Language Change and Functional Explanations 9783110813753, 9783110149135

Citation preview

Language Change and Functional Explanations

W DE

G

Trends in Linguistics Studies and Monographs 98

Editor

Werner Winter

Mouton de Gruyter Berlin · New York

Language Change and Functional Explanations

edited by

Jadranka Gvozdanovic

Mouton de Gruyter Berlin - New York

1997

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

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

Library of Congress Cataloging-in-Publication-Data

Language change and functional explanations / edited by Jadranka Gvozdanovic p. cm. - (Trends in linguistics. Studies and monographs ; 98). Selected papers of a conference on Functional Explanations of Language Change, Sept. 2 and 3, 1994, University of Amsterdam. Includes bibliographical references and index. ISBN 3-11-014913-3 1. Linguistic change-Congresses. 2. Functionalism (Linguistics)-Congresses. I. Gvozdanovic, Jadranka. II. Series. P142.L255 1996 417-dc21 96-46266 CIP

Die Deutsche Bibliothek — Cataloging-in-Publication-Data

Language change and functional explanations / ed. by Jadranka Gvozdanovic. - Berlin ; New York : Mouton de Gruyter, 1997 (Trends in linguistics : Studies and monographs ; 98) ISBN 3-11-014913-3 NE: Gvozdanovic, Jadranka [Hrsg.]; Trends in Linguistics / Studies and monographs

© Copyright 1997 by Walter de Gruyter & Co., 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. Diskconversion: Fotosatz-Service Köhler OHG, Würzburg. Printing: Ratzlow-Druck, Berlin. Binding: Lüderitz & Bauer, Berlin. Printed in Germany.

In memory of Simon C. Dik (1940-1995) who was still with us at IFOTT when this work came into being.

Preface

The papers in this volume are a selection of those presented at a conference on Functional Explanations of Language Change, held on September 2 and 3, 1994, at the Institute for Functional Research of Language and Language Use (IFOTT) at the University of Amsterdam, in part sponsored by the Dutch Graduate School in Linguistics (LOT).1 They reflect the comprehensive and interdisciplinary orientation of the organizing institute, which has at the same time benefitted from the invited foreign expertise. As stated by Lachlan Mackenzie, chairman of the IFOTT Board in his opening words at the conference, we have a better understanding of what 'function' is now than a few decennia ago, and we have various theories in which this notion plays a crucial role. Even if we grant that explanation in the very strongest sense - as prediction in reverse — is unattainable in a science that is dealing with human beings and the vicissitudes of their behavior, individually and in groups, we must feel confident that we now have ways of making language change intelligible. The present volume will hopefully provoke future discussions, further deepening our understanding of the field.

1

Special thanks are due to Casper de Groot, Louis Pols, and Yvonne Sanders of the IFOTT Directorial for their support in the organization of the workshop, and to Mouton de Gruyter for the expedient and pleasant collaboration in the preparation of this book.

Contents

Jadranka Gvozdanovic Introduction

l

Functional and cognitive aspects of change Rudi Keller In what sense can explanations of language change be functional?

9

Erica Garcia Grasping the invisible hand

21

Stig Eliasson The cognitive calculus and its function in language

53

Jadranka Gvozdanovic Tracing the origin of a change

71

Lourens de Vries The rise of switch-reference in the Awyu languages of Irian Jaya

89

Linguistic activity and change Wolfgang U. Dressler "Scenario" as a concept for the functional explanation of language change

109

Cor van Bree Changes in diminutive formation in the eastern Dutch dialect ofTwente

143

X

Contents

Peter van Reenen Reformed versus Catholic: the origins of the [hu:s]/[hy:s] isogloss in eastern Netherlands at the beginning of the 17th century

179

Henk Haverkate Indirectness in speech acts from a diachronic perspective: some evolutionary aspects of rhetorical questions in Spanish dialogue

219

Typology and language change Kees Hengeveld — Gerry Wanders On the use of subjunctive and indicative verb-forms in adverbial clauses

249

Brigitte L. M. Bauer Nominal syntax in Italic: a diachronic perspective

273

Introduction Jadranka Gvozdanovic

This is a book about motivations for change, change processes and instantiations, and their functional explanations. Its main focus is not on structural properties and the corresponding rules, in relation to which change may be analysed as reanalysis, analogy, or borrowing (cf., i.a., Anttila 1977, Kiparsky 1992, Hopper and Traugott 1993, Harris and Campbell 1995), but rather on communicative and cognitive circumstances under which change may arise and on sociolinguistic circumstances under which it may win out. What is it that triggers a new analysis of extant forms and arrangements, or attracts forms and arrangements to other already existing ones? Which factors favor borrowing over system-internal innovation? When will a change spread throughout linguistic and socio-cultural contexts, and when will it be prevented? These and similar questions are discussed and partly answered in the present book. Innovative changes typically take place between generations of speakers, and adaptive changes within generations (concerning adaptive rules, cf. Andersen, e.g., 1973). Both types arise from language communication. Language users analyse the language manifestations perceived by them by matching them with the general laws and rule types, thus arriving at a particular analysis of these manifestations (cf. Andersen 1973 on "abduction"). However, as language systems are never fully deterministic, there is the inherent possibility of partial ambiguity. In such cases, language users come up with alternative analyses of the same manifestations. With reference to these analyses, they then produce new forms and arrangements, which only seldomly retain the same ambiguity. In most cases, they lead to overt disambiguation in the end, by which the forms and arrangements become manifestly different from the preceding ones. When this happens, the language community will either sanction the innovation or reject it by correcting it. Its evaluation and acceptance will depend on its functional impact given general axioms of communicative activity, and so in a relevant pragmatic and sociolinguistic setting. Processes of innovation are complex indeed. As

2

Introduction

Andersen (1989: 14) points out, "each and every step in such a development is an innovation, not only the initial act, through which a new linguistic entity comes into being. It is through innumerable individual acts of innovation — of acceptance, adoption and acquisition — that any new entity gains currency and enters into competition with traditional entities in the usage of a linguistic community". We may say that investigation of change mechanisms and processes has been function-oriented from the outset. Explicit discussions about explanation of change in functional terms are of a more recent date. In the history of our thinking about functional explanations of language change, two turning points may be mentioned. The first of these is marked by Lass's (1980) book on language change, in which the author shows correctly that change does not involve (conscious) human purpose, and that proposed functional explanations cannot predict change, only make us understand it a posteriori. From this he draws the conclusion that a functional explanation of language change is not possible. However, this conclusion is based on a definition of the object of investigation which is not shared by Keller (1990) or by the authors in this book. As shown by Keller (1990: 121 etc.), there is always a conscious purpose involved, as in any communicative activity, whereas change is its (usually) unintended cumulative effect. Change is a nonfunctional effect of functional activity, and the object of our investigation should in Keller's view comprise both the activity and its effect: a theory of language history is explanatorily adequate to the extent that it is able to correlate attested language data with types of language activity whose consequence they are. This is an adequate statement of the explanandum of historical linguistics, but as such only a first step towards a theory of language change. A theory of language change should be able to reconstruct and independently motivate each link in the chain from language activity via cognitive evaluation to instantiation and acceptance. The present volume aims at contributing towards such a theory. Its first part focuses explicitly or implicitly on the issues of function, evaluation, and relation between function and structure in change. Its second part contributes to our understanding of scenarios of change as part of a more general theory of activity. It also sheds light on a related area of diachronic pragmatics. Finally, the third part shows that change can be explained in relation to an independently established hierarchy in typology, especially if put in functional terms.

Introduction

3

Defining function in a principled way is a necessary prerequisite for any functional explanation. This is done by Keller, who proposes to distinguish function in the ideological sense of goal-directedness as found in language use (i.e., at the micro-level) from function in the logico-mathematical sense of mapping at the level of the system (i.e., the macro-level). The latter is a cumulative and often unintended result of the former and therefore possibly non-beneficial - a phenomenon itself explainable in terms of properties of the logico-mathematical mapping. The resulting language systems are phenomena "of the third kind", neither fully natural nor artificial, but spontaneous and unintended, as if brought about by an "invisible hand". The properties and impact of the "invisible hand" in change are further investigated by Garcia. This paper on a pronominal change in Spanish traces contextual evaluations and shows how a qualitative leap can be related to accumulation of quantity. There is an experimentally evidenced link between function in language use and the (retrieval or establishment of the) corresponding rule at the level of the system. This is shown by Eliasson's paper on the cognitive calculus, which leads to the conclusion that change does not proceed unmonitored. The essential role of function in change processes as accompanying, and enabling, transitions from one instantiation of the system into another, is pinpointed by Gvozdanovic on the basis of textual attestations of a change in progress. A plethora of possible developments starting from a common language ground shows that change need not be beneficial, provided only that its outcome complies with the general maxims of activity such as concerning clarity of expression. This is shown by the differentiation of switch-reference phenomena in the Awyu Papuan languages of New Guinea. The second group of papers discusses various aspects of "scenarios" of change, as a further extension of a theory of activity. The interactional notion of "scenario" is defined by Dressier, whose paper aims at showing that each of the stages of a "scenario" runs its own course in accordance with independently justifiable preferential developments. Diminutive changes in the eastern Dutch dialect of Twente discussed by van Bree show that relative stability of elements follows its own system-internal and sociolinguistic regularities, thereby substantiating Dressler's claims.

4

Introduction

Profound relevance of sociolinguistic evaluation of change is demonstrated by van Reenen's minute account of of the background of a major phonological isogloss in Dutch. A relatively new field of historical pragmatics is exemplified by Haverkate's investigation of the history of rhetorical questions in Spanish, against the background of a general pragmatic theory concerning the role and value of such questions. The third group of papers discusses independently motivated typological criteria as explanatory for language change. This concerns in the first place a hierarchy of functional differentiations which appears to be able to account for synchronic typological variations and explain diachronically attested developments, as demonstrated by Hengeveld and Wanders. In rather general terms, questions of language type and type consistency as underlying changes are discussed by Bauer in her analysis of Italic. What the various papers in the volume show is that change is not motivated by the internal set-up of the system. It is only enabled by the system, but its outcome need not be beneficial in any formal or functional way. Explanation of change requires reference to such independently motivated maxims of activity as clarity at minimum effort on the one hand, and independently motivated properties of language systems on the other. At the same time, it is clear by now that change is not a matter of flipping a coin at the level of the system, as had been assumed by early functionalism. System-internal factors such as asymmetrical implications (including hierarchies and rule orderings) on the one hand and naturalness priorities on the other, as well as pragmatic and sociolinguistic considerations, make change never fully gratuitous. Change begins and ends in language communication: it is brought about by functions in language use and their mapping at the systeminternal and system-external levels (including, i. a., sociolinguistic norms); it is sanctioned by the communicative activity where it has begun. If there is to be an explanation of language change, it has to be a functional explanation.

Introduction

5

References (other than to the papers in this volume) Anttila, Raimo 1977 Analogy. The Hague: Mouton. Andersen, Henning 1973 "Abductive and deductive change", Language 49: 765-793. 1989 "Understanding linguistic innovations", in: Leiv Egil Breivik and Ernst Häkon Jahr, eds., Language Change: Contributions to the Study of its Causes. Berlin: Mouton de Gruyter, 5-27. Harris, Alice C. and Lyle Campbell 1995 Historical Syntax in cross-linguistic perspective. Cambridge: Cambridge University Press. Hopper, Paul J. and Elizabeth Closs Traugott 1993 Grammaticalization. Cambridge: Cambridge University Press. Keller, Rudi 1990 Sprachwandel. Von der unsichtbaren Hand in der Sprache. Tübingen: Francke Verlag. Kiparsky, Paul 1992 "Analogy", in: International Encyclopedia of Linguistics I, 56-61. Oxford: Oxford University Press. Lass, Roger 1980 On Explaining Language Change. Cambridge: Cambridge University Press.

Functional and cognitive aspects of change

In what sense can explanations of language change be functional? Rudi Keller

1. Opening remarks The community of contemporary linguists is like Gaul divided into three parts. One group is principally not interested in language and thus not in language change. These are the representatives of the Chomskyan mentalism. "Language ... has no existence apart from its mental representation." (Chomsky 1968: 81) For those who adopt this Chomskyan ontology, language change is a term whose denotation is the empty class. Since the assumption of the non-existence of a language outside an individual brain would have disastrous consequences for the present discussion, we'll let it drop. The two remaining groups can be called functionalists and anti-functionalists. This classification is, of course, a bit simplistic, since one can belong to one group or the other due to widely varying reasons. Simon Dik can be regarded as a prototypical representative of the group of functionalists; Roger Lass, as a prototypical representative of the group of anti-functionalists. The views of these two groups stand in irreconcilable opposition. Their respective arguments seem unable to win over the other side. Such a situation is often an indication that unclear notions have crept into the dispute. I believe, in fact, that in the discussion "can or must the explanation of the state of a language or of language change be a functional one?" the expressions function and functional are unconsciously used with three different meanings. In the course of these pages, I will first briefly present the classical arguments for and against the functional viewpoint. Next, I will evaluate these arguments; finally, I will try to save the claim that an explanation of language change can be regarded as a form of functional explanation.

10

Rudi Keller

2. The classical arguments The basic problem was formulated with sufficient clarity by Friedrich August von Hayek in his little-known essay of 1950, entitled "Über den 'Sinn' sozialer Institutionen" where the word sense is in quotation marks to prevent misunderstanding: Denn der Gegenstand dieses Aufsatzes sind nicht die bewußt geschaffenen Einrichtungen der Gesellschaft, sondern vielmehr jene ohne Absicht entstandenen Bildungen wie Moral, Sitte, Sprache und der Markt, deren Entstehen und Funktionieren zu erklären der eigentliche Gegenstand aller sozialwissenschaftlichen Theorie ist. Wie aber, so mag der Leser mit einem gewissen Recht fragen, kann etwas, das nicht von menschlichem Verstand bewußt geschaffen wurde, einen Sinn haben?" (Hayek 1956: 512, emphasis original) [The object of this essay are not consciously created social institutions but rather those unconscious developments such as morals, customs, language and markets. The explanation of their emergence and functioning is the essential object of all theories of social science. But how, the reader may correctly ask, can something that has not been consciously created by human reason have a sense?]

It is, he continues, ... ähnlich wie in der Biologie, wo gewohnheitsmäßig von dem 'Zweck' oder der Funktion eines Organs u. dgl. gesprochen wird, obwohl der Biologe natürlich genau weiß, daß das im wörtlichen Sinn nicht richtig ist und der Gebrauch solcher Ausdrücke eine gewisse Gefahr in sich birgt. Und doch kann er diese 'teleological shorthand', wie es Julian Huxley einmal nannte, nicht entbehren. (Hayek 1956: 513) [...similar to biology, where the 'purpose' or the function of an organ is habitually spoken of, although the biologists know, of course, that this is not true in the literal sense and that the use of such expressions bears a certain danger. And yet he cannot do without this 'teleological shorthand', as Julian Huxley once called it.]

Does functional jargon in linguistics also have the status of a basically incorrect but indispensable "teleological shorthand", the status of a convenient metaphor? Functionalists would probably refuse such an interpretation. "It is not only a handy fagon de parier", a functionalist would say, "the language really does have a function. It is an instrument." Let's have a look at some of the programmatic propositions of Simon Dik.

In what sense can explanations of language change be functional?

11

- A natural language is an instrument of social interaction. That it is an instrument means that... it exists by virtue of being used for certain purposes. (Dik 1989: 4) - The primary aim of natural languages is the establishment of interhuman communication. (Dik 1986: 10) — The aims and purposes for which natural languages are used are commonly discussed under the heading of 'functions' of language. (Dik 1986: 11) — The main function of a natural language is the establishment of communication between NLUs [natural language users (R.K.)]. (Dik 1989: 4) — Any natural language can be considered as a particular solution to an extremely complex problem. (Dik 1989: 7) - In the functional paradigm ... a language is in the first place conceptualized as an instrument of social interaction among human beings ... Within this paradigm one attempts to reveal the instrumentality of language with respect to what people do and achieve with it in social interaction. (Dik 1989: 3) — Saying that a certain feature of linguistic design or change cannot be functionally explained is tantamount to saying that we have not yet been able to find a functional explanation for that feature. (Dik 1986: 11) Thus, the basic idea is that a language has an aim: solving the problems of establishing interhuman communication. The aim of a language, or that which a language is used for, is its function. A language is a problem-solving instrument. Functionalists regard language with its instrumentality in mind. There can be no explanation other than a functional explanation. At first glance, these claims are so convincing, that it is difficult to see where a mistake may have crept into the argument. Before we analyse it in more detail, let's take a look at the opposing view. Roger Lass formulates his objection to the promotion of functional explanations under the heading: "The teleology problem: can language change be 'functional'?" - The main one [i.e. objection (R.K.)] is that it is always irreducibly post hoc (in the sense of being totally non-predictive), and the functions invoked often seem rather fishy and devoid of principled support. (Lass 1980: 69)

12

Rudi Keller

- The 'theory of functional change' ...is apparently so constructed that almost anything can be a supporting example, and nothing can be a counter example. (Lass 1980: 71) According to Lass, the direction of a development is retrospectively reinterpreted as its goal: - This makes 'goal' an empty notion: a goal is simply something that stands at the end of a sequence that (apparently) leads to it... The invalidating fallacy is post hoc ergo propter hoc. (Lass 1980: 80) — Any teleological argument in which a goal is not identifiable independently of the behavioral manifestations leading to it is not explanatory. (Lass 1980: 83) — There are no functional explanations, because... we have no principled definition of'dysfunction'. (Lass 1980: 90) The objection, then, runs as follows: The end of a development is interpreted post festum as its goal. Function cannot be identified independently of its explanandum. So-called functional explanations are thus ad hoc, post hoc., and circular. Of course, both positions are here presented in shortened versions, but, I believe, in sufficient detail for purposes of discussion.

3. Who is right? The development of a language is not predictable. It has no goal. There is no state of a language from which a certain other state necessarily follows. All linguists seem to agree on this. On the other hand, it is true that language serves as an instrument of mutual influence. We use it in a functional way. Its change, and thus each of its states, is largely a consequence of this use. There also seems to be no disagreement about this. From the aimlessness of language change follows its unpredictability, as well as the fact that teleological explanations are out of place. From the instrumentality of language follows that its change has to be seen in the light of its functional use. So are the functionalists and the anti-functionalists both right? This cannot be. The solution to the riddle lies, in my opinion, in the fuzziness of both positions. Let's examine them again, keeping possible ambiguities in mind.

In what sense can explanations of language change be functional?

13

Even in his heading, Lass makes an equation that carries the weight of tradition and which he obviously supposes to be inquestionable: "The teleology problem: can language change be 'functional'?". Is teleology and functionality the same thing? First, let's have a look at the heading's question. If "Can language change be 'functional'?" should mean "Can language change be regarded in view of the functional use of language?", the answer must be "yes". If the question means,"Does language change have a function?", the answer must be "no". Only in the second case is the teleology problem involved. From the claim, that language change is a consequence of the purposeful use of a language, it does not follow that language change is purposeful in itself. Positively put, if we want to save the claim that language change must be functionally, and only functionally explained, we must develop a concept of functional explanation which is free of ideological pollution. The quotes from Simon Dik oscillate between two different positions, which correspond to the two readings of Lass' question. The first position is that humans use their language purposefully and functionally for certain aims. The second position is that language has an aim and a function. The first position is that speakers use their language as an instrument, while the second states that the language is an instrument. What or who has a goal? The speaker or the language? What has a function? The language or the use the speakers make of it? From the claim that speakers use their language purposefully and as an instrument does not follow that language has a purpose and is an instrument per se. From the fact that I can use the branch of a tree as a lever does not follow that branches are levers nor that I must regard them in view of their "leverishness". Of course, there is a significant difference between a branch used as an instrument and language used as a tool. Language came into existence as a consequence of the communicative endeavors of speakers, while the existence of a branch is independent of its possible instrumental use. It is, actually, as von Hayek writes in the quoted passage, similar to biology, when one speaks, for example, of the function of bodily organs. Let's consider the following two propositions. (i) The length of a giraffe's neck is a function of the fact that it lives from the leaves of trees. This proposition is completely acceptable and probably correct. (ii) The function of a giraffe's neck is to reach the leaves oftall trees. This proposition is, at the very least, misleading. It has an incorrect teleological reading; the giraffes' necks are not as long as they are in

14

Rudi Keller

order to reach tall trees. Rather, the individuals with longer necks had a higher average rate of reproduction and could pass this characteristic on to their offspring. I think it was Konrad Lorenz who said that evolution is not drawn from ahead but pushed from behind. Let's compare the following two linguistic propositions: (i') The development of New High German haben to Post High German ham is a function of the speakers' attempt to lower articulation costs. This claim is also acceptable and probably true (even if it is not the whole truth), while the following claim favors the unacceptable teleological reading. (ii') The development of haben to ham has the function of lowering the speakers' costs of articulation. Thus, the word function has at least two meanings. (We'll soon encounter a third one.) I will call them the teleological reading and the logico-mathematical reading. In its teleological reading, the word function means approximately the same as 'purpose' or 'goal'. A function in the logico-mathematical sense is a mapping. To say that the development of haben to ham is a function of the speakers' attempt to lower the costs of articulation, thus means the attempt to lower the cost of articulation is mapped in the change from haben to ham. To express it less technically, the speakers' attempt to lower articulation costs is one of the determining factors of the change from haben to ham. From this correct claim, it does not follow that the change from haben to ham has the aim or goal of making articulation easier for the speaker. As a first interim result, we can formulate two cautionary principles. 1. One should strictly differentiate between propositions about speakers and their way of using their language, and propositions about language. 2. One should strictly differentiate between the teleological meaning of the word function and its logico-mathematical meaning. The question "Who is right?" must, in my opinion, be answered in the following way. - The claim that speakers have goals is correct, while the claim that language has a goal is wrong.

In what sense can explanations of language change be functional?

15

- The claim that change is a function of use is correct, while the claim that change has a function is wrong. - The claim that speakers use their language functionally, that is, purposefully, is correct, while the claim that language has a function is questionable. Functionalists are right if they hold themselves to each of the first claims. Anti-functionalists are well advised to criticize the second half of each claim. The question whether a language can be ascribed a function tout court was discussed mainly by Martti Nyman. In the following discussion I will devote myself to this question.

4. Rules and tools In the book Sprachwandel I made, among others, three claims: first, that a so-called natural language is spontaneous order; second, that the adequate mode of explanation for its change is the so-called invisiblehand explanation; and third, that the invisible-hand explanation is a functional explanation. Spontaneous order is a macro structural system which comes into existence under certain framing conditions due to microstructural influences. Indeed, the influences are not aimed at the formation of the system at hand. This definition applies to any spontaneous order, from the dune structure in the West Sahara to the pricing system in a market economy or a so-called natural language. Spontaneous order in the realm of social culture differentiates itself from that in animate and inanimate nature, in that the structure-forming forces are intentional actions. The pricing system in a market economy, as well as the English of today, are nonintentional side effects of intentional actions. In other words, the English of today is a consequence of intentional communicative acts whose aim was not to produce the English of today. Because such phenomena are neither natural nor intentionally created artifacts, I called them echoing Steven Spielberg — phenomena of the third kind. Whoever accepts this claim must accept the second one as well, that the adequate mode of explanation for such phenomena is the invisiblehand explanation, for it is a logical consequence of the first claim. Both claims seem to be widely accepted by those linguists who are

16

Rudi Keller

acquainted with them. But difficulties arise with the third claim, that an invisible-hand explanation is a functional explanation. This claim is refuted with some extremely sophisticated arguments, which seek to prove that it is self-contradictory (cf. Pinto de Lima 1993 and Nyman 1994). I shall try to save my claim. The confusion arises because, in the course of an invisible-hand explanation, the word function can be used with three different meanings. What is an invisible-hand explanation? An invisible-hand explanation explains its explanandum, a spontaneous socio-cultural order, as the causal consequence of individual intentional acts which realize at least partly similar intentions. Simon Dik's example for a markedness shift can serve as an example for an invisible-hand explanation (cf. Dik 1989: 42). When the frequency of a polite expression increases, because of the intention to avoid impoliteness, it loses its markedness and thus its ability to express particular politeness. This is a well-known phenomenon of semantic inflation which occurs again and again in many languages. The individual speaker makes his choice from linguistic possibilities available to him under certain framing conditions. The aim to avoid impoliteness leads him or her to choose the somewhat more polite variant. When many people do this frequently, the result is that the expression under discussion loses its relative exceptionality and therefore its politeness, because politeness is a game in which exception is trump. This is a simple example of the structure of an invisible-hand explanation. It explaines the change of meaning of the English you or Dutch jij or German Ihr or Spanish vos as non-intended causal consequences of intentional communicative acts. We can use this example to discuss the question of in which sense such an explanation is functional. Let's suppose that the goal of explanation, that is, the explanandum, is the change of meaning (that is, the markedness shift). The change in meaning is not intended. It is not one of the goals of the communicating language users. It has no function. Since the change in meaning is a causal phenomenon, a functional explanation is inappropriate. This is the argument used by Pinto de Lima, for example. An invisible-hand explanation consists of two components: the micro-level explanation, which is actually an analysis of the speakers' rational choices in communicative behavior, and a macro-level explanation, which shows the unintended macro-structural consequences of the choices of the speakers. An invisible-hand explanation is, in Martti

In what sense can explanations of language change be functional?

17

Nymans words, a "combination of micro-finality and macro-causality". (Nyman 1994: 246) The speaker's choice is functional in the teleological sense. That his or her choice is functional means that the speaker makes a choice from the linguistic means available to him, from which he (subjectively) expects, that they will serve his communicative purposes under the given circumstances. A functional choice presupposes a goal and the intention to reach it. If the speaker is lucky, which is usually the case, he has made a choice which serves its purpose. A means that serves its purpose is an instrument. An instrument that is made to serve a certain purpose is a tool. Purposefulness is built into the concept of the tool. Are linguistic means tools? Much has been written on kinship between rules and tools. What both have in common is that they "provide standard solutions to recurrent problems", as Victor Vanberg put it. (Vanberg 1993: 23) This is exactly what language does as well, a fact to which Simon Dik refers in several places. (Dik 1986: 8f., 1989: 7) They also have in common the fact, that they are objects of evolutionary processes and bearers of cultural knowledge. "Tools come to incorporate the experience of generations of experimenters, without any of the persons using them being aware of, or being able to articulate, the knowledge that is embodied in the tool." (Vanberg 1993: 23) "Learning without insight" is what Victor Vanberg calls this form of so-called collective learning. He who has acquired a language participates, like the user of a trowel or plow, in knowledge that stems from experience that has been collected by previous generations. One need only consider the empirical and communicative experiences embodied and stored in the systems of concepts in natural languages. Society benefits not only from the division of labor, but also, thanks to language, from division of experience. But there is one difference between tools and linguistic rules which makes the equation of the two questionable. A tool, say, a certain type of plow, cannot (genotypically seen) lose its usability through use. The evolutionary development of a plow can, with respect to its purpose, only be progessively directed, whereas the evolutionary development of linguistic means is open to all possibilities. What is polite today can be insulting within a few years; what is colloquial or even vulgar today can soon be standard, and vice versa. That is, the functional deployment of a linguistic means by many can and often does have the consequence that this means loses or modifies its use for certain purposes. "Because [consequences arising from invisible-hand processes (R. K.)] are not results

18

Rudi Keller

of human design, IH-phenomena are only CONTINGENTLY functional," (Nyman 1994: 247) writes Martti Nyman. With this, he brings the third variation of meaning of the word functional into play. Here, functional evidently means something like "useful" or "beneficial". From this correct observation, Nyman draws what is in my opinion a completely unacceptable conclusion: "IH-phenomena become institutions and other cultural objects by an act of FUNCTIONALISATION... I... suggest that functional superimposition is a rational, finalistic act, by which a given phenomenon of the third kind becomes a social artifact. There is nothing uncanny here: Such a process is known as SEMIOSIS." (Nyman 1994: 247) Elsewhere, he says that "if the end result is found to be beneficial, it will get functionalized by a collective rational act consisting in an epidemic mimesis". (Nyman 1994: 254) This theory saves the proposition of the instrumentality and thus the functionality of language. But the price is very high. With the concept of a rational, finalistic act of semiosis, a rationalistic mysticism is brought into play, which creates more problems than it solves. Who, then, performs this rational act of semiosis, and how is it done? I don't want to go any further into this question and will instead return to the correct starting point of this theory. The results of the invisible-hand processes are only contingently functional. That means that some of them are non-functional. Dutch du als well als English thou, has completely lost its usability, and jij and you continue to be usable only for modified functions. Since language change is not predictable, the functionality of its outcome is also not predictable. One must always expect that the results of processes of language change are partially non-functional; that is, they can no longer, or hardly ever, be used as tools (cf. Dik 1986: 7). But the explanation of non-functional (in the sense of nonbeneficial) change must be functional (in the logico-mathematical sense) in the same way as explanations of functional, i.e., beneficial outcomes. This is the true core of the therapy metaphor, which is as widespread as it is misleading: On of its most pithy formulations runs something like "Grammars practice therapy rather than prophylaxis". (Lightfoot 1979: 149) Therapy is an intentional intervention with the goal of a cure. In regards to a language or its speakers, the concept of therapy is also a rationalistic mysticism. A language can neither think nor act. It does not recognize its weaknesses and therefore cannot perform therapy on them. Are the speakers therapists? Because the individual speaker

In what sense can explanations of language change be functional?

19

generally neither knows the past of his language, nor is interested in its future, he is not able to apply a therapy to the language, either.

5. Concluding remarks What really happens is the following: The -word functional is used in the Ideological, the logico-mathematical, and the beneficiency sense. 1. In a communicative situation, the speaker has certain communicative goals, communicative possibilities, and a certain appraisal of the framing conditions. In relation to these factors the speaker generally has more than one lingusitic means for the realization of his communicative goals. He chooses the means from which he expects the highest net benefits. That is, he makes a rational choice from the linguistic means available to him. This choice is functional in the teleological sense of the word. 2. A multitude of choices with unidirected aspects generates confirmation or modification of rules. This generative process, the so-called invisible-hand process, is causal. It it usually neither intended nor noticed by the speakers. The result of such a process, confirmation or change is a function of the speakers' choices, a function in the logicomathematical sense of the word. 3. The degree of functionality of the unintended results of the invisible-hand process is contingent. That means that they can be functional, less functional or even dysfunctional. Here, functionality means about the same as 'usefulness' or 'beneficiency'. 4. Some of these unintentionally generated rules or structures become a part of the individual competences of some, many or all speakers, and thus become available for the speakers' rational and functional choices. See no. 1. Forms which have lost their functionality in the sense of usability do not undergo a therapeutic treatment, but are simply ignored, because the speakers' choices are aimed at individual benefit, not linguistic health. Every explanation of language change that claims explanatory force must attempt to reconstruct this cyclical process. If the notion "functional explanation" is understood in this complex sense, I see nothing wrong with putting the invisible-hand explanation under this notion.1 Translated by Kimberley Duenwald.

20

Rudi Keller

References Chomsky, Noam 1968 Language and mind. New York: Harcourt Brace Jovanovich. Dik, Simon 1986 "On the notion 'functional explanation'", Working papers in Functional Grammar No. 11 and Belgian Journal of Linguistics 1: 11-52. Dik, Simon 1989 The Theory of Functional Grammar. Part I: The Structure of the Clause. Dordrecht, Holland/Providence, Rhode Island: Foris Publications. Hayek, Friedrich August von 1956 "Über den 'Sinn' sozialer Institutionen", Schweizer Monatshefte 36: 512-524. Keller, Rudi 1990 Sprachwandel. Von der unsichtbaren Hand in der Sprache. Tübingen: Francke. [Erweiterte 2. Auflage 1994.] [1994] [On Language Change. The Invisible Hand in Language. Transl. by Brigitte Nerlich. London-New York: Routledge.] Lass, Roger 1980 On Explaining Language Change. Cambridge: Cambridge University Press. Lightfoot, David W. 1979 Principles of Diachronie Syntax. Cambridge: Cambridge University Press. Nyman, Martti 1994 "Language Change and the 'Invisible Hand'", Diachronica XI. 2: 231-258. Pinto de Lima, Jose 1993 "Functionalism in Invisible-Hand Explanation of Language Change", Linguistic Agency University of Duisburg Series A, Paper No. 334. Duisburg. Vanberg, Victor 1993 Cultural Evolution, Collective Learning, and Constitutional Design. Center for Study of Public Choice. Georg Mason University. Fairfax, Virginia. [Unpublished MS.]

Grasping the invisible hand Erica C. Garcia Text signals its own structure 'For use almost can change the stamp of nature' (Hamlet Hl.iv)

1. Introduction The assumption that an invisible hand mechanism lies behind language change (Keller 1982: 12-14 et passim, 1990: 91-105) superbly straddles the synchrony/diachrony divide, for by appealing to the synchronic motivation of individual speech acts it allows functional accounts of change to dispense with teleology, i.e., a specifically diachronic motivation. Furthermore, the coupling of individual purposefulness with mechanical cumulation accounts naturally for the unidirectionality of change (Liidtke 1980: 13). In fact, the essence of an invisible-hand type explanation is to be found in the reciprocal causation it posits between quantitative and qualitative phenomena, so that, as in a Möbius ring, langue and parole give rise to each other in a never-ending cycle. This view of linguistic change constitutes a significant departure from the stance originally taken by Weinreich-Labov-Herzog (1968: 186 et passim), later endorsed in its essentials by Timberlake (1977: 141, 142, 168 et passim) and Kroch (1989b: 201-202), and now reiterated in Kale's recent claim (1994: 150) that "two quite distinct diachronic events [are involved]: change, to be attributed to reanalysis on the part of the grammar constructor, and diffusion, driven by sociolinguistic factors of the well-known type". How many events are in fact involved in linguistic change is a far from trivial issue. Should reanalysis and actualization in fact be distinct and independent, grammar construction need not in any way be constrained by the quantitative properties of actual language use. Contrariwise, should the two not be independent, the autonomy of syntax vis ä vis pragmatics is immediately called in question, since the latter is clearly relevant to the frequency characteristics of language use. In settling this issue it would be useful to know exactly how many fingers the invisible hand has in the diachronic pie: but this is just what

22

Erica C. Garcia

we do not yet know. Keller's discussion of how language is invisibly manipulated is suggestive rather than exhaustive, and his examples are largely confined to lexical change. It is not clear whether the proposed account extends to sound-change, nor, in particular, how it handles grammatical change. If the invisible hand is to be seen as the central mechanism in linguistic change, it is necessary to show the invisible hand in action. Not least because, though lip service is often paid to the importance of frequency of use, the causes of change, and in particular of reinterpretation, are often sought elsewhere, for instance in unspecified general tendencies (cf. Kroch 1989 a: 153-156 et passim). The purpose of this paper, then, is to identify the mechanics of syntactic change in the nature of synchronic use, particularly in its quantitative properties. We will support our general account of the process with actual data from a change still in course in Spanish, namely the loss of the so-called reflexive prepositional pronoun si (a reflex of Lat. sibi), an invariable form currently recessive (at least in the spoken language, Fernandez (1951: 222 #116), Ramsey/Spaulding (1956: 79#3.30)) as a result of competition with the so-called third person pronouns el, ella, ellos, ellas, ello, reflexes of Lat. ille1.

2. Reinterpretation and the invisible hand In our opinion linguistic change provides an even better example of invisible hand operation than such classical examples as the formation of a path across a lawn, or the development of a traffic jam. For these latter cases simply involve reiteration of the critical event (namely, crossing the lawn on a particular spot, stepping on the brake), which, while it surely becomes more and more probable the more it happens, nonetheless remains, in a significant sense, the same event. This, however, is not the case with language. Here too using X rather than Υ increases the probability that X rather than Υ will once more be used. And, once again, it is the new conditions resulting from the use of X (rather than that event by itself) that increase X's chances of being used as against Y's. But here the parallel with the deerpath or the traffic-jam crucially breaks down: for in the case of linguistic change what makes the choice of X or Υ more or less likely is not something external to that choice but a change in the value of X or Υ themselves.

Grasping the invisible hand

23

In the case of language the change in the conditions for the use of X or Υ is not distinct from the very value of X or Y. Successive uses of the "same" form do not, then, constitute the "same event": for the form that is (re)used after the first use may not be assumed still to be what it was before. The traditional view that a form is reshaped through use is thus essentially correct: not only can the phonological form of a word "change under constant wear" but, what is much more crucial, its semantic value will vary depending on how often and where it is used and — especially — on how often it is used where. If we are to see the invisible hand in action, then, we must abandon the standard view that reanalysis and diffusion are distinct, and try instead to understand how the cumulative use observable in the community relates to the reinterpretation that takes place in speakers' and especially in language learners' minds. Thus, if Keller is right, and the proper approach to the study of language is a dynamic one, what is needed in order to account for change should be present in synchronic use - particularly in variation.

3. The nature of "variation" Practically all language historians are in agreement with Bailey (1973: 82 et passim), Timberlake (1977: 141) and Stein (1990: 198-203 et passim) that an innovation spreads along a contextual continuum. The diffusion is characterized by the fact that the innovation (henceforth I) is more frequent, and observed earlier, in certain contexts, from which it invades others. This process is observed in change after change: its synchronic reflex is the unequal distribution of the I across different contexts. It is evident that unless insight is gained into that asymmetrical distribution of forms across contexts, any attempt to account for syntactic change will be in vain. Unless a substantive motivation is provided for the continuum of synchronic variation it will never be clear why, diachronically, some environments are "unmarked for the change", i.e., favorable to the innovative alternative (Timberlake 1977: 157). Nonetheless, linguistic history is generally attempted without any clear idea as to what underlies the synchronic choice of one variant over another, so that the crucial characterization of the continuum is either avoided or glossed over (cf. Timberlake 1977: 153, Kroch 1989a: 156, Stein 1990: 200).

24

Erica C. Garcia

Given its key role in any process of replacement, synchronic variation cannot be assumed merely to amount to "different ways of saying the same thing" (Labov 1970: 34 fh. 7, 1972: 271). Such a view may be admissible in the case of diacritic units (phonemes, or allomorphs) which, carrying no specific informative load of their own, may be preempted by extra-linguistic factors and thus acquire the value of social markers (Garcia 1985b: 198-199). But it is not justified in the case of meaningful units, which are resorted to for the sake of their specific and distinct communicative values. Syntactic and lexical variation involves saying different things about the same referent, since the alternating variants present the "same" referent from different perspectives, and under a different light (MacLaury 1991: 40 fh. 3, 42 et passim). This view allows us to understand why variation is neither pervasive nor uniform. For, clearly, certain contexts practically require a given variant, others just as clearly reject it, so that the two alternatives vary only in some contexts, and do so in different degree. The reason is, simply, that different contexts, too, embody diverse perspectives, and do so with varying precision. (Almost) categorical contexts can not only be assumed, but in fact shown, so manifestly to embody one given perspective (Joos 1972: 263), that the speaker has no choice: to resort to the other alternative would amount to contradiction. Contexts of variation, on the contrary, can profitably be viewed as neutral, i.e., as suggesting no particular perspective. Here the contextual support is minimal, with little risk of incongruence: whatever alternative is chosen will be tolerable in the neutral context. Furthermore, since it is not contradicted by other elements, the chosen alternative will spell out the perspective for the utterance as a whole2. In the categorical environments the context so literally overshadows the form, that a given alternative becomes more or less inevitable given what has already been said. In the neutral environments - in contexts of variation - it is the variant chosen that enlightens us as to the speaker's perspective. But hereby any aspect of the context congruent with that perspective is effectively highlighted. A variable context can be visualized, then, as a Necker cube, in itself neutral to either perspective, but capable of highlighting from inside. The viewer's - i.e., speaker's - perspective is projected not onto, but from within the cube, since the critical clue - the chosen alternative - appears in context. Contexts of variation will accordingly be heard and read differently, depending on which alternative has been chosen. Variable contexts are

Grasping the invisible hand

25

thus not variable merely because they allow different choices: they are variable in the much deeper sense that they themselves vary, becoming different contexts, depending on the alternative present3. It is this variability in the structure of context that underlies reinterpretation, and allows variable contexts to be the natural locus of ongoing change.

4. The reinterpretation of contexts We have argued that the choice of variant projects its own perspective on a neutral context, whereby one aspect of the Necker cube is highlighted and hence becomes more salient than the other. In other words, the cube (i.e., the context) comes to be seen as having one shape rather than the other. Specifically, we assume that language learners are more likely to recognize any X-favoring traits the context may contain when exposed to the variable context-cum-X than when confronted with variable context-cum-Y. "X favoring traits", in short, can be read not only out of, but also into, the context. But the more (often) the (neutral) context is seen to be X-favouring, the greater X's chances of being (re-)used in it. For the basic principle underlying all syntax — namely contextual support - favors the choice of the more congruent alternative: and a context that has been perceived as X-favoring is thus more likely to see X rather than Υ used in it. But this will in turn highlight X-favoring traits in the context, and so forth. In this way a variable context can cease to vary, and gradually become categorical for one of the alternatives. Qualitative reinterpretation of a context can thus be plausibly related to quantitative changes in the recourse to one variant or another within that context. Critical to the operation of the invisible hand, then, is the possibility - and necessity - to (re)interpret contexts in the light of the forms which they exhibit. But the same is true, of course, for the forms themselves: for a language learner can abduce the value of a form only by taking into account the contexts where it occurs (cf. Garcia 1985 a: 283 et passim). A form which is observed to occur in a great variety of environments must be assumed to be congruent with more (different and potentially incompatible) contextual presupposals than one that is limited to a single specific environment. The extension of a form to different con-

26

Erica C. Garcia

texts will thus inevitably result in so-called bleaching, i.e., gradual de-specification of its value. The reinterpretation of context and form just discussed is not confined, however, to the most neutral of variable contexts, or to an I which (for whatever reason) becomes more frequent than it was: for with the reinterpretation of a variable context as categorical for I, a context that previously was categorical for the Recessive variant (henceforth R) will be reinterpreted as variable with I 4 . The reason is clear: if I varies with R in a given context, R necessarily varies with I in that "same" (neutral) context. But since R projects one and the same perspective in the variable and in (its own) categorical contexts, I can become an alternative for R in any context that does not too blatantly contradict the perspective suggested by I. Now it is clear that an increase in frequency of I, even within the most neutral of environments, will inevitably bring the form in contact with a greater variety of non-immediate contexts. Diffusion to a variety of (larger) contexts will naturally result in a less specific value being abduced for I than formerly. Any expansion of I will consequently have the effect that some contexts, which previously only tolerated R, might, at a pinch, now admit I 5 . And in those environments I can now begin to vary with R, since thanks to its partial bleaching, it is now perceived as more neutral6. I's expansion thus bleaches I, and this allows the reinterpretation of formerly categorical R contexts as newly neutral contexts. But as soon as I begins to alternate with R in that newly variable context (which now takes on the role of neutral Necker cube) we are both back where we started and one step further. In this way the reinterpretation of form and context feed each other, and a change in frequency in a variable environment is likely to result in double reinterpretation of form and context. There are, then, not two "distinct events" in linguistic change, but two cognitive forces invisibly steering the evolution of language. One is the language learner's search for the greatest generalization; the other the language user's dependence on contextual congruence: it is possible that the two may prove interdependent. That indeterdependence may perhaps be established deductively, on the basis of psychological insight into cognitive processes and properties of the mind generally. But it can in any case also be approached inductively, on the basis of empirical analysis. As linguists it behooves

Grasping the invisible hand

27

us to ascertain, at least, whether the various stages posited above are indeed observed in specific cases of linguistic change.

5. A case-study: the replacement of si by el The loss of si to be discussed here involves two interconnected processes of replacement: for historically, si Oneself yields not only to el, ella, etc., 'he, she, etc.' but also to the syntagm si + mismo, misma, etc., 'same', a reflex of Lat. metipsissimus (Fernandez 1951: 222#116, Schmidely 1983: 178). As might be suspected, the two processes interact diachronically, in truly "invisible-hand" fashion. The three relevant alternatives (si, si mismo, and el) stand in a double relation represented graphically in Fig. 1. Si mismo resembles el more than its base pronoun in that, like el, it morphologically specifies the gender and number of the referent. But from a semantic/referential point of view, si mismo is a sort of double si, since the lexical addition echoes the base pronoun reference to the self-same referent — namely, to that referent of whom "self-same" can most truly be said. Since si mismo so obviously appears to be a stronger version of si, a plausible hypothesis suggests itself: namely, that competition of si with el resulted in reinforcement of the former by means of mismo. This account might invoke Lüdtke's remarks on syntagmatic reinforcement under phonological attrition (1980: 15). The two replacements could Morphological Neutral

Specific for Gender & Number si mismo, a

si

el, ella, etc. Figure 1. Relation of si, si mismo and el

Semantic Maximum

si mismo, a

s a m e n e s s Minimum

si

el, ella, etc.

28

Erica C. Garcia

accordingly be related as a push-chain reaction between two invisiblehand processes. If this is how el and si mismo between them dislodged si, the alternation between si and el must have antedated the appearance of si mismo. Furthermore, si mismo would have to arise in those contexts where si alternated with el, and where reinforcement was hence most called for. But this is not very likely, since (as we shall see) the prototypical si mismo context precisely excludes el. It is advisable, then, to stop speculating and turn to the texts in order to date both developments, i.e., the addition of mismo, and the alternation with el. The competition between si and el or si mismo can be gauged by confining ourselves to the relative frequency of the different alternatives overall; alternatively, we can examine that relative frequency in specific environments. Since the relative frequency of si (with or without mismo) is at all periods extremely low in comparison to the massive recourse to the reflexes of L. ille (for Modern Spanish cf. Schmidely (1983: 178)), the overall figures are not likely to be very revealing. We must therefore examine the relative proportion of the relevant alternatives in specific environments. A selection of these, however, is impossible unless we have some idea of the difference in value between the two forms and of the distribution that can be expected. Elsewhere (Garcia 1991 a: 39-44) we have tried to specify that difference for the modern language; there is no reason to suppose that it was essentially different in earlier stages. We shall accordingly first present our analysis of si vs. el in Modern Spanish, and then illustrate it with a "minimal" pair from an Old Spanish text. In its failure to express gender and number si resembles personal pronouns, which express plurality through the root form itself. Si further resembles personal pronouns — particularly the core 1st and 2nd p. sg. — in that it fits the general pattern of differentiation for prepositional, verbal clitic, and possessive form, as can be seen in Figure 2. Finally, and for our purposes the most important trait, si resembles the prepositional 1st p. sg. and 2nd p. sg. mi and ti in that its referent is context-given. The relevant context, in the case of the speech-participants, is the speech-situation: in the case of the 3rd person, the content of the discourse. El, on the other hand, patterns like the demonstratives ese, este, aquel in showing different endings for feminine and plural, which further coincide for the four sets of forms. The specificity of these forms —

Grasping the invisible hand Strong

Person 1st 2nd 3rd Personal Deictic 1st pi. 2nd pi. 3rd pi.

Subject yo tu -

Prepositional mi ti si el, ella nosotros vosotros ellos, ellas

29

Weak Vb. Clitic Possessive me mi te tu se su le, la, lo nuestro nos OS vuestro les, los, las

Figure 2. The Modern Spanish pronoun system

including el - with regard to the "referential" categories of gender and number makes them eminently appropriate when the intended referent must be singled out among other comparable entities - i.e. when the entity is not uniquely self-evident, since given by the context. Our analysis, then, is that si is a personal pronoun, while el, in comparison, is more/still a deictic. Though both forms can be used in reference to the same entity, and even in - by and large — the "same" context, they convey different categorizations of the referent. A particularly clear - and elegant - illustration of the difference is provided by the following Old Spanish examples: PLA 23b46 Despues que el rey llego aquel sabado a Burgos, ouo su conseio e dixeronle algunos que Garci Laso tenia muchas conpannas consigo e ponian grandes escandalos en la su corte e en el su regno. After that the king arrived that Saturday to Burgos, had his counsel and said-to-him some that Garci Laso had many companies with self and put-3pl. great scandals in the his court and in the his kingdom. 'After the king arrived at Burgos that Saturday, he called together his council, and some of them told him that Garci Laso had many troops with him and that they caused grave disturbances in his court and in his kingdom' PLA 86b34 E partio el rey de Castilla de Tara?ona con todas sus conpannas que alii tenia con el, e fue a Borja, una villa del rey de Aragon.

30

Erica C. Garcia And departed the king of Castille from Tara^ona with all his companies that there (he) had with him, and went to Borja, a village of the king of Aragon. 'And the King of Castille left Taragona with all the troops he had there with him, and went to Borja, a village of the King of Aragon.'

At first blush, the two contexts appear to be identical, and one fears free variation. But in the case of PLA 86b34 the larger context makes clear that the King had been met, at Taragona, by different noblemen who came accompanied by their respective suites of followers, all of which were put at the King's disposal. The troops the King had 'with him' in PLA 86b34, then, were not, strictly speaking, his own troops. Nor could it hence be said that he had them consigo 'with him': for each set of followers answered, in principle, and in the first place, to its respective liege lord. Between the troops and the King there stood, then, a set of intermediary figures, and it is against this (linguistically implicit, but politically critical) background that the King is pointed out with el1. We accordingly expect that there should be a difference with respect to the contexts where si and el are more likely to occur: the question, now, is to identify those environments, and to trace the differential competition between si and its rivals.

6. The diachronic data Given the different value of si and el, three different types of environment are of interest: i) The intended referent is the only one possible in the context, being immediately accessible in the item with which the prepositional phrase (henceforth PP) is associated. This is the prototypical context for si, where in principle only this form is to be expected. This type of context will be labelled Immediate; it is illustrated by the examples from PLA quoted above. ii) The intended referent is one among many, not at all given in the context. It must hence be sought for in the crowd: this is the ideal el context, and here, in principle, we expect no si at all. The form of course does occur in this type of context; this distribution has recently been characterized as the "logophoric" use of an anaphor (Reinhart - Reuland 1991: 283, 311-317). This type of context will be labelled Remote: examples (of si, the minority form) follow:

Grasping the invisible hand

PCG 628a39 caualgo el Cid, et con el toda su companna muy noblemente vestidos, sus cauallos ante si et sus escuderos que leuavan las armas. rode the Cid, and with him all his company very nobly dressed, his horses before self and his squires who carried the weapons. 'the Cid rode, and with him all his company very nobly dressed, his horses before himself, and his squires who carried the weapons.'

Corvacho 166/17 Empero, la su pare9iencia e saber es en dos maneras: la una quanto al saber que es gerca de sy, e esto es yncomudible; la otra es quanto en esguarda de la criatura. However, the his foresight and knowledge is in two manners: the one what to-the knowledge that is about of self, and this is unchangeable; the other is what in regard to the creature 'However, His foresight and knowledge is in two guises: the one with regard to the knowledge which is about Himself, and this is unchangeable; the other is that which concerns the creature.'

Quij. 1084/12 y aunque don Gregorio cuando le sacaron de Argel fue con häbitos de mujer, en el barco los troco por los de un cautivo que salio consigo. and though don Gregorio when him took out-3pl. of Algiers went with dress of woman, in the ship them changed for the of a captive which went out with self. 'and though Don Gregorio, when they took him out of Algiers, wore the garments of a woman, in the ship he exchanged them for those of a captive who left with him(selfy

Masas 256 Porque asi como la mujer no puede en ningun caso ser definida sin referirla al varon, tiene este el privilegio de que la mayor y mejor porcion de si mismo es independiente por completo de que la mujer exista no. Because thus as the woman not can in no case be defined without refer her to the male, has this the privilege ofthat the greater and better portion of self same(masc.) is independent by complete of that the woman exists or not.

31

32

Erica C. Garcia 'Because just as woman can in no way be defined without reference to the male, the latter has the privilege that the greater and better portion of himself is entirely independent of whether woman exists or not.'

Cortäz. Oct. 123 Entonces ni siquiera asi, ni siquiera en el amor se abolia ese espejo hacia aträs, el viejo retrato de si mismo joven que Lina le ponia delante acariciandolo. Then not even thus, not even in the love itself abolished that mirror towards back, the old portrait of self same(masc.) young that Lina him put before caressing him. 'Not even in this way, then, not even in love was that backwards mirror abolished, the old portrait of himself young that Lina put before him (while) caressing him.'

iii) Though the intended referent is given in the context he is (indirectly) accessible only via a different referent. This type bridges the properties of the other two; it is labelled Mediate*. Examples follow: a. PP grouped with the Infinitive in an Accusative cum Infinitive (ACI) construction; the pronoun in the PP refers to the subject of the Main verb. Quij. 835/18 la cual, con el gusto que tenia de oirle, le hizo sentar junto silla baja.

si en una

the which, with the pleasure that had of hear-him, him made sit next to self in a low chair 'who, with the pleasure she had in hearing him, had him sit next to herself on a low chair'

b. PP grouped with a non finite verb whose (inferred) subject is not that of the main verb; the pronoun in the PP refers to the subject of the Main verb. Quij. 417/3 Dorotea la tomo por la mano y la llevo a sentarywwto se quitase el embozo.

si, y le rogo que

Grasping the invisible hand

33

Dorotea her took by the hand and her took to sit next to self, and her asked that self should remove the shawl. 'Dorothea took her by the hand and led her to sit next to herself, and asked her to remove her shawl'

c. PP grouped with an Adjective or other Modifier; the entity to which the Adjective applies is not the subject of the verb; the pronoun in the PP refers to the subject of the Main verb. PCG505al8 Et envio [Almenon] por el rey don Alfonso, et demandole quel fiziesse seguro de si et quel yurasse que mientre que el visquiesse que nin fuesse contra el nin contra sus fijos. And sent [Almenon] for the king don Alfonso, and asked him that him should make sure of self ana that to him should swear that while that he lived that nor should go against him nor against his sons 'And Almenon ι called for the king don AlfonsOj, and asked him that hej should give hinii security [lit. make him ( safe of himself^] and that he; should swear to hinij that while hej lived hej would not go against hinij nor against his ι sons'

From the very earliest texts both si and el are observed in all three types of context: change thus manifests itself only in the relative frequency of el. In the Remote contexts si is always very rare, though it actually never disappears (cf. Garcia 1983: 192-199). These contexts are thus of relatively little diachronic interest. The Mediate contexts are more promising, for they provide a bridging Necker cube, liable to be seen from different angles. For while the referent is readily accessible, as it is in the Immediate contexts, it is so only through a rival entity: in the presence of this rival the Mediate contexts resemble Remote ones, where the relevant referent must be pointed out among many other irrelevant ones. It is here, then, that the alternation between si and el must have begun and been strongest. For the same reason, mismo can be expected to have been disfavored in Mediate contexts, since the rival introduced by the grouping of the PP makes an assertion of sameness less plausible than in the Immediate contexts, which give (direct) access exclusively to the intended referent.

34

Erica C. Garcia

Table 1. Total si, si mismo, el and el mismo in Immediate vs. Mediate contexts Text

PCG ECL Corv Exem Quij Jov Reg Masas Cortaz

Immediate

Mediate

si mo

si

7 11 14 26 35 40 81 99 135

129 71 107 106 207 57 64 49 49

el

el mo

si mo

4 1

10 5 2 4 6 1

1 2 1 1 71

si

1 2 27

el

el mo

2 2

2 1 4 9 20

4

Table 2. Overall percentage of el, and percentage of mismo in si uses Text

Immediate

Mediate % mismo

PCG ECL Corv Exem Quij Jov Reg Masas Cortaz

3 1 0 1 1 0 1 2 35

5 13 11 20 14 41 56 67 73

% mismo 17 28 0 0 25 50 100 100 100

0 0 0 0 0 0 -

We have accordingly analyzed a sample of representative Spanish texts, dating from the 13th to the 20th century9, and have examined for each the occurrence of si (mismo) and el for Immediate10 and Mediate contexts''. Table 1 presents the data, which are reduced to percentages in Table 2 12 . The data of Tables 1 and 2 are just what is synchronically to be expected: the percentage of el is regularly higher in Mediate contexts, while the percentage of mismo is regularly higher in the Immediate

Grasping the invisible hand

35

ones13. Though the diachronic interpretation of the data is not obvious, three things are clear: i) variation between si and el antedates generalized addition of mismo, ii) el was I in Mediate contexts; (si) mismo in Immediate environments. iii) as I, si mismo begins earlier but takes longer, the replacement by el starts later, but goes faster. This suggests that though the two changes are somehow connected, the push-chain hypothesis must be rejected as clearly untenable.

7. Reanalysis of Mediate contexts As already suggested, the hybrid character of Mediate contexts allows a double perspective on the referent: as given within the local context (which justifies recourse to si), or as distinct from the entity with which the PP is directly grouped (thus enforcing recourse to el). Now it is important to note that Mediate contexts are, in and of themselves, far less frequent than either Immediate or Remote environments (the latter comprising the overwhelming majority of pronominal use). The relative rarity with which speakers have occasion to refer to a locally prominent, but not directly accessible, referent does not favor the development of a distinct referential strategy: on the contrary, it may facilitate the identification of Mediate contexts with either Immediate, or Remote contexts, with both of which they have something in common. Recourse to the deictic el sets up the pattern of variation summarized in Figure 314. The two I's are naturally distinct, since they present the referent from opposite perspectives. This distinctness gives each an edge on their common rival, si, and at the same time induces differentiation of Immediate and Mediate contexts, by highlighting the (respectively opposite) properties of the two sets of contexts. For the more si mismo is used in Immediate contexts, the less these will be perceived to resemble the Mediate contexts, just as the more el is resorted to in Mediate contexts, the more distinct these will appear from Immediate contexts.

36

Erica C. Garcia

Context

Variants Immediate, - human DO Immediate, + human DO

non Remote

si mismo vs. si si

Mediate

si

Remote

el

vs. el

non Immediate

Figure 3. Variation in Immediate/Mediate contexts

Mutual differentiation of the contexts, furthermore, is self-reinforcing: for it favors (even) greater recourse to si mismo in Immediate contexts, and to el in Mediate environments. The inevitable result is that si, which is neither fish nor fowl, is displaced in both sets of environments. Not surprisingly, the change goes faster in the (relatively rare) Mediate contexts, which stand at the greatest quantitative disadvantage with respect to both Immediate and Remote environments. Our hypothesis, then, is that el did not force the reinforcement of si: it was rather the rise in the frequency of si mismo that brought about invisibly — the polarization of the two domains, which increasingly adapted to the respective favoured variant. Mediate contexts, which in Old Spanish, as in Latin, seem to have been non-distinct from Immediate environments, appear to have switched allegiance from the mid-17th century onwards, coming to be seen as essentially Remote. The quantitative effect on the use of si was minor, since as already pointed out, the Mediate contexts are very infrequent. The qualitative consequences, however, were of the greatest importance, for 57 lost considerable ground in terms of privilege of occurrence, coming to be practically confined to reference within the clause.

8. Reinterpretation of si mismo The only difficulty with our account is its failure to be fully supported by the data of Table 2. The course of events just suggested presupposes

Grasping the invisible hand

37

a high frequency of si mismo in Immediate contexts: but this I exceeds 50% only at the beginning of the 20th century. Could an even lower percentage of I in the 17th century have brought about the hypothesized polarization of contexts? What we would like to find in the texts is really high use of si mismo in Immediate contexts: high enough to allow reinterpretation of the originally reinforced syntagm as a unit, paradigmatically equivalent, yet semantically stronger, than the single base pronoun si. The modern language in fact provides evidence for such a reinterpretation: the disappearance of the syllable -go in the conjunctive form consigo (Garcia 1988b: 201-205; 1989b: 113-116), particularly frequent in the presence of mismo, suggests that the two - si and mismo - have coalesced as a unit15. The question, then, is whether there is some environment that i) shows high frequency of mismo in combination with si; ii) can plausibly be argued to be prototypically Immediate; iii) is reasonably frequent and, furthermore, iv) is syntactically comparable to the Mediate contexts, so as to justify the reciprocal differentiation. The only referent with any (serious) chance of rating si in Mediate contexts is the Subject of the main verb; we must therefore concentrate on the Immediate contexts where the PP is grouped with a verb, the intended referent likewise being the subject of that verb. We furthermore exclude from consideration those cases where reference is made, before the relevant PP containing the critical reference to the subject, to another entity involved in a comparable prepositional relation. The cases that concern us, then, can be characterized as instances of "first mention". In the (very infrequent) "second mention" contexts pragmatic considerations affect the choice of form. Once again it will be necessary to distinguish different sub-contexts according to their relative congruence with (si) mismo; the four environments listed below are ranked in order of (decreasing) contextual salience of the subject of the verb: i) ii) iii) iv)

lacking a human Direct Object: duplication of se ditto: after a ditto: other prepositions with a specific human Direct Object

38

Erica C. Garcia

Examples follow: i) No human DO; clitic duplication Quij. 995/7 y se prometio a si mismo de hacer maravillas en el caso and himself promised to self same(masc.) of do marvels in the case 'and he promised himself to perform marvels in the case'

ii) No human DO; after a (not in clitic duplication) Quij. 955/17 contra el buen decoro que las doncellas principales deben guardar a si mesmas. against the good decorousness that the damsels principal(pl.) must keep to self same (fern. pi.). 'against the proper decorousness which noble damsels must keep with regard to themselves''

iii) No human DO; after prepositions other than a Quij.243/-7 y quedo con mäs deseo de saber quien era el desdichado loco, y propuso en si lo mesmo que ya tenia pensado and (he) remained with more wish of know who was the unhappy madman, and determined in self the same that already had thought 'and he remained with greater desire to know who the unhappy madman was, and determined within himself that which he had already thought'

iv) with a human DO Quij. 57/6 El labrador, que vio sobre si aquella figura llena de armas blandiendo la lanza... the laborer, who saw above self that figure full of arms brandishing the lance... 'the laborer, who saw above him that fully armed figure brandishing a lance...'

Grasping the invisible hand

39

In context i) the PP takes the form a + Pro, and duplicates the clitic se which, in turn, serves to double-mention (Garcia 1975: 115 —165 et passim) the subject, thus emphasizing its exclusive relevance to the event. At the other extreme, in context iv), the event involves, besides the subject, a specific human Direct Object. Both in terms of inherent salience, and because of its central involvement in the event, a human Direct Object can be expected to detract attention from the intended referent (the subject), and thus to diminish the claim of the latter to unique contextual givenness. The remaining two contexts differ in that ii) involves the preposition a, while iii) includes all other prepositions. The difference between a and other prepositions is critical: only a can introduce (at all stages of the language) a direct participant in the event (Garcia 1975: 92—95 et passim). Association with the verb via a is, then, more direct than via another preposition; it follows that the subject is more directly accessible in the former than in the latter case. In Table 3 we indicate the percentage of cases of si that appear with mismo in each environment16. The pattern is unmistakable: from the beginning the addition of mismo is favoured by a single context, the preposition a 'to', grouped with the verb. Around the turn of the 16th century, mismo appears to become categorical with si after a and, particularly, in clitic duplication.

Table 3. Percentage of mismo with si in Immediate (1° mention) contexts (PP grouped with the verb, reference to the subject) Text

Clitic dupl.

PCG ECL Cor Exem Quij Jov Reg Masas Cortaz

N 2 0 1 0 12 3 44 39 30

A

% mo 0 — 0 —

83 100 100 100 100

a N 1 8 8 14 3 64 0 3 1

Other preps.

Human DO

N 70 58 79 95 186 7 62 82 88

N 54 9 11 11 32 0 14 1 4

»

% mo 100 25 12 86 100 41 100 100

&

% mo 3 5 1 8 9 0 22 50 75

% mo 0 0 0 0 6 — 0 0 25

40

Erica C. Garcia

This, then, is the prototypically Immediate environment where si mismo is the prototypical choice. And in this context mismo becomes categorical in the course of the 17th century17 —just in time to trigger the mutual differentiation of Mediate and Immediate contexts. Our claim, then, is that the reinterpretation of si mismo as a unit allowed the two alternatives — single vs. reinforced si — to compete along the continuum of contexts illustrated in Table 3, with I spreading from its original core territory to other environments via the usual process whereby stronger variants displace weaker ones (Garcia 1991b: 31, 46—49). Table 3 indeed shows that by the 20th century si mismo has become the majority choice in all Immediate contexts, save in the presence of a human Direct Object.

9. Understanding the past: predicting the future? There seem then to have been three critical (turning) points in the two replacements of si: i) ca. 1500: mismo becomes obligatory after si in the environment la . This environment critically includes clitic duplication, the prototypical Immediate context; it motivates polarized recourse to el vs. si (mismo) in Mediate and Immediate reference; ii) ca. 1800: el becomes the majority option in Mediate environments; this confines si (mismo) to Immediate contexts; iii) ca. 1900: si mismo becomes the majority option in Immediate grouping with the verb, reference to the subject, prepositions other than a; el becomes obligatory in Mediate environments. This achieves the polarization begun in the 17th-19th centuriea. Now the last development - the diffusion of si mismo in Immediate contexts — had a critical invisible-hand by-effect, namely a considerable blurring of the distinction between the competing alternatives, now si (mismo) and el. For thanks to the regular recourse to mismo in conjunction with si, the gender and number of a context-given referent came to be specified in the great majority of references. Lexical emphasis on the self-evidence of the referent by means of mismo in fact neutralizes the

Grasping the invisible hand

41

morphological difference between the two variants, as is clear from Fig. 1. If the difference is indeed so small, and number and gender of the referent are practically always specified, why bother with si at all? Might we not expect el, which - starting out from massive use in Remote contexts - has taken over Mediate environments, to displace 51 in Immediate reference as well? The answer is: yes, and the beginnings of the change are already to be discerned in our data. Time and space limitations make it impossible to give here a full account of the development, but two points can be anticipated: i) the environment that takes over the role of neutral terrain where si and el vary and from which el penetrates Immediate contexts is, naturally, the sub-context where mismo penetrates last, namely, reference to the subject in the presence of a human DO; ii) the definitive evidence for el's take-over of si's domain would be provided by the recourse to el in the very homeland of si mismo, namely in clitic duplication of se. We have so far found a single (modern) case of this latter phenomenon18, and, not surprisingly, in a contrastive environment. Most gratifyingly, the example involves duplication of a se which neutralizes the opposition between double mention and the (oblique reference) Dative le: Cortäz. Fuegos 153/-1 "Soy yo", dice Jeanne, pero se lo ha dicho mäs a ella misma que a ese silencio opuesto en el que bailan, como en un telon de fondo, algunas chispas de sonido. "It's me", says Jeanne, but self it has said more to she same than to that silence opposed in the which dance, as in a curtain of back, some sparks of sound. '"It's me", says Jeanne, but she has said it more to herself than to that opposed silence where there dance, as if against a backdrop, some sparks of sound.'

This example allows us to surmise that the competition between si and el may end in the development of the compromise solution el mismo, whose beginnings can in fact already be observed in Table I. 19

42

Erica C. Garcia

The ultimate prediction, then, is that by ca. 2100 the grammatical categorization now conveyed by si will have been lexicalized in mismo, a development quite in line with the so-called drift to analytic expression that characterizes the history of the Romance languages. As for si, it will survive only in frozen phrases, in "grammatical serfdom" to specific lexical items (Garcia 1985a:297, 1991b:50, Schmidely 1983:179). At that stage, a (truly personal) pronoun el should contrast only with the syntagm el/ella mismo/a, etc., whose modifier highlights (whenever necessary or desirable) the contextual self-evidence of the referent. Should this development take place, the pronoun system sketched in Fig. 2 will attain even greater symmetry, the tonic functions being filled only by /- forms, while the s- forms (the clitic and possessive pronouns) would be confined to unstressed functions20.

10. Discussion and conclusions What is the tangible result of our groping attempts to take hold of the invisible hand? Perhaps the most important conclusion to be drawn is that not only forms, but environments too, change value in the course of time. This double diachronic reinterpretation follows, inevitably, from the synchronic interrelatedness of the two: contextual support guarantees that there should always be a certain indeterminacy as to the exact contribution of either party to the overall coherence of the message. The issue whether reinterpretation and diffusion constitute distinct events must thus receive an essentially negative answer: the two are indissolubly linked aspects of linguistic change, and the link is rooted in the synchronic dependence of communication on syntagmatic congruence. Our account in terms of an invisible hand mechanism has been mainly quantitative: it lacks, however, a clear specification of the qualitative shifts that presumably accompany and motivate the obvious shifts in relative frequency in different environments. Note, in particular, that we have assumed a stable and constant difference in value between the competing alternatives, and that at no point did we need to posit a shift in value for the two chief protagonists, namely si and el. And indeed, the fundamental difference between si and el unquestionably remains basically the same, in the sense that si was always 'more' personal than el, while the latter was always 'more' deictic than

Grasping the invisible hand

43

the former. The fact remains, however, that el can hardly be said to be the same in the 14th as in the 20th centuries. It is now obviously much closer than six centuries ago to being a 'real' personal pronoun. The problem is that it became so gradually: not at one fell swoop. One reinterpretation, i.e., that of the syntagm si mismo, we were forced to posit. But for the rest, examination of the texts reveals only a subtle, very gradual shift in both the nature and the relative proportion of the different environments, just as in the relative proportion of the forms. We have no way, as yet, of specifying, for any given stage, the qualitative synchronic reflex of the diachronic shift. For instance, how did the value of el change when, at the end of the 18th century, it for all intents and purposes became categorical in Mediate environments? Our account, then, still remains regrettably static and essentially structural. We have no way of indicating, at one and the same time, the stable (not static!) opposition of si to el, and the dynamic relation of either form to the cognitive substance to which it is syntagmatically applicable. It is the latter that 'places' a given alternative on the semantic-pragmatic continuum along which replacement takes place. And indeed: the former dimension guarantees and underlies the directionality of change; but insight into the second would allow us to actually track the use of a form along the centuries. Though neither the 'variable rules' of the sociolinguist, nor the recourse to 'prototypes' solve the problem, they clearly reflect scholars' awareness of the need to capture, in some fashion, speakers' evaluation of what communicative needs a given form is usable for. If we had the requisite conceptual tool, both historical linguistics and typologicalcomparative research would stand on much firmer ground. But perhaps the matter is even more complex: it is not merely that we have no way of indicating how exactly si changed in meaning (so that it could be used for fewer and fewer messages), or how el 'bleached'. How sure are we that communicative needs are constant, and comparable, across time and across languages? Our account rests on an extremely gross categorization of contexts: it thereby necessarily fails to do justice to the historic truth. For instance: the Remote uses of si that were possible in the 14th century differ greatly from what — though equally remote - is possible today. The finer contextual sub-classes certainly do not - and indeed cannot remain constant across the centuries.

44

Erica C. Garcia

Not only, then, do we not know how to express the shift in value of si and el: far less do we know how to indicate the change as to the messages conveyed in different centuries. The implications of this last fact for the nature of grammatical categories are as obvious as they are critical: "replacement" is possible without equivalence, pace Benveniste (1968: 86). An Innovation is not a Recessive category decked out in new finery, nor can contexts be regarded as a fixed set of buckets, waiting to be filled by an Innovation. Such views, compatible only with an inherently static conception of language, which is itself based on rigid and universal underlying categories, are given the lie by the protean changeability of language revealed in living use. Forms, thanks to their stable outline, enjoy a salience that obviously commands attention from both the analyst and the language learner. But it is the creative primeval syntactic soup out of which forms crystallize that really deserves the most careful consideration. Discovering the key to syntagmatic congruence remains the central issue of synchronic syntactic analysis, and a sine qua non for historical linguistics. Here, without question, we will find the answer to the two specifically diachronic questions: What quantitative shift in use triggers qualitative change? What quantitative change is to be expected from qualitative reinterpretation? If the Devil finds work for idle hands to do, he must surely feel frustrated when tackling linguists: the invisible hand certainly takes care that ours should be kept very busy. Notes 1. The same process took place in French, in whose modern stage soi is confined to a limited set of uses (cf. Brandt 1944: 273-286 et pass.). 2. It is not surprising, then, that the quantitative preference shown by a given form for different (linguistic) contexts (Garcia 1988a: 11-12, 1990: 303 et passim, 1991 a: 38-39) reflects a clearly qualitative motivation. 3. The inherent "circularity" of (con)textual interpretation is hardly news (Wootton 1975: 51-58 et passim, Sperber - Wilson 1986: 137) and results necessarily from two properties of language use:

Grasping the invisible hand

4.

5. 6.

7.

8.

9.

10.

45

i) its coherence and mutual relevance; ii) its variable precision: speakers are not always congruent to the same degree. It remains a question whether all cases of synchronic variation between meaningful variants need result in change, i.e., whether there may not be such a thing as historically stable variation (Romaine 1980: 237). The boundary between "categorical" and "variable" contexts is, clearly, itself anything but categorical, (cf. Stein 1990: 202). Note that an apparently monotonic change need not take place on a single (semantic) continuum: it may very well involve different cycles that take place along diverse dimensions since, depending on the relevant trait, it is possible to line up contexts along a variety of clines. The reason is that syntactic contexts necessarily represent a coincidence - or intersection - of diverse semantic-pragmatic dimensions which thus provide more or less obvious avenues of independent expansion for an innovation (Timberlake 1977: 160-168; Garcia 1989a: 138-145). There will accordingly always be some dimension along which an R-favorable context resembles the environment where I varies with R. The first passage, namely PLA 23b46, is the first sentence in the chapter, entitled "How Garci Laso and others of the city were killed in Burgos". The second, i.e., PLA 86b34, occurs two thirds of the way down in the chapter, entitled "How the King Don Pedro arrived at Borja where was the power of the King of Aragon and of the Count Don Henry"; the preceding text essentially consists of an enumeration of the forces that arrive at Taracona. In our analysis we ignore not only Remote contexts but also exclude a) reference to the subject a coordinate structure after entre 'between', where the coordination is itself a factor favoring use of el. These constructions are exceedingly infrequent; b) the use of pronouns with nominalizations or nouns that are liable to that interpretation, since in these uses the intended referent is never to be found in the context, but is actually inferred from the choice of pronoun, as shown by the pair el amor de si 'the love of oneself el amor de el 'the love of him; his love' We leave for another occasion the analysis of these constructions, where, in any case, the intrusion of el is, understandably, minimal. The corpus is comprised of prose classics, addressed - in their respective times - to a wide reading public. This guarantees not only their representativeness for the standard language but also their value as models for the prose of their time. Under this type we include the following contexts: i)

PP is grouped with the verb: reference to either the subject of the verb (whether finite or non finite) or to its direct object; this specifically

46

Erica C. Garcia

includes reference to the Accusative subject of an Infinitive in ACT constructions; ii) PP is grouped with a noun or modifier: reference to the entity referred to by the word which the PP is grouped; iii) PP is part of a comparison involving i) or ii); iv) non prepositional predicative use of si. 11. Since it is important to ensure direct comparability with Immediate contexts as regards the function of the referent, we take into account only Mediate contexts where the referent of the prepositional pronoun is either the subject of the main verb, or a verbal clitic, specifically the Dative object. The nature of the Accusative is much more directly presupposed by the event than that of the Dative, which is essentially independently addable: the Accusative (Direct Object) thus is more context-given than a Dative. This judgement is based on independent properties of the two cases (Garcia 1975: 461, 501, 98-102); it is supported by experimental evidence showing that an Accusative Object is referred to with si proportionately more than a Dative (Garcia 1991 a: 45). 12. We have counted occasional instances of modifiers such as solo 'alone' and propio Own' under mismo, since they also specify gender and number of the referent, and highlight either the sameness, or the unicity of the referent. Under el we have also counted other alternatives to si, such as uno One' for indefinite reference, and Usted 'You', since polite address resorts to grammatically third person forms. Uno and Usted begin to alternate with si very much later than el; the number of cases involved is about 1 for every 50 cases of el. 13. This latter relation is true "up to a point": for several 20th-century texts (e.g. Masas) show si regularly accompanied by mismo in the Mediate contexts. The development, however, is observed only after the generalization of mismo in the Immediate contexts. The diffusion of mismo in Immediate and Mediate contexts appears to provide a clear counterexample to Kroch's implicit claim (1989a: 158, 1989b: 205-207 et passim) that the rate of diffusion is constant in all contexts. 14. We indicate only the chief alternatives in either type of context, and identify the I variant by means of boldface. 15. For what the judgment of a grammarian may be worth, Fernandez (1951: 222, #117), categorizes mismo as "reflexive and intensive" pronoun. Furthermore, such a unit reinterpretation patterns with the development of nosotros and vosotros as personal pronouns (Garcia et al. 1990: 70). 16. Recall that the data in Table 3 involves only cases of "first mention", i.e. those where the PP considered is either the only one, or precedes any reference to other entitites involved in a comparable relation.

Grasping the invisible hand

47

17. By "categorical" we mean that the composite, rather than the single form, is resorted to in cases of duplication, not that every single case of se can or must necessarily be duplicated by an α-phrase (cf. Otero 1969: 1147-1148 et passim). 18. A second example occurs in the PCG - an early text. The example follows: PCG 600bl Et de la otra parte el Cid Ruy Diaz en Valencia gyusosse a el et a su conpanna; et la estoria cuentalos assy los que fueron con el aquellas vistas ... And on the other side Cid Ruy Diaz in Valencia readied self to him and to his company; and the story tells them thus the who were with him those sights ... 'And on the other side the Cid Ruy Diaz in Valencia readied himself and his company; and the story enumerates thus those who went with him to that meeting ...' Note that i) in the 14th century clitic duplication had not the compulsory character it has in the 20th (Rini 1990); ii) the use of el occurs in coordination: the α can thus be taken to introduce the Noun Phrase el et su companna, something not only allowed but favoured by the deictic nature of el. Both facts, as well as the extreme rarity of clitic duplication of se in 14th century texts suggest that the recourse to el had a very different value from what it has now. 19. In the competition between si mismo and el (mismo) recourse to si certainly adds a "self-evident" flavor to the reference: but it represents no (cognitive) saving in terms of lesser precision. 20. A similar suppletion is observed in the paradigm of voseo (Garcia 1994).

References Bailey, Charles-James 1973 Variation and linguistic theory. Washington, DC: Center for Applied Linguistics. Benveniste, Emile 1968 "Mutations of Linguistic Categories", in: W.P. Lehmann - Y. Malkiel (eds.), 83-94. Brandt, Gustaf 1944 La concurrence entre Soi et Lui, Eux, Elle(s). Lund: H. Ohlsson. Fasold, Ralph W. - Deborah Schiffrin (eds.) 1989 Language Change and Variation. Amsterdam: Benjamins.

48

Erica C. Garcia

Fernandez, Salvador 1951 Gramatica espanola. Madrid: Revista de Occidente. Garcia, Erica C. 1975 The role of theory in linguistic analysis. Amsterdam: North Holland. 1983 "Context-dependence of language and of linguistic analysis", in: Flora Klein-Andreu (ed.), 181-207. 1985 a "Quantity into quality: synchronic indeterminacy and language change", Lingua 65:275-306. 1985 b "Shifting variation", Lingua 67:189-224. 1988 a "Lingüistica cartesiana el metodo del discurso", Lenguaje en Contexto 1: 5-36. 1988b "-go, cronopio entre los morfemas: consigo contrastado con 51 mismo", Neuphilologische Mitteilungen 89: 197-211. 1989 a "Quantitative aspects of diachronic evolution: the synchronic alternation between O. Sp. y, alii 'there'", Lingua 77: 129149. 1989b "La historia se repite con sigo", in: Sebastian Neumeister (ed.), 113-124. 1990 "A psycholinguistic crossroads: frequency of use", Journal of Semantics!: 301-319. 199 la "Grasping the nettle: variation as proof of invariance", in: Linda R. Waugh - Stephen Rudy (eds.), 33-59. 1991b "Morphologization: a case of reversible markedness?", Probus 3: 23-54. 1994 "Una casilla vacia en el paradigma pronominal del voseo: convusco", in: Jens Lüdtke (ed.), 13-38. Garcia, Erica C. - Robert de Jonge - Dorien Nieuwenhuijsen - Carlos Lechner 1990 "(v)os-(otros): /r/ between vowels) in same subject and different subject transitions; in same subject transitions, t-forms occur alongside non-finite Same Subject forms. The use of t-forms in subject discontinuity conditions is relatively frequent compared to its same subject use in Yonggom texts because in subject continuity conditions, t-forms can be replaced by non-finite forms. This relative frequency of t-forms in Different Subject conditions in Past tense sequential-action chains probably is the basis for the reinterpretation of medial t-forms as Non-Future/DS forms in Digul-Wambon, assuming that the Yonggom dialect without switch-reference represents the older stage. The Future/Different Subject forms of Digul-Wambon (Future verbstem plus person-number), called zero-forms by Drabbe (1959: 127) because they lack tense and mood slots, occur also in Yonggom. Chainfinally, the zero-forms have adhortative and optative meanings in Yonggom; chain-medially, the zero-forms have Future Indicative meaning and occur in both Same Subject and Different Subject transitions. Chain-finally, Future Indicative is expressed by tensed forms with the Future tense marker -in. Thus Yonggom has a special chain-medial Future tense form, the zero-form. Although the chain-medial Future form occurs in both Same Subject and Different Subject transitions, in Same Subject conditions non-finite forms can replace the zero-forms. Thus the Different Subject use of zero-forms is relatively frequent, and in this way they could be reinterpreted as Future/DS forms in Digul-Wambon. Summarizing:

96

(12)

Lourens de Vries

Yonggom dialect t-form : final&medial/SS&DS; realis zero-form : final&medial/SS&DS; Future (medially), adhortative and optative (finally) Digul-Wambon dialect t-form : medial/DS; Non-Future zero-form : medial/DS; Future

Assuming that Yonggom-Wambon represents the older stage, we have some grounds to say that the -ί-morpheme which marks Different Subject in Digul-Wambon developed out of a realis marker. The narrative chain-medial use of this untensed realis form, relatively frequent in subject discontinuity conditions, as we can observe it in the Yonggom dialect, is a good starting point for a reinterpretation as a DS marking form in Non-Future narrative texts. The change from realis status to Non-Future tense is not unexpected, given the close association between realis as a status and Non-Future as a tense distinction (see Foley 1986: 159). There is still a trace of the realis-marking function of the morpheme -t in Digul-Wambon: it occurs as a part of the ending of Past finite forms in Wambon (de Vries and Wiersma 1991).

4. Aghu Aghu is spoken by about 2,000 persons in the area between the Digul and the Mappi (see 10. Maps). The Aghu language was described by Drabbe (1957). Aghu has non-finite medial verbs which signal Same Subject transitions. These verbs consist of the verb stem plus either -b or -d. According to Drabbe (1957: 18, 19), the b-forms were originally Future Same Subject forms and the d-forms Non-Future Same Subject forms but the tense distinction is fading out in Aghu. In Sjiagha and Yenimu the bV-forms and dV-forms still mark the Future/Non-Future tense opposition (Drabbe 1950: 109). The non-finite Same Subject forms, especially the d-forms (formerly Non-Future), can optionally be marked for sequence by the suffix -ke. If unmarked for temporality, the non-finite forms occur in both sequence and simultaneity conditions. The second option for Aghu Same Subject transitions is to use finite forms, so-called zero-forms, to which the SS/sequence suffix -ku is

The rise of switch-reference in the Awyu languages of Irian Jaya

97

added. These forms are called zero-forms by Drabbe (1957: 10) because they consist of the verb stem plus a person-number ending, without other morphological slots. The term zero-forms is perhaps better than untensed forms since zero-forms express the Non-Future and Future tense opposition via their suppletive stems: primary verb stem for NonFuture and Future stem for Future tense. The zero-forms (verb stem plus person-number) occur also chain-finally. If so, they have optative and adhortative meaning when they have a Future stem, and a NonFuture meaning with the primary stem. For Different Subject transitions, Aghu uses zero-forms plus the Different Subject/sequence suffix -ne or just zero-forms. The latter option, zero-forms without a switch-reference suffix, is temporally neutral. We do not know what the origin is of these Different Subject/ sequence and SS/sequence suffixes. A possibility is that they originated as interclausal conjunctions. It is remarkable that in Same Subject sequence conditions, usually seen as the most unmarked and continuous condition, a finite form can be used with a Same Subject/ sequence suffix. Notice that non-finite Same Subject forms can also be marked for sequence (by -ke). Aghu has grammaticalized the finite versus non-finite opposition into the Same Subject/Different Subject opposition: finite zero-forms (verb stem plus subject person-number) contrast with non-finite forms in a Different Subject/Same Subject opposition. But the language has a device by which these same finite zero-forms can be used in Same Subject conditions: by adding the Same Subject/sequence suffix -ku. Consider the following examples (from Drabbe 1957: 18, 36—38) of the use of -ne (13), dV-forms (14), -ku (15), bV-forms with the sequence suffix -ke (16) and the use of zero-forms without temporality suffixes as Different Subject forms (17): (13)

Tamo-gho-ne gho-de write-3SG-SS/SEQ go-ISO 'He wrote and then I went.'

(14)

Tame-de gho-de write-SS go-ISO Ί wrote and went.'

98

Lourens de Vries

(15)

Tamo-ku aghi-je write. 1SG.F-SS/SEQ go-lSG.F shall write and then go.'

(16)

Tame-bu-ke aghi-je write-SS-SEQ go-lSG.F shall write and then go.'

(17)

üföo-ghena hit-3Pl

efe go his blood all

isiom midiki-ke flow.away-3SG

kü-ngge/kü-n ia die-3SG/die-3SG.RemotePast 'They hit him, he lost all his blood and died.' Notice that it is just the zero-forms, and no other final forms, which are used medially in Aghu. Thus Aghu has a certain specialization between final-only forms and final forms which also occur medially; the Same Subject and Different Subject suffixes cannot be suffixed to any finite form. Only to the zero-forms which have just a person-number slot and no tense and mood and aspect slots. This property, the absence of tensemood-aspect slots, is what facilitates their development into dependent, medial forms. We turn now to another Awyu language, Korowai, which also uses Same Subject and Different Subject suffixes or rather Same Subject and Different Subject conjunction clitics, but which imposes no restriction on which final forms may be used chain-medially with the switch-reference clitics.

5. Korowai Korowai (de Vries and Van Enk f.c) is spoken in the area between the upper Becking and Eilanden rivers (see 10. Maps). The dialect described here is that of the clans living on the western banks of the Becking River in the proximity of Yaniruma. Korowai has one dependent (medial) verb form and many independent formations which function both medially and finally in clausechains. The dependent verb form is the non-finite Same Subject medial form consisting of the verb stem plus the optional Same Subject suffix -ne. Examples:

The rise of switch-reference in the Awyu languages of Irian Jay a

(18)

...mebol damilmo le-ne lu-ba-le ...grave open.SS come-SS ascend-PERF-lSG.NF '... I opened the grave and came up (the stairs) ...'

(19)

l-ne khami-bo see.SS stay-sit.3SG.NF 'He was looking.'

(20)

...dal-khal lop-tena-lena fu ...tree-bark hole-little-little put.SS

99

wap-ta ye-le that-on sleep-1PL.NF '... we covered the little holes with treebark and slept on that (treebark covered place).' Independent verb forms are switch-reference neutral, i.e., when they are used medially, they can be followed by clauses with the same subject and by clauses with a different subject. When independent verb forms function medially in clause chains, in the great majority of cases there is a switch-reference conjunction cliticized to the verb to indicate whether the next clause has the same subject or a different subject. The switch-reference conjunctions we found in the data are listed in (21): (21)

-do(n) -dakhu(l) -anggu -tofekho (or: -top)

'Different Subject' 'Same Subject' 'Same Subject/Intentional' 'DS/adversative'

The conjunctions -do(n) OS' and -dakhu(l) 'SS' are the most frequently used of the set (21). The conjunctions -anggu 'SS.intent' and -tofekho 'DS.but' have a more marked meaning and are accordingly more restricted in occurrence, with -anggu used in intentional contexts and -tofekho in adversative contexts. First an example with -do(n): (22)

Nu lep-telo-do yu be-lai-da I ill-3SG.NF-DS.and he NEG-come.3SG.NF-NEG Ί am ill, and he does not come (= Because I am ill, he does not come).'

100

Lourens de Vries

Example (23) shows -dakhu(l) 'SS.and': (23)

Nu khomile-le-dakhu khosü kha-le I die-lSG.NF-SS.and there go-lSG.NF died and went there (= to the place of the dead).'

The marker -anggu is a Same Subject clitic that we have found only in intentional contexts: (24)

ge-lal-to your-daughter-FOC

fedo-m-do give-2SG.IMP-DS.and

fo-p-anggu take-1 SG.INTENT-SS.and

ne-mom my-uncle

kholop-fuda-mo-p replace-compensate-SUPP-1 SG.INTENT 'You must give your daughter, and I want to marry her, and I want to replace my uncle (= you must give your daughter in marriage to me as compensation-gift for my (dead) mother's brother).' The marker -tofekho 'DS.but' is a Different Subject clitic that seems to be restricted to adversative contexts (centra-expectation, contrast): (25)

...nu khomile-le-dakhu ...I die-lSG.NF-SS.and

khosü kha-le-lofekho there go-lSG.NF-DS.but

debülop-tale walüp-ta lül aule-te-do... road-big middle-in barrier close-3PL.NF-DS... '... I died and went there, but on the big road they stopped me halfway and...' (26)

Khakhul nu ne-mom yesterday I my-uncle

dodepa-le-lofekho call-1SG.NF-DS.but

be-lai-da NEG-come.3SG.NF-NEG 'Yesterday I called my uncle, but he did not come.' We do not know what the origins of the switch-reference conjunctions are.

The rise of switch-reference in the Awyu languages of Irian Jaya

101

6. Summary The switch-reference systems discussed in this article can be summed up as follows: 1. Kombai All finite forms are chain-medially reinterpreted as Different Subject forms: non-finite verb finite verb

= Same Subject = Different Subject

2. Wambon (Digul dialect) Certain finite forms, the realis t-forms and the Future medial zeroforms, are reinterpreted as medial Different Subject forms: non-finite verb = Same Subject finite t-form = Different Subject/Non-Future finite zero-form = Different Subject/Future 3. Wambon (Yonggom dialect) Finite forms occur both chain-medially and chain-finally; there are special chain-medial Same Subject forms but there is no grammaticalized switch-reference: non-finite verb finite verb

= Same Subject = switch-reference neutral

4. Aghu A specific type of finite forms, untensed forms, are medially interpreted as Different Subject forms, but Same Subject and Different Subject clitics can be attached to these same forms and turn them into Same Subject or Different Subject forms: non-finite verb untensed finite verb untensed finite verb + -ku untensed finite verb + -ne

= Same Subject = Different Subject = Same Subject/sequence = Different Subject/sequence

102

Lourens de Vries

5. Korowai Interclausal conjunctions cliticize to finite forms as Different Subject and as Same Subject markers: non-finite verb finite verb + -ndakhu finite verb + -do

= Same Subject = Same Subject/sequence = Different Subject

The grammatical function of chain-medial finite forms varies unpredictably from one Awyu language to another. In Kombai, chain-medial finiteness is assigned the grammatical meaning 'Different Subject'. In Korowai, its neighbor, chain-medial finite forms are switch-reference neutral and may be marked for either S S or D S by clitics. In Aghu, a chain-medial finite form codes discontinuity if not accompanied by a switch-reference clitic, but if there is a clitic the same finite form may code either Different Subject or Same Subject. Digul-Wambon recruits certain less finite verb formations to serve as special medial Different Subject forms: t-forms and zero-forms. A high degree of arbitrary coding is the outcome of linguistic change in the Awyu family: the meaning 'Different Subject in next clause1 and the verb forms expressing that grammatical meaning came to be unpredictably and arbitrarily linked in the process.

Abbreviations 1 2 3 ADJ CONN COORD DEM DS DUR F FOC

first person second person third person adjective connective coordinator demonstrative different subject (switch-reference) durative future tense focus

INTENT NEG NF PERF PL REAL SEQ SG SIM

ss TR

intentional mood negation non-future tense perfective plural realis sequence singular simultaneity same subject transitional sound

The rise of switch-reference in the Awyu languages of Irian Jaya

103

Notes 1. Irian Jaya is the easternmost province of Indonesia, the western half of the island of New Guinea. There are about 1,000 languages spoken on New Guinea; about 250 of these belong to the Austronesian family. The remainder, 750 languages, often called Papuan languages, cannot be said to derive from one ancestral language and are organized in more than sixty language families (Wurm 1982, Foley 1986). 2. The Awyu languages, especially those in the foothills of the central New Guinea mountain range, are spoken in an extremely isolated area of Irian Jaya. Till today some Awyu groups live in isolation. Yet we have quite extensive data on this family because of the long-term research of Drabbe and myself. Father Drabbe did linguistic work in the 1940s and 1950s. I lived and worked as a linguist in the area for more than nine years between 1982 and 1991. 3. In this article we ignore two important issues in the area of switch-reference. The first is the issue of what counts as same referent and different referent in a given language (see Reesink 1983); the second is, what it is that switchreference monitors in a given language: (semantic) actor continuity, (pragmatic) topic continuity, (syntactic) subject continuity, or combinations of these (see Foley 1986, Roberts 1988). 4. See for a detailed treatment of clause combining in Awyu clause-chains, De Vries(1993b). 5. The Awyu verb forms which play a role in our discussion, may be divided in terms of degree of finiteness into three categories. We call the verb root plus optional suffix formations non-finite. Finite forms are the fully inflected independent forms. The term "less finite forms" will be used for independent forms with just a subject person-number slot or a subject person-number slot and the realis suffix -t (the f-forms and zero-forms of Aghu and Wambon). We restrict our attention to the best documented members of the Awyu family: Kombai, Digul-Wambon, Yonggom-Wambon, Aghu, and Korowai. 6. The term tail-head linkage (t-h linkage) is from Thurman (1975). Longacre (1972: 45) has described the phenomenon as follows: "Commonly the function of the first base in such chaining units is to refer back to the last base of the previous chain. If the chain structures as a paragraph, then such back-reference or recapitalation joins paragraph to paragraph". See also Healy (1966) for clause chaining and tail-head linkage. 7. The gloss REAL (realis) is mine. Drabbe (1959) does not gloss the -t and calls these forms consistently t-forms. Concerning the use of the t-forms, Drabbe (1959: 155) says that they are used in present and past tense conditions and frequently as narrative forms. My interpretation of the -t morpheme as a marker of reality status rather then as a tense marker is based on the fact that the -t also occurs in tensed Past forms preceding the Past marker -an.

104

Lourens de Vries

References Drabbe, Father P. 1950 Twee dialecten van de Awju-taal. [Two dialects of the Awyu language]. The Hague: Nijhoff. 1957 Spraakkunst van het Aghu-dialect van de Awju-taal. [Grammar of the Aghu dialect of the Awyu language]. The Hague: Nijhoff. 1959 Kaeti en Wambon. Twee Awju-dialecten. [Kaeti and Wambon. Two Awyu dialects]. The Hague: Nijhoff. Foley, William A. 1986 The Papuan languages of New Guinea. Cambridge: Cambridge University Press. Givon, Talmy 1983 "Switch-reference and topic continuity", in: John Haiman Pamela Munro (eds.), 50-82. 1990 Syntax. Volume II. Amsterdam: Benjamins. Haiman, John 1983 "On some origins of switch-reference marking", in: John Haiman - Pamela Munro (eds.), 105—128. 1985 "Introduction" in: John Haiman (ed.), 1-7. Haiman, John (ed.) 1985 Iconicity in syntax. Amsterdam: Benjamins. Haiman, John - Pamela Munro 1983 "Introduction", in: John Haiman - Pamela Munro (eds.), ix-xv. Haiman, John - Pamela Munro (eds.) 1983 Switch-reference and universal grammar. Amsterdam: Benjamins. Healy, Phyllis 1966 Levels and chaining in Telefol sentences. Pacific Linguistics, Series B, no. 5. Heine, Bernd — Ulrike Claudi — Friederike Hünnemeyer 1991 Grammaticalization. A conceptual framework. Chicago: The University of Chicago Press. Longacre, Robert E. 1972 Hierarchy and Universality of Discourse Constituents in New Guinea Languages: Discussion. Washington: Georgetown University Press. Reesink, Ger P. 1983 "Switch-reference and topicality hierarchies", Studies in language 7: 215-246.

The rise of switch-reference in the Awyu languages of Irian Jaya

105

Roberts, John 1988 "Switch-reference in Papuan languages: a syntactic or extrasyntactic device?", Australian Journal of Linguistics, 8: 75-117. St. Lucia, Queensland, Australia. Silzer, Pete - Helja Heikkinen 1991 Index of Irian Jaya languages. Jayapura: Summer Institute of Linguistics and University of Cenderawasih. Thurman, Robert C. 1975 "Chuave medial verbs", Anthropological Linguistics 17: 342352. Voorhoeve, Clemens L. 1975 Languages of Irian Jaya: checklist, preliminary classification, language maps, wordlists. Canberra: The Australian National University Press. Vries, Lourens de 1989 Studies in Wambon and Kombai. Aspects of two Papuan languages of Irian Jaya. PhD dissertation University of Amsterdam. 1993 a Forms and functions in Kombai, an Awyu language of Irian Jaya. Pacific Linguistics. Series B, no. 108. Canberra: Australian National University Press. 1993b "Clause combining in oral Trans-New Guinea languages", TEXT 13:481-502. Vries, Lourens de - Gerrit J. van Enk (f.c.) The Korowai of Irian Jaya. Aspects of their language and oral tradition. Oxford/New York: Oxford University Press. Vries, Lourens de - Robinia Wiersma 1992 The morphology of Wambon of the Irian Jaya Upper-Digul area. Leiden: KITLV Press. Wurm, Stephen A. 1982 Papuan Languages of Oceania. Tübingen: Gunter Narr.

Linguistic activity and change

"Scenario" as a concept for the functional explanation of language change Wolfgang U. Dressler

1. Introduction Let me start with the confession that I have never been in favor of distinguishing sharply between grammar on the one hand, and philology, pragmatics, discourse, style, etc., on the other hand, and that I have never believed that just one of the two areas is really worthwile of linguistic investigation (cf. Dressier 1995 a). As a result, I oppose mutual exclusion of functional and formal explanation also in the field of diachronic linguistics (cf. Hall 1992; Newmeyer 1992; 1994). I consider formal explanation rather as part of functional explanation. My main reason is (cf. Dressier and Merlini 1994: 1.2-1.6) that I belong to those who see human language as a specific type of human interaction1 so that pragmatics (particularly text pragmatics) is to be seen as the all-encompassing highest level of which grammar is just a subordinate subpart. Now for explanation in pragmatics, discourse, and sociolinguistics only functional explanation is adequate. And, as I will try to show, since scenarios of change always involve pragmatic and sociolinguistic aspects and function in discourse, at least some elements of functional explanation are inevitable. This does not mean, however, that we have to reject formal explanation per se, be it of a structuralist or of a generativist format (cf. Kiparsky 1988) or in terms of my own framework (Natural Phonology, Natural Morphology, etc.) or in terms of Andersen's typologies of changes (e.g. 1980, 1988) or in terms of grammaticalization theory (cf. Hopper and Traugott 1993) or of Labov's (1994: 2) "constraints and transitions", etc. But formal explanation of diachronic change can aspire only to the status of partial and very incomplete explanations in the sense of skeletal schemata of possible changes — although it proves to be very hard even to define the dividing lines between possible and impossible changes (cf. Morin 1994; Thomason and Kaufman 1988: 13ff; Boretzky 1993: 1). But the conditions under which a formal scheme of change

110

Wolfgang U. Dressler

may apply are often scarcely investigated; they are rather hidden or even forgotten behind the smoke-screen of the "instantiation/actuation problem" (cf. Weinreich, Labov and Herzog 1968: 186f; Romaine 1982: 269 ff), for example, whether simplification or elaboration occurs, or whether markedness is increased or decreased. In contrast, functional explanation tries to include as many of these omitted conditions as possible within the given approach and thus leads to "new frontiers" of explainability. In other publications I have dealt with the rationale of functional explanation in linguistics in general and in various areas including diachrony (1995b; 1985: 26Iff; Dressier and Dziubalska 1994). Here I can only briefly indicate: 1) that I do not equate function with communicative content or meaning, which lies at the basis of Labov's (1994: 40ff, 118, 547 ff) narrow interpretation of functional explanation of diachronic change, e.g., of phonological chain shifts; 2) that functional explanation is also applicable to linguistic structure and structural change, where we may abstract from speakers' intentions; 3) that functional explanation is not reducible to causal explanation (in the sense of Hempel-Oppenheim type deductive-nomological explanation); 4) that apparent weak points of functional explanation (such as goal conflicts, multiple strategies, plurifunctionality, competition of alternative explanations) are the necessary reflections of the complexity and multicausality of linguistic phenomena, particularly of diachronic change; 5) that the identification of functional analysis with a caricature of teleological explanation is confused; 6) that the important methodological concept of asterisks in generative grammar has a correspondent in functional analysis, namely the notion of dysfunction or functional deficiency and of dysfunctional change. As a result, functional explanation of diachronic linguistic change must try to consider which types of speakers or of groups of speakers are the likely innovators, in which speech situations and types of discourse, under which social conditions, in sum: in which scenario of change. Despite Andersen's (1989: l l f f ) insightful critique of the term change, I will stick to "scenario of change", considering it - in contrast to Andersen's term "innovation" — as an interactive phenomenon. Whereas Andersen's innovations may be envisaged as, so to say, social atoms, change and scenario of change may be envisaged as social molecules. In this way I hope to continue and promote work in a socially constituted and embedded theoretical diachronic linguistics2. In other

"Scenario " as a concept for the functional explanation

111

words, I propose that the concept of "scenario" may be an important help for interrelating the often distinguished "internal" and "external" factors of language change (cf. Giacalone Ramat 1989) in a more systematic way. Obviously this is an enormous task to which I can do justice only very imperfectly.

2. Scenario Among historical linguists I dare to confess the reason why I chose this term: it has been its etymology. "Scenario" comes from It. scenario which designated the scheme of a commedia dell'arte, i.e., of its plot and of its scenery with the dramatis personae, the props, the technical devices, and the transformations of the scene. And since I envisage language use in terms of an impromptu theater (cf. Dressier and Merlini 1994: 1.2), the theatrical concept of scenario fits this view very well. After the commedia dell'arte, scenario came to mean in several languages an outline of a hypothesized and predictable chain of events.3 Sociologists and sociolinguists often use the term scenario, but rarely define it, e.g. in interaction theory as a "model for explaining actions of actors in recurrent situations" or as "typical starting-situations with a set of fixed parameters" (translated from Fuchs-Heinritz et al. 1994: 666). In economic decision theory, scenarios are "focused descriptions of fundamentally different futures presented in coherent script-like or narrative fashion" (Shoemaker 1993: 195); they "bound the uncertainty range but do not give it probabilistic prominence" (ibid. 196). If we locate, as I propose (cf. Dressier and Merlini 1994: 1.2, 1.5), scenarios within action theory, then functional explanation becomes a necessity (cf. Anttila 1992). Functional explanation then needs for all scenarios of change respective bridge-theories (in the sense of Botha 1979) which connect action theory (including pragmatics and sociolinguistics), linguistic theory proper, and a theory of change. For lack of space, I can do this here only very sketchily. It is customary to distinguish between origin and spread or diffusion of linguistic innovations. But this, I find oversimplificatory, first, because the various phases of spread of an innovation may be of very different types, and, second, because the individual origin alone appears of little importance, since most innovations have no chance to persist. Therefore I think that a scenario of diachronic change should consist of

112

Wolfgang U. Dressler

at least five chronologically ordered phases: 1) origin with one or several individuals (independently) — this is termed innovation by Milroy (1992, 1993); 2) adoption by some other individual(s) as a result of negotiation4; 3) spread to one social group (in small tightly-knit networks, cf. Trudgill 1993; Milroy 1992: 184), thus establishing a new lect defined by this isogloss accepted and imitated by its speakers5; 4) competition with one or more older variants, their evaluation by speakers6, and the chances that the innovation wins over its competitors; 5) possible ways of further sociolectal and dialectal spread,7 where this fifth part takes the longest time and is the less fixed part of the scenario, i.e., later sociolectal and dialectal spread may be modelled by various scenarios of their own (cf. Milroy 1993; Labov 1994; including lexical diffusion). Although some types of origin might be fairly mechanistic, explainable by a formal account of language performance (cf. Berg 1988), adoption and spread of all sorts — without which no diachronic change would have occurred — needs functional analysis (cf. Cooper 1982). All five phases of each scenario may have among their properties (incl. controlling factors) a) defining properties, b) default properties, c) prototypical properties, and d) alternative properties, all of them to be distinguished from accidental properties (cf. Andersen 1988: 40). Properties of type b)-d) include what is covered by the notion of "locus of change" in Hall (1992: 35), viz., the determination "at what stage in the development (i.e. acquisition) of language" change takes place, and it goes beyond Hall's (1992: 33) concept of "linkage", i.e., the demonstration of how a universal preference (e.g., of universal markedness theory) is translated into diachronic change.8 The introduction of default, prototypical, and alternative properties precludes deductivenomological explanation, but allows functional explanation and accounts for the multicausality of historical change.9 Often it may prove useful to subdivide a scenario into subscenarios when the majority of properties is identical, but some crucial properties diverge systematically. A final remark on the instantiation/actuation problem: actuation belongs to the level of communicative performance (cf. the compatible view of Milroy 1993: 22Iff), whereas the concept of scenario belongs to the level of communicative competence, i.e., scenarios are types of chains of interactive events, of which specific instantiations are tokens.10 The aims of this line of work are 1) to establish different scenarios, seil, as types," 2) to exemplify them with tokens, i.e., both with well-

"Scenario " as a concept for the functional explanation

113

observable chains of events in ongoing change or recent, welldocumented change and with distant changes which lack detailed documentation, 3) thus to avoid the necessity to construct again and again a separate functional explanation for each case of change, i.e., to reduce it to equating a token of a change with its type, adding non-prototypical and counter-default properties, i.e., pointing out which default or prototypical properties must be modified. 4) A final aim is to justify linguists who are not interested in sociolinguistics, pragmatics, and extralinguistic history but want to concentrate on formal explanation of linguistic change. This narrow-mindedness is justified, if we can supply them with scenarios of linguistic change which allow them to assume that a formal type of change X is liable or likely to occur in scenarios under conditions Z. Of course this paper represents only a small step towards these goals, I can only discuss a tiny part of the immense area of problems involved, but I hope that at least I can make myself understood and arouse interest in this research program. In order to facilitate my task I will concentrate on phonology (section 3) and morphology (section 4).12

3. Specific scenarios, mainly phonological 3.1. Sociophonological scenario: natural backgrounding processes I want to start with the scenario that we have worked on for a long time in Vienna (since Dressier 1973). This scenario of propagation of phonological backgrounding or lenition processes from casual to formal speech styles is based on an integrated sociopsycholinguistic bridge theory (cf. Dressier and Wodak 1982; Dressier and Moosmüller 1991, Moosmüller and Dressler 1992) whose linguistic ingredient is a semiotically-based model of Natural Phonology. Already explanation on this phonological level is functional (cf. Dressier and Moosmüller 1991).13 Defining properties are a) the set of phonological backgrounding processes with their hierarchies of application, including generalization of application in casual speech according to these hierarchies, b) the characterization of speech situations, speech events, and psychological factors which favor phonological casualness,14 c) the propagation of generalized application of such processes from more casual to less casual speech.15

114

Wolfgang U. Dressler

As an example of well-observed ongoing/recent change I take monophthongization of /ae, ao/ in Viennese German (cf. Moosmüller 1991: 66ff and passim), as in: (1)

/vaet/ 'wide' > [vaet, vaEt] > [vaet, vaEt, vas:t] > [vaEt, vae:t] > /vae:t/

Gradual, partial, and optional monophthongization was first observed at the very end of the 19th century in the speech of lower-class youth, i.e., when first sociolectal spread up to acceptance by this group had been terminated. Later on it spread to other sociolects and notably from the Viennese basilect to the Viennese acrolect, and finally it has been spreading from this acrolect to other Austrian acrolects, both in use and positive evaluation. As long as monophthongization is an optional rule, it follows the defining phonological default hierarchy of unstressed before stressed diphthongs (and other rhythmic conditions) and the prototypical priority of frequent before infrequent words,16 of grammatical words before lexical words (cf. Moosmüller and Dressler 1992).17 In the Viennese basilect, however, monophthongization has become total and obligatory, thus the sound change /ae/ > /ae:/, /ao/ > /D:/ has been completed. As to its origin, monophthongization must have started with young people (in accordance with observations at the time). Since young children are observed to monophthongize spontaneously adult diphthongs they are exposed to,18 we may hypothesize that the change started in child-adult communication when adults did not correct monophthongization in the primary linguistic context, i.e., in unstressed positions of casual speech where monophthongization is difficult to perceive.19 Only later, this change generalized sociolectally and from preferred to dispreferred contexts, i.e., to stressed positions, to infrequent words, to lexical words, to formal speech. Thus finally this change has become a "regular sound change" in the Viennese dialect. Not only in the origin, but also in adoption and early spread of monophthongization and of similar changes, young children may play a crucial role, because young children are generalizing one of several variants before they learn to master the sociolinguistic conditions for the complementary use of variants (cf. Moosmüller and Vollmann 1994). The locus of later sociolectal and dialectal spread, however, is rather adolescents and adults. In the case of Viennese monophthongization,

"Scenario " as a concept for the functional explanation

115

the relatively rapid spread had to do with the emancipation and social rise of lower classes. The spread into the acrolect was accelerated after the second world war due to intentional national dissociation from Germany and its standard language.20 In many other cases spread within this scenario of "change from below" (Labov 1994: 78; cf. also Kiparsky 1988: 374ff) is slower, because social rise is slower and involves much smaller groups. This is, so to say, a normal course of events - in the absence of great social or demographic upheavals or of the sudden impact of puristic reactions. The chance of adoption and speed of spread may be influenced by the controlling factor of the relative degree of openness of a community, but not only in the sense of Andersen (1988: 40, 60ff, 71 f, 77): he defines an open community by a fair amount of interdialectal communication. But this should be expanded from the dialectal to the sociolectal dimension, as indicated by Trudgill (1993: 7): "tightly-knit communities may be more likely to occur in the case of smaller, more isolated, low-contact communities". But this general controlling factor holds for innovations in general, although it seems to be most important for the secondary phases of spread. In the change from below of such phonological backgrounding processes regular sound change may be achieved, due to the relatively lower level of monitoring of phonological backgrounding processes (cf. Labov 1994: 39). We may reanalyze, with much confidence, many ongoing changes described by many sociolinguists in terms of this sociophonological scenario of backgrounding processes, e.g., loss of intervocalic /h/ in Ardenne Liegeois (cf. Labov 1994: 444) and, context-free, in Southern British English (cf. Fonagy 1957: 208); fusion of/n j/ in union, panier —> palatal [ji] in colloquial French (cf. Fonagy 1957: 209); consonantcluster simplifications collected from many languages by Fonagy (1957); fricativization of Panama Spanish > sh studied by Henrietta Cedergren in 1969 and restudied in 1982 (cf. Labov 1994: 94 ff); gliding of the palatal lateral in Charmey French as observed by L. Gauchat and E. Hermann (cf. Romaine 1989: 200 f).21 An example of distant, less-documentable change I want to refer to is the loss of final -s and -m in Latin (after Dressier 1973, cf. Joseph and Wallace 1992), as in: (2)

Lat. amico/us and amico/um > It. amico vs. Lat. quern > Sp. quien

116

Wolfgang U. Dressler

This change can be studied both in real time and in apparent time (by comparing stylistic levels of Old Latin documents). Puristic Classical Latin curbs on the spread of this change represent a prototypical case of Standard resistance against terminal sociolectal and dialectal spread (cf. Trudgill 1993). One might object that at least some of these cases might be explained by Ohala's (1992: 324ff, 338f; 1993: 246ff, 259 ff) perceptual model of hypo-correction, but as long as this phonetic model is not elaborated into a full-fledged scenario, it is difficult to confront; in the long run the two approaches may prove to be largely intertranslatable. Our scenario of backgrounding processes may be assumed as a default scenario for those phonological changes that 1) correspond to synchronic backgrounding processes, 2) take considerable time for completion, particularly if there are traces of variation, and 3) do not co-occur with social or demographic counterindications. Functional explanation applies first of all to backgrounding processes as serving the function of phonology to be easily pronounceable, second to the preference for backgrounding processes in informal speech situations (cf. Dressier 1984); third, the generalization of optional processes can be explained as due to the parameters of biuniqueness and indexicality (cf. Dressier 1984: 43), insofar as an obligatory process results in less ambiguity and thus represents a more reliable sign, and more reliable verbal signs are necessary when a change spreads from informal to formal speech;22 fourth, the direction of generalization from informal to formal speech situations finds its explanation in social change (see above). Since casual speech is the most usual way of speech and abounds in backgrounding processes, our scenario should account for a large portion of phonological change. Such diachronic changes are, e.g., consonant lenitions in Romance and Celtic languages (cf. Back 1991: 117ff, 129ff); Latin monophthongization of /ai, ei, oi, au, eu, eo/ (cf. Back 1991: 147ff; Joseph and Wallace 1992); Ancient Greek palatalizations (cf. Brixhe 1979: 24 Iff); various consonant assimilations (cf. Back 1991: 182ff); several changes involved in Forner et al.'s (1992: 86 ff) list of "rapid speech" originated changes (although they deal with diachronic correspondences rather than with actual changes).

"Scenario " as a concept for the functional explanation

117

3.2. Phonological part of a scenario: natural foregrounding processes In contrast to backgrounding processes, natural foregrounding processes play a very restricted role in sociophonology, e.g. in emphasis, in motherese, in superlento speech. Neither have I observed any ongoing change involving the generalization of such processes nor am I confident in reanalysing cases reported in sociolinguistic literature in these terms. But good candidates are, for example, raising of long vowels, identified as fortitions by Donegan (1978), attested in emphatic motherese (cf. Locke 1993: 203) and studied in its diffusion (in chain shifts) by Labov (1994: 31, 116, 122ff, 145if, 176ff, 285ff); polarization of diphthongs, i.e., the falling of the nuclei of upgliding diphthongs (cf. Labov 1994: 116, 122, 124f, 144f, 167f, 285ff); fortition of Ancient Greek word-initial /j/ > = [dz] as in zygon 'yoke' (Brixhe 1979: 249ff). The phonological part is often clear enough, as in diphthongization of stressed or long high vowels (cf. Andersen 1972; Donegan 1978) or in the fortition of word-initial consonants (cf. Dressier 1980) as in: (3)

Lat. iam, iuvenis > It. gia, giova/ine, Fr. (da)ja, jeune; Lat. iugum > aragon. chugo (Back 1991: 136fF)

For similar Middle Iranian fortitions cf. Back (1981); cf. Andersen's (1988: 62ff) study on parasitic consonants whose origin he traces to phonostylistic variation. What is often offered as examples of diachronic fortition processes (e.g., in TrudgilFs (1993) "non-natural sound change") are mere sound correspondences difficult to identify with synchronic foregrounding processes.

3.3. Scenario of phonological replacement due to perceptual reanalysis Some phonological changes imputed to the above scenario 3.2 may, instead, go back to perceptual reanalysis in language transmission. Many cases have been studied notably by Ohala (e.g. 1992: 317ff and

118

Wolfgang U. Dressler

1993: 239ff with references), for example regular dissimilation (e.g. 1992: 328ff; 1993: 249ff). In these changes, the speech signal becomes more distinctive due to perceptual reanalysis, in line with the general perceptual goal of phonology, viz., of optimizing distinctiveness of the phonological input (phonemes as sound intentions).23 The locus of improving perceptual distinctiveness is either the generalization of synchronic foregrounding processes, as in the previous scenario 3.2, or perceptual mis-analysis (3.3). Mis-analysis is likely to occur on a word-for-word basis, as in the case of χ >/in E. laughter vs. G. lach-en or sporadic 0>/in dialectal English (e.g., [flrj] < thing, cf. Ohala 1993: 242) and similar non-gradual consonant changes (cf. Labov 1994: 539ff) or in spontaneous nasalization in Indo-Aryan languages (cf. Ohala 1993: 240f). Such mis-analysis is most probable to occur with speakers who do not fully master phonological norms. Thus we may assume two alternative scenarios: a) mis-analysis by young children in child-adult interaction, which may result in phonological change; b) mis-analysis when learning words of an imperfectly mastered foreign language or dialect, i.e. contact-induced change, which I will discuss in 3.5. Faber (1992) imputes reanalysis to the interplay of interindividual articulatory variation and categorical speech perception. Perceptive distinctiveness comes in with her remark (p. 68) that variants must be salient in order to spread.

3.4. Scenarios of hyperaccommodation Although Ohala (1992) assigns his cases of scenario 3.3 to "hypercorrection", the term hypercorrection should better be reserved to cases which coincide as much as possible with the traditional use of the term hypercorrection (cf. Hock 1986: 9.3 p. 11 Of; Orioles 1989). Yaeger-Dror (1992) has shown convincingly that hypercorrection is just one type of what she calls "hyperaccommodation", defined as "shift toward or away from a given speech variety ... creating linguistic patterns which overshoot the speech in that target variety" (p. 184, cf. Giles and Williams 1992: 343, 345). Socially, hyperaccommodation is "a form of selfconscious/insecure behavior" (p. 188). Of this scenario, hypercorrection is a subscenario defined as conscious change, in contrast to other, unconscious hyperaccommodative changes.

"Scenario " as a concept for the functional explanation

119

The locus of unconscious hyperaccommodative changes seems to be with adolescents or adults and neither in the highest nor in the lowest social stratum (cf. Janda and Auger 1992: 199), but particularly in those strata of society which are most insecure or self-conscious about their speech norms (Yaeger-Dror 1992: 188). Hypercorrection occurs also with children, when they have become aware of sociolinguistic differentiation, e.g. (4)

Vienn. G. Lobe stoolke > stooltje and steultje > stooltje. The highest scores are 40.1 in 65 for steulke, 30.5 in 55 for steultje, 17.9 in 25 for stoolke and, unexpectedly, 44.3 in 45 for stooltje (see above). We can ignore forms with medial schwa. In the case of steulke we have to take into account a possible hyperdialectic reaction of the middle-aged informants (compare the high score of 35.6 in 35).

Category 6 Category 6: the root word ends in r (but not in a short vowel plus r). Here, also, it is useful to assume two old forms: höärke and höärtje for hoar. Thus, we expect forms with umlaut and they-suffix. Ten words were tested (four with r following schwa in weakly stressed syllables): hoar 'hair', poar 'pair', oar 'ear', boor 'drill', snoar 'string', koar 'car', akker 'field', dokter 'doctor', bakker 'baker', koffer 'trunk, box'. See table 6. Table 6 (Category 6)

65 55 45 35 25 0

höärke

höärtje

hoarke

hoartje

köäreke

koareke

koaretje

40.3 27.5 30 25.8 16.1 33.1 22.8

1.7 1 0.9 0.6 5.4 1.3 2.6

2.8 2.7 3.3 8.8 10.1 2.9 8.2

46 53.8 54.5 50.9 56.4 50.8 53.8

7.4 14.3 9.4 11.9 8.7 10.6 10.1

1.7 0 0.5 1.3 2 0.8 1.4

0.6 0.5 1.4 0.6 1.3 0.7 1

162

Cor van Bree

The forms höärtje, hoarke, höärke and hoartje occur to a small, reasonable, large, and very large extent respectively. The low score of höärtje stands out. We might assume that only höärke may be considered an old form. But then höärtje should not occur, and the results are contradictory in this respect. Perhaps it is best to ascribe höärtje to an old increase of the y'-suffix. But it is still striking that the form increases from old to young. If we look closely at the age groups, however, we see that the form decreases from 65 to 35, as we would expect. The percentage of 5.4 for 25 is based on eight forms, of which four are occurrences of höärtje. Was it significant that here one often speaks of höäre (in the plural)? Discounting these four forms, we arrive at a percentage of 2.7. An increase from 0.8 (35) to 5.4 (25), however, is within the limit of 5%. Höärke decreases as expected from old to young, without a clear hyperdialectic increase in the middle-aged groups. Hoarke and hoartje increase from old to young. Hoartje already exhibits a high score in 65; compare the score of stooltje in cat. 5, namely 29%, which, however, is considerably lower. Should we think once again of an early loss of the umlaut in a form which already has the y'-suffix? The effect could have been so strong in this case that it replaced höärtje (see above). The number of forms with medial schwa also stands out: 102 in total, of which 60 for koar. This is understandable since this word also has a medial schwa in the standard language, e.g., karretje. In hoar and oor we also find a medial schwa fairly often: 28 and 13 times respectively. The form köäretje with umlaut and a y-suffix does not occur at all. In the absence of köärtje we must assume the following development for koar: köärke > köäreke or koarke (produced) > koareke > koaretje. Here, also, it would seem that schwainsertion can occur before disappearance of the umlaut (compare tungeke in category 3). Forms with s only occur four times, for poar (2x), snoar, and dokter. The old form höärke does not hold a strong position. Further lexical analysis reveals that forms without umlaut are infrequent in the case of hoar, oor and koar, but do occur frequently for akker, dokter, bakker and haar (see 4.3). We conclude: the two lines of development in the direction of the standard language are höärke > hoarke > hoartje and höärtje > hoartje. Höärtje, however, is a problematic form (see above). The highest scores are: 40.3 in 65 for höärke, 10.1 in 25 for hoarke, 56.4 in 25 for hoartje and, unexpectedly (see above), 5.4 in 25 for höärtje. Forms with medial schwa are rather frequent, especially in the case of koar.

Changes in diminutive formation in the eastern Dutch dialect ofTwente

163

Category 7

Category 7: the root word ends in an underlying nd. Here, we assume three old forms, namely tendke, tendeke (with old schwa-insertion: see 2-3) and tendje (with an old change from -k to -j) for tand. The following seven words were tested: tand 'tooth', hand 'hand', land 'land', mand 'basket',20 hand 'dog', mond 'mouth', pond 'pound'. See table 7. The forms tendke and tendje are present to a reasonable extent and tendeke and tandje to a large extent. The scores for tandeke and tandke are, however, low. Tendetje and tandetje are absent. Significantly, tendke increases to 35 after a drop in 55 and then decreases in 25. Globally, it even increases from old to young. Tendeke, which still scores highly in 55, decreases gradually. The intermediate forms tandeke and tandke and the standard form tandje increase. The relatively high score for tandje in 65 is again striking (cf. stooltje in category 5 and hoartje in category 6). Tendje increases slightly, but the difference between young and old is minimal. There is a peak at 55 or, more generally, in the middle-aged groups. We might outline the development in the following way. The old form tendke (65: 14.6%) loses ground, especially to tendje (55: 11.2%); 55 shows an old increase of this form. The development continues further to tandje (highest score in 25: 62). Tendeke, which may also be considered an old form (65: 55.4%), remains stable up to and including 55 (55: 53.7%). It develops without schwa according to the standard language via tendke (35: 19.1%), tandke (25: 3.7%) and tandje (25:

Table 7 (Category 7)

65 55 45 35 25 0 Υ

tendke

tendeke

tandeke

tandje

tendje

tandke

14.6 3.7 14.9 19.1 10.2 10.3 14.7

55.4 53.7 34.4 31.8 13 50.5 24.8

3 0 0 0 6.5 1.2 2.6

22.3 31.3 39.6 36.4 62 29.4 47.3

3.8 11.2 9 10.9 4.6 7.8 8

0.8 0 1.9 1.8 3.7 0.7 2.6

164

Cor van Bree

62 %).21 Thus, tendke can be an old form and an intermediate form as well. Another minor development is from tendeke via tandeke (25: 6.5%) to tandje. The schwa may be lost before the umlaut (compare the form tendke and compare also krelke in cat. 1). A form with s, pundske, occurs only once in 35. Hundke and hundeke occur with comparatively greater frequency for hand, and landje occurs more frequently for land. The old forms together (tendeke, tendje, tendke) do not hold a very firm postion now. We can ignore forms with umlaut plus schwa plus j (tendetje) and without umlaut plus schwa plusy (tandetje).

Category 8 Category 8: the root word ends in a t or an underlying d. Here, we can also expect forms with umlaut not only with k but also withy (after a dental): hood — heudke, heudje as the old sources show. The thirteen words which were tested are: hood 'hat', brood (or stoef) 'loaf, (boter)vloot 'butter dish', voot 'foot', blad 'leaf, bot 'bone',poi 'pot', slot 'lock', droad 'thread', noad 'seam', plant 'plant', krant 'newspaper', kont 'arse'. See table 8. The forms heudke, heudje, and hoodje occur frequently. Hoodke does not occur often, but clearly most often in 25. A form with schwa occurs nineteen times for blad, three times for droad and four times for kont. Forms with s occur nine times across the range. Contrary to what we would expect, heudke shows a slight increase from old to young, but this is largely due to the low score in 55. In any case, 25 scores lower

Table 8 (Category 8)

65 55 45 35 25 Ο Υ

heudke

heudje

hoodke

hoodje

heudeke

34 15.6 22.6 30.4 25.3 24.4 26.8

40.4 56.9 48.3 44.4 29.4 48.6 39.2

1.6 0.4 1.4 1.4 4.6 1.1 2.7

22.4 25.3 27.4 21.5 38.7 24.6 29.6

1.6 1.9 0.3 2.3 2.1 1.5 1.8

Changes in diminutive formation in the eastern Dutch dialect ofTwente

165

than 65. Mark the peak in 35: 30.4. Heudje increases from 65 to 55 and decreases thereafter to 25. Globally, it decreases from old to young. Hoodke and hoodje increase from old to young. The relatively high score of hoodje is again striking in 65 (cf. stooltje, hoartje and tandje in categories 5, 6 and 7). Striking is also the decrease of hoodje in 35. The results indicate a development analogous to that of category 5. The high score of heudje in 55 indicates an old increase at the expense of heudke. However, in the middle-aged groups, especially 35, a hyperdialectic response is evident as a result of which heudke increases again. This response is also at the expense of hoodje. The development is from heudke (highest score in 65:34) via hoodke (highest score in 25: 4.6) to hoodje (highest score in 25: 38.7) or from heudje (highest score in 55: 56.9) to hoodje. The forms with umlaut are still prevalent. For kont, plant, and krant, forms without umlaut (and with a y-suffix), occur often. For blad, the form with umlaut and the ^-suffix occurs with comparative frequency (see further 4.3). Forms with medial schwa occur very infrequently. Category 9 Category 9: the root word ends with n (but not a short vowel plus n). On the basis of the old sources we expect forms with umlaut plus k or j (after a dental): haan — heenke, heentje. The ten words investigated are: haan 'cock', baan 'job', zwaan 'swan', kraan 'tap', moan 'moon', troan 'tear', boon 'bean', kroon 'crown', loon 'wages' and toon 'tone'. See table 9.

Table 9 (Category 9) heenke 65 55 45 35 25 0 Υ

23.8 10.8 16.9 25.5 12.6 17.2 18.6

heentje

37.6 44.6 33.8 30.3 15.9 39.6 25.2

haanke

haantje

heneke

haneke

2.6 0 2.2 1.2 7.3 1.5 3.8

36 42.6 44.9 40 59.6 40.4 48.8

0 1.5 2.2 3 4.6 1 3.5

0 0.5 0 0 0 0.2 0

166

Cor van Bree

The forms heenke, heentje and haantje occur frequently or reasonably frequently. Haanke, in comparison, is infrequent but clearly present in 25, scoring 7.3%. Forms with schwa occur 21 times, of which 11 times for maan and six times for haan. Forms with s occur occasionally (8x), especially in 25 (7x). Once again, it is conspicuous that the form with umlaut plus the &-suffix increases slightly from old to young. Between 65 and 25, however, there is a marked difference. The increase is attributable to the low score in 55 and the relatively high score in 35. Heentje shows a clear decrease from old to young. As far as the age groups are concerned, first there is an increase in 55 and then a decrease to 25. Note the analogy with categories 5 and 8 for both forms, i.e., steulke, steultje; heudke, heudje. Here, also, we can think of an old increase of heentje in view of its peak at 55 and of a hyperdialectic use of heenke in the middle-aged groups. This hyperdialectic use is also at the expense of haantje: mark the (slight) decrease in 35. This form and haanke both increase from old to young. Haantje already has a notably high score for 65; there is a correspondence in this respect with stooltje, hoartje, tandje and hoodje in categories 5, 6, 7 and 8. The old forms together do not hold a strong position any more. Forms without umlaut occur for baan and zwaan in particular and also for loon and haan (see 4.3 for possible explanations). Thus, the two developments are: heenke (high score in 65: 23.8, but the hyperdialectic score in 35 is slightly higher: 25.5!) > haanke (highest score in 25: 7.3) > haantje (highest score in 25: 59.6); heentje (highest score in 55: 44.6) > haantje. Forms with medial schwa are very unfrequent.

Category 10 Category 10: the root word ends in a vowel (but not a schwa). There were only three words investigated: stro 'straw', koe 'cow', vlo 'flea'. See table 10. We consider streuke the old form (with umlaut and &-suffix). The forms strootje and streuke are well to very well represented. Stroke, however, scores low. Streutje is present to a reasonable extent and this already applies in the 65 group. There are three forms with s (2 for vlo and 1 for stro). Streuke decreases from old to young (with a significant increase in 35), stroke, strootje and also streutje increase

Changes in diminutive formation in the eastern Dutch dialect ofTwente

167

Table 10 (Category 10)

65 55 45 35 25 0

streuke

stroke

strootje

streutje

77.1 60.5 51 61.8 16.7 65.2 41.6

0 0 2.1 2.9 13.9 0.4 7.1

14.3 26.3 36.2 20.6 55.6 23.5 37.7

8.6 13.2 10.6 14.7 13.9 10.8 13.6

(strootje with a significant decrease in 35). Strootje, the form with the y-suffix and without umlaut, already has a relatively high score in 65 (compare this with stooltje, hoartje, tandje, hoodje, and haantje in the categories 5 to 9). We should probably consider both streuke and streutje to be old forms (compare the percentages for 65), although a slight increase of streutje in the direction of the young is curious. As far as this is concerned, the form reminds us of höärtje (cat. 6). The peak of streuke in 35 can be an indication of a hyperdialectic reaction; the decrease of strootje for this age group shows that this reaction was also at the expense of this form (compare the scores of stooltje, hoodje and haantje in the categories 5, 8 and 9). Thus, we assume the following developments: streuke (highest score in 65: 77.1) > stroke (highest score in 25: 13.9) > strootje (highest score in 25: 55.6); streutje (high score in 55: 13.2, but the highest score in 35: 14.7, is rather remarkable) > strootje. We should keep in mind, however, that the conclusions are based on only three words. 4.3. Individual words In some categories, we have referred to individual words which behave in a somewhat deviant way. This was especially the case with words in which conspicuously few umlauts occurred. In category 4, it concerned haas and also, to a lesser extent, gans and tas; in category 5 vaandel, dadel, zadel, appel, and schaal; in category 6 akker, dokter, bakker, and poar; in category 8 kont, plant, and krant and in category 9 baan and zwaan and to a lesser extent also loon and haan. How can these dif-

168

Cor van Bree

ferences be explained? The occurrence of schwa plus / or r is not a clear factor. For example, in schaal the umlaut is often absent, while this is not the case for vogel. Neither is the nature of the vowel (a, aa) a clear factor: there is no umlaut in dokter, kont, and loon either, while this does not apply to other words with a or aa. Another possibility is that this involves words which are not readily used in the dialect. But it does not apply, for example, to a word such as appel. Finally, it is a consideration that we are dealing here with words which a Twent is not very likely to put into the diminutive, and if he does then he does so using the standard language form. Whatever the explanation may be, the character of individual words can clearly play a role.

5. Conclusions and final remarks 5.1. Comparison between young and old The first hypothesis (see 3) is that there will be a decrease in dialect forms from young to old. If we take category by category all the old forms which we have encountered, there is no one category which forms an exception: young scores lower than old in each case. On the other hand, the forms corresponding to the standard language always increase from old to young. As regards the intermediate forms, the hypothesis was that they will increase or decrease from old to young, or first increase and then decrease. As far as I can see, there is no clear exception in this respect. Moreover, if we look at the umlaut and take not only the old, but also the hyperdialectic and intermediate forms for each category, we observe that young consistently scores lower. If we look at the &-suffix in the same way, young appears to score lower each time with the exception of three categories: 5, 8, and 9. For categories 8 and 9, however, 25 scores lower than 65, and for category 5 both of these age groups show an equally high percentage. Categories 5, 8, and 9 are categories in which we thought we could observe a hyperdialectic increase of the &-forms in the middle-aged groups. As far as the occurrence or non-occurrence of a medial schwa is concerned, it can be established for all categories that young dialect speakers behave more in accordance with the standard language: categories 1 and 7 (tendke can also be an old form), tungeke, etc. (cat. 3), höäreke, etc. (cat. 6). This can also be established for the absence of the medial s in category 3. We can draw

Changes in diminutive formation in the eastern Dutch dialect of Twente

169

the general conclusion that the young respond with fewer dialect forms than the old.22

5.2. Development per category In this section I will discuss the explorative aspects of the investigation, concerning the interdependence of part changes within forms. All the developments which we have established are given below. A double arrow divides the old developments from the more recent developments which have taken place under the influence of the standard language. Alternative developments are given after the slash. Forms for which a temporary hyperdialectic increase has been observed appear in italics. (In all cases more direct developments are also possible.)23 cat. 1. kenke > kenneke ^> kanneke > kannetje cat. 2. köpke > kopke > kopje cat. 3. stökke > stökske > stökke > stokke > stokje / stökke > stökske §> stökske > stokke > stokje cat. 3. tungke > tungske §> tungke > tungeke > tongeke > tongetje / tungke > tungske > tungke > tongke > tongeke > tongetje / tungke > tungske ^> tongske > tongke > tongeke > tongetje cat. 4. reuske §> rooske > roosje cat. 5. steulke > stoolke > stooltje / steulke > steultje > stooltje cat. 6. höärke > hoarke > hoartje / höärke > höärtje (?) > hoartje cat. 6. köärke §> köäreke > koareke > koaretje / köärke ^> koarke > koareke > koaretje cat. 7. tendke > tendeke > tendke > tandke > tandje / tendke > tendeke ^> tandeke > tandje / tendke > tendje ^> tandje cat. 8. heudke > hoodke > hoodje / heudke > heudje §> hoodje cat. 9. heenke > haanke > haantje / heenke > heentje > haantje cat. 10 streuke ^> stroke > strootje / streuke > streutje > strootje24 The following developments are remarkable: l. Stokke is not only an old form, but also an intermediate form which has developed via the unexpected form stokke in the direction of stokje. Thus, the s may be lost before the umlaut. However, there is also the possibility that this occurs in the opposite direction.25

170

Cor van Bree

2. An intermediate form, tungeke, with umlaut and medial schwa is also possible. Compare this also with köäreke.26 3. In 55 an old increase of steultje, heudje and heentje (and also, less clearly, of tendje and streutje) seemed to emerge. 4. In the middle-aged groups there is (probably) hyperdialectic use of steulke, heudke, heenke, and streuke. 5. Höärtje (old form?) occurs infrequently. Moreover, the high score of hoartje already evident in 65 is striking. Although less striking, high scores are also displayed in 65 by stooltje, tandje, hoodje, haantje, and strootje. It would seem that in a form which already had the y'-suff ix in the old dialect, we can speak of an early disappearance of the umlaut. 6. Tendke should be interpreted both as an intermediate form and an old form. The schwa may also be deleted before the umlaut. Compare also krelke in category 1. 7. Streutje should also be considered an old form. 8. A medial s also occurs hyperdialectically following other consonants as well as velars. Overall, unless there is already an old form with umlaut plus they-suffix, development occurs in such a way that first the umlaut disappears and then the ^-suffix gives way to the y-suffix. Thus, there is a clear implicational hierarchy. Loss of s and loss or occurrence of schwa may precede disappearance of the umlaut. But the reverse is also possible. In this case there is no clear implicational hierarchy. In each case, however, the development can also be direct, e.g., from kenneke to kannetje or from stökske to stokke. An interesting question is the relationship of the various categories to each other. Is it also possible to distinguish implicational hierarchies in the sense that someone who uses kopke (cat. 2), for example, will also use kanneke (cat. 1), while the reverse does not necessarily apply? (Bickerton (1973) would say that the principle of sub-categorization applies in this case.) In order to answer this question, a fresh analysis of the material would be necessary. Here we limit ourselves to a global comparison of the categories. We can establish that the categories 1, 2, 3 and 4 (belke, köpke, stökke, reuske) are the most conservative, with 57.9, 90, 64.1 and 71.9% old forms used by young; contrary to our expectations, category 8 (heudke, etc.) also belongs to this group, with 66% old forms used by young. The other categories show a much

Changes in diminutive formation in the eastern Dutch dialect ofTwente

171

sharper decrease in the dialect forms. It is a striking fact that the sharpest decrease is shown by category 6 (höärke etc.). The percentages are: 38.8 (category 5: steulke etc.), 21.5 (category 6), 27.8 (category 7: tendke etc.), 43.8 (category 9: heenke etc.) and 30.6 (category 10: streuke etc.). The umlaut seems to disappear rapidly in the position before r (cat. 6) but maintains itself fairly well before the dentals t/d (cat. 8).

5.3. Final remarks An intriguing question is how to explain the order of loss we have observed. This is the question of the stability of language elements and language characteristics. In answering it, we need to delve into the psyche of the language user and learner and the situation in which he uses and acquires his language, which requires an alternative approach to the one we have used so far in our contribution. Up to now, we have thought in terms of forms which change: the umlaut may disappear, the ^-suffix may give way to the y'-suffix, etc. In fact, it is not the forms which change but the language users and language learners who, albeit unintentionally, change the language by means of an invisible-hand process (see Keller 1990). They use different forms to those they used previously, or they acquire different forms. These forms can differ from the previous ones in one feature or in a number of features at once. Between the two groups — language users and language learners — there is no watertight division: the language learner acquires his language by using it, by using it the language user is in a process of acquiring his language. There are two situations from which we can depart. In the first situation a Twente child, entirely or partly on the basis of the standard language, must try to learn the language without sufficient correction on the part of his elders (imperfect learning). Which dialect elements and characteristics does he learn easily on the basis of the imperfect language input, which with difficulty? The problem is separating the various factors involved and weighing them up against each other. Compared with the &-suffix, the umlaut has the advantage of occurring in stressed syllables (and there is also a striking difference with the standard language27), but there are three disadvantages. First, the umlaut is not such a simple phenomenon: there are a number of vowel alterna-

172

Cor van Bree

tions (seven to be exact) for which there is a different umlaut vowel for each root vowel. There are only two alternatives for the suffix, -k or -j, -j being optional after certain consonants: t, d, n, l and possibly r as well. Second, the umlaut has the disadvantage of occurring less frequently. The suffix occurs in all diminutives, including those where the umlaut cannot occur, e.g., beek - beekske. Third, analogy (generalization) of the suffix works in favor of the A:-forms — one is always correct if one uses this form in all cases — while it does not work in favor of the umlaut by precluding analogous enlargement to cases without umlaut. The advantages of the suffix are apparently so great that they more than compensate for the disadvantage. This would explain why the umlaut is learned less well than the -kl-j suffix. The disadvantage of occurrence in non-accented syllables applies to both medial s and medial schwa. A marginal advantage of medial schwa is that an extra syllable is created as a consequence. The disadvantage is not clearly compensated for in these cases. Medial s only occurs after k or g, which is relatively straightforward, and there is a large difference with the standard language where this s does not occur. However, it has the disadvantage that it is fairly rare. As far as medial schwa is concerned, there is closer correspondence with the standard language, and the rules are also more complex: medial schwa occurs in the dialect after a short vowel plus n, m, l and r (but not after ng as in Standard Dutch) and also following nd (again contrary to Standard Dutch) (we will leave aside infrequent cases such as kral, kar). Medial schwa has also the disadvantage that it is fairly rare. Frequency is probably the factor which weighs most heavily. As a consequence, medial s and medial schwa are not observably more stable than umlaut. The second situation from which we can approach the question is that in which young dialect speakers at regional schools or adult speakers at work come into contact with speakers from other places. In this situation, regiolect forming (see Hoppenbrouwers 1982, 192ff) can occur, a particular form of regional diffusion in which marked dialect characteristics (Schirmunsky 1930: primary dialect characteristics) can be suppressed. In the long run these characteristics can disappear in the language use of certain individuals. This can also lead to imperfect dialect input for the children: cf. situation 1. The question is whether this situation is indeed of such great influence. All the characteristics mentioned are widespread anyway and not bound to specific locations. As far as we can speak of influence, we think especially of suppression

Changes in diminutive formation in the eastern Dutch dialect ofTwente

173

of the umlaut. As a sign of the accented syllable, this is most obvious in the interaction of young people or adults. For the umlaut, therefore, two sorts of changes (cf. 1) can very well converge: the influence of standard language and the influence of neighboring dialects. To summarize: older people either avoid to speak dialect to the children or suppress striking forms in their dialect. Of course, they have their reasons for it: the children should learn the standard language or they themselves do not want to attract attention by deviant language use. In any case, the social or communicative functions of language are here involved. It is not the older generation's intention that the dialect should disappear or change. However, their behavior leads, via an invisiblehand process, to a defective dialect input and a decrease of correction for the children, and this causes imperfect dialect acquisition. That the dialect is still acquired, means that the children have their reasons for acquiring it, for example because of their wish to communicate with the peer group (the other children in the street). Because of frequency of elements and other factors they will learn some elements and aspects sooner than others. The whole process is an interplay of factors inherent to language in general or a special variety (frequency of elements etc.) and human intentions and behavior. The older people create unintentionally conditions for change, the change itself should be located in the new generations. A number of the factors mentioned above manifested themselves in previous research: see Van Bree 1990, Van Bree 1992 (also for theoretical considerations) and Van Bree 1994. These and other factors are an object for further research with different data (e.g., more spontaneously produced data)28, other varieties and other phenomena. In any case, I am convinced that we can only obtain more insight into the regularities which I have uncovered by immersing ourselves in the psyche of the language user and language learner and his reaction to various situations. What is perhaps the most interesting result of the research is that the development does not take place in a chaotic manner, but in a way that is clearly characterized by regularities. Notes 1. In this research, which I have been carrying out since 1981, I was assisted by Rob van der Zalm (Amsterdam) and by groups of Dutch and German students (from Leiden and Bonn respectively). In the final preparation of

174

2.

3.

4. 5. 6.

7.

8.

9.

10.

Cor van Bree the material, the students Eric Prillwitz and Auke Woldhuis together with Rob van der Zalm also lent their assistance. The original Dutch text was translated by Drs. Jacqueline Burger (Leiden). I thank all those persons I have mentioned and also those I have not, as well as all the informants, for their co-operation. (In 1992 additional research was carried out with the help of a group of students from Leiden in Weetselo, Twente. The results of this research were comparable to the results accounted for in this paper.) An attempt was made to keep gender constant by only sending the questionnaire to males. This was not entirely successful: the informants include six females. At the time when they completed the questionnaire, most informants were beween 60 and 70 years of age. Three were over 80 and, as far as we know, 5 were between 50 and 60. No differentiation was made with respect to social position, the criterion being that an informant claimed to speak the dialect regularly. Another rule is that (as in Standard Dutch) a final obstruent is always voiceless immediately before they of the suffix. Thus, vlag 'flag' with underlying voiced velar fricative should have the diminutive form vlag/e with a voiceless fricative. Of course, this is also the case before k or sk in the suffix, e.g., vleg(s)ke with voiceless g. A special case is a short vowel plus b: slab(be) 'bib'. A special case is a short vowel plus g: vlag(ge) (see note 3). As a consequence of Dutch spelling conventions, an underlying voiced consonant is not always reflected in the spelling: compare (plur. heude) with (plur. rozen). In both cases final devoicing takes place: hoot, roos. For regional variations with final schwa such as kanne and sparre we assume that, according to the general rule already mentioned, a schwa is first deleted and subsequently reinserted. We make the same assumption for vlagge and slabbe (cf. vleggeke and slebbeke). See Nijen Twilhaar 1990: 26 for this rule. The variation also occurs after a short vowel plus g or b: compare (vlag-) vleggeke with vleg(s)ke and (slab-) slebbeke with slebke. In Standard Dutch we also find this variation: vlagje, vlaggetje; slabje, slabbetje. We can ignore the position immediately after a short vowel plus m, n, l or r (lam - lemke) because this position is restricted to the eastern part of Twente. (It always implies the ^-suffix.) The written questionnaire was also completed by a number of young people in the area indicated: 31, including 7 women, mainly between 20 and 30 years of age (7 were between 30 and 40 years of age). With the exception of one, they all spoke the dialect on a regular basis. From the research, it is clear that the changes mentioned, including ^-insertion, have gained ground.

Changes in diminutive formation in the eastern Dutch dialect of Twente

175

11. Indications for implicational hierarchies (disappearance of umlaut before adaptation of the suffix) can be found in Van Bree 1986 (concerning noun plurals in the dialect of Twente), Hoppenbrouwers 1990: 99 (concerning diminutive forming in Brabant dialects) and Vousten 1995: 5.2.3.3 and 6.3.2 (concerning diminutive forming in the Limburg dialect of the region of Venray). 12. A few examples follow. Originally categories 2 and 3 were combined into one category with the view that labials and velars could be considered the same element. Because s is often inserted after a velar and not after a labial, the category was later subdivided. An example of a word wrongly categorized is tong 'tongue', which was first assigned to cat. 1 (cf. Stand. Du. tongetje} but for old Twents had to be transferred into cat. 3 (cf. tung(s)ke without the possibility of medial schwa). The historical explanation for the Twente form is that the velar nasal in tong originated in a combination of this sort of nasal with a voiced velar occlusive. It was difficult to find sufficient examples for category 10. 13. Sometimes the number of unusable responses was high. There are various possible explanations for this. 1. The informant assumed a totally different word, e.g., bees(t) instead of koo 'cow' (stoet instead of brood 'loaf was accepted because both words fall under the same category). 2. The informant assumed a deviant form, e.g., veut 'foot' instead of voot. In the singular, veut has been influenced by the umlaut vowel of the plural. An umlaut vowel, at least a palatal vowel, was often used in the singular as well (cf. also duuf 'pigeon' instead of doef). Another example is the old form taske 'bag' which was occasionally given instead of tas. 3. The informant did not respond with a diminutive. Informants often claimed that a certain diminutive is absent in the Twente dialect, e.g., for vlo: "a flea is already small". It would seem - and informants even confirm this themselves - that Twents use fewer diminutives than people from the western Netherlands. Forms such as doeve - duifke and doem 'thumb' - duimke were also unusable. If an informant responded with more than one form, the form thought to be the oldest was taken, e.g., krelke (and not krelleke or kreltje) for kral 'bead'. Some forms ending in -ie have been interpreted as forms having they-suffix. 14. See Vousten 1995: 5.2.3.3 for similar forms with unexpected medial s. 15. A complete survey of the scores, including each separate word, is available. Address of the author: Vakgroep Nederlands, Rijksuniversiteit Leiden, P.O. Box 9515, NL-2300 RA Leiden. 16. Short vowel plus b is also excluded. For this case the word slab was tested. Here Standard Dutch has both a form with schwa and a form without, i.e., slabje, slabbetje. As the results are very unclear with respect to this word, I will only mention the forms produced: slebke, slebbeke, slabbeke, slabbetje, slabje.

176

Cor van Bree

17. Short vowel plus g is also excluded. Vlag was tested for this. In Standard Dutch both a form with or without schwa may occur, i.e., vlagje, vlaggetje. As the results with respect to this word are not very clear, I will limit myself to a summary of the forms produced: vlegke, vlegske, vleggeke, vleggeske, vlagske, vlaggeke, vlagje, vlaggetje. The occurrence twice of vleggeske with schwa and s is interesting. 18. Another interpretation of stökke can be that this is a hyperdialectic form of the middle-aged groups. Similar forms occur in categories 5, 8, 9, and 10. Of course, this interpretation implies that there is no connection with hyperdialectic s. The difficulty, however, is that a hyperdialectic stökske, the form most deviant from the standard language, would be more probable. See further notes. 19. We already find a preference for they-suffix after g in Wanink 1948. 20. In the case of mand, it is doubtful whether the word belongs in this category. In the singular (/-deletion has also taken place, with the result that this word appears as mane or (with schwa-deletion) maan. Thus, category 9 is also possible. As far as the frequent occurrence of medial schwa is concerned, however, this word behaves as a member of category 7. 21. Tendke of the middle-aged groups might also be a hyperdialectic form, but a hyperdialectic occurrence oftendeke, the form most deviant from Standard Dutch, would be more probable. On the other hand, it is interesting to see the same pattern as in category 5: steulke (and in 8, 9, and 10: heudke, heenke, streuke): an increase to 35 and a decrease of the standard language form in this age group. This can lead us to assume a similar explanation in all cases. Compare also stökke in cat. 3 (see note 18). 22. If we look at the relationship between k and j in terms of the consonant immediately preceding it, the impression we get for 65 is the same as that which we get on the basis of the written questionnaire (see 2). If for the 65 group we divide the percentage of forms with k by those with j, we get results lower than 1 for d/t (0.6), n (0.8), / (0.9) and r (0.9). In between are the positions following a vowel (3.4) and schwa (4.7), followed by the velar (9.4), the s (13) and, with a very high result, the labial (70.4). For 25 we get a result lower than 1 for d/t (0.3), n (0.3), r (0.4), a vowel also (0.4) and the 1 (0.9). Slightly higher is the schwa (1.5) followed by the s (2.6), the velar (4.9) and the labial (12.3). By and large, therefore, we find the old situation despite all the changes in 25. 23. Stökke and tendke are interpreted as mesolectic, not as hyperdialectic forms. See note 8 and 21. 24. The developments given here do not imply that the ultimate outcome is the standard language. It is possible that a new form of dialect (or regiolect) comes into being, e.g., containing diminutives without umlaut but with the &-suffix. Or more realistically, that a (dialectal or standard-dialectal) conti-

Changes in diminutive formation in the eastern Dutch dialect ofTwente

25. 26. 21.

28.

111

nuum comes into being, reaching from old dialect forms on the one hand to forms originating from the standard language on the other hand. In any case, the development can stop at a certain point. Also in the case of vlag a development from (old) vlegke via vlegske to vlegke is apparent, but the scores are not very high. We can also refer to slebbeke for slab and vleggeke and vleggeske for vlag. It can be assumed that the greater the difference with the standard language, in other words, the more salient the element or aspect in question, the sooner it will be acquired. Of course, the problem is how far data, collected through questionnaires and interviews, reflect real language usage. For example, are the hyperdialectic forms steulke, etc., forms of the linguistic reality or merely artifacts of the interview? Therefore it is important to collect (more) spontaneously produced data. However, this has the disadvantage to be expensive and to take a lot of time.

References Bezoen, Herman L. 1938 Klank- en vormleer van het dialect der gemeente Enschede. Leiden: E. J. Brill. Bickerton, Derek 1973 'The nature of a Creole continuum', Language 49, 640-669. Hoppenbrouwers, Cor 1982 Language change. Meppel: Krips Repro. 1990 Het regiolect. Muiderberg: Coutinho. Keller, Rudi 1990 Sprachwandel. Tübingen: Francke Verlag. Kloeke, Gesinus G. 1952 Ostniederländische Diminutiva', in: Verzamelde opstellen, Assen: Van Gorcum & Comp. - G. A. Hak & Dr. H. J. Prakke, 69-96. (1st printing: Jahrbuch des Vereins für Niederdeutsche Sprachforschung, Heft 55, 1929). 1952 a 'De taal van Overijssel', in: Verzamelde opstellen, Assen: Van Gorcum & Comp. - G. A. Hak & Dr. H. J. Prakke, 97-120. Nijen Twilhaar, Jan 1990 Generatieve fonologie en de Studie van oostnederlandse dialecten, Amsterdam: P. J. Meertens-Instituut. Pee, Willem 1936-1938 Dialectgeographie der Nederlandsche diminutiva, Tongeren: G. Michiels Breeders.

178

Cor van Bree

Schirmunsky, Viktor 1930 'Sprachgeschichte und Siedlungsmundarten', GermanischRomanische Monatschrift 18: 113-122, 171-188. Schönfeld Wichers, Karel D. 1970 Woordenboek Nederlands-Twents, Hengelo, etc.: Broekhuis. Van Bree, Cor 1986 'De meervoudsvorming bij de Twentse substantieven vroeger en nu', in: Cor Hoppenbrouwers et al., Proeven van Taalwetenschap, Tabu 16. 2: 315-328. 1990 'De stabiliteit van de syntaxis en andere taalsectoren. Bewustzijn omtrent twentismen bij groepen Twenten en uit het westen afkomstige sprekers', in: Taal en tongval, themanummer 3: Dialectsyntaxis, 186-210. 1992 'The stability of language elements, in present-day eastern Standard-Dutch and eastern Dutch dialects', in: J. A. van Leuvensteijn - J. B. Berns (eds.), Dialect and Standard Language... in the English, Dutch, German and Norwegian Language Areas, Amsterdam etc.: Royal Netherlands Academy of Arts and Sciences, 178-203. 1994 'The Development of so-called Town Frisian', in: Peter Bakker - Maarten Mous (eds.), Mixed languages, Amsterdam: IFOTT, 69-82. Vousten, Rob 1995 Dialect als tweede taal. Linguistische en extra-lingmstische aspecten van de verwerving van een Noordlimburgs dialect door standaardtalige jongeren, diss. University of Nijmegen. Wanink, Ger H. 1948 Twents-Achterhoeks woordenboek met grammatica, Zutphen: W. J. Thieme & Cie.

Reformed versus Catholic: the origins of the [hu:s]/[hy:s] isogloss in eastern Netherlands at the beginning of the 17th century1 Pieter van Reenen

Introduction Hieronymus Heroldt, Reformed minister in Den Ham, was not a favorite. He and his wife had discredited the reputation of two of his parishioners, Janneken Sticken, the wife of Gert Costers, and his departed sister. Testimony to this effect had been given by Gert Costers' witnesses in the provincial synod in Zwolle. Heroldt maintained his ignorance concerning some of the statements; for others he offered his sincere apologies. As punishment for his gross transgression, Heroldt's confession of guilt was read from his own pulpit. Were there to be a recurrence of the slandering, he would be immediately removed from his position. The matter certainly did no good to the authority of Heroldt over his congregation, which was not so good to start with. Heroldt had been a minister at least since 1586, but nine years previously, he resumed his role as Reformed minister in Den Ham, after an interruption due to the Spanish invasion in 1605. For his new position in 1606, he was indebted to a Catholic priest who had intervened on his behalf, a strange order of events within the sphere of Reformed influence within the new Republic of the Netherlands. Apparently, the congregation and the freeholders in Den Ham had not agreed to his reappointment. This was a minister without prestige. The case was quite different for the priest Frederick van Delden, who arranged the reappointment of Heroldt in Den Ham. In gratitude for his services in Den Ham, the Reformed synod in turn allowed Frederick van Delden to function as priest in Ootmarsum from 1606 onwards. Ootmarsum was then within the Catholic sphere of influence, situated close to Oldenzaal where the occupying Spanish army was stationed. Seen in this light, the gratitude of the Reformed synod did not, evi-

180

Pieter van Reenen

dently, lead them to be overly generous. On Pentecost Day in 1609, four Dutch horsemen complete with horses charged into the church in Ootmarsum during a mass and threatened Frederick van Delden with rifles and pistols. His parishioners came into action and managed to confiscate one of the horses, take a gun from one of the horsemen and began shooting at the corporal. The matter blew over without further consequences, but it was clear that the parishioners of the priest Frederick van Delden were ready to withstand hell and high water for him. He was a beloved and respected shepherd of his flock, with great authority and much prestige. Prestige is a key word in the present paper. Histories from the Dutch province of Overijssel, in the border area of the counties of Twente and Salland (see map 2), illustrate how Reformed ministers and Catholic priests contributed to the formation of the best known dialect isogloss in the Netherlands, discovered by, and discussed in, Kloeke (1927), and known world-wide thanks to Bloomfield (1935: 328-339). Den Ham is situated in the county of Salland, and Ootmarsum in the county of Twente (see map 2). Even before the period of 1606 to 1614, the influence of the new Republic of the Netherlands was particularly noticeable in the county of Salland, and the church classis in Deventer instituted the Reformed church there. The county of Twente, on the other hand, with an occupying Spanish army from 1605 onwards, was under Catholic influence. At this time, Rovenius was appointed vicargeneral of the diocese of Deventer and dean of Oldenzaal. It is primarily due to him that the Catholic population of Twente was not converted to the Reformation. The distribution of religions in the province of Overijssel shows even today the county of Twente to be mainly Catholic and the county of Salland to be mainly Reformed, as can be seen on a map of the distribution of religions in the Netherlands. Map 1 shows the situation as given in the oldest known data, from 1809, at which time the pattern was clearer than it is now. In this study a connection will be made between the distribution of religions in Overijssel and the pronunciation of a series of six words. I will attempt to make it plausible that these six words functioned as a shibboleth between 1606 and 1614, and that in the border area between the counties of Twente and Salland an isogloss was created which is still there today. An isogloss is a line on a map separating two territories which differ in a linguistic feature. On one side of that line, people say something

Reformed versus Catholic

181

75 -100% 50 - 75% 25- 50% 0 - 25% E3 0%

Map 1. The percentage of Roman Catholics in Overijssel and in Gelderland in 1809 given by municipality (division of 1849). Source: Knippenberg (1992: 29) who bases his figures on De Kok (1964: 350-362).

different than they do on the other side of that line, and they think that is how it should be: "this is us and that is them", "they" are different than we are and that should be audible. A true isogloss is also a boundary between two shibboleths, whereby the speakers consciously make a difference to define their identity.2 The isogloss on the border between the counties of Twente and Salland is a real one, and such are unusual. That an isogloss involves a list of six words is even more uncommon. In the dialects to the west of the dividing line, in the county of Salland, huis 'house', buiten Outside', uit Out', sluiten 'to close; shut', duizend

182

Pieter van Reenen

is9W.SD B > C > D

This hierarchy says that a language with feature C also has features A and B; a language with feature B also has feature A; a language which does not have feature B neither has features C and D; etc. The relationship between synchronic and diachronic variation may now be seen as related in the way indicated in Table 1. In Table 1 synchronic variation is shown horizontally, diachronic variation vertically. Languages 1-6 start out using the same configuration (A), and implement the various configurations permitted by the hierarchy in (1) at different time rates, language 1 showing the lowest, language 6 the highest time rate. The result in this idealized represen-

250

Kees Hengeveld — Gerry Wanders

Table 1. Synchronic and diachronic variation Language

1

2

3

4

5

6

Diachrony

A A A A A A

A A A A A A/B

A A A A A/B A/B

A A A A/B A/B A/B/C

A A A/B A/B A/B/C A/B/C

A A/B A/B A/B/C A/B/C A/B/C/D

Synchrony

tation is a perfect match between the vertical and horizontal dimension. For a more concrete example, see Hengeveld (1991). The aim of this paper is to put Greenberg's hypothesis to the test by investigating the ways in which adverbial clauses are expressed from both a typological and a diachronic perspective, thereby concentrating on the use of subjunctive and indicative verb-forms. For the typological part of this paper we make use of the results of a research project on the expression of adverbial clauses carried out within the context of EUROTYP (see Hengeveld 1993, fc.; Hengeveld ed. 1993). For the diachronic part of this paper we will concentrate on the history of the use of the Subjunctive and the Indicative in a well-delimited set of adverbial clauses in Spanish. A central idea behind this paper is that there is a systematic correlation between the semantic type of an adverbial on the one hand, and its expression on the other. In section 2 semantic types of adverbials are defined in terms of semantic primitives which allow for generalization across specific adverbial relations. Expression formats of adverbial clauses concern, at the highest level, the selection of non-finite versus finite verb-forms and, at more specific levels, the selection of specific non-finite forms (infinitives, converbs, etc.) within the non-finite domain, and the selection of specific mood-forms (indicative, subjunctive, etc.) within the finite domain. It is the distribution of the latter type of forms in a subset of adverbial clauses that we study from a typological perspective in section 3. In section 4 we then look at the development of the use of Subjunctive and Indicative in Spanish, where we will show that the parameters responsible for the typological variation

On the use of subjunctive and indicative verb-forms

251

in this domain of grammar are the same as those operative along the diachronic axis. Section 5 presents our conclusions.

2. Semantic types of adverbial clauses With respect to the semantic type of adverbials, four classifying parameters are relevant. A full motivation for the classification of adverbial clauses resulting from these parameters cannot be given here for reasons of space. Evidence for this classification can, however, be found in Hengeveld (1993, fc.).

2.1. Entity types The main classifying parameter concerns the entity types that adverbial clauses designate. This classification of entity types is a direct reflection of the various layers distinguished within the hierarchical structure of the clause in Functional Grammar (Dik 1989, Hengeveld 1989). Extending the analysis proposed in Lyons (1977: 442-447), linguistic units may be said to refer to entities of four different types, as listed in Table 2. An individual is a first-order entity. It can be located in space and can be evaluated in terms of its existence. A state of affairs, or event, is a second-order entity. It can be located in space and time and can be evaluated in terms of its reality. A propositional content is a third-order entity. It can be located neither in space nor in time and can be evaluated in terms of its truth. A speech act is a fourth-order entity. It locates itself in space and time and can be evaluated in terms of its felicity.2

Table 2. Entity types Entity type

Description

Evaluation

First order Second order Third order Fourth order

Individual State of affairs Propositional Content Speech Act

Existence Reality Truth Felicity

252

Kees Hengeveld — Gerry Wanders

Within the context of adverbial subordination first-order entities do not play a part, since they can only be expressed by means of noun phrases, not by means of clauses. The remaining three types do show up, however, in the form of adverbial clauses. Consider examples (2)-(4): (2) (3) (4)

The fuse blew because of our overloading the circuit (Cause — second order) Jenny went home because her sister -would visit her (Reason — third order) Jenny isn't here, for I don't see her (Evidence - fourth order)

Examples (2)-(4) are all of a causal nature, yet they are all of different types. The difference between Cause and Reason is one that Lyons (1977) adduces to substantiate the distinction he makes between secondand third-order entities, that is, between states of affairs and prepositional contents. In (2) the subordinate clause describes the event causing the main-clause event, without there being any intentional involvement on the part of an agent. In (3) the Reason adverbial does not cause the main-clause event in any litteral sense, but represents the consideration, idea, i.e., the propositional content that led a participant in the mainclause event to engage in the main-clause event. There are a number of differences in the behavior of (2) and (3) that reflect the differences between the entity types they designate. To give one example, Reason clauses being propositional, they admit the expression of a propositional attitude, whereas Cause clauses do not: (5)

Jenny went home because her sister might visit her

(6)

* The fuse blew because we might have overloaded the circuit

There are many differences between Reason (3) and Evidence (4) clauses as well. Several of these are discussed in Bolkestein (1991). Whereas the source of the reason in (3) is the main-clause participant Jenny, the source of the evidence in (4) is the speaker. Consequently, the adverbial clause cannot be interpreted as the reason for which the mainclause event took place. Rather, it presents the considerations that led the speaker to arrive at the conclusion contained in the main clause, and can thus be seen as constituting a separate speech act. Evidence clauses having an illocutionary component, illocutionary modifications may be expressed within them, whereas this is not the case with Reason clauses:

On the use of subjunctive and indicative verb-forms

(7) (8)

253

Jenny isn't here, for, honestly, I don't see her *Jenny went home because, honestly, her sister would visit her

The classification of entity types given above forms the basis for the implicational hierarchy predicating the distribution of expression formats given in (9): (9)

Entity-Type Hierarchy Second order > Third Order

>

Fourth order

This hierarchy interacts crucially with the ones to be described below.

2.2. Time dependency Within the class of second-order adverbials a further subdistinction can be made as to their time dependency (see Noonan 1985). Consider the following examples: (10) (11)

The streets are wet because it is raining (Cause - second order) He cut himself while shaving (Simultaneity - second order)

The adverbials in (10) and (11) both describe events taking place simultaneously with the main-clause event. They differ in the fact that this simultaneity of main and subordinate event is obligatory in (11) whereas it is not in (10), witness the following examples: (12) (13)

The streets are wet because it has been raining * He cut himself while having shaved

Thus, Simultaneity clauses have dependent time reference (DTR), whereas Cause clauses have independent time reference (ITR). We may now formulate a second hierarchy describing the distribution of expression formats in adverbial clauses, which is given in (14): (14)

Time-Dependency Hierarchy Dependent Time Reference (DTR) > Independent Time Reference (ITR)

254

Kees Hengeveld — Gerry Wanders

Note that even in cases of independent time reference there may be logical restrictions on the possible combinations of temporal specifications in the main clause and the adverbial clause.

2.3. Factuality A third parameter along which adverbial clauses may be compared concerns their factuality. Those that have been presented so far are all factual, i.e., they present pieces of information that are presented by the speaker as real (second order), true (third order), or asserted (fourth order). Apart from these factual clauses there is a class of adverbials which may be characterized as non-factual. These present pieces of information presented by the speaker as unreal (second order) or with respect to which he expresses no truth commitment (third order). The following examples illustrate: (15) (16)

He won't get the job if he has no qualifications (Potential Condition - third order) I'll come tomorrow in case Ann wants me (Potential Circumstance — second order)

The difference between these two sentences is that, whereas in (15) the adverbial clause describes a condition on the validity of the main clause, the one in (16) describes an event potentially accompanying the main-clause event. One of the effects of this difference is that the event described in the main clause in (16) occurs independently of whether or not the potential circumstance occurs, whereas in (15) the main clause is only valid if the condition is valid as well. Within the class of non-factual adverbial clauses designating secondorder entities a further subdistinction can again be made between those with independent time reference, such as Potential Circumstance adverbial clauses, illustrated in (17) and (18), and those with dependent time reference, such as Purpose adverbial clauses, illustrated in (19) and (20): (17) (18)

I'm wearing my boots in case it rains (Potential Circumstance second order ITR) I'm wearing my boots in case it has rained

On the use of subjunctive and indicative verb-forms

(19) (20)

255

I left early to catch the train (Purpose - second order DTR) * I left early to have caught the train

The differences as regards the expression of factual and non-factual adverbial clauses is captured by the Factuality Hierarchy given in (21): (21)

Factuality Hierarchy Factual > Non-factual

2.4. Factuality dependency A final parameter along which adverbial clauses can be classified concerns their factuality dependency (cf. Noonan's (1985: 92) "truthvalue dependency"). A factual clause with a dependent factuality value is factive, i.e., describes a state of affairs presupposed by the speaker to be real or a prepositional content presupposed by the speaker to be true. A non-factual clause with dependent factuality is contra-factive, i.e., presupposed by the speaker to describe an unreal state of affairs or a false proposition. Examples of factive adverbial clauses are the following: (22) (23) (24)

After doing the cooking I looked after the garden (Anteriority — second order DTR) Apart from doing the cooking I look after the garden (Addition — second order ITR) I looked after the garden even though I had been doing the cooking (Concession - third order)

The events described in the adverbial clauses in (22) and (23) are temporally and logically presupposed. The concessive adverbial clause in (24) describes a piece of information which the speaker presupposes to be true and in view of which the information contained in the main clause would not be expected. The diffences between factual adverbial clauses with dependent and independent factuality values come out most clearly under modalization. Compare the examples in (25)-(27) of factual adverbial clauses with independent factuality value discussed above with the corresponding examples in (28)-(30) of factual adverbial clauses with dependent factuality value:

256

Kees Hengeveld — Gerry Wanders

(25) (26) (27)

He probably cut himself while shaving The fuse probably blew because we had overloaded the circuit Jenny probably went home because her sister would visit her

(28) (29)

He probably looked after the garden after doing the cooking He probably looked after the garden apart from doing the cooking He probably looked after the garden even though he had been doing the cooking

(30)

In (25)-(27) the adverbial clause may fall within the scope of the modal adverb probably. The content of the adverbial clause is part of the modalized information. In (28)-(30), on the other hand, it is just the content of the main clause that is modalized. This difference follows directly from the fact that these adverbial clauses have a predetermined factuality value which does not permit further modalization. Examples of contra-factive adverbial clauses are the following: (31) (32) (33)

She left without saying goodbye (Negative Circumstance second order DTR) She always greets me as if I were her best friend (Unreal Circumstance - second order ITR) He wouldn't get the job if he had no qualifications (Unreal Condition — third order)

The events described in the adverbial clauses in (31) and (32) are presupposed not to be real. The Unreal Condition in (33) describes a proposition presupposed by the speaker to be false, and thus forms the opposite of the concessive clause in (24). The differences between non-factual adverbial clauses with dependent and independent factuality values can be demonstrated more easily than those that obtain in the case of factual clauses. In (31)—(33) the reality or truth value assigned to the content of the adverbial clause is always the opposite of what the adverbial clause expresses. Thus, according to the speaker, in (31) it is not the case that she says goodbye, in (32) it is not the case that / am her best friend, and in (33) it is not true that he has no qualifications. The differences as regards the expression of adverbial clauses with dependent and independent factuality is captured by the FactualityDependency Hierarchy given in (34):

On the use of subjunctive and indicative verb-forms (34)

257

Factuality-Dependency Hierarchy

Dependent factuality value

>

Independent factuality value

2.5. Integration The various hierarchies distinguished in the previous sections may now be combined as in Table 3. Apart from summarizing what has been said above, Table 3 captures the various interactions between the hierarchies distinguished. The Entity-Type Hierarchy and the Factuality Hierarchy fully interact, and are given as the main horizontal and vertical parameters in Table 3. The Factuality-Dependency Hierarchy operates within each of the two domains defined by the Factuality Hierarchy. The Time-Dependency Hierarchy, finally, operates within each of the four domains defined by the Factuality Hierarchy and the FactualityDependency Hierarchy.

Table 3. Semantic classification of adverbial clauses Second order Factual

NonFactual

Independent Factuality Value

ITR Cause

Dependent Factuality Value

ITR Additon

Independent Factuality Value

ITR Pot. circ.

Dependent Factuality Value

ITR Unr. Circ.

Third order

Evidence Reason

DTR Simultaneity

Concession DTR Anteriority

Pot. Cond. DTR Purpose

Unr. cond. DTR Neg. Circ.

Fourth order

258

Kees Hengeveld - Gerry Wanders

3. Typology

The hierarchies constituting the grid in Table 3 define the crosslinguistic distribution of expression formats in adverbial clauses. We will here concentrate on the use of subjunctive and indicative verb-forms in a limited set of adverbial clauses, identified by double lines in Table 3. The adverbial clauses which form part of this set share two features: (i) they designate second-order entities, and (ii) they are factual. They differ from each other with respect to two other features: (i) time dependency, and (ii) factuality dependency. Thus, two of the four hierarchies distinguished earlier, the Entity-Type Hierarchy and the Factuality Hierarchy, are used here to delimit the set of adverbial clauses to be investigated, in order to allow for a detailed study of the effects of the remaining two hierarchies: the Time-Dependency Hierarchy and the Factuality-Dependency Hierarchy, as indicated in Table 4. A study of the distribution of indicative and subjunctive verb-forms from a typological perspective shows that the hierarchies involved operate in the way indicated in (35) and (36): (35)

Time-Dependency Hierarchy Dependent Time Reference (DTR) > Independent Time Reference (ITR) Subjunctive > Indicative

Table 4. Second order factual adverbial clauses Independent Factuality Value (IFV)

Dependent Factuality Value (DFV)

Independent Time Reference (ITR)

Cause

Dependent Time Reference (DTR)

Simultaneity

Independent Time Reference (ITR)

Addition

Dependent Time Reference (DTR)

Anteriority

On the use of subjunctive and indicative verb-forms

(36)

259

Factuality-Dependency Hierarchy Dependent factuality value (DFV) > Independent factuality value (IFV) Subjunctive > Indicative

Table 5 shows some of the possible outcomes of the combined operation of these two hierarchies in a number of European languages. Modern Greek has subjunctive verb-forms but does not use them in the adverbial clauses under consideration. Albanian uses subjunctive verb-forms in both Simultaneity and Anteriority clauses, i.e., those clauses which have dependent time reference and are therefore most likely to be expressed by means of subjunctive verb-forms following the Time-Dependency Hierarchy. This language thus shows that the effects of the Time-Dependency Hierarchy are stronger than those of the Factuality-Dependency Hierarchy. Had the latter been stronger, then Addition clauses rather than Simultaneity clauses would have been selected for expression by means of subjunctive verb-forms. Spanish goes one step further in applying subjunctive verb-forms in Addition clauses as well as in clauses with dependent time reference, thus conforming to the Factuality-Dependency Hierarchy. In this language the use of the Subjunctive is furthermore predominant in Anteriority clauses, as will be shown below, which further confirms the effects of the Factuality-Dependency Hierarchy in this language.

Table 5. Indicative and subjunctive verb-forms in second order factual clauses: typology

IFV

DFV

Greek

Albanian

Spanish

ITR

Cause

Ind

Ind

Ind

DTR

Simultaneity

Ind

Ind/Sub

Ind/Sub

ITR

Addition

Ind

Ind

Ind/Sub

DTR

Anteriority

Ind

Ind/Sub

Ind/Sub

260

Kees Hengeveld — Gerry Wanders

4. Diachrony We now turn to an investigation of the changes in the use of indicative and subjunctive verb-forms in the history of Spanish in the four types of adverbial clause singled out in section 3. Three time periods will be looked at: Classical Latin (4.1), Old Spanish (4.2), and Modern Spanish (4.3). A summary of the results will be given in section 4.4.

4.1. Classical Latin Examples of the four types of adverbial clause are given in (37)-(40):3 Cause (37) Cetarius im trepidat quod fishmonger.NOM anger.ABL tremble.3.S.PRES.IND because fur se effugit. thief.NOM REFL escape.3.S.PRES.IND 'The fishmonger trembles with anger because of the escaping of the thief.' (Terentius) Simultaneity (38) Te dum vivebas noveram. you.ACC while live.2.S.IMPERF.IND know. 1 .S.PLUPERF.IND Ί knew you while you were alive.' Addition (39) At

id praeterquam quod fieri non apart CMPLR happen.PRES.INF not potuit, potest ne can.3.S.PERF.IND can.3.S.PRES.IND not fingi quidem. make_up.PRES.INF even 'But apart from the fact that this couldn't happen, it couldn't even be made up.' (Cicero, Div. 2.28)

On the use of subjunctive and indicative verb-forms

261

Anteriority (40) Postquam aurum abstulimus in after gold.ACC take_away. 1 .PL.PERF.IND PREP navem

conscendimus.

ship.ACC board. 1.PL.PERF.IND 'After we had taken away the gold, we boarded a ship.' (Plautus, Ba. 277) These examples show that in Classical Latin only indicative verb-forms are attested in the four types of adverbial clause under consideration. At first sight an apparent counterexample to this claim would be the use of subjunctive verb-forms in adverbial clauses introduced by the conjunction cum (see, e.g., Troll 1971). However, the conjunction cum is to such a degree polyfunctional that the adverbial clause introduced by it can hardly be placed unambiguously in one of the semantic categories (cf. Kroon 1995: 72-73). Therefore it could be concluded that in this case the use of the subjunctive seems to be provoked by the conjunction rather than being the result of the semantic function of the adverbial clause.4 4.2. Old Spanish In order to get a clear picture of the use of subjunctive verb-forms in Old Spanish at a particular point, we delimited the period of what we still will be calling "Old Spanish" to the 13th and 14th century. The different types of adverbial clause will be looked at in 4.2.1—4.2.4. For many of the examples given in these sections see Jensen and Lathrop (1973). 4.2.1. Cause In adverbial clauses with the semantic function Cause the indicative is the only verb-form found. (41)

Porque vyn' sin presente, because come.3.S.PAST without gift sana

la the

cresce.

anger grow.3.S.PRES.IND 'Because I came without a gift, your anger grows.' (Lib.Buen Amor, 1367)

vuestra your

262

Kees Hengeveld — Gerry Wanders

4.2.2. Simultaneity

In Simultaneity clauses indicative as well as subjunctive verb-forms are attested. The indicative is used when the state of affairs in the main clause is situated in the past or the present, as shown in (42)-(43): (42)

Finco el omne bueno mientre le kneel.3.S.PAST.IND the man good while him.DAT dio Dios uida. give.3.S.PAST.IND God life 'The good man kneeled while God gave him life.' (Apolonia)

(43)

Mientra que son pequennos non while CMPLR COP.3.PL.PRES.IND young not pueden defender sus cosas. can.3.PL.PRES.IND defend.INF their properties 'While they are young they cannot defend their properties.' (Juzgo, 136)

Both indicative and subjunctive verb-forms are used when the state of affairs in the main clause is situated in the future, as shown in (44)-(45): (44)

Mientre el siglo fuere no as_long_as the century COP.3.S.FUT.SUBJ not sera

oblidada.

COP.3.S.FUT.IND forget.PART.F 'As long as the century will be, she will not be forgotten.' (Milagros, 65) (45)

Mando... que ningun heredero o fijo require.l.S.PRES.IND CMPLR not_one heir or child parta con el romanient demientre share.3.S.PRES.SUBJ PREP the inheritance while que biura en esta uida. CMPLR live.3.S.FUTIND PREP this live Ί require that no heir or child shares out the inheritance while he (i.e. the longest living parent) is present in this life.' (Teruel, 450, 2)

On the use of subjunctive and indicative verb-forms

263

The indicative verb-forms seem to be used predominantly in what Lyons (1977: 680) calls "omnitemporal propositions", i.e. those "... whose truth-value is constant for all values ... in a finite or infinite set of time-points or time-intervals ...". This would explain why the use of the indicative future instead of the subjunctive in temporal clauses is particularly frequent in the Medieval fueros (law texts), a fact noted by Jensen and Lathrop (1973).

4.2.3. Addition In Addition clauses the indicative is the only verb-form attested. The earliest example of this type of adverbial clause that we found is from a 16th century text and is given below (see Keniston 1937: 353). (46)

Aliende que se signified claramente, apart CMPLR REFL manifest.3.S.PRES.IND clearly lo dizen muchos autores. it say.3.PL.PRES.IND many authors 'Apart from the fact that it manifests itself clearly, it is claimed by many authors.'

4.2.4. Anteriority In Anteriority clauses both indicative and subjunctive verb-forms are attested. The indicative is used when reference is made to the past or the present: (47)

E despues que Troya fue destroyda and after CMPLR Troy COP.3.S.PAST.IND destroy.PART.F salieron ende dos hermanos leave.3.PL.PAST.IND from.there two bothers 'And after Troy had been destroyed two brothers left from there.' (Primera Cronica General de Espana)

Subjunctive and indicative verb-forms are used when reference is made to the future:

264

Kees Hengeveld — Gerry Wanders

(48)

Despues que vos ayades fecho after CMPLR you have.2.S.PRES.SUBJ make.PART este sacrificio ofrecer-vos los he this sacrifice offer.INF-you them AUX.l.S.PRES.IND yo en gracias. I gratefully 'After you have made this sacrifice, I will gratefully offer them to you.' (Hita, 777 a)

(49)

Despues que el mes passa, non after CMPLR the month pass.3.S.PRES.IND not respondra el daynnador a-l answer. 3. S.FUT.IND the owner PREP-the messeguero. messenger 'After a month has gone by, the owner will not need to answer the messenger.' (Noverena, 168)

The use of indicative verb-forms is again restricted to omnitemporal statements. Note further that there is a difference in the use of the subjunctive in Simultaneity clauses and in Anteriority clauses in that in the latter the present subjunctive is used whereas in the former the future subjunctive is attested as well.

4.3. Modern Spanish 4.3.1. Cause In Cause clauses the situation is unchanged as compared to Classical Latin and Old Spanish: only indicative verb-forms are used. (50)

Las calles estan mojadas porque the streets COP.3.PL.PRES.IND wet because esta lloviendo. COP.3.S.PRES.IND rain.GER 'The streets are wet because it is raining.'

On the use of subjunctive and indicative verb-forms

265

4.3.2. Simultaneity In Simultaneity clauses subjunctive as well as indicative verb-forms are found, as shown in (51)—(53): (51)

(52)

(53)

Mientras hay vida hay esperanza. while exist.PRES.IND life exist.PRES.IND hope 'As long as there is life, there is hope.' Mientras fue nino tuvo while COP.3.S.PAST.IND child have.3.S.PAST.IND muchos problemas de salud. many problems PREP health 'When he was young he had many health-problems.' No la olvidare mientras viva. not her forget.l.S.FUT.IND as_long_as live.l.S.PRES.SUBJ Ί will not forget her as long as I live.'

The use of the indicative and subjunctive is almost similar to the situation in Old Spanish: the indicative is used when reference is made to the past or the present, the subjunctive is used when reference is made to the future. In the latter case, omnitemporal propositions may still contain an indicative verb-form. A difference between Old and Modern Spanish is, however, that the future subjunctive has gone out of use, its role having been taken over by the present subjunctive.

4.3.3. Addition In Addition clauses indicative as well as subjunctive verb-forms are used, as shown by examples (54), which is taken from the weekly Tiempo, and (55): (54)

Aparte de que nos hayamos apart PREP CMPLR REFL AUX.l.PL.PRES.SUBJ hecho mayores yo tambien me make.PART older.PL I also REFL he hecho mayor. AUX.l.S.PRES.IND make.PART older.S 'Apart from the fact that we became older, I also became mature.'

266

(55)

Kees Hengeveld - Gerry Wanders

No observe en la casa nada anormal, not notice.l.S.PAST PREP the house nothing abnormal aparte de que la puerta del jardin apart PREP CMPLR the door PREP-the garden estaba abierta. COP.3.S.PAST.IND open Ί didn't notice anything abnormal in the house, apart from the fact that the garden-door was open.'

The factors triggering the use of either of the two verb-forms are not entirely clear. Our impression is that the use of the subjunctive predominates when the speaker chooses to present the state of affairs described in the adverbial clause not as a fact not yet known but as a fact known to both speaker and hearer to be real. The use of the indicative would then predominate in cases in which this state of affairs is presented as a not yet known fact, as information new to the hearer.

4.3.4. Anteriority Examples (56)-(58) show that subjunctive as well as indicative verbforms are used in Anteriority clauses. (56)

Nada cambiara despues de que nothing change-3.SG.FUT after PREP CMPLR saiga su disco. come.out3.SG.PRES.SUBJ his record 'Nothing will change after his record will come out.'

(57)

Retorno a casa despues de que su return.3.SG.PAST to home after PREP CMPLR his tia se fuese. aunt REFL go.3.SG.PAST.SUBJ 'He returned home after his aunt had left.'

On the use of subjunctive and indicative verb-forms

(58)

267

Deja el programa despues de que leave.3.SG.PRES the program after PREP CMPLR varios personajes famosos se le various persons famous REFL she.DAT han atragantado. AUX.3.PL.PRES.IND choke.PART 'She leaves the program after having developed a disliking for various famous persons.'

As in Old Spanish, the present subjunctive is used if the state of affairs in the main clause is situated in the future. However, if the state of affairs in the main clause is situated in the past or present both indicative and subjunctive are used, the use of the subjunctive even being predominant. The use of the subjunctive in Anteriority clauses seems similar to the use of the subjunctive in Addition clauses (see 4.3.3), that is, it seems to be the predetermined factuality value of such clauses that allows the use of the subjunctive.

4.4. Summary As we have seen in the sections above, in the course of time many changes have taken place in the the use of indicative and subjunctive verb-forms in adverbial clauses in Spanish. Taking the situation in Latin, which used indicative verb-forms only, as our point of departure, the following observations can be made with respect to the development of the use of the subjunctive: (i) in Anteriority clauses, having both a dependent factuality value and dependent time reference, the use of the subjunctive has most strongly developed, with future-time reference being the prime motivating factor in Old Spanish, to which factuality dependency is added in Modern Spanish; (ii) in Simultaneity clauses, having an independent factuality value and dependent time reference, the situation is roughly the same as with Anteriority clauses in Old Spanish, but here factuality dependency cannot, and does not, come into play in Modern Spanish; (iii) in Addition clauses, having a dependent factuality value and independent time reference, the subjunctive is not used in Old Spanish but comes to be used in Modern Spanish, the motivating factor being, as in the case of Anteriority clauses, factuality

268

Kees Hengeveld - Gerry Wanders

Table 6. Indicative and subjunctive verb-forms in second-order factual clauses: diachrony

IFV

Latin

Old Spanish

Modern Spanish

ITR

Cause

Ind

Ind

Ind

DTR

Simultaneity

Ind

Ind/Sub

Ind/Sub

Addition

Ind

Ind

Ind/Sub

Anteriority

Ind

Ind/Sub

Ind/Sub

DFV ITR DTR

dependency; (iv) In Cause clauses, having an independent factuality value and independent time reference, the subjunctive is excluded. These facts are represented in Table 6. It is significant that in the type of adverbial clause which has been categorized as having both independent time reference and independent factuality value, i.e., Cause clauses, the situation remained the same from Classical Latin up to Modern Spanish, and that in the type of clause which has been categorized as having both dependent time reference and a dependent factuality value the subjunctive has most strongly developed. Table 6 furthermore shows that, as in the case of the typological facts presented in section 3, the effects of the timedependency hierarchy are prior to those of the factuality-dependency hierarchy.

5. Conclusion The typological and diachronic facts presented in this paper lend further support to Greenberg's (1978) hypothesis that there is a systematic relationship between typological variation and diachronic change. Not only were the same hierarchies, the Time-Dependency Hierarchy and the Factuality-Dependency Hierarchy, shown to be relevant from both a synchronic and diachronic perspective, but also the ordering of the two with respect to each other was shown to be equally relevant to both perspectives.

On the use of subjunctive and indicative verb-forms

269

Notes 1. We are grateful to Jadranka Gvozdanovic and Harm Pinkster for their comments on an earlier version of this paper. Gerry Wanders' research for this paper has been made possible by the Netherlands Organization for Scientific Research (NWO). 2. Hengeveld (1992) and Keizer (1992) furthermore recognize a class of zeroorder entities, designating properties and relations. This class of entities is irrelevant to the present discussion. 3. For (38) see Bolkestein (1992), for (40) see Lehmann (1988). 4. We believe that a full understanding of the use of the subjunctive in cumclauses would require further study of the functions of these clauses in discourse. It is furthermore interesting to note that the causal uses of cumclauses belong to the categories Reason or Evidence rather than to the category Cause.

References Badia Margarit, Antonio 1953 "El subjuntivo de subordination en las lenguas romances y especialmente en iberorromanico", Revista de Filologia Espanola^l: 95-129. Bolkestein, A. Machtelt 1991 "Causally related predications and the choice between parataxis and hypotaxis in Latin", in R. Coleman (ed.), New studies in Latin linguistics. Benjamins, Amsterdam, 427-451. 1992 "Limits to layering: locatability and other problems", in: Michael Fortescue - Peter Harder - Lars Kristoffersen (eds.), Layered structure and reference in a functional perspective, 387-407. Amsterdam: Benjamins. Croft, William 1990 Typology and universals. Cambridge: Cambridge University Press. Dik, Simon C. 1989 The theory of Functional Grammar. Part 1: The structure of the clause. Foris, Dordrecht. Falk, Johan 1979 SER y ESTAR con atributos adjetivales. Acta Universitatis Upsaliensis, Studia Romanica Upsaliensia 29, Uppsala: Almqvist & Wiksell.

270

Kees Hengeveld — Gerry Wanders

Greenberg, Joseph H. 1978 "Diachrony, synchrony, and language universale", in: Joseph Greenberg (ed.) 1978, vol.1: 61-91. Greenberg, Joseph H. (ed.) 1978 Universals of human language, 4 vol. Stanford: Stanford University Press. Hengeveld, Kees 1989 "Layers and operators in Functional Grammar", Journal of Linguistics 25.1: 127-157. 1991 "Tipologia, sincronia, diacronia", in: Henk Haverkate - Kees Hengeveld - Gijs Mulder - Hella Olbertz (eds.), Exploraciones semanticas y pragmaticas del espanol (Foro Hispanico 2), 81-94. Amsterdam: Rodopi. 1992 Non-verbal predication: theory, typology, diachrony (Functional Grammar Series 15). Berlin: Mouton de Gruyter. 1993 "Semantic type, factivity and the expression of adverbial clauses", in: Kees Hengeveld (ed.), 119-132. (fc.) "Semantic type and expression format: On internal structure of adverbial and complement clauses in the languages of Europe'. To appear in: Johan van der Auwera (ed.), Adverbial relations in the languages of Europe. Berlin: Mouton de Gruyter. Hengeveld, Kees (ed.) 1993 The internal structure of adverbial clauses (EUROTYP Working Papers V 5). Strasbourg: European Science Foundation. Jensen, Frede - Thomas A. Lathrop 1973 The syntax of the Old Spanish subjunctive. The Hague: Mouton. Keizer, M. Evelien 1992 "Predicates as referring expressions", in: Michael Fortescue Peter Harder - Lars Kristoffersen (eds.), Layered structure and reference in a functional perspective, 1—27. Amsterdam: Benjamins. Keniston, Hayward 1937 The syntax of Castilian prose; The sixteenth century. Chicago: University of Chicago Press. Kiss, Sandor 1982 Tendances evolutives de la syntaxe verbale en latin tardif. Debrecen: University of Debrecen. Kroon, Caroline 1995 Discourse particles in Latin; A study o/nam, enim, autem, vero and at. Amsterdam: Gieben.

On the use of subjunctive and indicative verb-forms

271

Lehmann, Christian 1988 "Towards a typology of clause linkage", in: John Haiman Sandra A. Thompson (eds.), Clause combining in grammar and discourse, 181-225. Amsterdam: Benjamins. Lyons, John 1977 Semantics, 2 vols. Cambridge: Cambridge University Press. Moreno Cabrera, Juan Carlos (1985) "Diacronia y tipologia: hacia una superacion del punto de vista sincronico", Revista espanola de lingüistica 15: 430-443. Noonan, Michael 1985 "Complementation", in Timothy Shopen (ed.), Language typology and syntactic description, Vol. II. Cambridge University Press, Cambridge, 42-140. Pinkster, Harm 1984 Latijnse syntaxis en semantiek. Amsterdam: Grüner. Troll, Paul 1971 Lateinische Sprachlehre. Frankfurt am Main: Diesterweg. Wanner, Dieter 1990 "Le subjonctif de subordination en latin vulgaire: Questions indirectes et adverbiales temporelles", in: Gualtiero Calboli (ed.), Latin vulgaire - latin tardif II, 249-280. Tübingen: Niemeyer.

Nominal syntax in Italic: a diachronic perspective Brigitte L. M. Bauer

In this paper, I will discuss a number of structures in Italic that traditionally have not been accounted for, and that may be explained from a diachronic and Indo-European perspective. The structures under discussion have been inherited from the protolanguage and share a number of archaic grammatical features. On the basis of these archaisms, and in line with recent findings in the reconstruction of Proto-Indo-European syntax, I will argue that these structures are residues of a different, non-nominative typology that characterized the protolanguage at an early stage. Their subsequent development is in accordance with the increasing importance of transitivity, which characterizes the history of Indo-European languages.

1. Introduction In the development of Indo-European languages, we observe a number of major grammatical changes that affected all languages of this phylum, independently of each other: the development of subordination, the shift from a verb system based on aspectual distinctions to a verb system based on temporal opposition, and the shift from left- to rightbranching structures (also referred to as the shift from OV to VO). Many of the structures in various Indo-European languages as well as their subsequent development fit the patterns indicated in the previous paragraph. An example is the emergence of habeo as a temporal auxiliary (cf., e.g., in French the past tense faifait replaces jefis, which goes back to perfective fed in Latin). Similarly, the increasing tendency to replace je chanterai with the analytic future form ye vais chanter, which features a preposed auxiliary (as opposed to chanterai < cantare habeo), is integrated in the comprehensive spreading of right-branching (or head-first) constructions that replace their left-branching (or headlast) archaic equivalents; these left-branching structures were inherited from the protolanguage (for analysis of this shift in Latin and French, cf. Bauer 1995).

274

Brigitte L. M. Bauer

Yet a number of structures in Latin have been very difficult to explain, and their development does not fit any of the major changes I referred to earlier; for example, the mihi est construction to indicate possession, impersonal verbs, or absolute and gerundive constructions. In addition, these structures do not fit the pattern of a nominative language. Latin indeed basically was a nominative language, as the fundamental opposition nominative-agent vs. accusative-patient shows, as does the distinction between active and passive paradigms in verbal conjugation. As a rule, transitive action involving a subject and a direct object is expressed in Latin by clauses featuring (1) a subject-Nominative referred to by a noun, pronoun and/or the personal ending of the finite verb; (2) a direct object-Accusative; and (3) a finite verb denoting the action. Whereas the aforementioned mihi est constructions, impersonal verbs, and absolute and gerundive constructions do not fit the patterns of a nominative language - as I will show in this paper, they have been replaced over time by structures that do fit these patterns: absolute constructions, for example, were replaced by subordinate clauses featuring a finite verb and a direct object, and mihi est constructions came to be replaced by habeo, which displays transitive syntax and combines a nominative-possessor with an accusative referring to the element in possession. From this perspective, the structures under consideration may very well be residues of the language system that characterized ProtoIndo-European at an earlier stage — Pre-Indo-European — before it became a nominative language. On the basis of lexical, morphological, and syntactic evidence, Indo-Europeanists recently have proposed that at an earlier stage Proto-Indo-European was typologically different. With time, a new - nominative - language system came into being, but residues of the former, non-nominative typology were still present (cf. Gamkrelidze and Ivanov 1984; Lehmann 1989, 1993). The replacement of these archaic residues therefore is no longer an isolated phenomenon, but it reflects the increasing importance of transitivity in Indo-European languages. In this paper I will examine mihi est constructions as well as gerundive and absolute constructions in Latin and in two other Italic languages, Oscan and Umbrian. I will point out and discuss the common grammatical features of these structures, and relate them to recent findings in Indo-European linguistics that suggest that the protolanguage at an early stage displayed a fundamentally different typology. The

Nominal syntax in Italic: a diachronic perspective

275

replacement of the structures under consideration then fits the increasing importance of transitivity. First I will discuss the approach in typology that put forth this recent hypothesis and that differs from other methods applied in typology in a number of respects (Sections 2, 3, and 4). Then I will present a grammatical analysis of the structures in the three Italic languages mentioned, arguing that we are indeed dealing with residues of an earlier, nonnominative typology (Section 5).

2. Typology and linguistics Typology originally was an approach in biology in the 18th century and remained one of the dominant methods of biological inquiry throughout the 19th century. I am referring here to the biological origin of this approach, not because I adhere to the concepts it implied, nor because I am convinced that linguists should borrow methods used in the sciences and apply them to the analysis of language, but because the early use of typology in biology is revealing for the usefulness of its application in linguistics. Aware of the mass of living animals and motivated by a desire for classification, biologists and naturalists in the 18th century set out to find order in what at first sight appeared to be total randomness. At first, their typological analyses aimed to discover common features in animals, and to use these as criteria for their classification; consider, for example, the series of invariable characteristics that are unique to birds: feathers, the specific arrangement of these feathers necessary to fly, a specific lung system and brain structure (Denton 1985: 107). It soon became clear, however, that specific features occur with others, that they are somehow related, and that the various parts of an organism form a functioning whole. Consequently, what started as an attempt at pure classification soon turned out to be also a means to reveal the internal correlation of features. Although the object of analysis is, of course, totally different in linguistics, the parallel with biological typology is interesting. Within the field of linguistic typology as well we find linguists who examine and classify a large number of languages on the basis of one structural or phonological feature - for example, Adelung's early classification of languages on the basis of morphological processes distinguishing (what

276

Brigitte L. M. Bauer

was later called) analytic, synthetic, and agglutinative languages (1806-1817), or Nichols's classification on the basis of head vs. modifier marking (1986). As long as cross-linguistic analysis of one specific feature is the basis for classification, it is theoretically possible that a language will be ascribed to different classes depending on the feature analyzed. Also, essentially different languages may belong to the same specific class on the basis of one common feature, yet belong to totally different classes on the basis of another feature. The classification may therefore change according to the criterion that is applied. In contrast, the more criteria that are applied, the less arbitrary the analysis — and hence the classification - becomes. This observation can be illustrated by referring to word-order typology. Although studies in word order were not new, Greenberg's article on "the order of meaningful elements" (1963) provided new impetus for word order typology. On the basis of correlations between syntactic phrases, Greenberg distinguished three basic types - or word order typologies - SVO, SOV, and VSO. In addition, he pointed out a correlation with patterning in morphology — namely Universal 41: "If in a language the verb follows both the nominal subject and nominal object as the dominant order, the language almost always has a case system" (Greenberg 1963: 75). This correlation between ordering patterns of several levels in language had been observed earlier by Wilhelm Schmidt, according to whom there is a relation between the preposing of the genitive, the occurrence of postpositions, and the morphological characteristics of the language in question: "steht der affixlose Genitiv vor dem Substantiv, welches es näher bestimmt, so ist die Sprache eine Suffixsprache eventuell mit Postpositionen, steht der Genitiv nach, so ist sie eine Präfixsprache eventuell mit Präpositionen" (Schmidt 1926: 382). ['If the genitive (without affixes) precedes the noun determined by it, then the language is a suffixal language possibly with postpositions; if it follows the noun, then the language is a prefixal language, possibly with prepositions.'] In line with these consistent and systematic quests for correlations and patterning, modern word-order typologists have been able — with evidence from cross-linguistic analyses also - to relate specific ordering patterns in the various syntactic phrases, to reveal their correlation with morphology, and to discover the underlying pattern common to both levels of language.

Nominal syntax in Italic: a diachronic perspective

277

Although the aim of this kind of analysis is to reveal the underlying regularity and correlation of word-order patterns across various syntactic and morphological structures, these studies deal with one feature only: the ordering of elements between which exists a hierarchical, grammatical relation. Yet, one of the unexpected but highly interesting consequences of typological analyses in biology was the observation that various features are correlated, that they are basic to the classification of the animal, and that their collaboration is essential for the functioning of the organism. If linguists as well could point out and explain the correlation of linguistic features, then we might better be able to understand the innermost structure of language. Discovering the "inner motivation" (Klimov 1983: 335) is one of the aims of an approach in language typology that owes much to Russian scholars, and which is called contentive or content-oriented or contentrelated typology. Contentive typology differs from other approaches in typology - referred to as "formal typology" - in a number of respects. Instead of examining a limited number of characteristics across nonrelated languages, and instead of classifying languages on the basis of surface "similarities," contentive (or, henceforth, content-related) typology "examines the formal side of languages from the point of view of the content conveyed by it" (Klimov 1983: 327). Consequently, the aim of this kind of analysis is to point out the correlation of linguistic features - not only of features on the same level (e.g., syntax) but also the correlation between features from different levels. In contrast with other approaches in typology, content-related typology typically is "allembracing" in that it analyzes the various levels of language, including the lexicon (cf. Klimov 1983). The underlying principle of this approach is the generally accepted conception of language as an organized system, to which for example Benveniste referred in the following way: "la langue constitue un systeme, dont toutes les parties sont unies par un rapport de solidarite et de dependence" (1966: 98). This conception does not mean that linguistics ought to be reduced to a set of sterile rules that at their best might describe (not explain) a number of phenomena. In contrast, it implies a careful analysis of structures that at first glance may seem to be totally different, but that share the same basic characteristics. The recognition of a language as representing a specific type is therefore based on the analysis of a wide range of related linguistic features from the various levels of language — lexicon, morphology, and syntax.

278

Brigitte L. M. Bauer

The basic feature of a language, according to content-related typology, is the way in which the relation between action, "agent", and "patient" is expressed; hence the classification of languages into major types, distinguishing nominative, ergative, and active languages. Since the aim of this approach is to analyze the "set of systematically dependent structural characteristics of languages," (Klimov 1983: 329) and to discover the underlying system, the features that are examined are no longer arbitrary, yet their occurrence and their correlation must also be accounted for. Consequently, the central issue of this approach in typology — the concept of language as a systematic organization — is the basis of the conception of language as a whole, and it avoids the pitfalls of applying arbitrary criteria. In the final analysis, in fact, it also implies the necessity of correlating content and structure: "The semantic characteristics of the structural features forming a language type make it possible to decide upon the specific features of its inner motivation, i.e. the semantic determinant" (Klimov 1983: 335). This typological approach seems promising, especially when we are dealing with structures that have not been accounted for. Within the field of Indo-European linguistics, where the genetic relation between the languages involved has been firmly established on the basis of the comparative method and the method of internal reconstruction, typological analyses of structures that are proven to be inherited and archaic may point to a different syntax in an earlier stage of the protolanguage. Those structures that do not fit the nominative pattern, but share the same underlying syntactic features and are inherited from Proto-IndoEuropean, may therefore be residues of an earlier non-nominative language system that characterized the protolanguage at a very early stage, presumably that of Pre-Indo-European. It is from this diachronic typological perspective that the Italic structures I mentioned earlier will be examined in this paper.

3. Typology and Proto-Indo-European The implications of typology for Indo-European linguistics and especially for the reconstruction of the Proto-Indo-European phonemic system are well-known (cf. the effect of Jakobson's report at the VHIth International Congress of Linguists in Oslo in 1957 [1962]; cf. also

Nominal syntax in Italic: a diachronic perspective

279

Lehmann 1993). Less well-known are the results of content-related typology with respect to the reconstruction of Proto-Indo-European syntax. As I mentioned earlier, linguistic features are, in the view of content-related typology, not isolated phenomena, but are interrelated, including various levels of language. On the basis of the way languages express action and the inherent relation between the object or persons involved in that action, content-related typology distinguishes three types of language: nominative, ergative, and active. Since there is not enough space in this article to discuss all three variants, I will limit my analysis here to active typology, for several reasons: (1) whereas the basic characteristics of nominative and ergative languages are fairly well-known, those of active languages are not (most linguists are unfamiliar with the very concept of active language); (2) the assumption that Proto-Indo-European at an earlier stage - presumably Pre-IndoEuropean - was an active language has recently been made, and is one of the major contributions of Gamkrelidze and Ivanov's work on IndoEuropean (1984); (3) the perspective of an active language might be promising for the explanation of the structures I want to discuss in this paper, and their evolution. Active languages display a number of typical features that distinguish them from ergative as well as nominative languages. The most important characteristic of an active language - the characteristic that determines the rest of the linguistic system — is the absence of transitivity as a grammatical feature. I do not argue that transitive verbs do not exist in active languages; they do, but the opposition transitive - intransitive is not grammatically marked. Instead, the basic distinction is between stative and active, as is shown, for example, in verbal morphology, where active verbs (transitive or not) display the same paradigms, distinct from stative verbs. We observe that this major distinction (active vs. stative) is reflected at all levels of the language in question. The lexicon, for example, features the opposition between active nouns, which refer to persons, animals, and plants, and inactive nouns, which are characterized by the absence of life activity. Verbs reflect the same fundamental distinction: active verbs, which express action or motion, are opposed to stative verbs, which refer to state and qualities — for example, 'kilF, 'go', 'run', 'burn' (transitive and intransitive), 'lead', 'come' as opposed to 'be', 'be red', 'be tired', 'be glad', and so on.

280

Brigitte L. M. Bauer

In line with these basic characteristics, active languages typically: (1)

do not have verba habendi, because these verbs express a subject — object relation. This characteristic is consistent with the absence of transitivity as a grammatical feature. Instead, possession is expressed either by a pronoun or by a construction featuring the verb 'to be,' cf. Latin ei filiae duae erant (PL, Such. 539) he-Dat. daughters-Nom. two-Nom. be-3pl.-Impf. 'he had two daughters' In addition, active languages typically display

(2)

the opposition between inclusive and exclusive pronouns, distinguishing between "we-inclusive" and "we-exclusive";

(3)

a distinction between the element to which the action is directed and the element that, roughly speaking, indicates the circumstance, instead of the distinction between a direct and an indirect object;

(4)

a distinctive class of impersonal verbs;

(5)

a basic, unmarked word order, which is left-branching: S O' O" V sequences in active constructions and S O7 V sequences in Stative constructions;

(6)

full-fledged verbal morphology, displaying active and inactive affixes. As a rule, active verbs typically have more extensive paradigms than stative verbs;

(7)

very poor noun inflection, if any. If there is noun inflection, there is an opposition between active and inactive cases, whereby the inactive case is unmarked. The active case refers to what from our perspective is the "subject" (or the element that has control) of an active verb, whereas the inactive case refers to the "subject" of an inactive verb or the "object" of an active verb;

Nominal syntax in Italic: a diachronic perspective

281

(8)

active verbs do not distinguish between active and passive, which are grammatical categories typical of a nominative language, but between centrifugal and centripetal forms, cf. the modern lexical distinctions 'lead' vs. 'bring,' or 'mourn' vs. 'weep.' This distinction is comparable to the opposition between active and medium in the protolanguage (cf. Klimov 1977); and

(9)

a verb system that is based on aspectual distinctions (Klimov 1977).

As the preceding characteristics already suggest, the sentence pattern of an active language is not based on the syntactic relation of government, which is typical for transitive-marked languages (either ergative or nominative). Instead, it is based on agreement, or lexical concord. In their strictest - and hence most archaic - variant, active languages typically combine active verbs with active nouns and inactive verbs with inactive nouns. This organization is based on the semantic (cognitive) feature [or animate]. Later this distinction may gradually develop into an agent — patient opposition. Consequently, one observes a shift from active typology to a linguistic system that features transitivity, whether ergative or nominative. There are theoretically three possible schemes: active becoming ergative; active becoming nominative; and active becoming ergative and, at a subsequent stage, nominative. (For the evolution of these systems, cf. Klimov 1983; K.H. Schmidt 1979.) The patterns I just mentioned are the result of careful analysis of still-existing active languages: a number of Indian languages in the Americas display active typology; early languages in the near East, such as Elamite also are assumed to have been active (Klimov 1977: 314). The characteristics that consistently come back in independent studies carried out by specialists in these languages point to a dramatically different language organization. Yet this awareness of fundamental and systematic differences is not new, as we may notice when referring to Sapir's typological analysis (1921 [1949]). It is well-known, of course, that one has to face a number of problems when dealing with Indian languages: the ongoing extinction of these languages or of their speakers and, even more importantly, the inconvenience of many studies carried out from a nominative perspective, generally that of English. Despite these inconveniences, it is possible to

282

Brigitte L. M. Bauer

reveal the basic characteristics and their correlation, cf. Klimov in a series of publications (see also K. H. Schmidt [1979]).

4. Active typology and Indo-European Whereas Klimov gives quite a detailed analysis of active languages in general, Gamkrelidze and Ivanov have identified Pre-Indo-European as an active language. Gamkrelidze and Ivanov on the one hand point out features in Indo-European languages that do not fit the patterning of a nominative typology; subsequently they relate these phenomena to characteristics that cross-linguistic analysis of non-nominative, active languages has pointed out. Carrying out this analysis, they not only find a way to explain the earlier mentioned "idiosyncrasies," but they are also able to account for phenomena that, while they do not seem to be alien, are nevertheless difficult to account for fully. An example is the defective /H'-inflection in Hittite, which goes back to the former stative verb class (cf. the characteristic under [6] in Section 3) - as opposed to the full-fledged inflection of the wz-paradigms, which go back to active verbs (Gamkrelidze and Ivanov 1984; Lehmann 1993). Gamkrelidze and Ivanov support their hypothesis with a number of arguments, which cover various levels of language. I will briefly refer to some of them. (For a full account cf. Gamkrelidze and Ivanov 1984 and Lehmann 1989, 1993.) At a very early stage, Indo-European was characterized by the fundamental distinction in the lexicon between animate and inanimate (or stative vs. active), pointed out by Meillet (1926a) in his analysis of words that appear in doublets, such as the words for fire or water. Comparative analysis of Indo-European languages has shown that a number of concepts are referred to by two words, one referring to the animate item, and one referring to the inanimate element. Each of these words continued to exist in the various daughter languages. An example is Latin ignis, which goes back to the animate noun, and whose cognate in Sanskrit, Agnis, denotes the god of fire; also, Sanskrit ap- 'water' referred, originally, to the animate element (cf. Ap m Nap t [RV 2.35] 'son of the waters') as opposed to the inanimate watar in Hittite, which is marked by the inactive suffix -r (cf. also the Greek word for (inanimate) fire, πυρ [Gamkrelidze and Ivanov 1984; Lehmann 1958]).

Nominal syntax in Italic: a diachronic perspective

283

The distinction animate vs. inanimate is still present in Hittite, where, in the absence of feminine gender, there is only a distinction between neuter and animate (Friedrich 1974). In addition, this basic distinction between animate and inanimate is reflected in the binarism of IndoEuropean morphology, cf. for example the adjectival inflection, where we find evidence for the original distinction between two paradigms, one for masculine/feminine, one for neuter; this pattern can still be found in the third-declension adjectives in Latin. Similarly, the distribution of the genitive marker is revealing. The marker of the genitive singular in Latin, for example, goes back to an ending *-os, whereas the plural ending goes back to *-om. We find the same endings also in other Indo-European languages, but Hittite evidence is particularly interesting. The Hittite ending -as is the suffix of plural and singular genitives. In addition, there is also a genitive plural in -an, "which is found... only with words for living beings, in the singular less frequently than in the plural" (Sturtevant 1951: 91). Consequently, the original distinction was not between singular and plural, but between animate and inanimate (see Gamkrelidze and Ivanov 1984: 268-271). Other features that refer to a different, active typology at an earlier stage are the residues of inclusive and exclusive pronouns. As Prokosch pointed out much earlier, "[t]here seems to be a connection between the plural (and dual) forms of the 1st and 2nd person" (1939: 282). He observed that the root we- survived in some languages as a firstplural pronoun nominative (cf. German wir [< Germanic wis], Sanskrit vayam), but as a second-person plural in other languages (cf. Latin vos, Slavic vy, and so on). Prokosch interpreted this evidence by positing originally two forms for the first person plural, one inclusive (referring to ego and tu, and possibly a third person), and the other exclusive (referring to ego and (a) third person(s)) (Prokosch 1939: 282; see also Gamkrelidze and Ivanov 1984; Lehmann 1993). In addition, we find impersonal verbs in all early Indo-European languages although their number and distribution vary from dialect to dialect. They cover a wide range of semantic values: meteorological expressions (e.g., Latin pluit 'it is raining'), expressions of emotions (e.g., Latin me pudet Ί am ashamed'), and expressions indicating necessity, possibility, and so on (Latin mihi licet Ί am allowed'). The typical occurrence of third-person verb forms, added to the fact that these structures despite their morphological differences cover the same

284

Brigitte L. M. Bauer

semantic values, allows us to assume that we are dealing with a pattern that has been inherited from Proto-Indo-European. In syntax, we observe the absence of the verb habeo 'have'. The variety of lexical verbs expressing possession in the Indo-European languages (Latin habeo, Greek έχω, and so forth) shows the absence of a common etymological origin, hence their non-inherited character. Instead, these verbs developed separately in the various dialects. Originally, possession was expressed by constructions featuring the verb 'to be', and a possessor referred to in an oblique case. Since these structures can be found in all early Indo-European languages, it is legitimate to assume that they are inherited from the protolanguage (cf. also Section 5.2).

5. Archaic structures in Italic Italic is a branch of closely-related Indo-European languages that were spoken in what is today Italy. The best-known representatives are Oscan, Umbrian, and Latin. Although evidence from Oscan and Umbrian is rather limited compared to Latin, both languages include valuable information on the early stages of Italic. The archaic constructions of Latin, Umbrian, and Oscan that I will analyze in the following pages are mihi est constructions to indicate possession, absolute constructions, and constructions featuring a gerundive. These three types of structure all share a number of specific characteristics: (1)

(2)

they denote an action or a situation that in our modern perspective would be expressed by transitive syntax; i.e., a construction featuring a subject-agent and object-patient; they all underwent a comparable development during which the archaic syntax came to be replaced with verbal — hence, transitive - syntax.

5.1. Constructions featuring a gerundive/gerund The gerund, gerundive, and absolute constructions show a common characteristic which affects the way action is conveyed. The gerund is a verbal noun and it is traditionally assumed that "le gerondif sert donner une flexion l'infinitif: signum recipiendi 'le signal de se

Nominal syntax in Italic: a diachronic perspective

285

retirer' [the sign to withdraw] (Caes., BG 7.52.1)" (Meillet and Vendryes 1924: 335). The nominal character of this form indeed shows quite obviously in the fact that the form displays declension (genitive, dative, accusative and ablative singular). We therefore find it functioning as the complement of verbs, prepositions, adjectives, or nouns, as in: denegandi pudor (Ten, And. 630) 'lack of courage to say no' peccandi finem (Ter., Phorm. 23) 'end to the offending' inpotiundopericlum (Ter., Heaut. 323) 'risk to be master of...'

The nominal nature of the gerund also shows in a perhaps less obvious, but nonetheless indisputable way, in another characteristic: as a rule the gerund does not combine with a direct object. We generally do not find structures of the type: tempus oleam time-Norn, olive tree-Ace. 'time to plant an olive tree'

serendi plant-Ger.-Gen.

Instead, we find a construction featuring a gerundive: tempus oleae serendae (Cato, R.R. 61.2) time-Norn, olive tree-Gen. plant-Gerv.-Gen. 'the (proper) time to plant an olive tree'

This construction is characterized by agreement, hence nominal syntax. I do not argue that instances of a gerund with a direct object did not occur at all; they did occasionally, even in early times. I am arguing, however, that the alternative solution, the gerundive construction, was much more common throughout the Latin period, as we see in Latin texts and in Early Latin inscriptions., cf. aedem faciendam coeraverunt temple-Ace. construct-Gerv.-Acc. take-care-Pf.-3pl. 'they took care of the construction of the temple' (Corpus Inscriptionum Latinarum 1511; Warmington 1979: 186)

Moreover, whereas there is no evidence of gerunds — with or without a direct object - in Oscan and Umbrian, we have examples of gerundive

286

Brigitte L. M. Bauer

constructions. The Oscan and Umbrian gerundives display the same form as their Latin equivalent, although the sequence -nd- has undergone the regular (in Oscan and Umbrian, that is) assimilation to -««-, cf. Umbrian pihaner, which corresponds to Latin piandae 'purify,' or Oscan sakrannas, which corresponds to Latin sacrandae 'consecrate'. Syntactically, we are dealing with the same construction, as in: arsfertur trebeit ocrer peihaner adfertur-Nom. remain-3sg.-Pr. mount-Gen. purify-Gerv.-Gen. 'the adfertur remains for the purification of the mount' (Iguvian Tables VIA-8; Poultney 1959: 232-234)

These instances show the archaic nature of this type of structure. The archaic character of this type of structure also shows in its parallel structure, the so-called ab urbe condita construction: ab incenso Capitolio (Sail., Cat. 47.2) from set-fire-Ppf.-Abl.sg. Capitol-Abl. 'since the burning of the Capitol'

This type of structure is the perfective variant of the gerundive construction. The distinction between these constructions is aspectual, which points to archaism. The basic characteristics of gerundive constructions in Oscan, Umbrian, and Latin are the occurrence of a verbal adjective denoting a transitive action — instead of a finite verb denoting this action — and the absence of an accusative referring to what from our modern perspective would be a direct object. Instead of a finite verb governing a direct object, we find a verbal adjective that agrees in number and in case with the noun that refers to what from our linguistic perspective is the direct object of the construction, hence: tempus oleae serendae (Cato, R. R. 61.2)

and not, as a rule: tempus oleam serendi Syntactically, the noun is head of the phrase [noun + gerundive], whence adjectival agreement. In contrast, in our modern perspective -

Nominal syntax in Italic: a diachronic perspective

287

and this also shows in our translation - the verb is head of the phrase and the noun is direct object, hence a complement. With time, however, a dramatic change took place, when the gerund came to govern, as a rule, a direct object. This is a late development that occurred after Oscan and Umbrian had disappeared as written languages. Once this change had occurred, the replacement of the gerund by the infinitive and of the ablative of the gerund by the participle became possible. Consider, for example, the well-known sentence from the Peregrinatio that features a gerund, where one might expect a participle: redirent... dicendo psalmos (15.5; Maraval 1982) return-Subj.-Impf.-3pl. say-Ger.-Abl. psalms-Ace, '(that) they returned saying psalms'

It is interesting to note that not only did the verbal form come to govern a direct object, but it also lost its potential for declension: the infinitive as well as the ablative of the gerund, which became a fossilized form, are not declinable. (For an analysis of the gerund/gerundive and its evolution, cf. Bauer 1993.) The gerundive also occurred in another context, combined with a noun in the nominative and a third person form of the verb esse: frater est exspectandus brother-Norn. be-3sg.-Pr. wait-Gerv.-Nom.sg. Ί have to wait for my brother'

mihi (Ter., Phorm. 460) me-Dat.

Here again the verbal adjective denotes a transitive action and displays agreement of case and number with the noun in the nominative, which in our modern linguistic perspective refers to the direct object of the action expressed by the verbal adjective, legendus. In addition, the "agent" of the action is referred to in a casus obliquus instead of in a nominative, as one might expect in a nominative language. This structure also has a perfective variant of the type: epistulae mihi letters-Nom. me-Dat. Ί have read the letters' fides est faith-Norn. be-3sg.-Pr. 'we have come to know

lectae sunt read-Ppf.-Nom.pl. be-3pl.-Pr. nobis cognita (Cic., Div. in Caecil. 20) us-Dat. know-Ppf.-Nom.sg. his fidelity'

288

Brigitte L. M. Bauer

In both types of structure that feature the gerundive (with and without est), we observe a syntactic organization different from modern syntax: (1) (2)

(3)

instead of a finite verb form, we find a verbal adjective; instead of a relation of government, whereby the verb governs a direct object, typical of verbal syntax, we find adjectival agreement, characteristic of nominal syntax; the head of the construction is not the verb, but the noun, with which the verbal adjective agrees in case and number: tempus \\legendarum} epistularum}. These characteristics are not isolated phenomena, as I will attempt to show.

5.2. Mihi est constructions Constructions featuring a noun, a gerundive — or a perfective participle — an "agent" in the dative, and a form of the verb 'be' (third person), are closely related to mihi est constructions expressing possession, e.g.: cui sit who-Dat. be-3sg.-Subj.-Pr. (that) he may have a wife'

uxor (PI., Cas. 519) wife-Norn,

Hecyra est huic nomen fabulae (Ter., Hec. 1) Hecyra be-3sg.-Pr. this-Dat. name-Norn. play-Dat. 'this play is called Hecyra'

These structures typically combine a form of the verb 'be' (third person), a noun in the nominative that refers to what from our perspective is the object in possession, and a noun or pronoun in the dative denoting the possessor. The possessive mihi est construction, therefore, typically displays the following structure: noun-Norn. + possessor-Dat. + the stative verb 'be'

We observe here basically the same type of syntax as in the gerundive construction discussed in Section 5.1. The verbal mihi est construction shows the following organization: noun-Nom. + Gerv.-Nom. + "agenf'-Dative + stative verb liber legendus mihi est

Nominal syntax in Italic: a diachronic perspective

289

This construction expresses what from our perspective would be a transitive action. As both types of example show, mihi est constructions are either verbal or possessive. In the later development of Latin and the Romance languages, we observe a shift towards transitive constructions. Gradually possessive mihi est came to be replaced by habeo, which combines with a possessor in the nominative and the element in possession in the accusative. In a parallel change, mihi est constructions combined with gerundives or perfective participles also came to be replaced with habeo, which governs a direct object: ibi agrum colendum habebat (Ter., Phorm. 365) there land-Ace. cultivate-Gerv.-Acc. have-3sg.-Impf. 'there he had land to farm' trecentos nummos numerates habet (PL, Poen. 594) 300-Acc. sesterces-Ace. count-Ppf.-Acc.pl. have-3sg.-Pr. 'he has (his) three hundred sesterces well counted'

I will not discuss here the specific use of possessive mihi est and its replacement by habeo in the Romance languages (cf. Bauer, in press), because I want to focus on the similarities between possessive and verbal mihi est constructions. The verbal mihi est construction represents a pattern that was inherited: the combination of an "agent" in an obliquus (generally the dative), a third person of the verb 'be,' and a nominal form of the verb is common in early Indo-European languages. These forms as such may vary from one language to the other, but they are nominal forms of the verb. It is therefore legitimate to assume that we are dealing with an inherited pattern. The same observation can be made with respect to the possessive mihi est construction. In our modern linguistic system possession is rendered by the verb habeo 'have,' which displays transitive syntax. Comparison with other Italic and Indo-European languages shows that the mihi est construction indicating possession - despite its variation with the genitive in some languages — is inherited from the protolanguage (cf. Bauer, in press), cf. Hittite or Umbrian, for example (Iguvinian Tables VIIB-4; Poultney 1959: 294):

290

Brigitte L. M. Bauer

fratreci motor sins a. CCC brother-superior-Dat. fine-Gen. be-3pl.-Subj. asses-Norn. 300 'the brother-superior will have a fine of 300 asses'

In addition to being archaic and inherited, possessive and verbal mihi est constructions share a number of characteristics: 1. a (third) person form of the verb esse, which is a stative verb; 2. a noun in the nominative that refers to what from our perspective is a direct object; 3. agent/possessor referred to in an obliquus in Italic as in most early Indo-European languages; and, 4. whereas in possessive constructions the object in possession determines the subject-verb agreement, it is the "object" in verbal constructions that determines, on the basis of adjectival agreement, case and number of the verbal form, the gerundive. Although the examples of verbal mihi est I have discussed until now all feature transitive verbs, we also find instances with intransitive verbs, cf. mihi saltandum est (PL, Stich. 754) Ί have to dance' vivendumst mihi (Ten, And. 152) Ί have to live...' The structures we are dealing with therefore not only show a different way of expressing action, they also turn out to be transitivity independent.

5.3. Absolute constructions The specific characteristics of mihi est and gerundive constructions show not only in their grammar, but also in their development. Similarly, the grammatical patterns of absolute constructions as well as their development reveal the specific characteristics of these structures. Absolute constructions were inherited from the protolanguage as their archaisms and their occurrence in all early Indo-European languages show (cf. Meillet and Vendryes 1924:556 and Coleman 1989). The

Nominal syntax in Italic: a diachronic perspective

291

variation in case (absolute dative in Germanic and Slavic, absolute genitive in Greek, and so on) is language-specific, and is related to the development of the case system in the individual languages. Consequently case variation is not an argument against the assumption that absolute constructions are inherited. Absolute constructions are aspectually marked: they appear in Latin with perfective as well as imperfective participles. Yet the imperfective variant is less frequent - it was rather rare - and in early times only combines with intransitive and Stative verbs. Similarly the following example from Oscan also displays an imperfective absolute construction featuring a stative verb: toutad praesentid (Tabula Bantina 21; Buck 1979: 233) people-Abl. be-present-Ppr.-Abl. 'the people being present'

Only at a later stage do transitive verbs also occur in this context, subsequently with a direct object. Whereas imperfective absolute constructions typically display the "agent" of the action, the agent is absent in perfective absolute constructions, which feature, instead, the locus of action, cf.: insciente domino (Cato, R.R.. 5.4) ignore-Ppr-Abl. master-Abl. 'the master not knowing, without the knowledge of the master'

vs.

the following example from Umbrian aves anzeriates (Iguvinian Tables la-1; Poultney 1959: 158) birds-Abl. observe-Ppf.-Abl.pl 'after having observed the birds'

Yet here again there is no verbal syntax: the verb does not govern what from our modern perspective would be its direct object; instead, the noun is head of the construction, as adjectival agreement shows. Consequently, as in the gerundive constructions discussed in the previous section, the locus of action determines the morphological form of the verbal adjective. Furthermore, the relation between the verbal form and the noun is not one of government, but of agreement. The specific, archaic characteristics of ablative absolute constructions therefore show (1) in the aspectual opposition between both

292

Brigitte L. M. Bauer

variants; (2) in the occurrence of a nominal form of the verb; (3) in the nominal declension: we are dealing with ablatives; and (4) in the adjectival agreement between the verbal adjective and the noun. Consequently, the locus of action is not referred to in an accusative, as one might expect in a nominative language. Instead, the transitive action is expressed by a verbal adjective that agrees in case and number with the noun that in our modern linguistic perspective is the direct object of the action expressed by the verbal form. This brief sketch of the characteristics of this construction not only shows its archaic nature, but also its non-transitivity. Instead of government, hence verbal syntax, we observe adjectival agreement, hence nominal syntax. The subsequent development from Latin to Romance supports this interpretation: while in the end absolute constructions came to be replaced with subordinate clauses featuring a finite verb and a direct object, Late Latin displays an interesting development. With time, we observe that absolute perfective constructions introduced a direct object, showing that transitivity was spreading into nominal constructions. In later texts, we therefore find numerous instances of a perfective participle in the ablative with a direct object in the accusative, as in: conpleto matutinas... procedunt complete-Ppf.-Abl.sg. morning-prayers-Ace. go-3pl.-Pr. 'having completed the morning prayers, ... they go on ...' (Itinerarium Antonini Placentini 11, cf. Löfstedt 1911: 292)

Since the participle varied in this context, there is no reason to assume that the participle at that stage had become a mere preposition (as suggested by e.g. Väänänen 1981: 167—168). Unfortunately, this development was interrupted when Latin became a learned language and stopped being a mother tongue: this phenomenon coincided with the spreading of the absolute accusative (displaying perfective participles) and absolute nominatives (displaying the imperfective participle), cf: frater manus ... extensas... coepit brother-Nom. hands-Ace, stretch out-Ppf.-Acc.pl. begin-Pf.-3sg. 'the brother his hands stretched out started...' ("Visio Baronti", Monumenta Germaniae Historica, Scriptorum Rerum Merovingicarum 5.378.8, Biese 1928: 27)

Nominal syntax in Italic: a diachronic perspective

293

Since absolute accusatives, as opposed to absolute nominatives, were typically perfective and displayed specific syntactic patterns, as I pointed out earlier (Bauer 1994), this development cannot be ascribed to the mere disintegration of the case system. Yet these accusative constructions, which survived in the Romance languages, were limited to a restricted number of verbs, and were merely appositives indicating the circumstances in which the action expressed by the main, finite, verb of the clause was taking place (cf. Bauer 1994). Absolute constructions had therefore become fossilized forms; they had given way to subordinate clauses featuring an explicit subject, a finite verb, and a direct object. This fits the shift towards transitivity.

6. Conclusion The grammatical patterns of mihi est constructions, and of gerundive and absolute constructions in Italic, reveal the nominal nature of these structures. The development they underwent in a later stage is characterized by the replacement of (1) mihi est by habeo, which displays transitive syntax; and (2) the introduction of transitive syntax in nominal constructions (gerunds and perfective participles eventually governing a direct object). Consequently, non-transitive structures characterized by stative verbs and adjectival agreement came to be replaced with constructions based on government, hence verbal syntax. Gamkrelidze and Ivanov (1984) pointed out the importance of possessive mihi est constructions, emphasizing their stative character and the implications of their existence for the reconstruction of the protolanguage. Yet as I have attempted to show, non-transitivity is also apparent in verbal mihi est constructions (perfective as well as imperfective) and in the syntax of the nominal forms of the verb that do not govern a direct object, but instead display adjectival agreement. The correlation between the structures I analyzed in this paper seems to be clear: they all display nominal syntax. In addition, the parallel development they underwent marks the spreading of transitivity: the last residues of non-transitive syntax thus disappeared. Consequently, those structures that at first sight seem to be aberrations turn out to be related, as we can observe in their grammatical patterns and in the parallel development they underwent. The replacement

294

Brigitte L. M. Bauer

of the possessive and verbal mihi est constructions and the development of government in structures that at an earlier stage were characterized by nominal agreement, show the increasing importance of transitivity in Italic. The same development can be observed for other Indo-European languages. Just as nominal structures were not a phenomenon restricted to Italic, their replacement with transitive structures can be observed in other Indo-European languages also; consider, for instance, the verbs expressing 'have' that replace mihi est constructions in the various languages. In diachronic linguistics we observe two types of change: change that is due to external factors and that is language-specific, and inherent language change, which occurs in all languages independently of each other and at various rates; this last type of change is not due to external factors, although external factors may delay or accelerate the inherent change that is taking place (cf. Meillet's earlier distinction between "lignes generates" and "divergences" [1926b] or Bopp's notion of language change [Lehmann 1991]). At the very beginning of this article (Section 1) I referred to this kind of change, mentioning the development of subordination and the shift from left to right branching as telling examples. Since the development of transitivity is a wide-spread overall change that takes place in all Indo-European languages, independently of each other, it is legitimate to assume that it is an inherent language change also. Inherent language change is rather difficult to explain, because its causes, which are not external, are difficult to identify. In addition, the very beginning may be difficult to pinpoint, especially when we are dealing with long-term changes, which start very early. Since ProtoIndo-European was a nominative language, the shift to transitivity occurred before that stage, that is before the stage at which the protolanguage is usually reconstructed. Similarly, the change in word order in Indo-European languages started before historical times. Earlier (Bauer 1995) I demonstrated that the grammatical left-branching structures of Latin, which were inherited from Proto-Indo-European, were replaced over time by their right-branching equivalents. Whereas this change took place in all structures, it occurred in some earlier than in others. Consequently, a number of structures underwent the shift before Latin was documented. It is on the basis of relics of the former structures we find in Latin, e.g., that we know that there used to be a different ordering pattern. Their specific use, their subsequent develop-

Nominal syntax in Italic: a diachronic perspective

295

ment as well as comparative analysis enable us to evaluate these forms and to interpret them as residues. Residues inform us, therefore, on the previous stage of the language. This can be easily illustrated by referring to adpositions in Latin. Although Latin featured prepositions, we observe a number of constructions that display postpositions, such as mecum, literally 'me-with.' Their specific context — that of relative and personal pronouns — their occurrence as full-fledged postpositions in other early Indo-European languages (cf. Umbrian -ku(m); see also Hittite's postpositions), and their subsequent replacement with a preposition (avec in French [< apud hoc]) allow us to interpret these structures: they are residues of an earlier stage characterized by left-branching structures (cf. Bauer 1995). Similarly, the structures I analyzed in this paper can be considered residues of a language type whose binding factor was not transitivity, or more precisely the opposition agent-Nominative vs. patient-Accusative. The definite stative character of these structures as shown in the occurrence in the verb 'be' as well as in nominal syntax may point to active typology. The grammatical patterns of the structures I analyzed in the preceding pages therefore give further evidence for the importance of agreement at an earlier stage of the protolanguage. On the basis of their characteristics, agreement - hence nominal syntax - and their stative nature, these structures support the assumption that Indo-European at an early stage was characterized by different, presumably active, typology. The results of the analysis of nominal structures in Italic are, therefore, twofold: (1) these structures inform us on the earlier stages of Proto-Indo-European; and (2) their subsequent development can be accounted for as being in accordance with the ever increasing importance of transitivity. One of the main questions this analysis brings up and which needs further research is, therefore, why transitivity did develop in the first place, or put differently, why the protolanguage at some point switched to a nominative system. In order to answer this fascinating question it will perhaps be necessary to refer to other fields, such as psycholinguistics, anthropology, or psychology. Relating diachronic linguistic evidence with findings from other (related) fields can be illuminating, as I showed in an earlier study on diachronic wordorder change and its correlates with acquisition patterns in child language (Bauer 1995). The evidence we now have on the current topic and its correlates would put us in a position of speculation, which we do not

296

Brigitte L. M, Bauer

want. In addition it would push us beyond the scope of this paper, which deals with the spreading, rather than the development of transitive patterns. This brings us to the second question brought up by the analysis of nominal structures in Italic and their development: why have they been replaced? As pointed out earlier, the preceding analysis of nominal structures has allowed us to identify the true nature of mihi est constructions, and gerundive and absolute constructions. Their very identification as residues of an earlier stage allows us at the same time to explain their eventual disappearance. Proto-Indo-European was definitely a nominative language and structures of the type analyzed in this paper were as it were corpora aliena since they did not fit the typical pattern of the language, that is a verb combined with a subject-agent and a direct object-patient. These patterns were obligatory for contexts featuring a finite verb from Proto-Indo-European onward. The replacement of the archaic constructions therefore shows the spreading of transitivity: nominal structures, which do not fit the predominant patterns, are gradually ousted out of the language. The new structures fit a language in which the binding factor, or "inner motivation," is transitivity, rather than nominal agreement. Syntactic regularity therefore seems to have been the driving force behind the spreading of transitivity as reflected in the evolution of nominal syntax in Italic. Abbreviations used in this article: Abl. Ace. Dat. Gen. Ger. Gerv. Impf. Nom. Pf. PL Ppf. Ppr. Pr. Sg. Subj.

ablative accusative dative genitive gerund (Latin gerundium) gerundive (Latin gerundivum) imperfective nominative perfect plural participle perfective participle present present singular subj unctive

Nominal syntax in Italic: a diachronic perspective 3sg. 3pl.

297

third person singular third person plural

Authors and works: (Cato) R.R.

De Re Rustica

Cic. Div. in Caecil.

Cicero Divinatio in Caecilium

PI. Cas. Poen. Stich.

Plautus Casina Poenulus Stichus

Sail. Cat.

Sallust Catilina

Ter. And. Heaut. Hec. Phorm.

Terence Andria Heautontimorumenos Hecyra Phormio

Acknowledgements I am much indebted to the Niels Stensen Foundation, whose grant allowed me to spend one year (1992-1993) at the University of Texas at Austin (Prof. W.P. Lehmann) and the University of Cambridge (Prof. R. Coleman). I owe much to the fruitful and stimulating discussions I had with Professor Lehmann and Professor Coleman on the subject of my new research project on archaic structures and their development. Currently a grant from the Netherlands Royal Academy of the Arts and Sciences allows me to continue my research in the Netherlands and (during extended stays) at the University of Texas at Austin and at the University of Cambridge, for which I am most obliged. Finally, I also thank the editor of this volume, Dr. Jadranka Gvozdanovic, for her valuable comments on an earlier version of this article.

298

Brigitte L, M. Bauer

References Aalto, Pentii 1949

Untersuchungen über das lateinische Gerundium und Gerundivum. Annales Academiae Scientiarum Fennicae 62—63. Helsinki: Suomalainen Tiedeakatemia. Adelung, Johann C. 1806-1817 Mithridates, oder allgemeine Sprachenkunde mit dem Vater Unser als Sprachprobe in bey nahe fönfhundert Sprachen und Mundarten. 3 Vol. Berlin: Voss. Bauer, Brigitte L. M. 1993 "The Coalescence of the Participle and the Gerund/Gerundive: An Integrated Change", Historical Linguistics 1989. Papers from the 9th International Conference on Historical Linguistics. Henk Aertsen - Robert J. JefFers (eds.). Amsterdam: Benjamins. Pp. 59-71. 1994 "The Development of Latin Absolute Constructions: From Stative to Transitive Structures", General Linguistics 33: 64-83. 1995 The Emergence and Development of SVO Patterning in Latin and French. Diachronie and Psycholinguistic Perspectives. Oxford and New York: Oxford University Press. In press "Non-nominative Syntax in Latin: The Mihi Est Construction." Historische Sprachforschung 109. Benveniste, Emile 1966 Problemes de linguistique generale. I. Paris: Gallimard. Biese, Yrjoe Moses 1928 Der spätlateinische Akkusativus Absolutus und Verwandtes. Helsinki: Suomalainen Tiedeakatemia. Buck, Carl Darling 1979 (1904) A Grammar ofOscan and Umbrian. Hildesheim: Georg Olms. Coleman, Robert 1989 "The Rise and Fall of Absolute Constructions: A Latin Case History", in: Gualtiero Calboli, (ed.), Subordination and other Topics in Latin. Amsterdam: Benjamins. Pp. 353-374. Conway, Robert Seymour 1897 The Italic Dialects. Cambridge: Cambridge University Press. Croft, William 1990 Typology and Universals. Cambridge and New York: Cambridge University Press. Delbrück, Berthold 1900 Vergleichende Syntax der indogermanischen Sprachen. 3. Syntax. Strassburg: Trübner.

Nominal syntax in Italic: a diachronic perspective

299

Denton, Michael 1985 Evolution: A Theory in Crisis. London: Burnett. Ernout, Alfred - Francois Thomas 1951 Syntaxe latine. Paris: Klincksieck. Foulet, Lucien 1923 Petite syntaxe de rancienfrangais. Paris: Champion. Friedrich, Johannes 1974 Hethitisches Elementarbuch. 1. Teil. 3 ed. Heidelberg: Winter. Gamkrelidze, Thomas - Vjaceslav Ivanov 1984 Indoevropejskij jazyk i indoevropejcy. [Indo-European and the Indo-Europeans.} Tbilisi: Tbilisi State University. Gougenheim, Georges 1973 Grammaire de la langue francaise du XVle siede. Paris: Picard. Greenberg, Joseph H. 1963 "Some Universals of Grammar with Particular Reference to the Order of Meaningful Elements", Universals of Language. Cambridge, Mass.: The M.I.T. Press, pp. 58-90. Jakobson, Roman 1962 "Typological Studies and their Contribution to Historical Comparative Linguistics", Selected Writings I. 's-Gravenhage: Mouton, pp. 523-532. Klimov, Georgij 1977 Tipologija jazykov aktivnogo stroja ["The Typology of Languages of the Active System"]. Moskva: Izdatel'stvo "Nauka." 1983 "On Contentive Typology." Lingua e stile 18: 327-341. 1986 "On the Notion of Language Type", in: Winfred P. Lehmann, (ed.), Language Typology 1985. Amsterdam: Benjamins, 105-110. 1990 "Integral vs. Partial Typology", in: Werner Bahner - Joachim Schildt - Dieter Viehweger (eds.), Proceedings of the Fourteenth International Congress of Linguists. HI. Berlin: Akademie Verlag. Pp. 149-154. Kury-fowicz, Jerzy 1964 The Inflectional Categories of Indo-European. Heidelberg: Winter. Lehmann, Winfred P. 1958 "On Earlier Stages of the Indo-European Nominal Inflection", Language 34: 179-202. 1989 "Problems in Proto-Indo-European Grammar: Residues from Pre-Indo-European Active Structure", General Linguistics 29: 228-246. 1991 "Franz Bopp's Use of Typology", Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 44: 276-284.

300 1993

Brigitte L. M. Bauer

Theoretical Bases of Indo-European Linguistics. London and New York: Routledge. Leumann, Manu - Johann Baptist Hofmann 1974 Lateinische Grammatik. München: Beck. Löfstedt, Einar 1911 Philologischer Kommentar zur Peregrinatio Aetheriae. Uppsala: Almquist & Wiksell. Maraval, Pierre 1982 Egerie. Journal de voyage. Paris: Cerf. Marchello-Nizia, Christiane 1979 Histoire de la langue francaise aux XIV et XVe siecles. Paris: Bordas. Martinet, Andre 1962 A Functional View of Language. Oxford: Clarendon. Meillet, Antoine 1926 a "La categorie du genre et les conceptions indo-europeennes", Linguistique historique et linguistique generale. Paris: Champion, pp. 211-229.. 1926b Linguistique historique et linguistique generale. Paris: Champion. Meillet, Antoine - Joseph Vendryes 1924 Tratte de grammaire comparee des languages classiques. Paris: Champion. Nichols, Johanna 1986 "Head-Marking and Dependent-Marking Grammar", Language 62:56-119. Poulmey, James W. 1959 The Bronze Tables of Iguvium. Oxford: Blackwell. Prokosch, Eduard 1939. A Comparative Germanic Grammar. Philadelphia: University of Pennsylvania Press. Sapir, Edward 1949 (1921) Language. An Introduction to the Study of Speech. New York: Harcourt. Schmidt, Karl Horst 1979 "Reconstructing Active and Ergative Stages of Pre-Indo-European", in: Frans Plank (ed.), Ergativity. London: Academic Press, pp. 333-345. Schmidt, P. Wilhelm 1926 Die Sprachfamilien und Sprachenkreise der Erde. Heidelberg: Winter.

Nominal syntax in Italic: a diachronic perspective

301

Sturtevant, Edgar H. 1951 A Comparative Grammar of the Hittite Language. New Haven and London: Yale University Press. Väänänen, Veikko 1981 Introduction au latin vulgaire. Paris: Klincksieck. Warmington, Eric Herbert 1979 Remains of Old Latin. IV Archaic Inscriptions. Cambridge, Mass.: Harvard University Press and London: Heinemann.

Bibliographical references of Latin texts mentioned in this article: Caesar. Cato.

Cicero. Plautus. . . Sallust. Terence. . . .

De Bella Gallico. H.J. Edwards, ed./transl. (1986). Cambridge, MA: Harvard University Press; London: Heinemann. De Re Rustica. William Hooper and Harrison Boyd, eds./transls. (1979). Cambridge, MA: Harvard University Press; London: Heinemann. Divinatio in Caecilium. H. De La Ville de Mirmont, ed./transl. (1960). Discours II. Paris: Les Belles Lettres. Casina. Paul Nixon, ed./transl. (1988). Cambridge, Mass.: Harvard University Press; London: Heinemann. Poenulus. Paul Nixon, ed./transl. (1959). Cambridge, Mass.: Harvard University Press; London: Heinemann. Stichus. Paul Nixon, ed./transl. (1960). Cambridge, Mass.: Harvard University Press; London: Heinemann. Catilina. J.C. Rolfe, ed./transl. (1985). Cambridge: Harvard University Press; London: Heinemann. Andria. John Sargeaunt, ed./transl. (1986). Cambridge: Harvard University Press; London: Heinemann. Heautontimorumenos. John Sargeaunt, ed./transl. (1986). Cambridge: Harvard University Press; London: Heinemann. Hecyra. John Sargeaunt, ed./transl. (1983). Cambridge: Harvard University Press; London: Heinemann. Phormio. John Sargeaunt, ed./transl. (1983). Cambridge: Harvard University Press; London: Heinemann.

Index of subjects

abduction 25 f. abstractness 57 accomodation 122 acquisition 295 lexical 58 acrolect 153 actualization 21 adoption 2, 112, 114 f., 120 f. agent in passive clauses 71-75, 79-81, 83 f. Aghu 96ff., 101, 102 alternative congruent 25 ambiguity resolution 54-57 analogy 1 surface 128 anaphor logophoric use of 30 aspect 273, 291 assertive(s) 225 Awyu 3, 89 f., 92 anti-functionalist(s) 9 backgrounding 113, 115f, 132 basilect 153 beneficial 3, 18 bilingualism 120f. bleaching 26,43 borrowing 1, 120f. branching left- 273, 294 f. right- 273,294 calculus cognitive

54-58, 60ff, 65, 67

cause 73, 75, 84 causer 73, 75, 82, 84 Celtic 116 change analogical 125f. inherent 294 linguistic 60 locus of 112 structural 78 choice rational 16, 19 circumstance 73 accessory 79 f. conditioning 79 f. class inflectional 126 classification typological 275 clause adverbial 250 ff. semantic type of 250 ff. expression format of 250 finite vs. non-finite 90, 92, 97 clause-chaining 89 ff., 99 coding arbitrary 102 cognition 54f, 67 competition 22, 27f, 30, 40, 47 condition 79 congruence contextual 26 syntagmatic 42,44 constraint on underlying representation(s) 57 construction absolutive 274 gerundive 274

304

Index of subjects

contamination 128 context(s) categorical 24, 26 continuum of 40 neutral 24,26 variable 25f., continuum contextual 23 semantic-pragmatic 43, 45 contra-factive 255 f. creolization 121 dialect 143 primary 172 diffusion 23, 26, 40, 42, 46 lexical 202,204 diminutive 129 diphthongization 117 disagreement 237 document(s) 14th century 203 Dutch 143, 181-207 effect perlocutionary 225 Elamite 281 English 55, 115, 118f, 124 entity type(s) 251 f., 257 first order 251 second order 251, 254 third order 251,254 fourth order 251,254 explanation formal 109 functional 2ff, 9, 110-131 invisible-hand 15 f., 19 exponence fused 63 expressive(s) 225 face 226 negative 221 positive 221

factive 255 f. feedback-elicitor(s) 233 foregrounding 117,132 form hyperdialectic 156, 158, 160-170, 176f. intermediate 168 fortition 132 French 115,119,273 frequency of use 22, 26 of words 186 Frequency Actuation Hypothesis 204, 206 function 3, 9, 14, 16, 18 functionalist(s) 9ff., 15 functionality 13, 18 German 14, 114, 118f, 124-128 Germanic 291 West 182, 186, 206 gerund 284 grammaticalization 90, 97, 101, 124 Greek Ancient 116, 128,282 hierarchy 2, 4, 77, 113 implicational 153, 155, 157, 170, 249, 253 Hittite 282 hyperaccomodation 118 f. hyperbole 222 hypercharacterization 127 hypercorrection 118 f. implicature conversational 222 indicative 250, 258 ff. Indian 281 Indo-Aryan 118 Indo-European 294 infinitive 284, 287

Index of subjects influence surface phonetic 204 innovation If., 112, 134 instrument 17 instrumentality 18 interjection 236 Iranian Middle 117 irony 222, 243 Italian 125, 128f. Italic 273 Javanese

62

Kombai 90, 92ff., 10If. Korowai 98, 102 language active 278 f. agglutinative 276 analytic 276 ergative 278 nominative 274,278 standard 143, 145, 152, 168 synthetic 276 language contact 119f. language decay 122 language shift 121 Latin 115ff., 127f., 273f., 282 Latvian 5 5 ff. learning imperfect 145, 171, 173 without insight 17 lenition 113, 132 lexicalization 42 linkage 112 maxim(s) Gricean 219 mesolect 153 metanalysis 61 metaphor 222

305

mis-analysis 118 modal deontic 243 monophthongization 114,116 morphologization 122 ff. morphology extragrammatical 127, 129 grammatical 134 inflectional 125 motivation 21, 23, 44 neologism 129 f. non-factual 254,256 neutralization 55ff. occasionalism 130 occurrence categorical 39 f., 43 omnitemporal 263 f., 267 order spontaneous 15 origin 90, 11 If. Oscan 274 Papuan 3, 89 paradigm inflectional 125f. participle 287 perspective 26,35, on a neutral context 25 speaker's 24 phenomena of the third kind 15 Polish 59, 79-81, 122 f. pollution teleological 13 polyfunctional 261 possession 280 postposition 276 Pre-Indo-European 274 preposition 276 prestige 179f., 184, 187, 194 f., 197

306

Index of subjects

problem actuation 110, 112 process invisible-hand 18, 171, 173 pronoun inclusive 280 exclusive 280 Proto-Austronesian 62 Proto-Indo-European 273 Proto-Polynesian 62 psycholinguistics 53-70 quality maxim of 222 question(s) batteries of 228 echo 228 informative 221 non-polar 223 polar 223 reading logico-mathematical 14 teleo logical 14 reanalysis 1,21,23,35 perceptual 118 recovery cognitively-aided 67 structural 66 regiolect 145, 172 reinforcement syntagmatic 27 f. reinterpretation 22 f., 26, 37, 40, 42 ff., 46, 92, 95 f. qualitative 25 reiteration 22 replacement 27, 40, 44 reproach(es) 231 residue 295 Romance 116,292 rule 17 Russian 59, 79

Samoan 6 If. Sanskrit 282 scenario 111-131 sense beneficiency 19 logico-mathematical 3, 18f. teleological 3, 19 Serbo-Croatian Croatian and Serbian varieties of 72-84 shibboleth 180,205 shift gradual 183 Slavic 72, 291 Slovene 78 f. solidarity in-group 233 sound change physiologically motivated 204 source 73, 75, 82, 84 Spanish 3, 4, 27-47, 115, 127, 219-243 speech casual 113 f. speech act direct 221 indirect 219 metadiscursive 225 speech processing 58 spread 112,114 stability of language elements 143,171 subjunctive 250, 258 ff. future 265 subordination 273 support contextual 24 f., 42 Swedish 60f., 63 ff. switch-reference 89 f., 92f., 99 f. syncretism 76, 78, 85 syntagm reinforced 37

Index of subjects syntax nominal 273,288 verbal 288

unpredictability 12 use cumulative 23

teleology 10 f., 13 tellability 225 theory invisible-hand 130 politeness 219 therapy metaphor 18 time reference 253, 258, 267 tool 17 transitivity 273 typology 249 f. contentive 277

value diference in 28, 42 factuality 255 ff, 267 variation 23 f., 45 contexts of 24 typological 249 verb impersonal 280 vocative 232

Umbrian 274 univerbation 125

Wambon 92-96, 101 word order 276, 294

307