Bisang, Walter 1998 Grammaticalization and language contact, constructions and positions. In The Limits of Grammaticalization, Anna Giacalone Ramat and Paul Hopper (eds.), 13–58. Amsterdam and Philadelphia: Benjamins. Bybee, Joan L., Revere Perkins, and William Pagliuca 1994 The Evolution of Grammar. Tense, Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press. Croft, William 2001 Radical Construction Grammar. Oxford: Oxford University Press. Diewald, Gabriele 2002 A model for relevant types of contexts in grammaticalization. In New Reflections on Grammaticalization. International Symposium, Potsdam, 17–19 June, 1999, Ilse Wischer and Gabriele Diewald (eds.), 103–120. (Typological Studies in Language 49.) Amsterdam/Philadelphia: Benjamins. 2006a Discourse particles and modal particles as grammatical elements. In Approaches to Discourse Particles, Kerstin Fischer (ed.), 403–425. (Studies in Pragmatics 1.) Oxford: Elsevier. 2006b Konstruktionen in der diachronen Sprachwissenschaft. In Konstruktionsgrammatik. Von der Anwendung zur Theorie, Anatol Stefanowitsch and Kerstin Fischer (eds.), 79–103. Tübingen: Stauffenburg. 2006c Context types in grammaticalization as constructions. In Constructions. Special Volume 1: Constructions all over – Case Studies and Theoretical Implications, Doris Schönefeld (ed.), http://www. constructions-online.de/articles/specvol1/. Diewald, Gabriele, and Gisella Ferraresi in print Semantic, syntactic and constructional restrictions in the diachronic rise of modal particles in German: A corpus-based study on the formation of a grammaticalization channel. In Theoretical and Empirical Issues in Grammaticalization, Elena Seoane, and María José López-Couso (eds.), in collaboration with Teresa Fanego. (Typological Studies in Language). Amsterdam/Philadelphia: Benjamins.

The catalytic function of constructional restrictions in grammaticalization 237 Fillmore, Charles J., Paul Kay, and Mary Catherine O’Connor 1988 Regularity and idiomaticity in grammatical constructions: The case of let alone. Language 64: 501–538. Fried, Mirjam, and Jan-Ola Östman 2004 Construction Grammar: A thumbnail sketch. In Construction Grammar in a Cross-Language Perspective, Mirjam Fried and Jan-Ola Östman (eds.), 11–86. (Constructional Approaches to Language 2.) Amsterdam/Philadelphia: Benjamins. Goldberg, Adele E. 1995 A Construction Grammar Approach to Argument Structure. Chicago: Chicago University Press. 2006 Constructions at Work. The Nature of Generalization in Language. Oxford: Oxford University Press. Helbig, Gerhard 1988 Lexikon deutscher Partikeln. Leipzig: Verlag Enzyklopädie. Kay, Paul, and Charles J. Fillmore 1999 Grammatical constructions and linguistic generalizations: The What’s X doing Y? construction. Language 75: 1–33. Lehmann, Christian 1992 Word order change and grammaticalization. In Internal and External Factors in Syntactic Change, Marinel Gerritsen and Dieter Stein (eds.), 395–416. Berlin/New York: de Gruyter. 1995 [1982]. Thoughts on Grammaticalization. Revised and expanded version. First published edition. München [usw.]: Lincom Europa. Meibauer, Jörg 1994 Modaler Kontrast und konzeptuelle Verschiebung. Studien zur Syntax und Semantik deutscher Modalpartikeln. (Linguistische Arbeiten 314) Tübingen: Niemeyer. Michaelis, Laura 2004 Type shifting in construction grammar: An integrated approach to aspectual coercion. Cognitive Linguistics 15: 1–67. Searle, John 1979 Expression and Meaning. Studies in the Theory of Speech Acts. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs 2003 Constructions in grammaticalization. In The Handbook of Historical Linguistics, Brian Joseph and Richard D. Janda (eds.), 624–647. Oxford: Blackwell. [to appear] The grammaticalization of NP of NP Patterns. In Constructions and Language Change, Alexander Bergs and Gabriele Diewald (eds.), Berlin: De Gruyter. Thurmair, Maria 1989 Modalpartikeln und ihre Kombinationen. (Linguistische Arbeiten 223). Tübingen: Niemeyer.

238 Gabriele Diewald B. Corpora (a) General Grimm, Jakob and Grimm, Wilhelm 1854ff. Deutsches Wörterbuch. 33 Bde. Leipzig: Hirzel, ff. [Nachdruck München 1984]. Akademie der Wissenschaften der DDR, Akademie der Wissenschaften zu Göttingen (eds.) 1965ff. Deutsches Wörterbuch. Neubearbeitung. Leipzig: Hirzel. Paul, Hermann (Henne, Helmut) 1992 Deutsches Wörterbuch. [9., vollständig neu bearbeitete Auflage von Helmut Henne und Georg Objartel unter Mitarbeit von Heidrun Kämper-Jensen]. Tübingen.

(b) Basic corpus 14th – 18th century Bibliotheca Augustana http://www.fh-augsburg.de/~harsch/augustana.html (BA) Bonner Früneuhochdeutsches Korpus http://www.ikp.uni-bonn.de/dt/forsch/fnhd/ (B) Gloning Korpus http://staff-www.uni-marburg.de/~gloning/etexte.htm (G)

(c) Additional texts 16th c.: Martin Luther’s Bible, 1545. Digitale Bibliothek 29.

18th c.: Christian Fürchtegott Gellert, Fabeln und Erzählungen, 1746–48, Digitale Bibliothek, Bd. 1: Deutsche Literatur

19th c.: Keller, Gottfried: Der grüne Heinrich (1853–55). Digitale Bibliothek, Bd. 1: Deutsche Literatur Fontane: Cécile Stechlin; Effi Briest; Irrungen, Wirrungen; Die Poggenpuhls; Autobiographie (http://gutenberg.spiegel.de)

The catalytic function of constructional restrictions in grammaticalization 239

20th c.: Kafka, Franz: Amerika (written 1911–1914, first published 1927), Digitale Bibliothek, Bd. 1: Deutsche Literatur. IDS-Korpora of Written German, EDV-verfügbares Textkorpus des Instituts für deutsche Sprache, Mannheim.

(d) Spoken language IDS-Korpora, Freiburger Korpus EDV-verfügbares Textkorpus des Instituts für deutsche Sprache, Mannheim, Freiburger Korpus gesprochener deutscher Standardsprache. IDS_DSAV “Datenbank Gesprochenes Deutsch”, EDV-verfügbares Textkorpus des Instituts für deutsche Sprache, Mannheim. Keil, Martina 1990 Analyse von Partikeln für ein sprachverstehendes System – am Beispiel telefonischer Zugauskunftsdialoge. Magisterarbeit in der Philosophischen Fakultät II (Sprach- und Literaturwissenschaften) der Universität Erlangen. [typoscript].

Grammaticalization of periphrastic constructions Ilse Wischer

1. Introduction Grammaticalization is most generally defined as a shift of “a linguistic expression further toward the functional pole of the lexical-functional continuum” (Haspelmath 1999: 1044). Within this range of changes the most widely studied processes of grammaticalization focus on lexical items turning into grammatical morphs. This tradition is reflected in one of the earlier definitions of grammaticalization given by Kurylowicz (1965: 69): “Grammaticalization consists in the increase of the range of a morpheme advancing from a lexical to a grammatical or from a less grammatical to a more grammatical status.” Thus we find topics in grammaticalization studies like ‘the grammaticalization of verbs in Mandarine Chinese’, ‘the grammaticalization of “punya” in Malay’, ‘the grammaticalization of auxiliaries in sign language’, ‘the grammaticalization of English adjectives of difference’, ‘the grammaticalization of concessive conditionals’, etc. Although it has long been known that “grammaticalization does not merely seize a word or morpheme … but the whole construction formed by the syntagmatic relations of the element in question” (Lehmann 1992: 406) a real interest in the role of constructions in grammaticalization processes has only developed quite recently (e.g. Hopper 2001; Traugott 2003; Diewald 2006). At this point it must be admitted that the evolution of function words (e.g. prepositions, modals, pronouns), from lexical categories in their relevant contexts is indeed a central aspect of grammaticalization processes. But it should also be considered that many linguistic items do not develop a grammatical function by themselves, but must occur in particular constructions to express this function (e.g. be+V-ing or have+Ven in English). These ‘periphrases’ existed in earlier stages of English without being markers of aspect or tense, although the auxiliary had already lost its lexical status. My paper will take a closer look at grammaticalization of such periphrastic constructions focussing mainly on examples from the history of English.

242 Ilse Wischer 2. Periphrases According to Haspelmath (2000: 654) a periphrasis “in its most general sense refers to the use of longer, multi-word expressions in place of single words”, thus literally it is a “circumlocution”. This does not say anything about grammatical or lexical status. However, in linguistics the term is usually applied to grammatical constructions. Crystal (1980: 262) defines it as “a term used in grammatical description to refer to the use of separate words instead of inflections to express the same grammatical relationship.” Similarly Haspelmath (2000: 655) describes it in a more narrow, grammatical sense as referring “to a situation in which a multi-word expression is used in place of a single word in an inflectional paradigm.” He distinguishes three ways in which periphrastic constructions may enter an inflectional paradigm: (a)

to restore paradigm symmetry by filling a gap (e.g. Latin future subjunctive: facturus sit)

(b)

to achieve inflectional generality by affecting members of a word class that are otherwise not included in an inflectional paradigm (e.g. English analytical comparison: more beautiful)

(c)

to express a (new) grammatical category (e.g. English progressive: is working)

The gap-filling type (a) is basically restricted to situations in which a language expresses combinations of categories (present indicative / present subjunctive, past indicative / past subjunctive, future indicative / −). In the second type (b) a new layer emerges in the same functional domain of an already existing inflectional category. In contrast to Haspelmath I would not consider this an achievement of inflectional generality, but rather see it as a typical instance of the principle of layering described by Hopper (1991: 22−24). However, it is a specific type of layering, displaying a complementary distribution with regard to the application to individual class members, comparable to the emergence of the weak verb past morphology in addition to the old layer applying to the strong verbs, which form their past by ablaut. The “categorial periphrases” (c) are multi-word expressions that are interpreted in such a way that they express a new semantic distinction and finally may come to mark a new grammatical category in a language or form an alternative layer to an already existing category, but applying to all members of the same word class. The latter type will be the topic of my further discussion.

Grammaticalization of periphrastic constructions 243

3. Grammaticalization of periphrastic constructions Haspelmath does not give a detailed account of the grammaticalization of periphrases. He just argues that “the more grammaticalized a construction is, the more it can claim to have periphrastic status” (Haspelmath 2000: 661). If we consider that in his theory a periphrasis is by definition grammatical then such a statement is not very sophisticated. However, it implies that periphrastic constructions undergo a process of grammaticalization, albeit, according to Haspelmath, for its full understanding we still need a comprehensive theory of grammaticalization. In the following I will assume a wider concept of periphrases, in the sense that a periphrastic construction will be understood as a linguistic construct consisting of a function word which is related to lexical items in a systematic and conventional way. That means that there may be periphrastic constructions which have not yet been specified for a particular grammatical category, although they contain a fully grammaticalized word. At this stage the whole construction is still optional and displays a considerable degree of variation. In the process of its grammaticalization a specification for a particular grammatical function takes place, which may even result in the introduction of a new grammatical category, or it may just constitute a new layer in an already existing category, as mentioned above. For this reason it is appropriate to assume three types of grammaticalization that are involved in the evolution of periphrastic constructions: (a) lexical word > grammatical word (e.g. grammaticalization of have) (b) functionalization/specialization of the whole construction (e.g. grammaticalization of have+V-en) (c) emergence of a new category (e.g. secondary tense/correlation in English) or a new layer to an already existing category.

In the following it will be shown that stage (a) and (b) are not parallel but consecutive processes.

4. The perfect construction in English In Modern English the perfect1 construction have+V-en is highly grammaticalized and functions as ANTERIOR marker in relation to the

244 Ilse Wischer primary tenses. As such it is part of a secondary tense system besides an unmarked SIMULTANEOUS and a partly grammaticalized POSTERIOR form.2 The question is now: How did this periphrastic construction grammaticalize? Scholars largely agree that a periphrastic construction have/be+V-en existed from the earliest accessible times: “The ‘Perfect’ construction made up of an auxiliary and the past participle form of a main verb existed in English from the earliest recorded data” (Berndt 1982: 154). Berndt, however, also argues that at that time the construction had not yet been used as an ANTERIOR and it was not obligatory at all in the contexts where it is required today. This is evidence for the consecutive stages of the grammaticalization of the auxiliary and the whole construction. 4.1. Stage 1: The grammaticalization of have3 This process must have started in untypical contexts4, i.e. where the object of have was modified by a participle. In such contexts a stative resultative meaning was pragmatically inferred. This stage must have existed in prehistoric times. The next step occurred when have was used in critical contexts. This was the case when the participle had a zero inflection and the subject of the sentence was coreferential with the agent of the participial predicate. In such contexts the meaning of possession was bleached and the conversational implicature of a RESULTATIVE sense was conventionalised. At the same time a syntactic reanalysis took place, in the sense that the participle lost its adjectival character and was reinterpreted as a verb, and the object was no longer seen as governed by have, but as dependent on the participle. (1)

ic hæfde hit gewriten I had it written S V O Part-Adj ‘I possessed it in a written form.’

(2)

ic hæfde hit gewriten I had it written S Aux O Part-V ‘I had written it.’

Grammaticalization of periphrastic constructions 245

By analogy the new reading was now extended to isolating contexts, where a possessive interpretation was no longer possible. This final step had been reached in Old English, cf. example (3):5 (3)

æfterþæmþe hie gesyngod after they sinned ‘after they have sinned.’

habbaþ have

4.2. Stages 2/3: The grammaticalization of have+V-en as ANTERIOR The construction have+V-en was not restricted to one single function in Old English and it was by no means obligatory in any particular context. In the Old English part of the Helsinki Corpus, which contains texts from various regions and genres and comprises about 400,000 words, I have been able to identify: - ANTERIOR (correlation) functions (example (4)); - RESULTATIVE functions (example (5)); - PAST (temporal) functions (example (6)); - PERFECTIVE (aspectual) functions (example (7)).

(4)

he cwæð: ðæt sindan ða ða ðe mid wifum ne beoð besmitene, & hira mægeðhad habbað gehealdenne; ‘he said: that are those who are not dishonoured by women, & have kept their virginity;’ (Alfred’s Cura Pastoralis, 52.409.5)

(5)

ðonne hæbbe we begen fet gescode suiðe untællice ‘then have we both feet shod very blamelessly’ (Alfred’s Cura Pastoralis, 45.10)

(6)

Him þa Libertinus andswarode, leof fæder, ic hæfde gyrstandæge gecweden,… ‘Libertinus answered him, dear father, yesterday I said, …’ (Gregory the Great, Dialogues, MS H, 5.21.15)

(7)

þa hæfdon þa welisce menn gewroht ænne castel on Herefordscire ‘then (that year) the Welsh people had built a castle in Herefordshire’ (Chronicle Ms E (03/4); (04), 1048.43)

246 Ilse Wischer Although a distinction between ANTERIOR (correlation) and PERFECTIVE (aspectual) functions is not always easy since especially in their resultative uses they partly overlap, there is nevertheless a major difference: If the construction is used in a PERFECTIVE (aspectual) function the speaker/writer focuses on the internal structure of the event, i.e., the result or the completion as such is of primary relevance, without any reference to a point of orientation in time. Thus in example (7) the completion of the construction of the castle is in the centre of interest. If, however, the construction is used in an ANTERIOR (correlation) function, the speaker/writer signals that the event is related to a particular point of orientation in time, i.e., the internal structure of the event is not relevant. It may be completed, as in resultative uses, but need not be so, as in continuative uses, cf. ex. (4), where the event of keeping their virginity is not over, but has lasted up to the moment of speaking and may even continue beyond it. The quantitative distribution of the four functions mentioned above is shown in Table 1. Table 1. The functions of the perfect construction in Old English Function ANTERIOR RESULTATIVE PAST PERFECTIVE

Total

O2 (850−950) ca. 90,000 w. 70 55% 26 21% 17 13% 14 11% 127 100%

O3 (950−1050) ca. 250,000 w. 99 45% 31 14% 52 23% 39 18% 221 100%

O4 (1050−1150) ca. 60,000 w. 47 69% 13 18% 5 7% 4 6% 71 100%

The data reveal a considerable variety of uses throughout the Old English period. The RESULTATIVE uses first decline and then increase again a little bit. The PAST (temporal) and PERFECTIVE (aspectual) uses first increase and then decrease towards the end of the period, while the percentage of ANTERIOR uses increasingly dominates. If the construction had been specialized as PAST marker, this would have created an alternative layer to an already existing past tense category; as PERFECTIVE marker it might have filled a gap left by the erosion of the Old English verbal prefixes, in the sense of a renewal process. However, a new aspect system was already about to emerge in the form of the PROGRESSIVE, which can be seen as a special IMPERFECTIVE marker. Thus, what can be observed as early as at the end of the Old English period is an increasing specialization of the construction have+V-en as

Grammaticalization of periphrastic constructions 247 ANTERIOR gram and with it the evolution of a new grammatical category of secondary tense (or correlation). However, the construction can only be considered fully grammaticalized with its obligatorification, i.e. when it stands in a marked − unmarked opposition within a grammatical category and when its non-use where it is required by the context is considered ungrammatical. This stage is perhaps not reached before the end of the 19th century.

5. Conclusion I have shown that the grammaticalization of a periphrastic construction involves three constituent elements: the grammaticalization of the function word; the functionalization/specialization of the entire construction; and the emergence of a new grammatical category or a new layer within an already existing grammatical category. With regard to the grammaticalization of the English perfect construction it became obvious that this process did not take place to fill a particular gap in the tense and aspect system of Old English. This is exactly what Lehmann (2004: 184) calls the creativity in grammaticalization. “Grammaticalization can create genuinely new grammatical categories in a language, it can press items into a grammatical category which did not exist in the system.” In Old English have had already acquired the status of an auxiliary, but the whole periphrasis was not grammaticalized yet. It served various functions. Although the further development of the construction in Middle English and Early Modern English has yet to be studied, at the end of the Old English period a tendency to specialize as an ANTERIOR gram is already noticeable. My analysis further contradicts Bybee et al.’s (1994: 105) proposed path of development for ANTERIORS cross-linguistically6, and supports Brinton’s (1988) conclusion that there cannot be established any clear chronology from stative resultative to dynamic perfective meaning. The whole process of specialization was rather determined by the interrelation of this construction with other verbal categories in the language system. In Old English there was still the option of its functionalization as aspectual or temporal category. The focus on an endpoint inherent in the original resultative meaning of the construction could have been interpreted aspectually or temporally. The specialization of this construction as

248 Ilse Wischer may have been related to the development of the PROGRESSIVE aspect, which now marked the feature INCOMPLETION, so that a focus on COMPLETION (an aspectual marker) was no longer necessary. The grammaticalization of the perfect construction was a long process and resulted in its obligatorification as ANTERIOR gram in Late Modern English. It seems as if this happened parallel to the obligatorification of the PROGRESSIVE form. One additional aspect of this grammaticalization process is the specialization of have as the only auxiliary after be had been abandoned in this construction. While the be-construction still marked a resulting state (he is gone), the have-construction (he has gone) became a clear marker of anteriority. ANTERIOR

Notes 1.

2.

3.

4. 5. 6.

I use the term ‘perfect construction’ for traditional reasons to refer to have/be+V-en constructions, though their real functions may, however, vary considerably. The status of the perfect construction in Modern English is a highly controversial issue. Some authors treat it as tense, others as aspect (cf. Quirk et al. 1991: 189; Givón 1993: 163−166). For the concept of secondary tense cf. Reichenbach (1947); Jespersen (1949); Paul (1880: 273−274). Graustein et al. (1980: 144−147) calls it anterior correlation. The concept of ANTERIOR is also discussed in Bybee, Perkins and Pagliuca (1994: 61-63). I will only concentrate on the auxiliary have. Be, which also occurred in perfect constructions, was grammaticalized in other contexts, but by similar mechanisms. For types of context in grammaticalization processes cf. Diewald 2002. The status of have in Old English is dealt with in more detail in Wischer (2004). They claim that there is a chain of developments from RESULTATIVE to ANTERIOR and then further to PERFECTIVE and/or PAST.

Grammaticalization of periphrastic constructions 249

References Berndt, Rolf 1982 History of the English Language. Leipzig: Enzyklopädie. Brinton, Laurel J. 1988 The Development of English Aspectual Systems: Aspectualizers and Post-verbal Particles. (Cambridge Studies in Linguistics 49.) Cambridge: Cambridge University Press. Bybee, Joan, Revere Perkins, and William Pagliuca 1994 The Evolution of Grammar: Tense, aspect and modality in the languages of the world. Chicago: University of Chicago Press. Crystal, David 1980 A first dictionary of linguistics and phonetics, London: André Deutsch. Diewald, Gabriele 2002 A model for relevant types of context in grammaticalization. In New Reflections on Grammaticalization, Ilse Wischer and Gabriele Diewald (eds.), 103−120. (Typological Studies in Linguistics 49.) Amsterdam: Benjamins. 2006 Konstruktionen in der diachronen Sprachwissenschaft. In Konstruktionsgrammatik. Von der Anwendung zur Theorie, Anatol Stefanowitsch and Kerstin Fischer (eds.), 79−103. Tübingen: Stauffenburg. Givón, Talmy 1993 English Grammar. A function-based introduction. Vol. 1. Amsterdam/Philadelphia: Benjamins. Graustein, Gottfried (ed.) 1980 English Grammar − A University Handbook. Leipzig: Enzyklopädie. Haspelmath, Martin 1999 Why is grammaticalization irreversible? Linguistics 37 (6): 1043−1068. 2000 Periphrasis. In Morphologie - ein internationales Handbuch zur Flexion und Wortbildung, 1. Halbband. Geert Booij, Christian Lehmann, and Joachim Mugdan (eds.), 654−664. Berlin: De Gruyter. Hopper, Paul J. 1991 On some Principles of Grammaticalization. In Approaches to Grammaticalization. Vol. 1, Elizabeth Closs Traugott and Bernd Heine (eds.), 17−35. Amsterdam/Philadelphia: Benjamins. 2001 Grammatical constructions and their discourse origins: prototype or family resemblance? In Applied Cognitive Linguistics I: Theory and Language Acquisition, Martin Pütz, Susanne Niemeier, and René Dirven (eds.), 109−129. Berlin: Mouton de Gruyter.

250 Ilse Wischer Jespersen, Otto 1949 A Modern English Grammar on Historical Principles. Part IV, Syntax, Vol. 3. Copenhagen: Einar Munksgaard; London: Allen and Unwin. Kurylowicz, Jerzy 1965 The evolution of grammatical categories. Diogenes 51: 55−71. Lehmann, Christian 1992 Word order change by grammaticalization. In Internal and external factors in syntactic change, Marinel Gerritsen and Dieter Stein (eds.), 395−416. (Trends in linguistics. Studies and monographs 61.) Berlin/New York: Mouton de Gruyter. 2004 Theory and method in grammaticalization. Zeitschrift für Germanistische Linguistik 32 (2): 152−187. Paul, Hermann 1880 Prinzipien der Sprachgeschichte. Leipzig: Niemeyer. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik 1991 A Comprehensive Grammar of the English Language. London: Longman. Reichenbach, Hans 1947 Elements of Symbolic Logic. New York: Free Press. Traugott, Elizabeth Closs 2003 Constructions in Grammaticalization. In The Handbook of Historical Linguistics, Brian D. Joseph and Richard Janda (eds.), 624−647. (Blackwell Handbooks in Linguistics.) Oxford: Blackwell. Wischer, Ilse 2004 The HAVE-perfect in Old English. In New Perspectives on English Historical Linguistics, Christian J. Kay, Simon Horobin, and Jeremy Smith (eds.), 243−255. Amsterdam: Benjamins.

Grammaticalization in constructions: Clitic doubling with experiencers in Modern Greek1 Elisabeth Verhoeven

1. Introduction Recent work on the diachrony of experiencer constructions has shown that oblique experiencers tend to develop to subject experiencers in the long run (Seefranz-Montag 1983, Cole et al. 1980, Allen 1995, Haspelmath 2001, Croft 2001: 159). Well-known examples for such a change can be found among others in the Indo-European languages, as evidenced by the Old-English example in (1) and its Modern English translation. (1)

Þam wife þa word wel licodon. the.DAT woman.DAT those.NOM words.NOM well liked.3.PL ‘The woman liked those words well.’ (Beowulf 639)

The change from oblique experiencers to subject experiencers may be explained by the application of functional parameters, i.e. the affinity of the experiencer to topicality being based on the fact that the experiencer is necessarily animate and preferentially definite (s. Verhoeven 2007). In as far as there is a tendency for human topics to be coded as subjects, the experiencer is morphosyntactically adjusted to this position (Haspelmath 2001: 78). Following data presented in Anagnostopoulou 1999, Modern Greek oblique experiencers show some properties that can be analyzed as reflecting subject or topic properties. Among these is the property that certain experiencer objects occur in a construction with topic marking by clitic doubling (2), while regular clitic doubling is licensed by a prominence condition (cf. e.g. Anagnostopoulou 1999), i.e. the object-NP to which the ‘doubled clitic’ refers must be prominent in the common ground.

252 Elisabeth Verhoeven (2)

Ta épipla DEF:NOM.PL.N furniture:NOM.PL.N *?(ton) eno≈lún ton pétro. 3.SG.ACC.M bother:3.PL DEF:ACC.SG.M Peter:ACC.SG.M ‘The furniture bothers Peter.’ (Anagnostopoulou 1999: 79)

Based on this observation of a special syntactic behavior of oblique experiencers, we will review their construction possibilities with different information structures focusing especially on their possible or necessary occurrence with the object clitics. The main question pursued is if the class of experiencer object verbs has developed a special argument structure construction in Modern Greek. This may go hand in hand with the grammaticalization of the object clitic in the observed constructions. Section 2 introduces the Modern Greek experiencer object (henceforth EO) constructions while section 3 gives an overview over the basic structural and pragmatic aspects of the Modern Greek object clitics. Section 4 investigates the occurrence of the object clitics in construction with EO verbs and compares it to that of canonical transitive verbs. Section 5 gives a systematic account of the empirical data in the light of a grammaticalization perspective. Furthermore, evidence from a corpus study is given, which investigates the occurrence of EO-Verbs with object clitics in the Hellenic National Corpus and compares it to that of canonical transitive verbs. Section 6 summarizes the main results of the study.

2. Experiencer object constructions in Modern Greek Modern Greek has EO verbs belonging to diverse valency frames. The experiencer argument may be coded in the genitive (3a) or in a prepositional phrase (3b), or it may bear accusative case (4), (s. Anagnostopoulou 1999, Kordoni 1999). (3)

a. To krasí tu DEF:NOM.SG.N wine:NOM.SG.N 3.SG.GEN.N tu pétru. DEF:GEN.SG.M Peter:GEN.SG.M ‘The wine pleases Peter.’

arési please:3.SG

Grammaticalization in constructions 253

b. To krasí arési DEF:NOM.SG.N wine:NOM.SG.N please:3.SG s-ton pétro. LOC-DEF:ACC.SG.M Peter:ACC.SG.M ‘The wine pleases Peter.’(Anagnostopoulou 1999: 69) (4)

Ton pétro ton anisi≈í DEF:ACC.SG.M Peter:ACC.SG.M 3.SG.ACC.M worry:3.SG i katástasi. DEF:NOM.SG.F situation: NOM.SG.F ‘The situation worries Peter.’ (Anagnostopoulou 1999: 68)

As in many languages, a number of Modern Greek verbs with accusative-marked EOs are systematically ambiguous between an agentive and a non-agentive/causative reading (see Verhoeven 2008). The reading of the verb eno≈lí in (5a) can be agentive, while in (5b), (repeated here from (2)), it has to be understood as non-agentive/causative. Note that the stimulus in subject function is inanimate in (5b) so that an agentive reading is excluded. (5)

a.

b.

I maría eno≈lí DEF:NOM.SG.F Maria:NOM.SG.F bother:3.SG ton pétro. DEF:ACC.SG.M Peter:ACC.SG.M ‘Maria bothers Peter (intentionally).’ Ta épipla *?(ton) DEF:NOM.PL.N furniture:NOM.PL.N 3.SG.ACC.M eno≈lún ton pétro. bother:3.PL DEF:ACC.SG.M Peter:ACC.SG.M ‘The furniture bothers Peter.’ (Anagnostopoulou 1999: 7879)

Following the analysis of Anagnostopoulou 1999, this difference between an agentive and a non-agentive/causative interpretation is syntactically implemented in the following way.2 Such experiencers show obligatory doubling of the object clitic with non-agentive readings of the experiential verb (5b), while clitic doubling is optional with the respective verbs in their agentive reading (5a). This means that eno≈lí in (5a) can only

254 Elisabeth Verhoeven be understood as agentive since the non-agentive reading of eno≈lí requires clitic doubling. However, the analysis of this data needs to be more precise. In what sense clitic doubling in (5a) is optional? To answer this question is must be clarified under which conditions clitic doubling of (direct) objects takes place in Modern Greek. On this basis, a putative special status of EO constructions with clitic doubling may be identified. Given that the analysis proposed by Anagnostopoulou can be confirmed, it still needs to be clarified what functional motivation explains the (obligatory) use of clitic doubling with non-agentive EO verbs (in contrast to agentive EO verbs). At the centre of the following investigation are the accusative marked EO verbs as introduced in example (5). EO verbs with dative experiencers will be observed in comparison. Next to EO verbs, which allow for an agentive and a non-agentive/causative reading (called ‘labile’ EO verbs in this paper), there are also EO verbs which are obligatorily non-agentive. (6) and (7) present some examples of both groups. The grouping is based on an experiment which tests the compatibility of EO verbs with volitionality adverbs such as epítiDes ‘intentionally’, eskeména ‘on purpose’, siniDitá ‘consciously’, carried out with 32 native speakers of Modern Greek (cf. Verhoeven, ms.). (6)

(7)

‘Labile’ EO verbs prokaló ‘provoke’, eno≈ló ‘bother’, kse©eláo ‘fiddle’, enTaríno ‘encourage’, eksor©ízo, ‘enrage’, tromázo ‘frighten’, apas≈oló ‘concern’, etc. Non-agentive EO verbs enDiaféro ‘interest, concern’, provlimatízo ‘puzzle’, ©oitévo ‘captivate, charm’, siginó ‘touch’, etc.

In the rest of the paper, Modern Greek EO verbs will be systematically investigated in their co-occurrence with object clitics in diverse information structural contexts. For this aim, section 3 introduces the main structural and pragmatic properties of the object clitics.

Grammaticalization in constructions 255

3. Modern Greek pronominal object clitics 3.1. Structural aspects The pronominal clitics of Modern Greek constitute a paradigm of nonempathic personal pronouns, which cannot bear a lexical accent (contrary to the emphatic personal pronouns). They occur with accusative and genitive marked constituents, which are part of the VP (i.e. direct objects, indirect objects (in the genitive) and genitive marked adjuncts (e.g. beneficiaries), however not temporal accusative or genitive marked adjuncts). They occur adjacent to the verb, i.e. they are proclitics to finite verb forms, and form a phonological word with it, as shown in (9). (8)

V obj.NP Diávasa to vivlío read:AOR:1.SG DEF:ACC.SG.N book:ACC.SG.N ‘I read the book’3

(9)

cl.obj V (*V cl.obj) Diávasa a. to 3.SG.ACC.N read:AOR:1.SG ‘I read it’ b. *Diávasá to

There are two syntactic construction types, which involve the combination of an object clitic with a coreferential lexical NP within the clause: ‘simple’ clitic doubling (10) and so-called clitic left dislocation (11). With clitic doubling (10), the coreferential NP follows the V. In (11), the coreferential NP is left dislocated. (10)

cl.objj V NPj (= clitic doubling) Diávasa to vivlíoj toj 3.SG.ACC.N read:AOR:1.SG DEF:ACC.SG.N book:ACC.SG.N ‘I read the book’

(11)

NPj cl.objj V (= clitic left dislocation) toj Diávasa to vivlíoj DEF:ACC.SG.N book:ACC.SG.N 3.SG.ACC.N read:AOR:1.SG ‘I read the book’

256 Elisabeth Verhoeven 3.2. Pragmatic aspects The object NP occurs with clitic doubling if its referent is prominent in the ‘common ground’ (Anagnostopoulou 1994). It represents given information and has ‘out-of-focus’ status in the utterance, i.e. it is part of the information structural background (Alexopoulou & Kolliakou 2002). On the contrary, clitic left dislocation requires a functional motivation for preposing the object. Different contextual conditions can license such a preposing, e.g. contrastive topicalization (cf. Iatridou 1995), anaphoric relations with respect to the preceding context (s. Alexopoulou & Kolliakou 2002) etc. Both constructions share the property that they generally do not occur in contexts that license object focus, as is shown in (12). (12)

Question inducing object focus A: ti diávases ? what:ACC.SG.N read:AOR.2.SG

Diávasa

to vivlío read:AOR:1.SG DEF:ACC.SG.N book:ACC.SG.N Diávasa to B´: #to 3.SG.ACC.N read:AOR:1.SG DEF:ACC.SG.N vivlío book:ACC.SG.N vivlío to B´´: #to DEF:ACC.SG.N book:ACC.SG.N 3.SG.ACC.N Diávasa read:AOR:1.SG

B:

In a question inducing subject focus (13), the object NP is given information. Then a construction with clitic doubling of the object is a possible answer (B’). The same applies to a construction with clitic left dislocation (B’’). (13)

Question inducing subject focus A: pios diávase to who:NOM.SG.M read:AOR.3.SG DEF:ACC.SG.N vivlío ? book:ACC.SG.N

Grammaticalization in constructions 257

o jánis Diávase DEF:NOM.SG.M Janis:NOM.SG.M read:AOR:3.SG to vivlío DEF:ACC.SG.N book:ACC.SG.N B´: o jánis to DEF:NOM.SG.M Janis:NOM.SG.M 3.SG.ACC.N Diávase to vivlío read:AOR:3.SG DEF:ACC.SG.N book:ACC.SG.N B´´: to vivlío to DEF:ACC.SG.N book:ACC.SG.N 3.SG.ACC.N Diávase o jánis read:AOR:3.SG DEF:NOM.SG.M Janis:NOM.SG.M

B:

Figures 1 and 2 illustrate the pragmatic properties of clitic left dislocation and clitic doubling. The occurrence of a coreferential object clitic is pragmatically licensed: the coreferential lexical NPs cannot be focused. Furthermore, all permutations of the focus feature are possible in these constructions: verb focus, subject focus, adjunct focus, V+S Focus, etc. pragmatics

out-of-focus

semantics

pati

pati

syntax

NPj

[cl.objj

V]

Figure 1. Pragmatic properties of clitic left dislocation

Prototypically, left dislocation occurs with definite NPs; in the case of ‘hanging topics’, however, indefinite NPs may be combined with a coreferential object clitic. In contrast, clitic doubling is exclusively compatible with definite NPs and not with indefinite NPs. pragmatics

out-of-focus

semantics

pati

syntax

[cl.obji

pati V]

Figure 2. Pragmatic properties of clitic doubling

NPi

258 Elisabeth Verhoeven 4. Experiencer object and object clitics 4.1. Introduction In this section, EO verbs are investigated in their interaction with object clitics. The resulting constructions are systematically checked for the information structural conditions associated with them. The behavior of the three types of experiencer verbs introduced in section 2 is compared with that of canonical transitive verbs. A number of test constructions are revised which vary concerning the following parameters: (a) verb type: EO verbs (agentive vs. non-agentive) vs. canonical transitive verbs, (b) animacy of the stimulus: animate vs. non-animate stimulus, (c) position of the EO-NP, (d) occurrence of a co-indexed clitic. Examples (14) - (17) illustrate the test sentences with a labile EO verb (14), a non-agentive EO verb with accusative EO (15), a non-agentive EO verb with genitive EO (16) and a canonical transitive verb (17). The a-sentences contain an inanimate subject (stimulus), the b-sentences an animate subject (stimulus). (14)

Labile EO verb (accusative object) Tórivos (to) a. O DEF:NOM.SG.M noise:NOM.SG.M 3.SG:ACC.N eno≈lí to korítsi. bother:3.SG DEF:ACC.SG.N girl:ACC.SG.N ‘The noise bothers the girl.’ TóDoros (to) b. O DEF:NOM.SG.M Thodoros:NOM.SG.M 3.SG:ACC.N eno≈lí to korítsi. bother:3.SG DEF:ACC.SG.N girl:ACC.SG.N ‘Thodoros bothers the girl.’

(15)

Non-agentive EO verb (accusative object) a. I istoría (to) DEF:NOM.SG.F story:NOM.SG.F 3.SG:ACC.N to korítsi. enDiaféri concern:3.SG DEF:ACC.SG.N girl:ACC.SG.N ‘The story concerns the girl.’

Grammaticalization in constructions 259

b.

I maría (to) DEF:NOM.SG.F Maria:NOM.SG.F 3.SG:ACC.N to korítsi. enDiaféri concern:3.SG DEF:ACC.SG.N girl:ACC.SG.N ‘Maria concerns the girl.’

(16)

Non-agentive EO verb (genitive object) a. I istoría (tu) DEF:NOM.SG.F story:NOM.SG.F 3.SG:GEN.N arési tu koritsiú. please:3.SG DEF:GEN.SG.N girl:GEN.SG.N ‘The story pleases the girl.’ b. I maría (tu) DEF:NOM.SG.F Maria:NOM.SG.F 3.SG:GEN.N arési tu koritsiú. please:3.SG DEF:GEN.SG.N girl:GEN.SG.N ‘The story pleases the girl.’

(17)

Canonical transitive verb a. O keravnós (to) DEF:NOM.SG.M thunder:NOM.SG.M 3.SG:ACC.N ≈tipái to korítsi. hit:3.SG DEF:ACC.SG.N girl:ACC.SG.N ‘The thunder hits the girl.’ b. O kóstas (to) DEF:NOM.SG.M Kostas:NOM.SG.M 3.SG:ACC.N ≈tipái to korítsi. hit:3.SG DEF:ACC.SG.N girl:ACC.SG.N ‘Kostas hits the girl.’

4.2. Clitic left dislocation If a lexical object NP is preposed, there are two constructions which are functionally complementary. If there is no coreferential object clitic the construction necessarily expresses object focus (18a).4 All other permutations of the focus feature are not possible, as (18b-e) show.

260 Elisabeth Verhoeven (18)

a.

b. c. d. e.

{To korítsi}FOC ≈tipái DEF:ACC.SG.N girl:ACC.SG.N hit:3.SG o keravnós. DEF:NOM.SG.M thunder:NOM.SG.M ‘The thunder hits the girl.’ *To korítsi {≈tipái}FOC o keravnós. *To korítsi ≈tipái {o keravnós}FOC. *To korítsi {≈tipái o keravnós}FOC. *{To korítsi ≈tipái}FOC o keravnós.

If a coreferential object clitic is present, the possible focus domains are illustrated in (19): excluded are those possibilities in which the preposed object is part of the focus domain. (19)

a.

b. c. d. e.

To korítsi to DEF:ACC.SG.N girl:ACC.SG.N 3.SG:ACC.N {≈tipái o keravnós}FOC. hit:3.SG DEF:NOM.SG.M thunder:NOM.SG.M ‘The thunder hits the girl.’ To korítsi to {≈tipái}FOC o keravnós. To korítsi to ≈tipái {o keravnós}FOC. *{To korítsi}FOC to ≈tipái o keravnós. *{To korítsi to ≈tipái}FOC o keravnós.

Topicalization of preposed and postposed objects is not restricted with clitic doubling, i.e. topicalization is compatible with the occurrence of a co-indexed object clitic. Therefore it will be observed in the following if the regularity shown in (18) and (19) with respect to focusing also applies to (clitic) left dislocation with EO verbs. Three possibilities of focusing are tested with the verb types introduced in section 4.1: object focus, verb focus and subject focus. The results are summarized in Table 1.

Grammaticalization in constructions 261 Table 1. Clitics and left dislocation canonical

experiencer agentive

OFVS

OVFS

OVSF

nicht-agentive accusative

genitive

with clitic

*

*

3

3

without clitic

3

3

3

3

with clitic

3

3

3

3

without clitic

*

*

*

*

with clitic

3

3

3

3

without clitic

*

*

*

*

turning point

On the basis of Table 1 the following observations can be made: If the object is preposed and not in focus (see lines 3, 4, 5, and 6 in Table 1) the clitic is obligatory. If the object is preposed and focused, the clitic is generally ungrammatical (20). However, the non-agentive experiencer verbs constitute an exception: With these verbs, the clitic is possible even with object focus (see examples (21a), (22) and (23)). (20)

{To korítsi}FOC DEF:ACC.SG.N girl:ACC.SG.N a. (*to) ≈tipái o keravnós. 3.SG:ACC.N hit:3.SG DEF thunder:NOM.SG.M b. (*to) ≈tipái o kóstas. 3.SG:ACC.N hit:3.SG DEF Kostas:NOM.SG.M ‘The thunder/Kostas hits the girl.’

(21)

{To DEF:ACC.SG.N a. (to) 3.SG:ACC.N b. (*to) 3.SG:ACC.N

korítsi}FOC girl:ACC.SG.N eno≈lí o bother:3.SG DEF eno≈líagentive o bother:3.SG DEF

Tórivos. noise:NOM.SG.M TóDoros. Thodoros:NOM.SG.M

262 Elisabeth Verhoeven (22)

(23)

{To DEF:ACC.SG.N a. (to) 3.SG:ACC.N b. (to) 3.SG:ACC.N

korítsi}FOC girl:ACC.SG.N i enDiaféri concern:3.SG DEF i enDiaféri concern:3.SG DEF

istoría. story:NOM.SG.F maría. Maria:NOM.SG.F

{Tu DEF:GEN.SG.N a. (tu) 3.SG:GEN.N b. (tu) 3.SG:GEN.N

koritsiú}FOC girl:GEN.SG.N arési i please:3.SG DEF arési i please:3.SG DEF

istoría. story:NOM.SG.F maría. Maria:NOM.SG.F

A similar phenomenon can be observed with preposed interrogative pronouns in object function. While in questions with canonical transitive verbs and agentive EO verbs there cannot be an object clitic (24), (25b), nonagentive EO verbs do not show such a restriction (25a), (26), (27). The results are presented in Table 2. (24)

a. Pion (*ton) ≈tipái Q:ACC.SG.N 3.SG:ACC.N hit:3.SG o keravnós ? DEF:NOM.SG.M thunder:NOM.SG.M ‘Whom does the thunder hit?’ b. Pion (*ton) ≈tipái Q:ACC.SG.N 3.SG:ACC.N hit:3.SG o kóstas? DEF:NOM.SG.M Kostas:NOM.SG.M ‘Whom does Kostas hit?’

(25)

a. Pion (ton) eno≈lí Q:ACC.SG.N 3.SG:ACC.N bother:3.SG Tórivos ? o DEF:NOM.SG.M noise:NOM.SG.M ‘Whom does the noise bother?’ b. Pion (*ton) eno≈líagentive Q:ACC.SG.N 3.SG:ACC.N bother:3.SG

Grammaticalization in constructions 263

o TóDoros? DEF:NOM.SG.M Thodoros:NOM.SG.M ‘Whom does Thodoros bother?’ (26)

a. Pion (ton) enDiaféri Q:ACC.SG.N 3.SG:ACC.N concern:3.SG i istoría? DEF:NOM.SG.F story:NOM.SG.F ‘Whom does the story concern?’ b. Pion (ton) enDiaféri Q:ACC.SG.N 3.SG:ACC.N concern:3.SG i maría? DEF:NOM.SG.F Maria:NOM.SG.F ‘Whom does Maria concern?’

(27)

a. Pianú (tu) arési Q:GEN.SG.N 3.SG:GEN.N please:3.SG i istoría? DEF:NOM.SG.F story:NOM.SG.F ‘Whom does the story please?’ b. Pianú (tu) arési Q:GEN.SG.N 3.SG:GEN.N please:3.SG i maría? DEF:NOM.SG.F Maria:NOM.SG.F ‘Whom does Maria please?’

Table 2. Clitics and preposing of interrogative pronouns canonical

experiencer agentive

OQVS

non-agentive accusative

genitive

with clitic

*

*

3

3

without clitic

3

3

3

3

turning point

264 Elisabeth Verhoeven Summarizing the results of this section, it can be concluded that, contrary to canonical transitive verbs and agentive EO verbs, non-agentive EO verbs allow for the occurrence of a co-indexed clitic if the left dislocated object NP is associated with initial focus or if it is represented by a question pronoun. That means that the pragmatic property of left dislocation identified in Figure 1, namely that the left dislocated NP necessarily constitutes background information (i.e. cannot be part of the focus domain), does not apply to non-agentive EO verbs, as indicated in Figure 3. pragmatics

out-of-focus

semantics

expi

expi

syntax

NPj

[cl.objj

V]

Figure 3. Pragmatic properties of left dislocation with non-agentive EOs

4.3. Relative clause In this section, the construction of EO verbs in relative clauses with a relativized object will be investigated. In such a relative construction, it can be observed that the object clitic is obligatory with non-agentive EO verbs. This does not only apply to genitive experiencers (28g,h) but also to accusative experiencers if the experiential verb does not have an agentive reading (28c,e,f). With EO verbs bearing an agentive reading (28d) and canonical transitive verbs (28a,b), co-indexed object clitics cannot occur in relative clauses under definite object extraction.5 (28)

to SibáTisa like:AOR:1.SG DEF:ACC.SG.N ‘I liked the girl, whom ...’ a. (*to) ≈típise 3.SG:ACC.N hit:AOR:3.SG ‘... the thunder hit.’ b. (*to) ≈típise 3.SG:ACC.N hit:AOR:3.SG ‘... Kostas hit.’

korítsi girl:ACC.SG.N

pu ... REL.PRON

o keravnós DEF thunder:NOM.SG.M o kóstas. DEF Kostas:NOM.SG.M

Grammaticalization in constructions 265

c. *(to) enó≈lise o Tórivos. 3.SG:ACC.N bother:AOR:3.SG DEF noise:NOM.SG.M ‘... the noise bothered.’ o TóDoros. d. (*to) enó≈liseagentive 3.SG:ACC.N bother:AOR:3.SG DEF Th.NOM.SG.M ‘... Thodoros bothered.’ i istoría. e. *(to) enDiéfere 3.SG:ACC.N concern:AOR:3.SG DEF story:NOM.SG.F ‘... the story concerned.’ i maría. f. *(to) enDiéfere 3.SG:ACC.N concern:AOR:3.SG DEF Maria:NOM.SG.F ‘... Maria concerned.’ g. *(tu) árese i istoría. 3.SG:GEN.N please:AOR:3.SG DEF story:NOM.SG.F ‘... the story pleased.’ h. *(tu) árese i maría. 3.SG:GEN.N please:AOR:3.SG DEF Maria:NOM.SG.F ‘... Maria pleased.’ For the genitive experiencers, it can be argued that the object clitic is obligatory since the relative pronoun is not case marked. For the accusative experiencers, however, this argument is not valid; rather the existence of the clitic pronoun distinguishes between an agentive and a non-agentive reading of the experiential verb. Table 3 summarizes the behavior of EO verbs in relative clauses: contrary to canonical transitive verbs and agentive EO verbs, non-agentive EO verbs require a co-indexed clitic with a relativized (E)O. Table 3. Clitics and relative clauses canonical

experiencer agentive

ORELVS

non-agentive accusative

genitive

with clitic

*

*

3

3

without clitic

3

3

*

*

turning point

266 Elisabeth Verhoeven 4.4. Clitic doubling Finally, the behavior of EO verbs with ‘simple’ clitic doubling will be investigated in this section. First, the possibility of focusing the EO (section 4.4.1) and then the interaction with topicalization will be analyzed (section 4.4.2). 4.4.1. Interaction with focus Similar to clitic left dislocation, ‘simple’ clitic doubling excludes object focus. In a construction without clitic doubling (SVO order), the prototypical reading is that of VP focus. However, all other possibilities of focusing exist as well, as shown in (29). (29)

a. {O keravnós ≈tipái DEF:NOM.SG.M thunder:NOM.SG.M hit:3.SG to korítsi}FOC. DEF:ACC.SG.M girl:ACC.SG.M ‘The thunder hits the girl.’ b. O keravnós {≈tipái to korítsi}FOC. c. O keravnós ≈tipái {to korítsi}FOC. d. O keravnós {≈tipái}FOC to korítsi. e. {O keravnós}FOC ≈tipái to korítsi. f. {O keravnós ≈tipái}FOC to korítsi.

In contrast, in a construction with clitic doubling the object cannot be part of the focus domain. This is illustrated in (30). (30)

a. *{O keravnós to DEF:NOM.SG.M thunder:NOM.SG.M 3.SG:ACC.N ≈tipái to korítsi}FOC. hit:3.SG DEF:ACC.SG.M girl:ACC.SG.M ‘The thunder hits the girl.’ b. *O keravnós to {≈tipái to korítsi}FOC. c. *O keravnós to ≈tipái {to korítsi}FOC. d. O keravnós to {≈tipái}FOC to korítsi. e. {O keravnós}FOC to ≈tipái to korítsi. f. {O keravnós to ≈tipái}FOC to korítsi.

Grammaticalization in constructions 267

Thus, again we can notice a functionally complementary distribution of the construction with and without clitic doubling: If the object is doubled by a clitic, object focus is impossible. If there is no clitic doubling, object focus is possible. However, contrary to the case of clitic left dislocation (cf. section 4.2) the possibilities of focusing with ‘simple’ clitic doubling are not influenced by the thematic properties of the object. Table 4 shows the results of the permutation of the focus feature in combination with and without clitic doubling. It is obvious that agentive as well as non-agentive EO behave like canonical transitive verbs in the conditions illustrated. The examples with object focus and an inanimate stimulus are shown in (31). (31)

a. *O keravnós (to) DEF:NOM.SG.M thunder:NOM.SG.M 3.SG:ACC.N ≈tipái {to korítsi}FOC. hit:3.SG DEF:ACC.SG.F girl:ACC.SG.F ‘The thunder hits the girl.’ Tórivos (to) b. *O DEF:NOM.SG.M noise:NOM.SG.M 3.SG:ACC.N eno≈lí {to korítsi}FOC. bother:3.SG DEF:ACC.SG.F girl:ACC.SG.F ‘The noise bothers the girl.’ c. *I istoría (to) DEF:NOM.SG.F story:NOM.SG.M 3.SG:ACC.N {to korítsi}FOC. enDiaféri concern:3.SG DEF:ACC.SG.F girl:ACC.SG.F ‘The story concerns the girl.’ d. *I istoría (tu) DEF:NOM.SG.F story:NOM.SG.M 3.SG:GEN.N arési {tu koritsiú}FOC. please:3.SG DEF:GEN.SG.N girl:GEN.SG.N ‘The story pleases the girl.’

268 Elisabeth Verhoeven Table 4. Clitic doubling canonical

experiencer agentive

SFVO

SVFO

SVOF

non-agentive accusative

genitive

with clitic

3

3

3

3

without clitic

3

3

3

3

with clitic

3

3

3

3

without clitic

3

3

3

3

with clitic

*

*

*

*

without clitic

3

3

3

3

Thus, as regards the properties of EO verbs with ‘simple’ clitic doubling, we can conclude that non-agentive EO verbs do not allow for clitic doubling if the respective NP (referring to the experiencer) is focused. In this respect, they totally conform to canonical transitive verbs and agentive EO verbs. This is represented in Figure 4. pragmatics

out-of-focus

semantics

expi

syntax

[cl.objj

expi V]

NPj

Figure 4. Pragmatic properties of clitic doubling with non-agentive EOs

Contrary to the analysis proposed in Anagnostopoulou 1999 (see section 2, esp. example (5b)), Table 4 shows that clitic doubling with EOs of non-agentive verbs is not obligatory. Next to elicitation the investigation of the Hellenic National Corpus (HNC) in section 5.2 produces the same result: non-agentive EO verbs are common without clitic doubling. 4.4.2. Interaction with topic However, also postposed experiencer objects of non-agentive verbs show a special behavior with respect to pragmatic licensing in comparison to the

Grammaticalization in constructions 269

other verb types tested: with canonical transitive verbs and agentive EO verbs, clitic doubling requires that the respective referent be mentioned or prominent in the preceding discourse, i.e. it has to be contextually given. This can be seen in example (32). (32a) and (32b) are investigated as possible continuations of the context given in (32). In (32a), clitic doubling of the object NP is not possible with a canonical transitive verb, since the object referent is not mentioned before, i.e. it is not prominent (enough). The same holds true for (32b), where an agentive EO verb is tested. (32)

Context: ‘Some time ago, I reviewed a new book on clitic doubling.’ a. #Ar©ótera ton sinándisa later.on 3.SG.ACC.M meet:AOR:1.SG ton sigraféa (...) DEF:ACC.SG.M author:ACC.SG.M ‘Later on, I met the author ...’(Anagnostopoulou 1999: 76) ton enó≈lisa b. #EpítiDes absichtlich 3.SG.ACC.M bother:AOR:1.SG ton sigraféa. DEF:ACC.SG.M author:ACC.SG.M ‘Deliberately, I annoyed the author.’ (Anagnostopoulou 1999: 79)

However, this prerequisite does not apply to non-agentive EO verbs, as the examples in (33) show. The same context as in (32) licenses clitic doubling of the experiencer objects with the non-agentive EO verbs in (33a-c). In these examples, the referent of the EOs is not mentioned previously, but is inferable from the context, i.e. accessible (cf. Lambrecht 1994). (33)

Context: ‘Some time ago, I reviewed a new book on clitic doubling.’ a. I kritikí mu DEF:NOM.SG.F criticism:NOM.SG.F 1.SG.GEN ton enó≈lise ton 3.SG.ACC.M bother:AOR:3.SG DEF:ACC.SG.M sigraféa (...) author:ACC.SG.M ‘My criticism bothered the author ...’ (Anagnostopoulou 1999: 76)

270 Elisabeth Verhoeven b. I kritikí mu DEF:NOM.SG.F criticism:NOM.SG.F 1.SG.GEN ton ton enDiéfere 3.SG.ACC.M concern:AOR:3.SG DEF:ACC.SG.M sigraféa. author:ACC.SG.M ‘My criticism concerned the author’ c. I kritikí mu DEF:NOM.SG.F criticism:NOM.SG.F 1.SG.GEN tu árese tu 3.SG.GEN.M please:AOR:1.SG DEF:ACC.SG.M sigraféa. author:ACC.SG.M ‘My criticism pleased the author.’ Thus, contrary to canonical transitive verbs and agentive EO verbs, nonagentive EO verbs can occur with clitic doubling, if the referent of the object NP is inferrable from the context. It does not have to be contextually given, a condition which has to be fulfilled with canonical transitive verbs and agentive EO verbs. This result is represented in Table 5. Table 5. Clitics and “Bridging Topics” canonical

experiencer agentive

Oaccessible

non-agentive accusative

genitive

with clitic

*

*

3

3

without clitic

3

3

3

3

turning point

Grammaticalization in constructions 271

5. Grammaticalization and the emergence of a construction 5.1. From pragmatics to syntax The phenomena presented show different types of licensing of the object clitics with respect to different constructions, as summarized in Table 6. The occurrence of the object clitics is completely licensed by the context with canonical transitive verbs and agentive EO verbs. However, nonagentive EO verbs optionally occur with an object clitic in left dislocation under object focus and in an object focus question as well as with clitic doubling if the object referent is accessible in discourse. Finally, in a relative clause construction with relativization of the EO, the object clitic is even obligatory with non-agentive EO verbs. This applies to genitive as well as accusative experiencers. Table 6. Types of licensing of the object clitic optional (dep. on context)

optional (free variation)

obligatory

canonical verbs & agentive EO verbs

non-agentive EO verbs

non-agentive EO verbs

- clitic left dislocation (O-focus/O-question) - clitic doubling (O-accessible)

- relative clause

Presumably, these types of licensing correspond to stages of a grammaticalization process which leads from pragmatics to syntax (see Table 7).6 With the canonical transitive verbs and the agentive EO verbs, the occurrence of the clitic is pragmatically licensed in all contexts investigated. With the non-agentive EO verbs, the occurrence of the clitic is just free variation in a number of contexts (see examples of the non-agentive verbs in the construction with a left-dislocated object in focus ((21a), (22), (23)) and in the context of preposed question pronouns ((25), (26), (27)) as well as with simple clitic doubling with an accessible object (33a-c)). This suggests that the use of the clitic is desemanticized in these contexts. However, in other constructions the object clitics are obligatory (see relativization of EOs in (28c), (28e-h)). This means that the occurrence of the object clitics in these contexts is syntactically licensed.

272 Elisabeth Verhoeven Table 7. Types of licensing and grammaticalization optional (dep. on context)

optional (free variation)

→ desemanticization pragmatic licensing

obligatory

→ grammaticalization syntactic licensing

Given this analysis, we can observe a double asymmetry in the data. On the one hand, there is a asymmetry with respect to the diverse constructions investigated. The supposed stages of grammaticalization do not concern the experiencer-related clitics in their entirety but dependent on certain constructions. This asymmetry is illustrated in (34), which has to be read as an implicative hierarchy. The step in the change from pragmatics to syntax which is reached in the right hand constructions is presupposed in the left hand construction. (34)

asymmetry I: surrounding constructions < clitic left dislocation with O-focus / O-question relative clause < clitic doubling (Oaccessible)

On the other hand, we can note an asymmetry in the two clitic constructions introduced in section 3, clitic doubling and clitic left dislocation (see (35)). From a functional point of view, it is expected that the effects of the thematic role peculiarities of EOs are clearer in those constructions, in which the experiencer is topical, i.e. in preposing. Our data confirms this intuition. While preposed object experiencers have already reached the step of desemanticization (of the “out-of-focus” properties), this does not hold true for postposed experiencers (cf. also Figures 3 and 4). (35)

asymmetry II: ‘clitic’ constructions clitic left dislocation > clitic doubling

The occurrence of the object clitics with non-agentive EO verbs may suggest (from the perspective of the verb group) that a new argument structure construction (of non-agentive EO verbs) is in the making. If this is indeed the case, we are still at the very beginning of such a process. Contrary to the analysis of Anagnostopoulou 1999, the data presented above

Grammaticalization in constructions 273

has shown that the occurrence of the object clitics with non-agentive EO verbs is optional rather than obligatory in most cases.

5.2. Corpus evidence The evidence presented in the previous sections is based on the intuition of acceptability of particular structures in particular contexts. In order to verify the results of this investigation in a naturalistic data basis, we undertook an investigation of the Hellenic National Corpus (HNC). HNC (developed by the Institute of Language and Speech Processing (ILSP)) is a large-size online corpus (currently 47.000.000 words) that comprises written discourse from the following sources: books (9,41%), internet (0,31%), newspapers (61,30%), magazines (5,90 %), and miscellaneous (23,08%). In order to test hypotheses resulting from the intuition data, we created a dataset containing all occurrences of some representative verbs in HNC. We selected eight canonical transitive verbs in order to estimate a baseline for the occurrence of clitics in a set of verbs that represent the typical alignment pattern agent/subject & patient/object. Furthermore, we selected eight EO verbs, which corresponded to the two basic categories dealt with in this paper: four non-agentive verbs and four labile verbs (see (36)). (36)

a. canonical transitive verbs spró≈no ‘push’, DiorTóno ‘correct, repair, fix’, vlápto ‘damage’, proskaló ‘invite’, klotsáo ‘kick’, Djó≈no ‘turn out/away, chase, kick out’, klévo ‘steal, rob’, ≈tízo ‘build, construct’ b. non-agentive EO verbs aréso ‘please’, enDiaféro ‘interest, concern’, ©oitévo ‘captivate, charm’, provlimatízo ‘puzzle’ c. labile EO verbs eno≈ló ‘bother’, kse©eláo ‘fiddle’, eksor©ízo, ‘enrage’, tromázo ‘frighten’

In order to exclude variation that is induced by the different categories of personal and temporal deixis, we restricted our data set to tokens that involve the above verbs in the third person singular of the indicative past forms (perfective and imperfective).

274 Elisabeth Verhoeven Our hypotheses relate to transitive verbs with two arguments, hence all occurrences of the above verbs without an object (see (37) for instance) were characterized as ‘non-valid’ and are excluded from consideration for the measurements below. (37)

I motosikléta árese. DEF:NOM.SG.F motorbike:NOM.SG.F please:AOR:3.SG ‘The motorbike pleased (People liked the motorbike).’

For each occurrence of a verb, the realization of the object was identified according to the following parameters: lexical or emphatic pronominal vs. clitical vs. both; NP vs. PP; definite vs. indefinite; preverbal vs. postverbal. Note that information about the definiteness of the NPs is necessary since clitic left dislocation and clitic doubling differ as to their compatibility with definite NPs: while clitic left dislocation is compatible with indefinite NPs, clitic doubling is exclusively compatible with definite NPs (cf. section 3.2). Finally, cases with object extraction in relativization using the relative pronoun pu (cf. section 4.3) were identified. Table 8 gives an overview of the numerical results for those constructions reported here. The first line indicates the total of verb-object constructions identified in the corpus for the different verb groups. The second line (CL+V) gives the occurrences of the verbs with an object clitic but without a lexically realized object NP. The third line (V+arg) indicates the total of those cases where the verb is followed by a VP internal argument, either an object NP or an oblique argument PP. These numbers include cases with definite and indefinite objects as well as cases with and without the simultaneous presence of an object clitic, i.e. it includes case of clitic doubling. The following three lines single out those cases where the postposed NP is definite. The first of these three lines (V+argdef) gives the total of the cases while the following two lines indicate the amount of postposed object NPs without clitic doubling (see ‘–CL’) and those with clitic doubling (see ‘+CL’). The next section of Table 8 (arg+V) features the cases of left dislocation of the object NP or PP, again including definite and indefinite NPs. Again, the numbers are given in more detail for the definite left dislocated NPs separating simple left dislocation (see ‘–CL’) and clitic left dislocation (see ‘+CL’). Finally, the last section of Table 8 (rel. Obj.) indicates the number of occurrences of relativized objects using the relative pronoun pu. Again the numbers for cases with a definite object are presented in more detail.

Grammaticalization in constructions 275 Table 8. Object constructions with some verb groups in HNC

Total CL+V V+arg V+argdef –CL +CL arg+V argdef +V –CL +CL rel. Obj rel. Obj def –CL +CL

Non-ag. V. n % 1646 1268 244 108 96 88,9 12 11,1 127 77 4 5,2 73 94,8 7 7 0 0,0 7 100,0

Labile V. n % 718 358 347 285 285 100,0 0 0,0 13 11 4 36,4 7 63,6 0 0 0 0

Canonical V. n % 1230 362 741 601 593 98,7 8 1,3 78 66 53 80,3 13 19,7 49 45 45 100,0 0 0,0

Sum n 3594 1988 1332 994 974 20 218 154 61 93 56 52 45 7

%

98,0 2,0

39,6 60,4

86,5 13,5

First, we calculate the percentage of pronominal object realizations (line CL+V of Table 8) out of the total of object constructions (first line in Table 8) for each verb group. The frequency of occurrence with object clitics results in a scale: non-agentive EO verbs > labile EO verbs > canonical transitive verbs (see Figure 5). 100

% of n occurrences

80 60 % clitics

40 20 0 Non-ag (n=1646)

Labile (n=718)

Figure 5. Pronominal realization of object in HNC

Canonical (n=1230)

276 Elisabeth Verhoeven Next, we will consider the cases of clitic doubling and clitic left dislocation in our corpus since we saw on the basis of elicitation that the use of the object clitic is desemanticized and no more pragmatically licensed with non-agentive EO verbs in clitic left dislocation and clitic doubling with an accessible object. Thus, we expect that clitic left dislocation and clitic doubling are more frequent in our corpus with non-agentive EO verbs than with the other verb groups. Furthermore, we expect that clitic left dislocation and clitic doubling are more frequent with labile EO verbs than with canonical transitive verbs, since the labile EO verbs also display a nonagentive reading next to their agentive reading. These expectations are confirmed in our corpus, as is shown in Figure 6. Figure 6 visualizes the section (argdef +V) of Table 8. 100

% of n occurrences

80 60 % clitics

40 20 0 Non-ag (n=77)

Labile (n=11)

Canonical (n=66)

Figure 6. Clitic left dislocation of argdef with some verb groups in HNC

However, as concerns clitic doubling, the results are less clear in our corpus. Figure 7 visualizes the section (V+argdef) in Table 8. While it is still obvious that clitic doubling is more frequent with non-agentive EO verbs (11,1% of all V+argdef cases) than with the other verb groups, the expected difference between the labile EO verbs (0,0% of all V+argdef cases) and the canonical transitive verbs (1,3% of all V+argdef cases) cannot be confirmed. This result might be due to the overall small number of tokens of clitic doubling in the corpus. Note that clitic doubling is a rather seldom phenomenon in general. In our corpus, the overall frequency of clitic doubling in relation to the total of verb occur-

Grammaticalization in constructions 277

rences is 0,56%. The corresponding percentage of clitic left dislocation is 2,73%. Crucially, this data clearly falsifies the analysis proposed in Anagnostopoulou 1999 (see section 2, esp. example (5b)): Table 8 and Figure 7 show that clitic doubling with EOs of non-agentive verbs is in no way obligatory but even relatively infrequent in discourse. 100

% of n occurrences

80 60 % clitics 40 20 0 Non-ag (n=108)

Labile (n=285)

Canonical (n=601)

Figure 7. Clitic doubling of argdef with some verb groups in HNC

Finally, a word has to be said about the corpus evidence for the relative clause construction with object extraction discussed in section 4.3. The section (rel. Objdef) in Table 8 clearly shows that the corpus data supports the elicitation results in section 4.3. With non-agentive EO verbs, all instances of a definite extracted object show the clitic following the relative pronoun pu. In contrast, with canonical transitive verbs no instance of a definite extracted object in construction with a co-indexed clitic occurs. Note that the corpus did not possess tokens of relative clauses with labile EO verbs.

6. Summary The investigation has shown that EOs of non-agentive verbs show a special behavior in constructions with a co-indexed object clitic, i.e. with clitic left dislocation, simple clitic doubling and relative clause formation. This behavior is in contrast to that of objects of canonical transitive verbs and

278 Elisabeth Verhoeven agentive EO verbs. The construction of non-agentive EO verbs with coindexed object clitics implies a desemanticization of the otherwise pragmatic licensing of the object clitics in the mentioned constructions. In relative clause formation with object relativization, the occurrence of the object clitic is even syntactically licensed. Therefore it is analyzed as grammaticalized in this construction. Those factors that have been identified as playing a crucial role in the change of oblique experiencers to subject experiencers in section 1, in particular their topicality properties, can be identified as functional motivation for the development described in the preceding sections, i.e. the construction formation with non-agentive EO verbs. The effects of the thematic peculiarity of EOs are more clearly shaped in those constructions in which the experiencer is topical, i.e. when it is preposed. While preposed object experiencers already reached the stage of desemanticization (of the “outof-focus” properties), this does not yet hold true for postposed experiencers. The corpus study undertaken supports the findings from elicitation and shows that non-agentive EO verbs occur much more frequently with object clitics than labile EO verbs and canonical transitive verbs. This difference in frequency favors the analysis that the development of grammaticalization and construction formation with non-agentive EO verbs is discourse based. Corpus frequencies of clitic left dislocation and clitic doubling also mirror the peculiarity of non-agentive EO verbs in their occurrence with object clitics. This investigation has focused on the behavior of agentive and nonagentive EO verbs and their occurrence with object clitics. In the future, it remains to be tested in how far non-experiential verbs with similar semantic properties (i.e. obligatorily animate objects, causative-stative event structure, etc.) show a similar way of construction to those verbs investigated here.

Grammaticalization in constructions 279

Notes 1.

2.

3. 4.

5. 6.

Work on this paper was financially supported by project 10/853/05 (University of Bremen). Part of it has been presented at the workshop Grammaticalization and grammatical categories held at the University of Bremen in June 2007. I thank the participants of this workshop and especially Christian Lehmann and Stavros Skopeteas for very helpful comments and discussion. Furthermore, the binding properties of EOs of non-agentive verbs are analyzed as a syntactic manifestation of the difference between agentive and nonagentive EO verbs. Contrary to EOs of agentive verbs (and canonical transitive verbs) EOs of non-agentive verbs are able to bind anaphoric pronouns belonging to the formal subject. This property has been analyzed as a subject property of such EOs (see Belletti und Rizzi 1988, Pesetzky 1987, 1995). If not otherwise indicated, Modern Greek examples stem from elicitation with native speakers. Some special cases of object topicalization without the occurrence of a clitic which are restricted to written discourse are left out of consideration (see Alexopoulou 1999:45). Note that the acceptability holds true for the restrictive interpretations of the relative clauses. Referring to Lehmann (1995: 13), the step identified in Table 7 as grammaticalization can also be called syntacticization, i.e. the first step in a grammaticalization from discourse to syntax. Within a complete cycle of grammaticalization this is only the initial phase, which is then followed by morphologization, demorphemicization, and, finally, complete loss. With respect to the empirical phenomenon of the Greek object clitics described here, it is not implied that a further grammaticalization on the described path necessarily follows.

References Allen, Cynthia L. 1995 Case Marking and Reanalysis: Grammatical Relations from Old to Early Modern English. Oxford: Clarendon Press. Alexopoulou, Theodora 1999 The syntax of discourse functions in Greek: a non-configurational approach. Ph.D. Dissertation, University of Edinburgh. Alexopoulou, Theodora, and Dimitra Kolliakou 2002 On Linkhood, Topicalization, and Clitic Left Dislocation. Journal of Linguistics 38 (2): 193–245.

280 Elisabeth Verhoeven Anagnostopoulou, Elena 1994 Clitic Dependencies in Modern Greek. Ph.D. Dissertation, University of Salzburg. 1999 On experiencers. In Studies in Greek syntax, A. Alexiadou, G. Horrocks and M. Stavrou (eds), 67–93. Dordrecht: Kluwer. Belletti, Adriana, and Luigi Rizzi 1988 Psych verbs and θ-theory. Natural Language and Linguistic Theory 6: 291–352. Cole, P., W. Harbert, G. Hermon, and S.N. Sridhar 1980 The acquisition of subjecthood. Language 56: 719–743. Croft, William 2001 Radical Construction Grammar: Syntactic Theory in Typological Perspective. Oxford: Oxford University Press. Grimshaw, Jane 1990 Argument Structure. Cambridge etc: The MIT Press. Härtl, Holden 2001 Cause and change: thematische Relationen und Ereignisstrukturen in Konzeptualisierung and Grammatikalisierung [studia grammatica, 50]. Berlin: Akademie. Haspelmath, Martin 2001 Non-canonical marking of core arguments in European languages. In Non-canonical Marking of Subjects and Objects. A. Aikhenvald, R.M.W. Dixon, and M. Onishi (eds), 53–83. Amsterdam, Philadelphia: Benjamins. Iatridou, Sabine 1995 Clitics and island effects. Penn Working Papers in Linguistics 2: 11– 31. Jackendoff, Ray 1990 Semantic Structures. Cambridge, Mass.: MIT Press. Klein, Katarina, and Silvia Kutscher 2002 Psych-verbs and lexical economy [Theorie des Lexikons, 122]. Düsseldorf: Universität Düsseldorf. Kordoni, Valia 1999 Lexical semantics and linking in HPSG: the case of psych verb constructions. Arbeitspapiere des SFB 340, 494–527. Lambrecht, Knut 1994 Information Structure and Sentence Form: Topic, Focus, and the Mental Representation of Discourse Referents. Cambridge: Cambridge University Press. Lehmann, Christian 1995 Thoughts on Grammaticalization. Unterschleissheim: Lincom Europa.

Grammaticalization in constructions 281 Pesetzky, David 1987 Binding problems with experiencer verbs. Linguistic Inquiry 18: 126–140. 1995 Zero Syntax: Experiencer and Cascades. Cambridge: The MIT Press. Seefranz-Montag, Ariane von 1983 Syntaktische Funktionen und Wortstellungsveränderung: die Entwicklung “subjektloser” Konstruktionen in einigen Sprachen [Studien zur Theoretischen Linguistik, 3]. München: Fink. Van Valin, Robert D., Jr., and David P. Wilkins 1996 “The case for ‘effector’: case roles, agents, and agency revisited”. In Grammatical Constructions: Their Form and Meaning. M. Shibatani and S. Thompson (eds.), 289–322. Oxford: Clarendon. Verhoeven, Elisabeth 2007 Experiential Constructions in Yucatec Maya. A Typologically Based Analysis of a Functional Domain in a Mayan Language. Amsterdam: Benjamins. 2008 (Non-)canonical marking of experiencer objects: A typological comparison of Chinese, Korean, Turkish, and Modern Greek. Language Typology and Universals 61 (1): 81-92. ms. Stativity and agentivity in experiencer verbs: Implications for the theory of predicate classes. University of Bremen: Manuscript.

Grammaticalization of honorific constructions in Japanese Yoko Nishina

1. Introduction Honorification is deeply entrenched in the grammar and lexicon of the Japanese language. Lexical as well as morphological means are used to express certain degrees of politeness. The present study describes a specific syntactic construction used to express politeness in terms of grammaticalization theory. The construction is a combination of two other constructions, the causative construction and the autobenefactive construction with the verb of receiving, as the following example illustrates: (1)

happyoo-o hazime-sase-te presentation-ACC begin-CAUS-SEQ ‘I begin my presentation.’

itadaki-mas-u receive:H3-H1-PRS

A literal meaning of this example would be: ‘I benefit from letting myself begin my presentation’. The honorific meaning is derived from the speaker’s attitude in assigning himself a certain role such as beneficiary and causee in the causative-permissive situation. The aim of this paper is to show how the causative and the autobenefactive constructions are grammaticalized to an honorific construction. As can be expected with regard to grammaticalization processes, the specific honorific usage of this construction is an additional function of this construction type leading a) to synchronic variation and b) to an enrichment of the Japanese honorification system. In the following sections, the blending of both types of constructions with the resulting politeness function will be presented in some detail. Firstly, a brief overview of the honorification strategies in Japanese (section 2) and the voice functions of the respective constructions are given (section 3 and section 4). Secondly, multiple participant roles expressed by these constructions are examined (section 5). Finally, the way in which these two constructions may acquire a politeness function will be explained functionally.

284 Yoko Nishina 2. Honorification in Japanese Two types of expression for honorification can be distinguished in Japanese: one type is the choice of an appropriate style depending on the situation and the addressee of the speech. The other type is the referent-oriented honorification, which is divided into two further types: to be respectful to the referent or to humble oneself or someone on one’s side in order to express respect towards the referent relatively. In traditional grammaticography these three types of honorification are 1 named teinei-go, sonkei-go and kenzyoo-go. The classification of honorification is illustrated in Table 1. H1 is situation-oriented and refers to the appropriate style. H2 is oriented to [-EGO], which represents the addressee of the speech act or a third person referent. H3 is oriented towards [+EGO], which refers to the speaker or speaker’s in-group, such as a family member or a colleague to be degraded. Examples (2), (3) and (4) exemplify these three linguistic honorification types. Table 1. Classification of honorification H1 H2 H3

Teinei (stylistic, situation-oriented, absolute) Sonkei (respectful, referent-oriented, relative, [-EGO]) Kenzyoo (humble, referent-oriented, relative, [+EGO])

(2)

Taroo-ga Taro-NOM ‘Taro came.’

ki-masi-ta come-H1-PAST

(3)

sensei-no imooto-san-ga irassyai-masi-ta teacher-GEN younger.sister-H2-NOM come:H2-H1-PAST ‘The teacher’s sister came.’

(4)

watasi-no imooto-ga 1.SG-GEN younger.sister-NOM ‘My sister came.’

mairi-masi-ta come:H3-H1-PAST

The examples show that stylistic and referential honorification is expressed mostly by lexical and morphological means, i.e. there are different verbal stems, nouns, prefixes and suffixes. Besides such lexical and morphologi-

Grammaticalization of honorific constructions in Japanese

285

cal means, there are also some syntactic strategies, which are presented in the remainder of this paper.

3. Causation and benefaction in Japanese Japanese has different morphological and syntactic means to manipulate the argument structure of a simple clause. One of these results in a causative construction, the other in a benefactive construction. Both construction types are illustrated in examples (5) through (8). (5)

Hanako-ga hon-o yon-da Hanako-NOM book-ACC read-PAST ‘Hanako read the book.’

(6)

Taroo-ga Hanako-ni hon-o yom-ase-ta Taro-NOM Hanako-DAT book-ACC read-CAUS-PAST ‘Taro made Hanako read the book’

(7)

Hanako-ga Taroo-ni hon-o yon-de Hanako-NOM Taro-DAT book-ACC read-SEQ ‘Hanako read the book for Taro.’

yat-ta give-PAST

(8)

Taroo-ga Hanako-ni hon-o yon-de Taro-NOM Hanako-DAT book-ACC read-SEQ ‘Taro got the book read by Hanako.’

morat-ta get-PAST

Example (5) represents a simple situation which is embedded in both causative and benefactive constructions in the subsequent examples. A causative construction in (6) is formed on the basis of the simple clause by means of the causative suffix -(s)ase attached to the base verb. Examples (7) and (8) represent a benefactive situation by means of the benefactive function verb ageru ‘give’ and morau ‘receive’. The embedded base verb is realized as a non-finite form. In (8), Hanako is encoded as a dative dependent, and the effect of promoting the beneficiary to the actor role, which has more control and agentivity, is an autobenefactive construction. Causative and (auto-)benefactive situations are produced by adding a participant (Taroo) to the subordinate proposition (sub-event) – beneficiary in the benefactive situation, causer in the causative situation. Hanako can be

286 Yoko Nishina called ‘realizer’, the one who carries out the sub-event in both situations – benefactor in the benefactive situation, causee in the causative situation. The structure of the constructions is illustrated in Table 2. Table 2. Structure of causation and benefaction construction category causative benefactive autobenefactive

NOM CAUS’ER (Taro) BEN’OR (Hanako) BEN’RY (Taro)

DAT CAUS’EE (Hanako) BEN’RY (Taro) BEN’OR (Hanako)

verb suffix/ function verb

example

-(s)aseru

(6)

yaru/kureru ‘give’ morau ‘receive’

(7) (8)

The participants in the autobenefactive clause have two semantic roles: Taro can be considered as the causer and the beneficiary of the situation, while Hanako can be considered as the causee and the benefactor. Thus this construction shows some similarity to a causative construction and can be regarded as a variation to express the causative situation (see section 3.2.).

3.1. Causative construction and pragmatic relation in Japanese There are some pragmatic restrictions in the usage of causative constructions in Japanese which cannot be found in the same way in European languages. In the following example (9) with the causative suffix -(s)aseru, the causer is marked by the nominative particle ga, the causee by the dative ni, and the direct object of the verb by the accusative o. (9)

Taroo-ga Hanako-ni hon-o yom-ase-ta Taro-NOM Hanako-DAT book-ACC read-CAUS-PAST ‘Taro made Hanako read the book.’

However, this rather socially neutral clause cannot be used with participants that have a certain social position with respect to the speaker. Compare the pragmatic unacceptability of the following clause in (9):

Grammaticalization of honorific constructions in Japanese

(10)

*Taroo-ga Taro-NOM

sensei-ni teacher-DAT

287

ronbun-o yom-ase-ta thesis-ACC read-CAUS-PAST

Imagine a situation where Taro, writing his thesis, had his supervisor read his thesis. If the causee (teacher, supervisor) is situated in a socially higher position, a morphological causative is not appropriate. A Japanese sentence such as (10) would sound rude because to order, to force, or even to allow one’s own teacher to do something is not polite behaviour. Only someone in a higher position is supposed to order you; you cannot command your boss and force him to obey you. Even in an equal or more intimate relationship, such as among family members and friends, this causative expression would certainly be avoided because this expression would have a connotation of force or an order. One of the expression strategies Japanese speakers have in order to avoid this conflict is the benefactive construction, as the following section illustrates.

3.2. Benefactive construction Benefactive constructions in Japanese result from the grammaticalization of verbs of giving and receiving to function verbs, which is in itself not unusual cross-linguistically. In benefactive constructions with the verb of giving, the benefactor appears in subject position and the beneficiary is marked by the dative, whereas in the benefactive construction with the verb of receiving it is the other way round (11): (11)

Taroo-ga sensei-ni ronbun-o yon-de Taro-NOM teacher-DAT thesis-ACC read-SEQ morat-/itadai-ta get-/get:H3-PAST ‘Taro got (his) thesis read by (his) teacher.’

In the grammaticalized construction with the verb of receiving, Taro does not receive a concrete object but merely has his thesis read by the teacher. The meaning of the verb is metaphorically extended, and the person who gives it is the benefactor. In (11), this construction shows the beneficiary’s respect and gratitude to the benefactor. The lexical alternative form of the verb of receiving, itadaku, expresses the speaker’s respectfulness to the benefactor.

288 Yoko Nishina This sentence can also be interpreted to mean that Taro asked or persuaded his teacher to read his thesis. To the extent that the subject referent strives for a benefit to achieve the event, the subject earns its agentivity and control over other participants. This construction is called in particular “autobenefactive” (Lehmann 2004: 12f), with the verb of receiving, in contrast to the benefactive with the verb of giving. In the case where the subject is more agentive, the construction can be regarded as a variation of causative constructions. In contrast to the coercive and permissive causative, which are both expressed by the causative suffix, this construction always has a benefactive meaning in the causative situation (Nishina 2001: 174). On the other hand, the benefactive construction with the verb of giving (12) cannot be interpreted in such a way that the beneficiary has a causative intention, because the beneficiary does not share control of the situation with the benefactor: (12)

sensei-ga ronbun-o yon-de teacher-NOM thesis-ACC read-SEQ ‘The teacher read the thesis for me.’

kure/kudasat-ta give-/give:H2-PAST

The alternation between the morphological causative construction and the syntactic autobenefactive construction can be understood pragmatically in that the interpersonal relation between causer and causee is reflected in the choice of the construction expressing the causativity. One of the motivations of the autobenefactive construction with the verb of receiving for the causative meaning is the similarity of the cognitive structure. Both causative and autobenefactive situations have a complex structure in which the event to achieve is subordinated. The central participant of both voices, the causer and the beneficiary, are encoded as the subject, and the agent of this event, who is the realizer of the sub-event, is embedded in the lower syntactic function (Table 2). As illustrated below in Figure 1 and Figure 2, the event to achieve is the causatum (CAUS’UM) in the causative construction and the benefactum (BEN’UM) in the benefactive construction. In the same way, the realizer of the event is the causee (CAUS’EE) in the causative construction and benefactor (BEN’OR) in the (auto-)benefactive construction. The head is the grammatical category, which is encoded by a verb suffix in the causative and by a function verb in the benefactive. These parallel structures are illustrated in Figure 1 (see (9)) and Figure 2 (see (11)).

Grammaticalization of honorific constructions in Japanese

289

CAUS -(s)aseru CAUS’ER Taroo

CAUS’UM Hanako reads the book CAUS’EE Hanako

Proposition read the book

Figure 1. Structure of the causative construction BEN morau BEN’RY Taroo

BEN’UM Teacher read the thesis BEN’OR Teacher

Proposition read the thesis

Figure 2. Structure of the autobenefactive construction

Due to this parallel structure a participant can have both roles in the autobenefactive situation; namely the beneficiary has the role of the causer and the benefactor functions as the causee, as follows: BEN’RY BEN’OR

= =

CAUS’ER CAUS’EE

Figure 3. Participant relations in the autobenefactive situation

This is why the autobenefactive construction functions as a polite version of a causative construction. Additionally, the agentivity of the central participant is decisive for the selection of the voice in Japanese. The participant would be willing to be agentive for an event which is advantageous and beneficial. This is the formation principle of the causative in Japanese in contrast to the passive as a “Leidform (suffering form)”. Table 3 illustrates the differentiation of the voices in Japanese based on the agentivity of Taro. The autobenefac-

290 Yoko Nishina tive has the benefactive meaning in the case of [-agentive], the causative meaning in the case of [+agentive]. Table 3. Voice and agentivity PASS ronbun-o yom-are-ta thesis-ACC read-PASS-PAST ‘Taro suffered that his thesis was read.’ [-agentive] BEN ronbun-o yon-de kure-ta thesis-ACC read-SEQ give-PAST ‘(Someone) read the thesis for Taro’ [-agentive]

CAUS ronbun-o yom-ase-ta thesis-ACC read-CAUS-PAST ‘Taro made (someone) the thesis read.’ [+agentive] AUTOBEN ronbun-o yon-de morat-ta thesis-ACC read-SEQ get-PAST ‘Taro got the thesis read.’ [+/- agentive]

4. Structure of the causative-autobenefactive construction This section deals with the combination of the causative construction and the autobenefactive construction, thereby forming the so-called causative2 autobenefactive construction. In Japanese it is possible to combine the verb attached by the causative suffix and the verb of receiving into one construction, as the following example demonstrates: (13)

tomodati-ni/no konpyuutaa-o tukaw-ase-te morat-ta friend-DAT/GEN computer-ACC use-CAUS-SEQ get-PAST ‘I was (grateful to be) allowed to use my friend’s computer.’

The speaker is the beneficiary and the BEN’UM refers to the proposition that the friend (BEN’OR) lets the speaker use his computer. The friend is the causer, and the CAUS’UM refers to the proposition that the speaker (CAUS’EE) uses his computer. The causative structure is subordinated under the autobenefactive as illustrated below:

Grammaticalization of honorific constructions in Japanese

291

BEN morau BEN’RY I

BEN’UM My friend made me use the computer BEN’OR friend

Proposition made me the computer CAUS

CAUS’OR friend

CAUS’UM I used the computer CAUS’EE I

Proposition use the computer

Figure 4. Causative-autobenefactive construction (see (13))

There are two positions in this participant structure which have to be commented on: (a) causee as a humble position As described above, the causative construction can be used only for a causer who is in a socially higher position than the causee. Thus, it is already polite for the speaker to appear as cause in (13). By playing the role of causee, the speaker humbles himself and shows his respect to the other participant in forming the causative construction. This is done by defining one’s own position in this construction according to the principle of H3. (b) autobenefactive through permissive Furthermore, the causee appears at the same time as beneficiary by combining the causative with the autobenefactive. In (13), the autobenefactive expresses the speaker’s benefit in being allowed to use his friend’s computer. Making himself the causee, the speaker defines his role for the event voluntarily. Thus, this type of causative is always permissive. Otherwise

292 Yoko Nishina the coercive causative would be combined with the passive morpheme tukaw-as-are-ta (use-CAUS-PASS-PAST), which implies that the causee does not want to carry out the event and suffers when forced to do so. The permission, in contrast, presupposes the will of the causee and leads to a benefactive situation. In this situation the causer (permitter) is identical to the benefactor, the causee to the beneficiary. This relation between the participants is illustrated in Figure 5. speaker: “friend”:

CAUS’EE = BEN’RY CAUS’ER = BEN’OR

Figure 5. Participant relations

The next section investigates the character and relation of these participants, showing how this multiple voice structure is grammaticalized to a simple honorific construction expressing a single event.

5. Character and relation of participants 5.1. Syntactic functions of participants The relevant participants of the constructions in the following examples – beneficiary in the benefaction (14), causer in the causation (15), and indirect object of the proposition caused (here: recipient in the transfer situation) – are all encoded in the dative, as the following examples show. (14)

sennsei-ni suisen si-te morat-ta teacher-DAT recommendation do-SEQ get-PAST ‘(I) got the recommendation of the teacher’ (benefactor = realizer)

(15)

musume-ni hanataba-o watas-ase-ta daughter-DAT bouquet-ACC hand-CAUS-PAST ‘(He) made her pass the flowers’ (causee = realizer)

(16)

sotugyoosei-ni hanataba-o watasi-ta graduates-DAT bouquet-ACC hand-PAST ‘(She) gave the graduates the flowers’ (recipient)

Grammaticalization of honorific constructions in Japanese

293

Imagine a situation where the speaker (beneficiary) wants his daughter to get the honored role of handing flowers to the graduates (recipient) in the graduation ceremony of a school. The teacher (causer) makes an effort so that she (causee) gets the role of handing out the flowers. In the following example all these participants are involved in a single event: (17)

watasi-wa sensei-ni yot-te 1.SG-TOP teacher-DAT cause-SEQ musume-kara sotugyoosei-ni hanataba-o daughter-ABL graduate-DAT bouquet-ACC watas-ase-te itadai-ta hand-CAUS-SEQ get:H3-PAST ‘I am honored that her teacher caused my daughter to give the bouquets to the graduates’

This sentence may be artificial to some extent, but it is quite possible in Japanese. The following constellation of semantic roles of the participants and their syntactic functions with regard to a syntactic function hierarchy emerge in this sentence: The beneficiary, the referent of the autobenefactive function verb which is the head of the structure, is marked by the topic marker wa as the subject being in the uppermost position in the hierarchy of the syntactic functions. As for the participants involved in the event, the recipient is marked by the dative, the direct object by the accusative, and the realizer of the proposition (causee) has the lower syntactic function marked by the ablative -kara. The causer is the most downgraded and is expressed by the complex postposition grammaticalized from a verbal phrase such as ‘caused by the teacher’. The participants are encoded hierarchically depending on their function, i.e. the lower the syntactic function, the more complex the marking according to the scale: case marker > postposition > complex postposition. In the hierarchy of the syntactic function the causer has the lowest position. This participant might be the first who is backgrounded, downgraded, and omitted from this construction. To support this hypothesis, the significance of the causer should be investigated. In the real language data of Japanese, it is not always possible to determine which participant is the causer in a certain causative clause, or the causer plays no longer a relevant role in the situation designated by the clause. It seems contradictory that the causer is one of the central participants creating the causative situation. The next three sections discuss the mechanism leading to the omission of the causer.

294 Yoko Nishina 5.2. Double agent Depending on how the autobenefactive construction is used for an agentive beneficiary as discussed in section 4, the causative-autobenefactive construction can also be interpreted as having an agentive subject. In fact, a sentence like (18) is possible. If the speaker of (13) is agentive and decided to use his friend’s computer by his own will, the adverb katteni ‘at my discretion’ or ‘without permission’ can be added, although here we have a morphological permissive structure: (18)

tomodati-no konpyuutaa-o katteni tukaw-ase-te friend-GEN computer-ACC at.my.discretion use-CAUS-SEQ morat-ta get-PAST ‘I allowed myself to use my friend’s computer.’

Note that the agent in this causative-autobenefactive construction should be the causer and the benefactor as well in the permissive situation. If there is no permission, this means that in a permissive construction the permitter loses his control and a reason to exist. The subject, which should be a causee and a beneficiary in this structure, has control over the situation. It is confusing when a sentence has two different agents for the same proposition. Thus the causer without agentivity can no longer exist in this situation. The friend appears in this sentence not as a dative-marked causer but in the genitive, as a possessor of the computer. The relation between these two participants is not permissive but possessive. The causativeautobenefactive construction expresses only the speaker’s gratitude to the possessor. The participant can be identified only as the possessor, and the slot for the causer remains unexpressed.

5.3. Animacy of the causer In example (18) from the previous section the causer lost his causing function, but it is still possible to identify this participant. In the following example it is rather difficult to identify a specific causer:

Grammaticalization of honorific constructions in Japanese

295

(19)

Itaria-de-wa oisii ryoori-o tabe-sase-te Italy-LOC-TOP delicious meal-ACC eat-CAUS-SEQ morat-ta get-PAST ‘In Italy, I ate a delicious meal.’

(20)

(Nomiya-ya karaoke-de sanzai-suru-yori) sono kane-de hon-o kat-te MED money-INS book-ACC buy-SEQ yon-da-hoo-ga yoppodo tanosim-ase-te read-PAST-CMPR-NOM much enjoy-CAUS-SEQ morae-ta get:POT-PAST ‘I would much rather enjoy buying and reading a book with that money (than drinking and singing karaoke).’

In these causative phrases the possible causer would be “Italy” in (19) or “book” in (20). These are not human, which would be a typical characteristic of the causer role, but rather are a locality, an occasion, or an instrument. A sentence such as ‘The book made me happy’ is an unremarkable causative construction in English in which a causer need not have intention. In Japanese, such non-human causers cannot appear as subject of a simple causative sentence (cf. 21). A causer like that and the causative phrase has to be subordinated in the causative-autobenefactive construction: (21)

*hon-ga watasi-o tanosim-ase-ta book-NOM 1.SG-ACC enjoy-CAUS-PAST (‘The book made me enjoy myself’)

A country or a book does not have an intention to cause the event. They offer only an opportunity which affects the participant indirectly. This type of non-human indirect causer, which is rather a stimulus for the experiencer, is possible only in the causative-autobenefactive construction. This is, so to speak, a mirror-reflecting mechanism to the case of a double agent. Because the indirect causer does not have an agentivity, it can exist in the causer position in the causative-autobenefactive construction which already has an agent.

296 Yoko Nishina 5.4. Absence of the causer There is a case in which it is rather ungrammatical to express the causer, even if the causer can be identified. As we have seen in section 5.2., it is possible to insert an adverb against the causative meaning. In that example (18) it is also possible to express the causer depending on the interpretation of who possesses the agentivity. In the following example it is strange to designate the causer of the construction: (22)

*mina-san-ni happyoo-o hazime-sase-te all-H2-DAT presentation-ACC begin-CAUS-SEQ itadaki-mas-u get:H3-H1-PRS

A literal translation such as ’I benefited to be allowed by all of you to begin my presentation’ would surely be too wordy and unnecessary. It would sound rather natural without the causer phrase: mina-san-ni (1), and one can use this construction quite freely for every utterance with a modest effect. If anyone should allow the beneficiary to carry out the event, it should be the beneficiary himself. In this way, the beneficiary becomes the agent of this sentence and achieves the single event, namely to begin his presentation. The combination of the causative and the autobenefactive has a reduced valency, as the slot for the causer is suppressed. The complex morphological and syntactic structure results in a predicate expressing a single agent and a single event. The new function of this complex predicate is to express honorification. The verb carries a suffix for the causative and the function verb for the autobenefactive, which then becomes an honorific version of the causative construction. The next step in the grammaticalization of this construction is -(s)aseteitadakimasu, where nothing intervenes. This form is then reanalyzed as a single predicate. However, the acceptability of this causative-autobenefactive construction differs among speakers. The following example (23) from a research 3 report by the Agency for Cultural Affairs was evaluated as “unnatural” by 41.4% of the persons tested.

Grammaticalization of honorific constructions in Japanese

(23)

297

(makotoni moosiwakenaku) hukaku hansei s-ase-te itadak-imas-u deeply regret do-CAUS-SEQ get:H3-H1-PRS ‘(I am very sorry and) I regret it deeply.’

In this example, there is no one to cause/allow the speaker to deeply regret, and the speaker does not receive any benefit of deeply regretting. Compare this with (16) without a causer, which has a higher acceptability rate (evaluated as unnatural by only 8.1%) in the same research report: (24)

kore-de kaigi-o syuuryoo PROX-INS conference-ACC end itadaki-mas-u get:H3-H1-PRS ‘Now I end the conference.’

s-ase-te do-CAUS-SEQ

Here, (23) is less acceptable than (24). This is evidence that the grammati4 calization has not been completed but is ongoing. The only thing that varies between the two examples is the semantics of the main verb: “I benefited from your letting me regret” is less acceptable not only because there is no causer, but because you cannot benefit from regretting. Now that the benefactive meaning of this construction is not relevant, it is to be expected that (23) is gradually becoming more acceptable. It illustrates the gradience of the grammaticalization, and the last type of usage is the most advanced in the grammaticalization stage of the honorific construction.

6. Honorific constructions The question may arise as to why this construction is necessary as an honorific construction when there is already a rich honorific system in Japanese, as seen in Table 1. It is commonly claimed that, apart from some sets of lexical verbs, the majority of verbs make use of a morphologically productive way of forming honorific expressions, e.g. kiki-masu (H1), o-kikini naru (H2) and o-kiki-suru (H3) for the verb kiku ‘ask, listen to’. The differences are explained as a variation based on the speakers’ social status. However, taking a closer look at various situations, the issue is not that 5 simple. The example of H3 above, o-kiki-suru, is a humble expression to

298 Yoko Nishina refer to the speaker’s own activity of asking and to express respect for the listener. But if the addressee the speaker is asking belongs to the speaker's in-group, this expression is ungrammatical. Besides, it is even impossible to derive this form with many of the native verbs. Since verbs with a Chinese origin behave differently morphologically, the issue has also to do with the origin and formation of the verb. In the following section I will discuss such differences between honorific expressions and the reason for the grammaticalization of the causative-autobenefactive construction.

6.1. Paradigm of honorification A verb of H3 could substitute for the causative-autobenefactive construction in (24) as in the following example: (25)

kore-de kaigi-o syuuryoo PROX-INS conference-ACC end ‘Now I end the conference.’

itasi-mas-u do:H3-H1-PRS

Itasu is a humble form of suru ‘do’ and refers to the activity of [+EGO]. If the subject is [-EGO], a respectful verb of H2 nasaru or a respectful verb suffix of H2 -(r)areru is used. Nouns such as syuuryoo ‘end’ in (25) or hansei ‘regret’ in (23) are loan words of Chinese origin and form compound verbs with the light verb ‘do’. In a similar way, the stem (V) of some native verbs can be attached to an honorific prefix and a light verb such as o-V-suru, as seen above. However, there are a number of native verbs which have neither this form with ‘do’ nor a lexical form of H3. For example, verbs such as hazimeru ‘begin’ in (22), tanosimu ‘enjoy’ in (20), tukau ‘use’ in (18), etc., do not have a corresponding verb of H3. Whereas the respectful verb form of H2 is productive (either by the form with a prefix o-V-ni naru or by the verb suffix of H2 -(r)areru) and able to derive almost all native verbs into an H2 verb, there are no morphological means to derive a verb into an H3 apart from some basic verbs such as itasu ‘do’ and itadaku ‘get’. In order to fill the gap in the Japanese three-way honorification paradigm, the grammaticalized unit -saseteitadakimasu has become a productive verb suffix of H3.

Grammaticalization of honorific constructions in Japanese

299

6.2. Target of honorification It is not only the subject of the clause that receives honorific status. Other participants can be marked respectfully as well. In the honorific verb forms each component has a different honorific function. In the following examples, the H3 verbs show respect to the listener. In addition, in (26) the H2 prefix shows respect with regard to the recipient: (26)

siken.kekka-o sensei-ni go-hookoku.itasi-mas-u exam.result-ACC teacher-DAT H2-report.do:H3-H1-PRS ‘I will report the result of the examination to my teacher.’

(27)

siken.kekka-o haha-ni hookoku.itasi-mas-u exam.result-ACC mother-DAT report.do:H3-H1-PRS ‘I will report the result of the examination to my mother.’

In (27), the H2 prefix cannot be used for the recipient who belongs to the speaker’s in-group. It is not polite to respect the speaker’s own mother in the presence of a respected addressee. Whereas the verb here has a Chinese origin (hokoku), the verb okuru in (28) is a native verb. As demonstrated above (section 6.1.), there are no productive morphological means to derive a native verb such as okuru in (29) into an H3 which would have made a counterpart for a humble recipient of (28). (29) is ungrammatical both with and without the prefix: (28)

siken.kekka-o sensei-ni o-okuri.itasi-mas-u exam.result-ACC teacher-DAT H2-send.do:H3-H1-PRS ‘I will send the result of the examination to my teacher.’

(29)

*siken.kekka-o haha-ni (o-)okuri.itasi-mas-u exam.result-ACC mother-DAT (H2-)send.do:H3-H1-PRS (‘I will send the result of the examination to my mother.’)

A possible way to solve this problem is the causative-autobenefactive construction, which can be used for both classes of recipients:

300 Yoko Nishina (30)

siken.kekka-o sensei/hah-ni okur-ase-te exam.result-ACC teacher/mother-DAT send-CAUS-SEQ itadaki-mas-u get:H3-H1-PRS ‘I will send the result of the examination to my teacher/mother.’

In this construction, the speaker humbles himself in any event, which means that a respectful distance between speaker and the addressee is always kept up. On the other hand, the relative politeness between participants (teacher or mother) and the speaker is neutralized in the expression. As was shown in section 4, the causer in the causative structure is suppressed, so that the internal participant relation of this construction is given up. This construction becomes fairly convenient, because the speaker does not care about the clausal participant relation being the target of honorification, about being humbled or respected, or about the formation class of the verb (section 6.1). Because of its high productivity, it facilitates the right use of honorifics.

7. Conclusion We have shown that the honorific construction -(s)ase-te itadaki-masu is the grammaticalized outcome of the combination of the causative and the autobenefactive constructions. The status of the participants involved and the relation between them, as well as the semantics of the main verb, play an important role in acceptability. The syntactic functions of the encoded participants are hierarchically organized, with the causer having the lowest position. This facilitates the suppression of its slot. The autobenefactive construction promotes the agentivity of the beneficiary, to the extent that the causer loses the agentivity in the causativeautobenefactive construction. Since the causer role is suppressed, the participant relation of the internal causative-autobenefactive structure is simplified, which creates a simple honorific construction expressed as a complex predicate. This grammaticalized construction fills a gap in the existing honorific paradigm.

Grammaticalization of honorific constructions in Japanese

301

Notes 1. 2. 3.

4. 5.

Japanese examples are transcribed phonologically. About -sasete itadaku see Hidaka (1995) and Yamada (2004). Bunkatyoo: “Kokugo ni kansuru seron tyoosa. heise 10nendo” [Agency for cultural affairs Japan: A Research Report on Attitudes toward Japanese Language. 1998]. Usui (2006) reports that the use of this construction in the Minutes of the National Diet has increased over the last ten years. About subclassifications by various researchers see Nakanishi (1995).

References Hidaka, Mizuho 1995 O/go~suru rui to ~itasu rui to saseteitadaku – kenzyoo hyoogen –. In Nihongo ruigi hyoogen no bunpoo (ge) fukubun – renbun hen, T. Miyajima and Y. Nitta (eds.), 676–784. Tokyo: Kuroshio. Lehmann, Christian 2004 Participant roles, thematic roles and syntactic relations. In Voice and Grammatical Relations. Festschrift for Masayoshi Shibatani, T. Tsunoda and T. Kageyama (eds.), 167–190. Amsterdam/Philadelphia: Benjamins. Nakanishi, Kumiko 1995 O/go~ni naru to ~reru/rareru – sonkei no zyodoosi/setuzi –. In Nihongo ruigi hyoogen no bunpoo (ge) fukubun – renbun hen, T. Miyajima and Y. Nitta (eds.), 676–784. Tokyo: Kuroshio. Nishina, Yoko 2001 Zyuyoo doosi no bunpooka – doitugo ni okeru bekommen ukemi to nihongo ni okeru morau sieki no taisyookenkyuu. In Sekai no nihongokyooiku, 11, Japan Foundation (eds.), 167–177. 2004 Satzverbindung und Satzreduktion am Beispiel der japanischen Konverbkonstruktionen. Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt 17. Usui, Yoshiko 2006 saseteitadakimasu no kuni – kokkaikaigiroku kennsaku sisutemu wo siyoosite – Syakaigengokagakkai dai 17kai taikai happyooronbunnsyuu, 182–185. Yamada, Toshihiro 2004 Nihongo no benefakutibu – ‚teyaru’, ‚tekureru’, ‚temorau’ no bunpoo. Tokyo: Meijishoin.

From cause to contrast: A study in semantic change* Anna Giacalone Ramat and Caterina Mauri

1. Introduction In this paper we will discuss a particular case of semantic change undergone by the adverbial coordinating marker però in the history of Italian.1 As shown in (1), però is nowadays used to code a contrast originating from the denial of some expectation and is roughly equivalent to aber in German (cf. Scorretti 1988: 230–231). (1)

Mario gioca bene però perde lose:3SG Mario play:3.SG well but ‘Mario plays well but always loses.’

in continuazione. continuously

However, the specialization of però as an adversative marker is relatively recent. From its earliest occurrences in the 12th century until the end of the 16th century, it rather had a causal (però che ‘since, because’) or resultive function (però ‘therefore’, see example (3)). It is only in the 17th century that this marker is attested for the first time with an unambiguously adversative meaning. The evolution of però thus shows a functional reversal: whereas in its first occurrences it introduced the cause or the result of a causal sequence, now it signals the denial of an expected causal sequence, that is, what is commonly classified as counterexpectative contrast (see Scorretti 1988: 260–263; Mauri 2007a: 186, 2007b: chapter 5). Previous studies on the grammaticalization of adversative markers mainly focused on two patterns of semantic change: on the one hand the SPATIO-TEMPORAL > ADVERSATIVE pattern (Heine and Kuteva 2002: 291; Traugott 1986) attested in the evolution of markers such as whereas and while (Traugott 1995: 39–42); on the other hand the COMPARATIVE > ADVERSATIVE pattern, exemplified by the evolution of Lat. magis into It. ma, Fr. mais (cf. Ducrot and Vogt 1979). By contrast, except for a few brief hints in the literature (see section 2.2), the development of an adversative function from a causal one has not

304 Anna Giacalone Ramat and Caterina Mauri been examined in detail yet, even though it is particularly intriguing due to the functional reversal it involves. In this paper we aim to describe such a semantic change by reconstructing the process through which the Italian adversative marker però developed its counterexpectative function from an original causal/resultive one.2 In section 2 a brief synchronic account of però and a summary of previous analyses will be given. In section 3 the development of the adversative function of però will be described through all its stages, which shows the crucial role played by negation. Finally, in section 4 we will argue that the semantic change from cause to contrast is context-driven and involves the subjectification of the relation involved.

2. Previous studies 2.1. A synchronic account of però In contemporary usage, però encodes a contrast generated by the denial of some expectation (‘contrasto controaspettativo’ according to Scorretti 1988: 230–231; Battaglia 1961: 73). Leaving the exclamative use aside (però! ‘Wow!’), counterexpectative contrast is the only function associated with però as a clause linkage device (see example (1)). Però is only partially equivalent to the general adversative marker ma ‘but’, which has a broader semantic domain and may express both counterexpectative and corrective contrast.3 Però is functionally rather similar to German aber and Spanish pero, and it shares with the latter its etymological origin (< Lat. per hoc). As argued by Scorretti (1988: 231–232), però may cooccur with another coordinating marker such as e ‘and’ (Mario gioca bene e però perde ‘Mario plays well and però loses’), whereas this is not possible for ma. Therefore, according to the basic criterion for distinguishing adverbial from pure coordinating devices (see Dik 1968: 34), ma is classified as a pure coordinating marker, while però is considered an adverbial coordinating marker. Moreover, besides occurring in the initial position in the second member of a coordinating construction, as exemplified in (1), this marker is often found either within a clause, as in (2a), or at the end, as an afterthought, as in (2b).

From cause to contrast

(2)

305

a. Mario gioca bene, perde però in continuazione. b. Mario gioca bene, perde in continuazione però.

As said above, the functional and syntactic properties just described for però in Modern Italian do not hold for its earliest occurrences in Old Italian. Yet, before describing in detail the semantic change undergone by this marker and its syntactic implications, let us first briefly examine how its development has been explained by previous scholars.

2.2. Earlier analyses The development of però has been mentioned only briefly in a few studies, none of which specifically focused on this topic. Marconi and Bertinetto (1984: 490, 507) make a short remark on the history of però at the end of a detailed account of the evolution of Latin magis into Italian ma. They argue that the adversative function of però has its origins in the frequent cooccurrence of this marker with the contrastive conjunction ma (in sequences like ma però). Such cooccurrence would have modified the semantics of però from an original resultive function to a later adversative one. However, as will be shown in section 3, the cooccurrence of ma and però was rather rare and its low frequency does not support the hypothesis proposed by Marconi and Bertinetto. Rohlfs (1969: 170) argues that the modern adversative function of però developed from the original resultive meaning through the intermediate phase ‘despite all this>nonetheless’, which would also be attested in Old Spanish. However, he gives no evidence for this intermediate stage and does not explain how the change from the resultive into the concessive function would have occurred. According to Corominas and Pascual (1997: 495) and García (1999: 3856), the adversative meaning of Spanish però arose in negative contexts, where pero acquired a concessive nuance.4 The interplay between negation and resultive function is a promising explanation, but neither author provides any evidence or examples for this hypothesis. The role of negative contexts is also discussed by Mazzoleni in a footnote (2002: 407), where he cites Mussafia (1983: 54) in stating that the scope of negation over the causal/resultive connective may give rise to an adversative interpretation of the clause linkage (as in English not for

306 Anna Giacalone Ramat and Caterina Mauri that…). It is in such ambiguous contexts that però would have developed its denial of expectation function, as will be shown in section 3.2. Interesting remarks on negated causality contexts are made by König and Siemund (2000) too, even though their study does not focus on adversative markers but takes the concessive relation into account. In fact, concessivity shares a central semantic aspect with counterexpectative contrast, that is, the denial of an expected causal sequence. The propositional content of Paul was ill but went to the cinema is indeed roughly equivalent to that of Even though Paul was ill, he went to the cinema, if we do not take into account the coordinate vs. subordinate nature of the two relations. Therefore, although they do not mention the specific case of però, their suggestions on causality and concessivity will turn out to be relevant to this analysis too, as will be shown in detail in section 3.2. Except for the few above mentioned studies, the semantic change from cause to contrast and in particular the development of the adversative function of però have received little attention in the literature. In what follows, we will reconstruct the stages through which this marker has evolved, showing in what way negative contexts have triggered a process of reanalysis.

3. A history of però From its earliest occurrences in the 12th century,5 però is attested with a resultive function, which directly continues the meaning of its late Latin antecedent per hoc. As is also pointed out by Corominas and Pascual (1997: 495), per hoc gradually replaced propter hoc in Pliny, Apuleius, Quintilian and, more generally, in texts from the 2nd century on. Christian Latin texts show that per hoc is especially frequent as a resultive clause linkage device in Augustine (4th-5th century), where it is often preceded by ac ‘and’. Its use continued into Medieval Latin. The first phase of grammaticalization from Latin to Romance thus involves phonological erosion of Latin per hoc to Old Italian però, which leads to loss of morphological transparency and univerbation (see Hopper and Traugott 1993; Lehmann 1995; Heine 2003). However, since our primary concern here is the semantic change from cause/result to contrast, we will not explore the issue of the Latin origin of però any further. In Old Italian, però was indeed characterized by two functions, a causal (‘since’) and a resultive (‘therefore’) one (see example (3)). The semantic

From cause to contrast

307

change we are interested in originated from the resultive meaning in very specific contexts, that is, after a negation that had scope over the causeeffect relation. We will first examine the semantic and syntactic properties of però in its original causal and resultive functions (section 3.1.). The following discussion is based on 41 texts from the 13th to the 19th century (see Text Sample at the end of the paper), which were searched by means of automatic tools (WordSmith, ConcGram). This made it possible to enrich the qualitative analysis with relevant quantitative data. In section 3.2. the semantic change from resultive to adversative will be described and exemplified in detail.

3.1. Però che and però: cause and result In Old Italian texts from the 13th until the 15th century, one of the most widespread constructions introducing a cause within a causal sequence is però che (Vignuzzi 1973b; Barbera, forthcoming), meaning ‘since, because’. In this function, però is always followed by che and no other linguistic material may be inserted in between. Thus, they constitute a unitary and inseparable block. The subordinate causal clause introduced by però che could either precede the main clause, as shown by the first occurrence in example (3), or follow it.6 The use of però che with a causal function became rarer during the 16th century and almost disappeared from the 17th century on, being replaced by perché and poiché. Thus, for some centuries, the causal value of però che coexisted with the resultive value of però ‘therefore’, as shown by (3). (3)

Dante Alighieri, Divina Commedia, (1305–1321) Inferno XXXI, 23– 28. Ed elli a me: «Però che tu trascorri and he to me since you pass:2.SG pe le tenebre troppo da la lungi, through DEF darkness:PL too.much from DEF far away avvien che poi ne maginare abborri happen:3.SG that afterwards in.DEF imagine err:2.SG

308 Anna Giacalone Ramat and Caterina Mauri Tu vedrai ben, se tu là ti congiungi, you see:FUT:2.SG well if you there 2.SG.ACC join:2.SG quanto 'l senso s' inganna di lontano; how much DEF sense itself deceive:3.SG from far away però alquanto più te stesso pungi». therefore somewhat more yourself sting:2.SG ‘And he to me: «Because thou peerest forth / Athwart the darkness at too great distance, / it happens that thou errest in thy fancy. / Well shalt thou see, if thou arrivest there, / how much the sense deceives itself by distance; / therefore a little faster spur thee on.»’. (English translation by Henry Wadsworth Longfellow, 1861) As a resultive marker, però is usually placed in initial clause-position often following the conjunction e, which signals continuity with what was said in the previous discourse (example (4)). (4)

Agnolo Ambrogini, alias Poliziano, Detti piacevoli, 38 (1479) Messer Rinaldo, io ho inteso che voi that 2.PL Sir Rinaldo I have:1.SG heard impazzaste una volta,e però vi prego go mad:PAST:2.PL one time and therefore 2.PL.ACC pray:1.SG che voi m' insegnate come voi faceste that 2.PL 1.SG.DAT teach:2.PL how 2.PL manage:PAST:2.PL a guarire […] to recover ‘Sir Rinaldo, I have heard that you once went mad, and therefore I pray you to teach me how you managed to recover.’

Resultive però is frequent in texts from the 12th century until the 17th century, becoming increasingly learned and rare after about 1700. In I Malavoglia, a novel written by Giovanni Verga in 1881, no resultive occurrence of però is attested. The resultive meaning is instead systematically conveyed by perciò, which shares with però the origin in the Latin construction per hoc, but did not develop any counterexpectative meaning. Since the 20th century, però has only had an adversative meaning.

From cause to contrast

309

3.2. (Non) però: from non-resultive to adversative As Vignuzzi (1973a: 427) points out, a few rare occurrences of però allowing an adversative interpretation are already attested in the Divina Commedia, as exemplified in (5). However, as he argues, the basic meaning of the marker in such contexts is ‘therefore, for that’ and the counterexpectative reading is only collateral and determined by the context.

(5)

Dante Alighieri, Divina Commedia (1305–1321), Paradiso XVII, 93– 96. Poi giunse: «Figlio, queste son le chiose then add:PAST:3.SG son these be.3.PL DEF.PL commentary:PL di quel che ti fu detto; ecco of that which 2.SG.DAT be.PAST:3.SG said here are le 'nsidie che dietro a pochi giri son DEF.PL snare.PL that behind to few round.PL be:3.PL nascose. Non vo' però ch' a' hidden:PL NEG want 1.SG for.that/however that DAT tuoi vicini invidie […]». your PL neighbour.PL envy:2SG ‘Then added: «Son, these are the commentaries/ On what was said to thee; behold the snares / That are concealed behind few revolutions; / Yet would I not thy neighbours thou shouldst envy, […]»’ [English translation by Henry Wadsworth Longfellow, 1861]

The contexts that allow an adversative interpretation are characterized by the presence of some wide scope negation, as in (5) and (6). In such cases, però introduces some consequence that does not take place despite expectations, determining a contrast between the cause and the denial of the expected effect. (6)

Leon Battista Alberti, I Libri della famiglia, Prologue (1433–1441) Si fu la loro immensa gloria spesso dalla from:DEF if be.PAST:3.SG DEF their immense glory often invidiosa fortuna interrupta, non però fu adverse fortune interrupted NEG for.that be PAST:3.SG

310 Anna Giacalone Ramat and Caterina Mauri denegata alla virtù negated to:DEF virtue ‘If their immense glory was often obstructed by adverse fortune, not for that was it denied to virtue.’ Sentences like (6) are ambiguous between a negative-resultive and a counterexpectative reading: both interpretations ‘not for that was it denied…’ and ‘but it was not denied…’ make perfect sense. However, the contrast generated by the frustration of the expected cause-effect sequence is not explicitly coded, but only inferred from the negation of the consequence. It is thus plausible that però, when preceded by a negation, came to be interpreted by speakers as an overt marker of contrast. In particular, it was reanalyzed as a marker of a specific type of contrast, i.e. the one generated by the denial of an expected causal sequence. The ambiguity of cases like (6) is due to the fact that the frustration of a causal sequence is a conceptual aspect shared by both negative causal and counterexpectative contexts. As already mentioned in section 2.2, negative contexts are viewed by König and Siemund (2000) as pointing out the logical equivalence between negated causality and concessivity. In fact, the concessive relation is semantically equivalent to the counterexpectative one, at least for what concerns the denial of an expected effect (see section 2.2). Therefore König and Siemund’s analysis turns out to be insightful for this study too, even though they make no explicit reference to però. By explaining why some contexts may be ambiguous between two apparently contradictory interpretations like causality and concessivity, it becomes clear why those contexts may be ambiguous also between a causal and a counterexpectative contrast reading. According to König and Siemund, the connection between two contradictory relations such as causal and concessive is evident if we take negative contexts into account. An external negation that has scope over a causal relation is equivalent to an internal negation in a concessive or adversative linkage. In other words, negated causality [ ¬ because of p, q ] (7a) is equivalent to a concessive construction containing a negation [ although p, ¬ q ] (7b) (König and Siemund 2000: 354). To these two formalizations we may further add a third one with internal negation, namely the counterexpectative relation [ p, however ¬ q ] (7c), which is still equivalent to negated causality.

From cause to contrast

(7)

311

a. This hotel is not [less comfortable because it is cheap]. This hotel is cheap, not [because of that is it less comfortable]. b. Although this hotel is cheap, it is not less comfortable = c. This hotel is cheap, it is not however less comfortable = This hotel is cheap, but it is not less comfortable.

The process of reanalysis of però is thus based on the semantic equivalence existing between (7a) and (7c): in ambiguous contexts, where a cause was followed by the denial of its consequence (e.g. (5), (6), (8)), negative-resultive non però ‘not because of that’ was reanalyzed as adversative non però ‘not however’. Although the same connective is rarely used to express both a causal and a concessive function (Kortmann 1997: 202), König and Siemund point out some cases of causal markers developing into concessive ones such as English for all (2000: 346). The pattern of development whereby però became an adversative connective is slightly different since it involves the development of a causal marker into a counterexpectative one, but still it is a remarkable development, which, as our data show, originated just in those negated contexts which are ambiguous between the two readings. Occurrences of non però are rather rare in the 13th and 14th centuries (in the Divina Commedia there are only 6 instances out of 163 occurrences of però) and become increasingly frequent from the 15th and 16th centuries in contexts which generally allow for an ambiguous reading. Especially during the 16th century, non però is very frequently used as a clause linkage device: in Ludovico Ariosto’s poem Orlando furioso (1516), 80 out of 148 occurrences of però (54%) show the sequence non però linking a cause to the denial of its consequence and, more in general, 114 out of 148 occurrences of però are preceded by negation. In such negative constructions, però may either occur right after the negation, as in (6), or after the verb, as shown in (8). (8)

Luigi Pulci, Il Morgante, XX (1478) E benché tutto il mondo qua in aiuto DEF world here in help and although all […] venga a mia vendetta[…] come SUBJ:3.SG to my revenge Non riarò però quell ch’ ho NEG regain FUT:1.SG for that/however that which have 1.SG

312 Anna Giacalone Ramat and Caterina Mauri perduto. lost ‘[…] and although all the world here to help […] comes for my revenge […], I will not for that/however have back what I lost.’ The negated resultive construction frequently occurs in the environment of other contrast markers, such as a concessive clause preceding the main one, as in (8), or, more rarely, a connective like ma. Such redundancy may depend on the semantics of però, which is still ambiguous during this phase and is therefore reinforced by other more clearly adversative means. The adversative reading of però was strictly connected with the presence of negation for a long period of time (about three centuries). The earliest, rare occurrences of però without a negation and with a clearly adversative function (see example (9)) are found at the beginning of the 17th century. Cases like (9) show that the process of reanalysis is complete and però has been re-semanticized as a marker of counterexpectative contrast. (9)

Paolo Sarpi, Istoria del Concilio Tridentino, I (1619) […] il quale, se ben mandato dal pontefice […] REL although sent by DEF pope intervenne però come mandato da Francia […] sent by France participate:PAST:3.SG however as ‘[…] who, even though sent by the Pope,[…] participated however as if he was sent by France.’

There is an interesting temporal coincidence between the development of però as an adversative connective and the disappearance of però che with a causal function. By the end of the 17th century, causal subordination is almost exclusively introduced by perché and poiché. This may be due to two reasons. First, però had increased the number of its functions, thus giving rise to the development of new senses, which in turn caused ambiguity and hence a reduction in functions. Second, the frequency of occurrence of però in main clauses increased as a result of the development of the adversative meaning, since both adversative and resultive uses can appear in main clauses. Consequently, the disappearance of però che was favored, in that it was less frequent, since the causal function occurred in subordinate clauses.

From cause to contrast

313

From the 17th until the 19th century, the resultive and the adversative functions of però coexist but show different syntactic distributions. In its resultive meaning, però always occurs in clause-initial position and is often preceded by e (cf. example (4)). By contrast, when it has a counterexpectative value, però tends to be postposed, either after the verb or at the end of the second clause, as shown in (9). In other words, the ambiguous semantics of the marker is disambiguated by the syntactic context. In the second half of the 19th century, però only rarely occurs with a resultive function and the syntactic contexts where it is used as a counterexpectative marker start to include also the initial position, where it is commonly found also in modern Italian (see example (1)). Therefore, the disappearance of the resultive function of però and the consequent specialization of this marker as an adversative coordinating marker is recent and must be located between the end of the 19th century and the beginning of the 20th.

4. Concluding remarks: semantic change and subjectivization As pointed out by Traugott (1995: 32), subjectivization in grammaticalization concerns the development of a grammatically identifiable expression of the speaker’s belief or attitude to what is said. Traugott provides a number of examples involving a shift from relative objective reference to functions based in speaker’s attitude, such as let us, let alone, I think, while, etc. (Traugott and Dasher 2002, Traugott 1999). The data discussed here suggest a similar development for the causal to adversative interpretation of the adverbial coordinating marker però in Italian. To sum up, we have shown that: Latin (ac) per hoc is attested as a connective element only from the 2nd century in contexts in which it provides the reason for some conclusion. In Old Italian texts, però normally has a resultive meaning (‘therefore, as a result’), while però che (together with perché) is used as causal connective ‘because’. Some examples, however, were provided of the kind of context in which an adversative inference might have originated (see examples (5), (6) and (8)). Such contexts appear to be constrained by the presence of negation. The inference of contrast does not seem to be dominant in earlier texts; rather, it seems a potential interpretation which has coexisted for some centuries alongside the meanings of causality and

314 Anna Giacalone Ramat and Caterina Mauri consequence. In the literature on grammaticalization this phase of coexistence is referred to as layering (Hopper 1991; Hopper and Traugott 1993). We have found evidence that the adversative interpretation of resultive però was grammaticalized in the period in which però che ‘because’ was replaced by perché (as we said, both connectives coexisted in Old Italian). In the same period, the resultive meaning of ‘therefore’ was consistently expressed by perciò and this provided an opportunity for però to extend the contrastive interpretation which conventionalized the unexpectedness of some situation. There was no change in grammatical status, since però maintained its status as an adverbial coordinating marker, but a process of semantic reanalysis focusing on the speaker’s attitude took place. A causal relation may have different degrees of subjectivity (cf. Pander Maat and Degand 2001; Pander Maat and Sanders 2001; Sweetser 1990) depending on the degree of speaker involvement. The relation may be objectively established between two states of affairs (on the content level, with a minimal degree of speaker involvement), between a state of affair and a speaker’s assumption (on the epistemic level, in this case the speaker is involved with an inferential process) or between a state of affair and a speech act (on the speech-act level, in this case the speaker is involved with an action). In other words, causality may be based on objective circumstances of the real world and does not necessarily imply the presence of an inferential process involving the speaker’s assumptions. In our sample, all the attested occurrences of resultive però are of the first type, that is, they belong to the content level (cf. examples (3), (4) and (10)). (10) Giovanni Boccaccio, Il Decameron, Third Day - Fifth Story (1478) […] Or qui non resta a dire al presente now here NEG remain:3.SG to say to:DEF present altro; e però, carissima mia donna, […] a Dio more and therefore dearest my lady to God v' accomando. you.ACC entrust.1.SG ‘Now there is nothing else to say here at present; and therefore, my dearest Lady, […] I entrust you to God.’ On the other hand, when però is preceded by a negation (cf. examples (5), (6) and (8)), the relation is ambiguous: if it is interpreted as belonging

From cause to contrast

315

to the content level, the reading is resultive, whereas if it is interpreted as involving a speaker’s assumption, the reading is adversative. At the final stage of this semantic change, when the presence of a negation is not compulsory any more, the only possible reading of però is counterexpectative and it necessarily involves the denial of some assumption. As argued by Lang (2000: 243–244), the interpretation of a counterexpectative contrast always implies inferring an assumption that is denied by one of the two statements (cf. sections 2.1. and 3.2.). This subjectivity of the adversative relation has already been pointed out in linguistic research. As Malchukov (2004: 183) observes, two states of affairs standing in a contrast relation (i) have some aspects in common, (ii) are different under some respects (otherwise they would be identical) and, most importantly, (iii) they are compared with respect to these differences. Rudolph (1996: 20) also describes the connection of contrast as the speaker’s opinion that two states of affairs are valid simultaneously and that the second state of affairs conflicts with some information given in the first one. In other words, contrast is characterized by the speaker’s evaluation of similarities and dissimilarities with respect to some previous expectations. As a consequence, contrast cannot be exclusively based on objective circumstances of the world independently of the speaker’s attitude, but it rather depends on the speaker’s inferential ability. Under this respect, counterexpectative contrast may be regarded as more subjective than causality. Thus the history of però is a further example of discourse-based development of connectives which come to express speaker-based functions out of some earlier, more objective meanings. The adversative/concessive interpretation developed in negative contexts and expressed the speaker’s assessment of an unexpected relation between two events or two propositions. A further step in the grammaticalization process of the adversative relation occurred when the negation ceased to be required, and però directly introduced the second proposition as the marker of some conclusion unexpected for the speaker. It is possible to analyze this change as a metonymical change favored by the contiguity of the negation and però: the negative force is transferred to the contrastive marker. It should be added that nowadays the old meaning of però is not recoverable by inference. This is what we should expect if the hypothesis of unidirectionality of inferences is correct (Traugott and König 1991:199). Let us conclude with some questions for grammaticalization theory: how crucial are semantic bleaching and phonological reduction to

316 Anna Giacalone Ramat and Caterina Mauri grammaticalization (the issue is discussed by Campbell 2001: 118, among others)? Phonological reduction in però (< Latin per hoc) is probably a motivating factor for the replacement of però as a resultive marker with more transparent, analytical constructions: perciò, per ciò, per questo (motivo) “for that/this reason”. By contrast, the notion of semantic bleaching does not seem to apply to the semantic change of però from a resultive to a counterexpectative meaning. Even though some parts of its meaning were lost, namely the causal and the resultive functions, a change in perspective took place which corresponds to the strengthening of the speaker’s point of view.

Text sample XIII: Novellino; Tesoretto (Brunetto Latini). XIV: - Divina Commedia (Dante Alighieri); Decameron (Giovanni Boccaccio). XV: Elementi di pittura, Uxoria, Sofrona, Naufragus, I libri della famiglia, Deifira, De Iciarchia, De amore, Cena familiaris (Leon Battista Alberti); Il Morgante (Luigi Pulci); L’Orlando Innamorato (Matteo Maria Boiardo); Trattato, Scritti, Aforismi (Leonardo da Vinci). XVI: Orlando Furioso (Ludovico Ariosto); La Cortigiana, Angelica (Pietro Aretino); Vita (Benvenuto Cellini); Gerusalemme conquistata, Discorso sulla virtù femminile e donnesca, Aminta (Torquato Tasso). XVII: La città del sole, Le lettere (Tommaso Campanella); Adone, Amori (Gianbattista Marino); Istoria del Concilio Tridentino (Paolo Sarpi). XVIII: Autobiografia, Principi di scienza nuova (Gianbattista Vico); Oreste, Vita (Vittorio Alfieri). XIX: I promessi sposi (Alessandro Manzoni); Zibaldone, Paralipomeni della Batracomiomachia, Guerra dei topi e delle rane (Giacomo Leopardi); I Malavoglia, Mastro Don Gesualdo (Giovanni Verga).

From cause to contrast

317

Notes *

1.

2.

3.

4.

5.

6.

This work is the result of a continuous exchange of ideas between the two authors. However, Caterina Mauri is responsible for the writing of sections 2 and 3 and Anna Giacalone Ramat is responsible for the writing of sections 1 and 4. In defining però as an ‘adverbial coordinating marker’ we follow Scorretti (1988: 230), who uses the label ‘operatore avverbiale di coordinazione’, literally ‘adverbial operator of coordination’. By coordinating marker it is meant here any overt marker coding a coordination relation (see Mauri 2007b: chapters 1 and 2 for a detailed discussion) and by adversative marker it is meant any overt marker coding a coordination relation of contrast. German dafür, etymologically the same as però, perciò, has undergone a similar semantic change from causal 'therefore' to contrastive ‘but, on the other hand’: Sie ist klein, dafür wohl proportioniert. ‘she is short, but well-proportioned.’ (Marconi and Bertinetto 1984: 507). Further research is however needed to reach a fuller understanding of the development of dafür. The term corrective contrast refers to the contrast generated by the negation of a state of affairs and by its substitution with another state of affairs (Anscombre and Ducrot 1977: 25, Rudolph 1996: 141, Abraham 1979: 92– 94), as in ‘He did not run upon the hill, but simply walked slowly and lazily following the rest of the group’. This type of contrast is coded in German by sondern, in Spanish by sino, in Italian by bensì. The evolution of Spanish pero is slightly different from that of Italian però, even though their semantics is basically the same. In Spanish pero does not have the syntactic freedom that is typical of Italian però and always occurs in initial position (Corominas and Pascual 1997: 495–496, cf. also García 1999). Moreover, the development of Spanish pero did not go through the causal meaning ‘since’, which remained instead one of the main functions of però (che) in Italian until the 15th century (see Section 3.1). One of the earliest occurrences of però is found in the Ritmo Laurenziano (1180 ca.), in the form peròe and with the meaning ‘for that’ (see Spitzer 1951 for further discussion). According to Corominas and Pascual (1997: 495) and García (1999: 3856), the causal construction with però che is not found in Old Spanish texts and would thus be a specific development of Italian però.

318 Anna Giacalone Ramat and Caterina Mauri

References Abraham, Werner 1979 But. Studia Linguistica XXXIII,II: 89–119. Anscombre, Jean-Claude and Oswald Ducrot 1977 Deux mais en Français? Lingua 43: 23–40. Barbera, Manuel forthc. Causali. In ItalAnt. Grammatica dell’Italiano Antico (http:// geocities.com/gpsalvi/konyv/), Lorenzo Renzi and Giampaolo Salvi (eds.). Bologna: Il Mulino. Battaglia, Salvatore 1986 Grande Dizionario della Lingua Italiana, Vol. XIII. Torino: Unione Tipografico-Editrice Torinese. Campbell, Lyle 2001 Grammaticalization: A critical assessment. Language Sciences 23. Corominas, Juan, and José A. Pascual 1997 Diccionario Crítico Etimológico Castellano e Hispánico, Vol. IV. Madrid: Editorial Gredos. Dik, Simon 1968 Coordination. Its implications for the Theory of General Linguistics. Amsterdam: North-Holland Publishing Company. Ducrot, Oswald, and Carlos Vogt 1979 De magis à mais: une hypothèse sémantique. Revue de Linguistique Romane 43: 317–341. García, Luis Flamenco 1999 Las construcciones concesivas y adversativas. In Gramática Descriptiva de la Lengua Española, Ignacio Bosque and Violeta Demonte (eds.), 3805–3878. Madrid: Espasa-Calpe. Heine, Bernd 2003 Grammaticalization. In Handbook of Historical Linguistics, Richard D. Janda and Brian D. Joseph (eds), 575–601. Oxford: Blackwell. Heine, Bernd, and Tania Kuteva 2002 World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Hopper, Paul J. 1991 On some principles of grammaticization. In Approaches to Grammaticalization, Elizabeth Closs Traugott and Bernd Heine (eds.), 17–36. Amsterdam/Philadelphia: Benjamins. Hopper, Paul J., and Elizabeth C. Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press. König, Ekkehard and Peter Siemund 2000 Causal and concessive clauses: formal and semantic relation. In Cause, Condition, Concession, Contrast, Elizabeth Couper-Kuhlen

From cause to contrast

319

and Bernd Kortmann (eds.), 341–360. Berlin/New York: Mouton de Gruyter. Kortmann, Bernd 1997 Adverbial Subordination. A Typology and History of Adverbial Subordinators Based on European Languages. Berlin: Mouton de Gruyter. Lang, Ewald 2000 Adversative connectors on distinct levels of discourse: a reexamination of Eve Sweetser’s three level approach. In Cause, Condition, Concession, Contrast, Elizabeth Couper-Kuhlen and Bernd Kortmann, (eds.), 235–256. Berlin, New York: Mouton de Gruyter. Lehmann, Christian 1995 Thoughts on Grammaticalization. Munich and Newcastle: Lincom Europa. (Original edition, 1982). Malchukov, Andrej 2004 Towards a semantic typology of adversative and contrast marking. Journal of Semantics 21: 177–198. Marconi, Diego, and Pier Marco Bertinetto 1984 Analisi di ma (Parte prima: Semantica e pragmatica). Lingua e Stile XIX/2: 223–258, (Parte seconda: Proiezioni diacroniche) Lingua e Stile XIX/3: 475–509. Mauri, Caterina 2007a Conjunctive, disjunctive and adversative constructions in Europe: Some areal considerations. In Europe and the Mediterranean as Linguistic Areas Convergencies from a Historical and Typological Perspective, Paolo Ramat and Elisa Roma (eds.), 183–214. Amsterdam-Philadelphia: John Benjamins. 2007b Coordination relations: a cross-linguistic study. Ph.D. Thesis, University of Pavia. Mazzoleni, Marco 2002 La ‘paraipotassi’ con MA in italiano antico: verso una tipologia sintattica della correlazione. Verbum. Analecta Neolatina IV/2: 399– 427. Mussafia, Adolfo 1983 Reprint. Il Decameron di Giovanni Boccaccio riscontrato coi migliori testi e postillato da Pietro Fanfani. In Scritti di Filologia e Linguistica, Antonio Daniele and Lorenzo Renzi (eds.), 1–94. Padova: Editrice Antenori. Original edition, in Rivista Ginnasiale IV: 733–766 and 857–908, 1857]. Pander Maat, Henk, and Liesbeth Degand 2001 Scaling causal relations and connectives in terms of speaker involvement. Cognitive Linguistics 12 (3): 211–245.

320 Anna Giacalone Ramat and Caterina Mauri Pander Maat, Henk, and Ted Sanders 2001 Subjectivity in causal connectives: an empirical study of language in use. Cognitive Linguistics 12 (3): 247–273. Rohlfs, Gerhard 1969 Reprint. Grammatica storica della lingua italiana e dei suoi dialetti Vol. III. Torino: Einaudi. Original edition, Historische Grammatik der Italienischen Sprache und ihrer Mundarten, Bern: Francke, 1954. Rudolph, Elisabeth 1996 Contrast. Adversative and Concessive Expressions on Sentence and Text Level. Berlin, New York: Walter de Gruyter. Scorretti, Mauro 1988 Le strutture coordinate. In Grande Grammatica Italiana di Consultazione Vol. I, Lorenzo Renzi (ed.), 227–270. Bologna: Il Mulino. Spitzer, Leo 1951 Notes to the Text of “Ritmo Laurenziano”, Italica Vol. 28, No. 4: 241–248. Sweetser, Eve 1990 From Etymology to Pragmatics. Metaphorical and cultural aspects of semantic structures. Cambridge: Cambridge University Press. Traugott, Elizabeth 1986 On the origins of “and” and “but” connectives in English, Studies in Language 10 (1): 137–150. 1995 Subjectification in grammaticalisation. In Subjectivity and Subjectivisation, Dieter Stein and Susan Wright (eds.), 37–54. Cambridge: Cambridge University Press. 1999 The rhetoric of counter-expectation in semantic change: a study in subjectification. In Historical Semantics and Cognition Andreas Blank and Peter Koch (eds), Berlin, New York: Mouton de Gruyter. Traugott, Elizabeth C., and Ekkehard König 1991 The semantics-pragmatics of grammaticalization revisited. In Approaches to Grammaticalization Vol 1, Elizabeth C. Traugott and Bernd Heine (eds.), 189–218. Amsterdam: John Benjamins. Traugott, Elizabeth, and Richard Dasher 2002 Regularity in Semantic Change. Oxford: Oxford University Press. Vignuzzi, Ugo 1973a Però. In Enciclopedia Dantesca vol. IV., Umberto Bosco (ed.), 425– 428. Roma: Istituto della Enciclopedia Italiana. 1973b Però che. In Enciclopedia Dantesca vol. IV, Umberto Bosco (ed.), 428–431. Roma: Istituto della Enciclopedia Italiana.

Index

accessibility hierarchy, 215 accusative, 36, 41, 252–271, 286 actor/agent, 136, 145f, 186, 189, 195, 273, 294 adjective, 80, 85, 87, 226–228 adjunct, 78f, 93, 255 adposition, 28–30, 32, 50, 77–102, 139–141, 150 adverb, 34, 45–50, 79, 86, 97–101, 159, 180, 227–229 adversative, 303–316 agentivity, 189–191, 194, 197f, 202f, 253–278, 285, 288–300 agreement, 80, 121–133 Akan, 98 analytic languages, 114 AnejomÕ, 162 animacy, 59–61, 64–71, 80, 251, 258, 267, 294f applicative, 135–151 apposition, 107, 112 argument, 80, 86, 94, 136, 140, 144, 252 aspect, 80, 169–182, 187, 245–248 assimilation, 146 associative, 157–160 attrition, 2f autonomy, 2, 63, 135, 181 auxiliary, 169–182, 186, 241, 244, 247f Basque, 209 beneficiary, 144, 255, 283–293 Biloxi, 145, 147, 149 bleaching, 25, 148, 244, 315 bondedness, 2, 4, 64, 131f Boruca, 126–129 Boumaa Fijian, 162 case, 33, 39–42, 80, 139, 160

Catalan, 209 causation, 186, 193–202, 285–292 cause, 303–316 causee, 190, 199, 283–294 Chantyal, 188–190, 193, 202f Chibchan languages, 121–133 Chinese, 163 cleft construction, 99 clitic, 81f, 122, 128–130, 251–278 coalescence, 2, 4 cohesion, 64, 130–132 collective, 158, 162 comitative, 90, 158 complement, 86, 94, 97, 142, 179, 196 compositionality, 21, 87–91, 220f condensation, 2, 4 conjunction, 94, 112f, 173, 305 construction, 6, 92f, 97–99, 173, 219–235, 241–248, 283–300 continuative, 170–182 contrast, 227, 303–316 conventionalization, 162 coordination, 92, 94, 105, 107, 111, 113, 304, 306 copula, 85, 159, 179, 197 coverb, 139–141 Crow, 147 Cuna, 129–132 Dakotan, 147 dative, 36, 39, 41, 93, 285 decay, 138, 150 degeneration, 159 degrammaticalization, 6f desemanticization, 3, 25, 33–35, 37f, 42, 47–49, 62, 67f, 71, 89, 91, 101, 142, 147f, 159, 177, 271f Dhegiha, 147

322 Index directional, 138, 160–162 directive (speech act), 224–234 discourse, 26, 121, 125, 131f, 137, 219–235, 304–315 dislocation, 255–278 ellipsis, 109 English, 28, 30, 84, 195, 198f, 209, 241–248, 251, 295, 305 erosion, 306 Ewe, 77f, 87, 92, 98, 102, 157, 186, 188–190, 195, 197, 202 exclamative, 222f experiencer, 187, 196, 251–278 fixation, 2, 4 focus, 126, 128, 133, 187, 222, 246, 256–278 fossilization, 142, 145, 147, 149 French, 189 function word, 241, 283 functional domain, 27–30 fusion, 128 genitive, 36, 87, 105–114, 210–215, 252–271 German, 105–114, 136f, 189, 202f, 219–235 grammaticalization cline, 114, 122 cycle, 8, 18, 25f, 50 multiple, 157–165 parameter, 2 path, 77, 83, 92, 101, 139, 219 source, 42f, 64–71, 77, 122, 139, 142 Greek, 33–55, 251–278 habitual, 187 Hausa, 188, 190f, 193f, 196f, 202 head, 80, 82f, 87, 92, 210, 288 head-marking, 132 Hidatsa, 145, 147f Hočank, 135–137, 145f, 148 honorific, 60, 283–300 ideophon, 193f, 199, 202f idiomaticization, 91, 220, 234f

imperative, 222–235 imperfective, 171 impersonal, 195 inflection, 80, 144, 242 information structure, 127, 131 innovation, 23 instrumental, 136, 138, 147 integrity, 2, 63, 150 interrogative, 222–224, 263 inversion, 122 Italian, 303–316 Jaminjung, 185, 188f, 192, 196–203 Japanese, 209–215, 283–300 Kalam, 185, 188f, 195f, 200, 202f Kansa, 149 Kham, 188, 190, 202 Korean, 59–72 Koromfe, 77f, 86 Latin, 313 lexicalization, 7, 135, 139–150, 175 localization, 27–30, 84 Malagasy, 209, 214f Maltese, 169–182 Manambu, 157–165 Mandan, 145 metaphor, 34, 43, 90 metathesis, 145 metonymy, 101 Mixtec, 89 mood, 80, 222–225 motion, 28f, 41, 163 mutation, 15–23 Nànáfwê, 28f, 77–104 negation, 17–20, 50, 95, 97f, 180, 304–315 noun, 80f, 83–86, 111 body part, 89, 101 common, 59–71, 85 locative, 83–91 relational, 77, 79, 139–141, 150 number, 63, 81, 123–129 numeral classifier, 59–72

Index object, 122f, 126f, 136–151, 195, 210–215, 244, 251–278, 286 obligatorification, 2f, 113, 131, 225, 247f, 271f, 277 Ofo, 145, 147, 149 Osage, 145 paradigmatic variability, 2, 30, 148, 150, 181 paradigmaticity, 2, 131f, 150, 225 paradigmaticization, 2f participle, 244 particle, 219–235 passive, 210–215, 290 perfect, 243–248 perfective, 245–247 periphrasis, 171, 187, 189, 241–248 person, 123–130, 145 polygrammaticalization, 157 portmanteau morpheme, 128 possession, 80f, 83, 86–91, 105–114, 129, 244f postposition, 77–91, 127 predicate, 80, 89, 198–200 preposition, 21, 28–30, 33–49, 84, 105–114, 241, 252 progressive, 246 pronoun, 80f, 87–91, 122–133, 145, 241, 251–278 purpose, 34 quantifier, 65 Quapaw, 149 Rama, 125f, 132, 141f reanalysis, 41, 97f, 101, 159, 244, 306, 311f, 314 recipient, 34, 144, 292f, 299 reciprocal, 157–165 reduplication, 159, 161, 163 reflexive, 164 region, 84, 90 reinforcement, 8, 25f, 49 reinterpretation, 162, 244 relative clause, 209–215, 264–266, 271f, 277f

323

renovation, 8, 25f, 49 resultative, 244–247 resultive (function), 304–316 Russian, 18–22 Samoan, 185, 188f, 192, 197, 202f selection, 15–23 semantic change, 228, 303–316 semantic map, 200–202 Siouan languages, 135–151 Spanish, 304f spatial region, 28f speech act, 223–234, 314 stimulus, 253–267 structural scope, 2 subject, 80, 82, 95, 123–132, 140, 186, 210–215, 258 subjectification, 304, 313–316 subjunctive, 226, 230–234 synecdoche, 61 syntagmatic variability, 2, 30, 65 synthetic languages, 114 Tagalog, 214f Tariana, 162 tense, 80, 95, 187, 241–244, 246f Teribe, 122–125, 132 Tongan, 162 topic, 128f, 187, 251, 256f, 268–272 transparency, 38f, 91, 148f, 220 Twi, 92 undergoer/patient, 136, 145f, 273 unidirectionality, 6, 15–23, 26, 135, 315 univerbation, 141, 181 universal, 45 valency, 42, 85, 252 variation, 38, 41, 87, 271 verb, causative, 186–190, 194, 253, 283–300 ditransitive, 136f experience, 251–278 inflection, 80, 95 of generalized action, 185–203

324 Index verb, of motion, 28–30, 40f, 85, 90, 93f, 97f, 147f, 157, 193 of posture, 88 of speaking, 179 quotative, 191–193 serial, 77, 81, 92–99 transitive, 77, 86, 135–151, 159, 195, 259–278

volitionality, 254 Welsh, 209 word order, 80, 99, 107f, 112, 122, 124–133, 140, 251–278 Yimas, 188, 190, 196f, 202f Yucatec Maya, 86 Yuman, 199 zero, 85, 122, 129, 244