Diachronic Clues to Synchronic Grammar (Linguistik Aktuell Linguistics Today)

Diachronic Clues to Synchronic Grammar Linguistik Aktuell/Linguistics Today Linguistik Aktuell/Linguistics Today (LA)...

Author: Eric Fuss | Carola Trips

23 downloads 1006 Views 1MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Diachronic Clues to Synchronic Grammar

Linguistik Aktuell/Linguistics Today Linguistik Aktuell/Linguistics Today (LA) provides a platform for original monograph studies into synchronic and diachronic linguistics. Studies in LA confront empirical and theoretical problems as these are currently discussed in syntax, semantics, morphology, phonology, and systematic pragmatics with the aim to establish robust empirical generalizations within a universalistic perspective.

Series Editors Werner Abraham

Elly van Gelderen

University of Vienna

Arizona State University

Advisory Editorial Board Guglielmo Cinque

Ian Roberts

University of Venice

Cambridge University

Günther Grewendorf

Ken Safir

J.W. Goethe-University, Frankfurt

Rutgers University, New Brunswick NJ

Liliane Haegeman

Lisa deMena Travis

University of Lille, France

McGill University

Hubert Haider

Sten Vikner

University of Salzburg

University of Aarhus

Christer Platzack

C. Jan-Wouter Zwart

University of Lund

University of Groningen

Volume 72 Diachronic Clues to Synchronic Grammar Edited by Eric Fuß and Carola Trips

Diachronic Clues to Synchronic Grammar Edited by

Eric Fuß J. W. Goethe-University, Frankfurt

Carola Trips University of Stuttgart

John Benjamins Publishing Company Amsterdam/Philadelphia

8

TM

The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.

Library of Congress Cataloging-in-Publication Data Diachronic clues to synchronic grammar / edited by Eric Fuß, Carola Trips. p. cm. (Linguistik Aktuell/Linguistics Today, issn 0166–0829 ; v. 72) Includes bibliographical references and index. 1. Linguistic change. 2. Grammar, Comparative and general. I. Fuß, Eric. II. Trips, Carola. III. Linguistik aktuell ; Bd. 72. P142.D497 2004 417’.7-dc22 isbn 90 272 2796 9 (Eur.) / 1 58811 587 9 (US) (Hb; alk. paper)

2004057651

© 2004 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa

Table of contents

Preface Introduction Eric Fuß and Carola Trips

vii 1

On the development of possessive determiners: Consequences for DP structure Artemis Alexiadou

31

Diachronic clues to pro-drop and complementizer agreement in Bavarian Eric Fuß

59

Syntactic effects of inflectional morphology and competing grammars Eric Haeberli Language change versus grammar change: What diachronic data reveal about the distinction between core grammar and periphery Roland Hinterhölzl

101

131

The EPP, fossilized movement and reanalysis Andrew Simpson

161

Restructuring and the development of functional categories Zoe Wu

191

Index

219

Preface

The contributions in this volume are a selection of papers that were originally presented at the workshop Diachronic Clues to Synchronic Syntax held in November 2002 in Stuttgart, which focused on the idea that sometimes synchronic phenomena can only be fully understood by looking at their diachrony. Of the six papers, the contributions by Zoe Wu and Eric Fuß were not presented at the workshop. The papers presented by Anthony Kroch and Barbara Vance could unfortunately not be included in the volume. We would like to take this opportunity here to thank the participants of the workshop for taking part in it and also for helping us to create this volume.

Introduction Eric Fuß and Carola Trips Universities of Frankfurt and Stuttgart

Sometimes, an old idea in a new context can provide the right insight. Luigi Rizzi

.

Generative approaches to diachronic phenomena

In the last twenty years, the investigation of diachronic phenomena has developed into a productive and well-established sub-discipline of generative linguistics (for an overview of more recent work cf. the volumes edited by Battye & Roberts 1995; Beckman 1995; van Kemenade & Vincent 1997; Pintzuk, Tsoulas, & Warner 2000). Following the pioneering work of scholars such as David Lightfoot, Anthony Kroch and Ian Roberts, there is by now general agreement on the notion that the study of language change can provide insights into the properties of Universal Grammar (UG) that cannot be gained from a purely synchronic perspective. This conviction stems from the insight that the triggers for language change are located in the workings of language acquisition, that is, the interaction between innate UG-principles (and the parameters associated with them) and the linguistic evidence the child encounters. From this point of view, language change reflects the “limits to attainable grammars” (Lightfoot 1991: 172) in the sense that language change reveals how parametric choices are selected on the basis of the evidence available. Accordingly, most diachronic generative work focuses on explanations for specific changes (e.g. the loss of verb movement in the history of English) in terms of parametric choices that at some point failed to be acquired or, for certain reasons, made their way into the grammar of the learner during language acquisition. This volume aims at adding a new aspect to the generative approach to language



Eric Fuß and Carola Trips

change by reviving a line of thinking which has its roots in the 19th century, namely the interest in diachronic explanations for synchronic phenomena. From the earliest days of comparative philology, it has been observed that languages change. In the 19th century, when linguistics came into existence as an academic discipline in its own right, linguists were primarily interested in the history of languages. The observation of systematic similarities between ancient languages such as Latin, Classical Greek and Sanskrit led to the idea that these similarities should be explained as the result of inheritance from a common ancestor, which is nowadays referred to as Proto-Indo-European (PIE) (Jones 1807; Schlegel 1808; Bopp 1816). Based on this general insight, the goal of 19th century linguistics was twofold: first, linguists aimed at reconstructing the best hypothesis about the sound system and the morphology (focusing on the inflectional system) of this ancestor language by comparing the relevant grammatical traits of the daughter languages (the so-called Comparative Method; cf. Szemerényi 1989 and more recently Fox 1995). Second, they described (and explained) the changes that led from PIE to the systems found in its descendants. As a consequence, all valid explanations were thought to be diachronic in nature. Thus, linguists mainly tried to establish the facts of language change, and they did so by comparing sound systems and the morphology of languages. They noted that the sounds of related languages corresponded to each other apparently systematically and therefore they called this phenomenon sound shift, e.g. the Indo European voiced plosive /b/ was transformed into the voiceless plosive /p/ in Germanic. They further noted that these shifts took place in an extremely regular way and that they could be seen as sound laws. The Neogrammarians adapted this view and claimed that all exceptions to sound laws could be explained by processes of analogy. So for the average 19th century linguist, the scientific study of language was essentially a study of its history. This credo is very explicit in one of the most influential (and comprehensive) publications of the Neogrammarian era, Hermann Pauls Principien der Sprachgeschichte (1880), in which he summarized and organized the work of contemporary Neogrammarians such as Brugmann, Leskien, Osthoff, Sievers and others (see e.g. Jespersen 1922 for similar statements): Es ist eingewendet, dass es noch eine andere wissenschaftliche Betrachtung der Sprache gäbe, als die geschichtliche. Ich muss das in Abrede stellen. (Paul 1880: 19) ‘It has been objected that there exists another form of scientific study of language apart from the historical one. I must deny that.’

Introduction

However, at the beginning of the 20th century, the field of linguistics underwent a major transformation, with the scientific focus shifting to the study of the structure of living languages. The most influential figure behind the rise of Structuralism was Ferdinand de Saussure. In his famous lecture Cours de linguistiques générale, Saussure was the first to distinguish systematically between the historical and the synchronic approach to language, highlighting the advantages of the latter over the former (cf. Sapir 1921; Bloomfield 1933 for related statements). He illustrates the synchronic approach to language by using an analogy with a game of chess: if we come into a room where a game of chess is underway, a study of the position of the pieces on the board suffices to assess the state of the game, without knowing what happened before. Within the branch of American Structuralism, the interest for theoreticallyinformed historical analyses began to falter even more with the rise of generative grammar in the 1950s. When Noam Chomsky established the linguistic intuitions of a native speaker as the major source of empirical data, the inquiry into the historical stages of a language lost its attraction within this new brand of linguistics, basically because native speaker intuitions for, say, Old English were hard to come by.1 Instead, researchers began to seek for purely synchronic explanations which are framed – using notions of the current Principles and Parameters framework – in terms of innate principles of Universal Grammar and parameters associated with these invariant principles that are fixed in the course of language acquisition. It was the work of scholars such as David Lightfoot (e.g. 1979, 1991, 1999), Anthony Kroch (e.g. 1989, 1994, 2001, Kroch & Taylor 1997) and Ian Roberts (e.g. 1993a, 1993b, Clark & Roberts 1993; Roberts & Roussou 2003) which stimulated a new interest in the study of language change within a generative setting, arguing forcefully for the idea that investigations into the nature of language change may provide new insights for the theory of grammar that are not available from a purely synchronic study of language. As David Lightfoot puts it: The properties of Universal Grammar and the way in which grammars are triggered shed light on catastrophic historical changes reflecting a new parameter setting, because the old parameter setting is no longer attainable. Conversely, the point at which parameters are set differently illustrates the limits to attainable grammars. [...] If changes over time can inform us about the limits to grammars, the study of diachronic change can show how idiosyncratic properties may be added to a grammar without affecting its internal structure, and diachronic study may show what it takes to drive a grammar to reanalysis, with a parameter set differently. Examining historical reanalyses






illuminates what kinds of triggering experiences elicit grammatical reanalyses and what kinds are not robust enough to have that effect. The point at which reanalyses take place sheds light on the load that can be borne by grammatical operations; the limits are manifested by the occurrence of the reanalyses; the reanalyses are manifested by the simultaneity of various surface changes and the other properties of parametric change (above). In this way research on historical change informs work on a restrictive theory of grammar and is fully integrated with that general enterprise. [...] If in historical change a parameter comes to be set differently, the precise information given by Universal Grammar for that parameter will have implications (often far-reaching ones) for what exactly will change in the surface structures. Examining the cluster of properties encompassed by particular reanalyses sometimes suggests things about how parameters should be defined. (Lightfoot 1991: 172f.)

Following Lightfoot, generative approaches to language change today generally agree on the notion that (the cause of) language change is deeply rooted in the process of language acquisition. In the Principles and Parameters framework, it is assumed that variation between individual languages results from different settings for a limited number of (binary) parameters that are associated with invariable principles of Universal Grammar (UG) (Chomsky 1981, 1986a). Thus, the task of acquiring a given grammar consists of filling in the gaps left open by the principles of UG, that is, detecting the values of the parametric choices that are reflected by the linguistic input data the child is confronted with. The stimulus that serves to set a given parameter one way or other is usually called trigger experience or cue (see Lightfoot 1991, 1999 for discussion). The interaction between principles and parameters can be neatly illustrated with the subjacency condition which can be (informally) characterized as a principle of UG requiring that syntactic movement may not cross more than one bounding node (cf. Chomsky 1981). The relevant parametric choice that is associated with the subjacency condition/principle concerns the set of bounding nodes for a given language. Whereas some languages such as English or German may choose {IP, NP} as bounding nodes (probably the default choice), in others like Italian the existence of sentences such as (1) causes the child to set the relevant parameter as {CP, NP} (cf. Rizzi 1982). domando [ CP che (1) tuo fratello, [CP [ a cui]i [ IP mi storie [IP your brother to whom myself ask-1sg which stories abbiano raccontato ti ]]]] have-3pl told ‘*your brother [to whom]i I wonder which stories they told ti ’

Introduction

In the relative clause (1), the relative pronoun a cui moves directly from its base position in the embedded indirect question to the SpecCP of the higher clause, thereby skipping the intermediate SpecCP which is occupied by the wh-phrase che storie. This movement operation crosses two IP nodes (underlined in (1)), but only one CP node (marked by double underlining). This configuration leads to ungrammaticality in a language where the bounding nodes are {IP, NP} (which becomes clear from the English translation), but is well-formed in Italian which is characterized by the choice {CP, NP}, in accordance with the subjacency condition. Apart from accounting for synchronic parametric differences between different individual languages (such as e.g. Present Day English and Italian), this approach can also be applied to diachronic parametric differences between different historical stages of a single language (such as e.g. Old English and Present Day English). Thus, syntactic change is analyzed as a change that affects the parameter settings for a given language. Due to the fact that a single parameter typically determines a whole complex of (syntactic) properties, a single parametric change manifests itself in a variety of complex effects and at times dramatic changes at the syntactic surface. In other words, parametric change is often abrupt and ‘catastrophic’ in nature (Lightfoot 1979, 1991, 1999). An instructive example of the workings of parametric change is given in Roberts (1993a) who shows that in the history of French, three apparently independent changes can actually be analyzed as reflexes of a single parametric change that affected the mechanism of nominative case assignment. Old French was a language which exhibited three phenomena which were lost in later historical stages: “simple inversion” in interrogatives, null subjects, and V2 order. The three changes are illustrated with the examples below: (2) Loss of ‘simple inversion’ in interrogatives OF Comment fu ceste lettre faitte? how was this letter made ModF *A Jean pris le livre? has Jean taken the book

(Clark & Roberts 1993: 319)

(3) Loss of null subjects OF Si firent pro grant joie la nuit. thus (they) made great joy the night ModF *Ainsi s’amusaient bien cette nuit. thus (they) had fun that night (Clark & Roberts 1993: 320)






(4) Loss of V2 OF Lors oïrent ils venir un escoiz de tonoire. then heard they come a clap of thunder ModF *Puis entendirent-ils un coup de tonnerre. then heard-they a clap of thunder (Clark & Roberts 1993: 320)

Roberts (1993a) shows that all three phenomena were lost in the sixteenth century, which already suggests that they are connected in some way or other. He then argues that these changes reflect an underlying change in the value of the parameter which determines nominative Case assignment. By following Koopman and Sportiche (1991) he assumes that nominative Case can be assigned in two ways, either under government or under agreement. He then shows that the loss of the parametric option to assign nominative Case under government entailed the loss of null subjects as well as the loss of V2 orders and simple inversion. Furthermore, Roberts links the syntactic change in question with morpho-phonological changes that led to the loss of distinctive inflectional endings. Thus, morphological properties of individual lexical items are assumed to play an important causal role for setting off changes that affect the syntactic component of the grammar. This account is in line with recent approaches to parameterization which assume that parameters are lexical, that is, they are associated with properties of individual lexical items (i.e. the Lexical Parameterization Hypothesis, cf. Borer 1984; Wexler & Manzini 1987). In the late 1980s, the Lexical Parameterization Hypothesis converged with another theoretical development within the framework of Principles and Parameters theory that aimed at implementing the traditional distinction between lexical and grammatical categories directly into the structure of the clause. Following work by e.g. Fukui (1986), Chomsky (1986b), Abney (1987), grammatical categories like determiners, conjunctions and inflections are associated with a closed class of functional heads that project their featural content in accordance with universal principles of phrase structure (e.g. X-bar theory or Bare Phrase Structure). Initially, in the verbal/clausal domain only two functional heads were distinguished, C (conjunctions, clause type, subordination) and Infl (verbal inflection, finiteness, nominative case assignment). In later work by Pollock (1989), Belletti (1990), Chomsky (1991), the category Infl is split further into distinctive Agrs (subject agreement, nominative case assignment), T (tense, finiteness), and Agro (object agreement, accusative case assignment) heads. In the 1990s, this line of research has led to a sudden increase of the number of functional categories (cf. e.g. Cinque 1999). By now, most researchers agree on the existence of a

Introduction

universal inventory of core functional categories that consists of the elements C (clause type, subordination), T (tense, subject-verb agreement, nominative assignment), ν (voice, transitivity, accusative assignment, object agreement) and D (nominal inflection, definiteness) (cf. Chomsky 1995, 2000, 2001a, 2001b). Note, however, that many scholars today argue for the existence of more elaborate systems that include a fine-grained structure of inflectional/functional heads in the Infl and C domain (cf. Rizzi 1997; Cinque 1999, among others). At least with respect to the set of core functional categories it is assumed that they are universally present in natural languages as basic ‘building blocks’ or ‘skeleton’ of clause structure and that they trigger syntactic operations in order to license their (abstract) morphological content (cf. Chomsky 1991, 1995, 2000). On these assumptions, it seems natural to analyze parametric variation between individual languages in terms of varying (morphological) feature specifications of (a universally present set of) functional categories (cf. e.g. Chomsky 1991, 1995; Ouhalla 1991): If substantive elements (verbs, nouns, etc.) are drawn from an invariant universal vocabulary, then only functional elements will be parametrized. (Chomsky 1991: 419)

In this approach to parameterization, the computational system (CHL ), that is, the mechanisms that build up syntactic hierarchical structures by combining individual lexical items, is universal and therefore cross-linguistically and diachronically invariable (Chomsky 2000, 2001a, 2001b). Thus, strictly speaking, there is no such thing as ‘syntactic change’: the properties of the syntactic component of grammar remain constant over time (cf. Hale 1996, 1998; Longobardi 2001; Keenan 2002). Apparent ‘syntactic change’ (and synchronic differences between languages) results from changes affecting the feature content of a closed class of functional categories such as C, T, ν and D (via phonological erosion or grammaticalization processes). It is expected that a single change in the featural properties of these core functional categories results in a complex of (at times dramatic) distinct changes on the syntactic surface, which is a hallmark of parametric change (cf. Lightfoot 1979, 1991, 1999). Moreover, if overt inflectional morphology is taken to reflect the (abstract) feature content of functional categories, it is possible to construe a correlation between parametric change and morphological change, that is, to provide a principled explanation for the traditional observation that changes affecting the inflectional morphology of a given language often go hand in hand with syntactic change (cf. e.g. Paul 1880; Sapir 1921). In the generative literature on language change, this line of research has proven to be very productive,






covering for example the historical impact of the loss of verbal inflection on the availability of verb movement (for the history of English cf. Roberts 1993a, for the Scandinavian languages cf. Platzack 1988 on Swedish) and pro-drop (for the history of French cf. Roberts 1993a; Vance 1997; for Swedish Falk 1993; for English Allen 1995; Haeberli 1999) or the relation between the loss of nominal inflection (i.e. case) and changes affecting argument order and the rise of ECM constructions (for the history of English cf. Lightfoot 1979, 1991, 1999; van Kemenade 1987; Roberts 1997; Kiparsky 1997; Haeberli 1999).

. The logical problem of language change Still, the generative approach to language change has not gone unchallenged. It has often been noted that the concept of parametric change conflicts with two widely-held views on language change and acquisition. First, it is a traditional observation that language change is a gradual process. A given change may go on over many centuries before it is completed (cf. e.g. the change from OV to VO in the history of English, Lightfoot 1979, 1991, 1999; Kemenade 1987; Roberts 1997; Pintzuk 1999; Trips 2002; Fuß & Trips 2002). At first sight, this trait of language change does not seem to be compatible with the abrupt character of parametric change. Another problem arises from the assumption widely held in generative linguistics that language acquisition proceeds in a flawless fashion. In other words, most researchers (working in the generative paradigm) agree that children always succeed in acquiring the target grammar that generates the linguistic data they are exposed to, even if this data is apparently flawed and insufficient (sometimes called the “the logical problem of language acquisition”, cf. Chomsky 1986a for discussion). This leads to the “logical problem of language change” as Niyogi and Berwick (1998) choose to call it: After all, if all children successfully attain the grammars of their parents and they continue to do this generation after generation, then the linguistic composition of every generation would look exactly like the linguistic composition of the previous generation and languages would not change with time. Yet they do. (Niyogi & Berwick 1998: 192f.)

Thus, more has to be said to reconcile the idea of parameter change with (i) the apparent gradualness of language change and (ii) the apparent perfection of language acquisition. To answer the latter problem, it is usually assumed that for some reason, the trigger experience that resulted in a given parame-

Introduction

ter setting in the parent grammar has become obscure or ambiguous in the output of this grammar (cf. Lightfoot 1979, 1991), due to factors such as language contact, (morpho-) phonological erosion or syntactic reanalyses that blur the evidence for certain parametric choices in the linguistic input. Of course, this raises a number of further questions concerning the nature of the trigger experience. Note furthermore that ambiguity of trigger experience is only a necessary condition for a change to take place. It is still left unclear what actually motivates parametric change. Here, a wide-spread line of thinking assumes that principles of economy take over and ultimately decide if the trigger evidence contained in the input is truly ambiguous (cf. e.g. Clark & Roberts 1993). In general, these economy considerations come in two varieties: first, some researchers assume that there are marked and unmarked (or default) parameter values and that the learner assigns a given parameter the unmarked value if no decision can be made based on the evidence available in the input (cf. Wexler & Culicover 1980; Berwick 1985 for discussion). A related idea lies behind the Subset Principle (Berwick 1985: 37; Wexler & Manzini 1987: 61):2 (5) The Subset Principle The acquisition device selects the most restrictive parametric value consistent with experience. (O’Grady 1997: 283)

Thus, if there are two possible settings A and B for a given parameter, and A generates a subset of the sentences generated by B, the learner will acquire the more restrictive setting A in the absence of clear trigger experience. In this sense, A can be viewed as the unmarked setting. Alternatively, the decision in question is assumed to be sensitive to the notion of derivational/representational economy (e.g. a ‘Least Effort Strategy’, Clark & Roberts 1993; Roberts 1993b), in the sense that the learner assigns a given input string the most economical representation/derivation that is compatible with the data. If the manifestation of a given marked parameter value is ambiguous in the input, the learner will choose the parameter value that guarantees the most economical derivation/representation. These considerations are usually employed to account for the diachronic loss of movement operations (if the data is compatible with a non-movement analysis), for example the loss of V-to-I in the history of English (Roberts 1993a; cf. Roberts 1999 for an account that combines markedness considerations and derivational economy). A different approach is presented in Lightfoot (1999) who argues for a ‘cue-based’ theory of language acquisition, originally developed by Dresher and Kaye (1990), Dresher (1999) for the acquisition of phonological properties






such as stress patterns. The basic assumption is that UG contains not only a set of parameters, but also specifies for each parameter a cue that serves to switch the parameter one way or other (cf. Fodor 1998 for a related approach).3 According to Lightfoot, a cue is not directly present in the input, but rather part of the abstract mental representation/structures derived by parsing the input (i.e. a piece of ‘I-language’ in the sense of Chomsky 1986a). For example, Lightfoot assumes that the cue for the V2 property is SpecCP [XP], XP being an arbitrary phrasal category that occupies SpecCP. This form of abstract trigger experience is contained in a (incomplete) structural description of an input string, but not in the input itself. In other words, a cue can only be detected if the learner has assigned a structure to a given input string. If the learner detects a cue that is attested robustly in these (initially incomplete) parses, this will activate a given parameter or syntactic operation in the learner’s grammar.4 In other words, cues are ‘points of variation’ the absence/presence of which determines differences between languages/grammars. Language change results either if a given syntactic operation fails to be cued or if it starts to be cued, in contrast to the target grammar. Thus, instead of assuming the existence of binary parameters, it is perhaps more appropriate to speak of ‘basic’ or ‘extended’ parametric choices, where the cues for the latter are present only in certain grammars. From this point of view, language change is a contingent and unpredictable process that depends on the availability of certain cues in the linguistic input the learner is exposed to.5 Hence, the learner does not try to match the input (in contrast to most other approaches, cf. e.g. Clark & Roberts 1993; Gibson & Wexler 1994), but rather scans structures for cues that trigger certain parametric choices, “without regard to the final result” (Lightfoot 1999: 149), that is, whether the new grammar will differ from the target grammar or not. The latter aspect fits in well with the view that language change is an abrupt process that may at times lead to catastrophic consequences, but seems at odds with the often noted apparently gradual nature of language change, which brings us back to the first problem noted above. It has been argued that the often gradual character of language change represents a serious problem for the concept of parametric change which predicts that languages should change rather abruptly (cf. Hopper & Traugott 1993; Harris & Campbell 1995 for discussion). In what follows, however, it is shown that the impression of gradualness arises from two factors that traditional approaches failed to recognize and that we can maintain the concept of parametric change. First, we demonstrate that the impression of gradualness disappears under a sufficiently constrained notion of ‘grammar’ (and ‘grammar change’) as the proper object of linguistic study. Second, it is claimed (following the pi-

Introduction

oneering work by Kroch 1989) that language change typically proceeds via a stage of internal bilingualism, where generations of speakers have command over more than one internalized grammar giving rise to a degree of linguistic variation which gradually disappears when one grammar wins out over the other over time. In generative linguistics, there is general agreement that the object of fruitful linguistic study is a special notion of ‘language’, namely the linguistic knowledge of an idealized speaker/hearer, cf. Chomsky (1965: 3) for a classic statement: Linguistic theory is concerned primarily with an ideal speaker-listener, in a completely homogenous speech-community, who knows its language perfectly and is unaffected by such grammatically irrelevant conditions as memory limitations, distractions, shifts of attention and interest, and errors (random or characteristic) in applying his knowledge of the language in actual performance.

On this conception of ‘language’, the proper object of linguistic study is generally referred to as linguistic competence or simply grammar. Accordingly, the proper object of the linguistic study of language change must be defined as a change between (individual) grammars, which we will refer to as ‘grammar change’ (cf. Lightfoot 1991, 1999; Hale 1996). It is fairly clear that in this sense, language change is necessarily an abrupt phenomenon, namely a clearly identifiable difference between a grammar A and a grammar B. Hence, the impression of gradualness is in fact a misperception that arises from confusing the actual change process with the (highly gradual) diffusion of this change within a population/linguistic community.6 The actual change is not a socio-linguistic phenomenon, but rather the result of cognitive processes that determine the process of language acquisition, resulting in a grammar in the mind of the individual speaker that differs from the target grammar. Only under this restricted interpretation of language change as grammar change can we hope that the inspection of language change will reveal something about the structure and the workings of the language faculty/UG. The issue on hand is also taken up in the work of Anthony Kroch, who tries to reconcile a formal approach in terms of parametric change with the apparent gradualness of language change. It is a well-known observation that language change usually proceeds via a (intermediate) stage where old and new (i.e. changed) linguistic forms coexist side by side, leading to a degree of variation which is not encountered in ‘stable’ linguistic communities.7 Furthermore, it has been noted that the loss of the old form and its eventual replacement by






the new form is a gradual process, which at times may extend over many centuries.8 In a series of publications (Kroch 1989, 1994 and subsequent work), Kroch has developed a formal account of these observations which is based on the Principles and Parameters framework. Here, the notion of Grammar Competition represents the core concept of an integrated theory of language change and variation. The basic idea of Kroch’s approach is that parametric change must always proceed via a stage where the speaker (or, a generation of speakers) of a language X has access to more than one internalized grammar (sometimes referred to as an instance of internal diglossia). The grammars in question may differ in a number of parametrical choices, giving rise to a wider range of linguistic variation. However, Blocking Effects imposed by UG (see Kroch 1994 for details; cf. Aronoff 1976 on Blocking Effects on morphological doublets) restrict the co-existence of grammars that differ only minimally with respect to a set of parameter doublets (i.e. co-existing ‘competing’ values for one parameter), thereby warranting that one grammar will eventually win out over its competitors, which completes the change process in question. We can therefore conclude that the apparent gradualness of language change does not create a problem for the concept of parametric change. Rather, change between grammars is necessarily abrupt, but may involve a stage where speakers have command over more than one internalized grammar, giving rise to an unusual degree of linguistic variation, and over time, the impression that language change is gradual when one parametrical choice wins out over the other.

. On the status of diachronic explanations In a set of recent papers, Alexiadou and Fanselow (2001, 2002) argue that certain properties of grammars or apparent synchronic generalizations do not reflect UG principles, but are rather the result of the way syntactic change proceeds (see below for details). In other words, certain grammatical laws hold in natural languages because grammars violating them simply cannot arise in the course of language change. This perspective is reminiscent of the Neogrammarian view briefly discussed above that central synchronic properties of natural languages are the result of diachronic processes – “synchrony is determined by diachrony.” In contrast to 19th century linguistics, however, where contemporary languages were studied primarily for the light they could throw on the past, present-day generative linguistics is interested in the insights which

Introduction

the study of diachronic data can add to the understanding of the structure of grammar and the make-up of the human language faculty. Actually, Alexiadou and Fanselow are not the first to revive the idea to look for diachronic explanations of synchronic phenomena.9 This line of thinking figures prominently in work on language typology and language universals (cf. e.g. Givón 1971, 1976; Greenberg 1978; Bybee 1985; Bybee et al. 1990) where it is often claimed that language universals follow at least in part from the way language change proceeds, cf. the following statement by Joseph Greenberg:10 Diachronic principles are involved in the explanation of both low and higher level synchronic generalizations. In so doing they often explain exceptions. They also go even further than synchronic typology in subsuming under general principles not only non-universal typological traits, but often even highly idiosyncratic language-specific rules which can be treated as evidence of transitions between less complex, more widely occurring types. (Greenberg 1978: 89)

To take a concrete example, Givón (1976) argues that a set of cross-linguistic (synchronic) generalizations on argument-predicate agreement can be explained as resulting from the way verbal agreement arises historically. Givón claims that cross-linguistically, agreement markers arise from the reanalysis of resumptive pronouns in topic left dislocation structures. This development goes hand in hand with another major reanalysis, where (due to an over-use) former topic left dislocation structures are reinterpreted as the ‘neutral syntax’, with the former topic becoming the new subject: (6) [The wizard]i , hei lived in Africa → The wizard he-lived in Africa topic pronoun subject agr

On this assumption, the observation that cross-linguistically, subject agreement is the most frequent case of grammatical agreement can be attributed to the prominence of subjects in the topicality hierarchy proposed by Givón. Furthermore, the fact that in many languages subjects must be either definite or generic (Givón mentions Mandarin, Kinyarwanda and Malagasy as examples. Similar restrictions can be found in many Austronesian languages such as Indonesian, Tagalog etc.) is taken to follow from the origin of these subjects as former topics, which are generally subject to similar restrictions (either definite or generic, but never referential-indefinite).11 Thus, it seems that the synchronic generalizations in question are at least partially reducible to the way argument-predicate agreement arises historically.






As already noted above, this line of thinking is taken up in recent generative work where it is also argued that certain apparent synchronic generalizations or universals are actually the result of the mechanisms of language change. Alexiadou and Fanselow (2001, 2002) (henceforth A&F) propose an interesting new perspective on the often noted correlation between rich inflectional morphology and V-to-I movement (cf. Kosmeijer 1986; Holmberg & Platzack 1991; Roberts 1993a; Rohrbacher 1994; Vikner 1994, 1995, 1997; Bobaljik 1995). They contradict the claim that there is a correlation between richness of verbal inflectional morphology and V-to-I movement. Instead, A&F show that this hypothesis is empirically not borne out and that in fact this type of movement is independent of morphology. In the group of Germanic SVO languages we find languages like English where the finite verb stays in V0 and follows a VP-adverbial like often, but we also find languages like Icelandic where the finite verb appears to be in I0 because it precedes the same kind of adverbial: (7) a. That John [VP often eats tomatoes] surprises most people. b. Að Jonas borðar [VP oft tómata] kemu flestum á óvart.

Apart from the fact that these languages differ in the relative order of finite verb and VP-adverb it has also been observed that these languages differ with respect to their verbal inflectional paradigms. In Icelandic finite verbs are marked for a range of distinctions for person and number, while in English an agreement ending (-s) is found only in the 3sg present indicative (see Table 1 below). These observations have led to the assumption that there is a direct correlation between syntactic head movement and properties of verbal inflection, that is, the presence of rich inflectional morphology is responsible for the presence of overt V-to-I movement in a language. It has even been proposed that this correlation is governed by UG (“Syntactic Movement is triggered by morpholTable 1. Verbal inflection (present indicative) in English and Icelandic (Vikner 1995: 133)

Infinitive 1sg 2sg 3sg 1pl 2pl 3pl

English

Icelandic

throw throw throw throw-s throw throw throw

kasta ‘throw’ kasta kasta-r kasta-r köst-um kast-ið kasta

Introduction

ogy”, Haegeman 1997). A&F claim that this is not the case and they strengthen their assumption by showing that the correlation between head movement and rich morphology cannot be found in any other domain of grammar. They show for a number of movement operations (V2, N-movement, NP-movement, A’movement) that there is no such correlation (see the data below for the V2 property): (8) a.

Måske har Peter læst denne bog. maybe has Peter read this book b. Vielleicht hat Peter dieses Buch gelesen. maybe has Peter this book read ‘Maybe Peter has read this book.’

In both Danish and German we find movement of the finite verb to C0 although (i) rich verbal agreement is found only in German and (ii) in both languages C0 , that is, the target of verb movement in main clauses, is not associated with any inflectional morphology whatsoever. This implies that the presence or absence of V2 does not predict whether inflectional morphology is rich in a given language and it can be concluded that the V2 property is unconnected to any morphological property. A&F further note that although there is no direct correlation between head movement and rich morphology we find evidence that there is still a connection between the two phenomena and that only an implicational relation (Vikner 1994; Bobaljik 2002) is defensible. By looking at data from a number of SOV, SVO and VSO languages they come to the conclusion that only the following generalization is tenable: (9) Suffixal rich inflection implies V-to-I-movement.

They note that properties such as (9) can be part of a natural language grammar without being a universal principle because “they are the products of diachronic processes, the results of a series of steps of language change.” (A&F 2002: 233). A&F further note that “it is at least conceivable that the mechanisms of language change are such that certain combinations of properties simply cannot arise as the product of a natural diachronic process” and claim that the V-to-I movement generalization exemplifies such a case. They argue that suffixal inflection can only arise if the verb precedes the element (subject pronoun) that will eventually develop into a suffix and that this can happen if either the verb is in I or in C: (10) a.

[Infl+verb] [vP subject .... ]






b. [C+verb] [IP subject .... ]

In these configurations, the subject pronoun can cliticize to the verb. Eventually, the sequence V + clitic will be reinterpreted as V + agreement. At this point, it is clear why (9) holds: rich suffixal agreement can only develop if there is verb movement. If there was no such movement, the clitic would not have a host where it could attach to. In this way, A&F show that the traditionally assumed correlation between V-to-I movement and rich morphological inflection can be reduced to the way suffixal agreement arises in the diachronic development of languages. A related avenue of research has to do with the question of how diachronic data can be taken into account to provide new insights for the analysis of individual present-day languages. Within a generative setting, Lightfoot (1979) was the first to emphasize the significance of diachronic data for the evaluation and construction of synchronic analyses. Lightfoot claims that a consideration of the diachronic perspective and thus the mechanisms of language change can help to decide on the correct analysis for certain synchronic facts. He points out that such a decision is based on “a theory of linguistic change motivated by historical data” and concludes that such a perspective “offers an approach to the old problem of the synchronic relevance of historical data and the possibility of a genuinely unified linguistic theory” (p. 1). More recently, Simpson and Wu (2002) argue for an analysis of negative concord in French that makes use of a two-shell (Larsonian) NegP (i.e. the structure in (11)), which is motivated by a thorough analysis of the historical development of this particular property of French grammar. (11)

NegP1 Neg1’

spec Neg10 ne

NegP2 spec

Neg2’

pas Neg2

ZP

Simpson and Wu argue that historically, AgrP/ConcordP develops out of a FocusP that is originally selected by some higher functional head. The morphological material contained in SpecFocusP (or Foc◦ ) serves to repeat and thereby emphasize the semantic content of the selecting functional head. Over

Introduction

time, the focus interpretation of the selected FocusP decays, giving rise to a two part shell structure (Larson 1988), where the lower shell is a simple AgrP or ConcordP that is parasitically licensed by the semantic content of the upper head of the shell structure. Later on, the lower shell may come to be reanalysed as the ‘real’, formerly selecting functional category, giving rise to Jespersen’s cycle effects. Concerning the development of negative concord in French, it is argued that the former direct object pas has been reanalyzed as a purely functional morpheme in a higher position. This higher position was originally associated with clear emphasis and focus on the negation and attracted the direct object to its Spec. Structurally, this is implemented by the assumption that ne (in Neg0 ) began to select for an (optional) FocP between Neg0 and VP that attracted pas to its specifier, SpecFocP. The continued association of pas with focus and negation led to a reanalysis of pas as base generated in SpecFocP. After this reanalysis, additional direct objects could co-occur with pas. Eventually, the emphatic force of pas underwent gradual weakening and was lost so that in modern French there is no longer any emphatic interpretation associated with the occurrence of pas, giving rise to the two-shell concord structure in (11) above. Thus, the particular synchronic analysis proposed by Simpson and Wu reflects (and is supported by) the diachronic development of negative concord, in line with the proposals by Lightfoot (1979). In a similar vein, Longobardi (2001) shows for the Modern Romance languages that the diverging syntactic properties of the reflexes for Latin CASA ‘house, home’ can only be understood if we look at their individual historical development, taking into account reconstructed properties of their common origin, that is the grammar of Proto-Romance. More specifically, Longobardi shows that an explanation of the peculiar diachronic and synchronic properties of the reflexes of CASA(M) in Romance can only be explained under the assumption that a construct state pattern must be reconstructed for ProtoRomance, similar to the well known Semitic construct state.12 (12) a.

D+CASA+PP (e.g. ‘the home/house of XY’, similar to the Semitic ‘absolute state’) b. CAS+DPposs (e.g. ‘home XYgenitive ’, CAS moves to D, similar to the Semitic ‘construct state’)

Longobardi shows that chez, the French reflex of the nominal CASA(M), exhibits a set of peculiar properties. In present-day French, chez has to be analyzed as a locative preposition, in contrast to the nominal reflexes of CASA(M) which are found in the other Romance languages. The examples below show






that chez behaves like a ‘proper’ preposition (e.g. it does not co-occur with a determiner nor is it followed by a PP): (13) a.

Je vais chez mes parents. I went to my parents b. *Je vais chez de mes parents. c. *Je vais chez les mes parents.

However, Longobardi notes that in colloquial French, chez can function either as a preposition or as an apparent construct state noun. In the latter reading, chez resembles the nominal reflexes of CASA in other Romance languages like Italian, Catalan or Spanish: (14) Chez Paul, c’est très beau. at Paul’s/Paul’s home, it’s very nice. (15) Italian a. Casa è sempre il posto migliore per rilassarsi. home (one’s home) is always the best place to relax Catalan b. Ca’ l mestre (no és lluny d’acquí). home the teacher (is not far away from here) ‘The teacher’s home is not far away from here.’ Spanish c. En ca’ Pedro. in home Pedro ‘in Pedro’s home’

According to Longobardi, the systematic syntactic similarities of the reflexes of CASA(M) found in other Romance languages (and in colloquial French) should be taken to indicate that a construct state must be reconstructed for Proto-Romance, the common ancestor of these languages (following the rationale of the Comparative Method that obvious similarities between genetically related languages are most probably features that are inherited from a common ancestor language). Interestingly, this assumption provides new insights into the special diachronic development of French chez as well. Longobardi shows that the development of French chez actually exhibits four diachronic changes: a change in the lexical domain (the loss of an allomorph), irregular phonological change, a categorical shift and a semantic shift:

Introduction

(16) a. b. c. d.

Lexical Phonological Categorial Semantic

= = = =

CASA(M)/CHIESE > ∅(i.e. loss of an allomorph)13 CASA(M) > *CAS > chies > chez N>P ‘home’ > ‘generalized and abstract location’

Other Romance varieties display similar irregular reflexes/phonological developments of CASA, but do not exhibit the other changes that took place in the history of French (i.e. (16a, c, d)). Longobardi argues that the peculiar diachronic and synchronic properties of chez can be explained if it is assumed that the development of chez started from a nominal element in a ProtoRomance construct state where a reduced form, that is CAS+DPposs , was used (see above). The reduced form CAS was an allomorph of CASA ‘hut, house, home’, which came into existence via phonetic/phonemic reduction of the head noun of the construct state, a process often observed in this particular nominal construction.14 Later in the history of French, the ‘full’ allomorph (i.e. chiese) of OF chies got lost, possibly due to lexical pressure exerted by the innovation maison for the meaning ‘house, home’. This in turn led to a categorial reanalysis of chies as a preposition for the following reasons: due to its reduced phonological shape (no inflectional endings) and its exclusive use in the construct state (Nto-D movement blocks the presence of an article and was mostly triggered in prepositional contexts, i.e. P+chies+DPposs ), the nominal status of chies could only be detected if the learner recognized chies as a regular allomorph of an otherwise unequivocally nominal form, namely chiese. Thus, when the allomorph chiese was lost (for independent reasons, change (16a)), the nominal status of chies could no longer be acquired, resulting in the changes (16c) and (16d) above. Note that this explanation becomes only available if it is assumed that chez developed from a reflex of CASA(M) formerly restricted to a contruct state in Proto-Romance. That is, again, the diachronic perspective (here: the reconstruction of certain syntactic traits of the ancestor language) is crucial for explaining properties of synchronic grammar.15 To sum up, the above discussion has identified a set of new directions for diachronic generative linguistics. First, we have to assess whether synchronic generalizations can be reconstructed as the result of specific diachronic processes or general properties of the way language change proceeds. Second, diachronic evidence should be used to re-evaluate and reshape synchronic analyses of certain syntactic phenomena that resist satisfying analyses in purely synchronic terms. Some of these research targets are tackled by the works collected in this volume, which are briefly summarized below.






. The contributions The paper by Artemis Alexiadou argues that an investigation into the diachronic development of possessives in English, German and French provides important insights into the structure of possessive DPs in general. More specifically, it is claimed that the different historical stages of the grammaticalization path from possessive adjectives (which can co-occur with determiners such as Italian il mio libro ‘the my book’) to possessive determiners (which cannot co-occur with determiners, cf. English *the my book) provide further evidence for a tripartite distinction for possessive pronouns along the lines of Cardinaletti (1998), where a structural difference is made between clitic possessors that move to D0 , and weak/strong possessors that are XPs and either move to the specifier of a DP-internal AgrP (weak possessors) or may stay behind in their base position (SpecnP). The contribution by Eric Fuß presents a diachronic explanation of the fact that in Bavarian, complementizer agreement and pro-drop are limited to 2nd person contexts (along with 1pl in some dialects). It is argued that this restriction follows from a conspiracy of morphological and syntactic factors that guided the reanalysis of subject clitics as markers of verbal agreement in the history of Bavarian. More specifically, it is shown that this change could only take place in inversion contexts, leading to partial pro-drop and the presence of an agreement morpheme attached to C0 . The person/number restrictions are then attributed to the workings of morphological blocking effects that confined the grammaticalization of new verbal agreement morphology to previously underspecified slots of the agreement paradigm. The diachronic analysis is based on a new approach to complementizer agreement that ascribes the presence of inflection in the C-domain to the post-syntactic insertion of a ‘dissociated’ Agr-morpheme. Eric Haeberli investigates the often discussed correlation between certain properties of inflectional morphology M and a certain set of syntactic properties S. Based on an analysis of a set of potentially problematic cases that lack M but show S (involving XP-subject order/rich verbal morphology in Germanic and word order variation/rich case morphology in historical stages of English) it is argued that the relevant synchronic generalizations concerning the relation between syntax and morphology (in the sense that M implies S) can be maintained if certain assumptions on the diachronic and synchronic relation between M and S are adopted. First, it is shown that S can continue to exist in the absence of M (as a transitional stage) if S is triggered by purely syntactic evidence in the input (e.g. word order). Second, Haeberli develops a new

Introduction

explanation for the observation that historically, the loss of M entails that S is eventually lost as well. More specifically, he assumes that the loss of M creates a mismatch between the morphology and the syntax if a set of syntactic properties S that were once triggered by M remains part of the input the learner is confronted with. By assumption, this mismatch gives rise to a new (more economical) grammar that lacks both M and S and competes (in the sense of Kroch 1989) with the older grammar that lacks M but continues to have S. In the course of time, then, the more economical variant drives out the old grammar. The contribution by Roland Hinterhölzl presents a new and challenging approach to word order changes in the history of English and German and the traditional observation that language change often proceeds in a gradual and slow fashion. To reconcile the latter observation with the abrupt nature of parametric change (see the discussion above), Hinterhölzl invokes the idea that the gradualness of language change should be analyzed as resulting from a gradual change in language use/stylistic rules that may grammaticalize over time, eventually leading to a change in the grammar. More precisely, it is proposed that grammars provide a limited amount of optionality that can be exploited by speakers to achieve certain communicative effects. This kind of optionality is modeled in terms of stylistic or peripheral rules (in contrast to rules of the core grammar) that may affect word order or prosodic phrasing to derive information-structurally marked forms. Over time, these stylistic rules may lose their stylistic force (due to an over-use) and become reanalyzed as (obligatory) rules of the core grammar. Based on this general idea, the paper develops a new account of word order changes in the history of English and German which makes available a diachronic explanation of the fact that event-related adverbs occur preverbally in the Germanic OV-languages but postverbally (in the exact mirror image) in VO-languages. Based on a set of data involving postverbal modals in a number of Southeast Asian SVO languages such as Thai, Taiwanese or Cantonese, Andrew Simpson argues that certain movement operations that apparently have no clear motivation in a certain synchronic stage of the grammar arise historically via a reanalysis of formerly pragmatically/semantically motivated operations (Focus etc.) as EPP-driven movement. Thus, this paper provides a diachronic explanation for the rise and existence of EPP-driven movement in general. More specifically, EPP features are assumed to be available for the language learner as a formal means to cope with movement operations encountered in the input where the original ‘substantial’ trigger (i.e. semantic, pragmatic or morphological) has been lost in the course of time. As a consequence, movement






operations are not lost from a structure if the original trigger disappears, but rather are converted into ‘fossilized’ movement triggered by EPP features. The general approach advocated in this paper leads to two very interesting claims: first, historically, there is no such thing as ‘negative reversal-type changes’ that is, loss/discontinuation of movement. Second, synchronically, the apparent lack of any understandable motivation for an assumed movement operation “should not necessarily cause one to doubt the hypothesis that movement does indeed take place.” Finally, Zoe Wu argues that an investigation of the diachronic development of resultative verb constructions (RVCs) in Mandarin Chinese provides important insights for the synchronic analysis of these constructions in the present-day language. More specifically, it is shown that diachronic data suggest that the second verb in RVCs of modern Mandarin Chinese – exhibiting the surface order V1 -V2 -obj. – should be analyzed as a completive aspect suffix that is base-generated with the verbal head V1 . It is shown that historically, the aspectual suffix developed from a full lexical predicate via two major reanalyses. First, it is observed that in the earliest records of the RVCs in question, the V2 and the object were in a predication relation. At some point in time, the predication relation between V2 and the DP waned, and the function of V2 reduced to providing a telic bound to the event described by V1 . This development went hand-in-hand with a structural change where the former predicate V2 was reanalyzed as a completive Aspect head that selected a VP complement in its specifier. Later on, this structure underwent a word order change from V1 -obj-V2 to V1 -V2 -obj. It is assumed that the latter change was triggered by a language-internal pressure that prefers the direction of selection to be uniform in a given language. Accordingly, the VP formerly selected to the left of the completive Aspect head V2 (as a specifier) was reanalyzed as a rightward complement of Asp0 (with V2 as an aspectual suffix base generated on V1 ) conforming to the canonical direction of selection in a VO grammar.

Notes . Note that actually there are some early generative approaches to language change, most notably in the field of phonology (Kiparsky 1965; Chomsky & Halle 1968; Postal 1968), but also in syntax (King 1969). Still, the analysis of language change in terms of changes affecting transformational rules (rule loss/addition/change) advocated in these works was rather descriptive in nature and accordingly did not stir much interest among fellow generativists (cf. Postal 1968 for some skeptical remarks).

Introduction . Note that an approach in terms of default settings is not always compatible with the Subset Principle. For example, Hyams (1986) claims that the default setting for the prodrop parameter is [+pro-drop], which is clearly in conflict with the Subset Principle, since [+pro-drop] permits a set of grammatical sentences that includes the set generated by the setting [–pro-drop]. . Lightfoot actually defends a somewhat more radical position, namely that there are in fact no parameters and “that cues which are realized only in certain grammars constitute the parameters, the points of variation between grammars.” (Lightfoot 1999: 149). . Concerning the question of when a given cue is ‘attested robustly’ factors such as ‘unambiguous expression’ and frequency play a role. In the case of V2, for instance, only those utterances that can only be analyzed as SpecCP [XP] are taken to express the cue in question. Furthermore, it has been observed that in the Germanic V2 languages only 30% of main clauses exhibit an arbitrary XP in clause initial position, whereas 70% are subject-initial. Lightfoot (1999) is therefore led to conclude that these 30% of nonsubjects represent the threshold that is sufficient to count as a ‘robust’ instantiation of SpecCP [XP]. . Often, a given parametric choice is associated with a set of cues that may be syntactic or morphological in nature. For example, it is usually assumed that the cue for V-to-I movement consists of (i) syntactic evidence that the verb has moved out of VP (i.e. the configuration Infl [Vfin ]) and (ii) morphological evidence, namely the presence of rich verbal agreement morphology. See Bobaljik (2002) and Haeberli (this vol.) on the relative significance of syntactic and morphological cues (for related considerations cf. Anderson 1980). . According to Hale (1996: 16) the difference between change and diffusion can be characterized as follows: “In the case of change, there has been imperfect transmission of some feature of the grammar. The acquirer’s input sources had features X, Y, and Z and the acquirer constructed a grammar which had features X, Y and W. The difference (W instead of Z) represents a change (Z>W). In the case of diffusion, the acquirer had input sources with features X, Y, Z and other input sources (or another input source) with features A, Y, W and constructed a grammar with features X, Y, and W. Note that there is no imperfect transmission of the relevant features: the child had feature W in his or her input sources and constructed a grammar with feature W. (Similarly for X and Y).” . See Labov (1994) for the claim that linguistic variation constitutes the origin of language change. . One of the most prominent of these long-term changes is the rise of do-support and the loss of verb movement in the history of English (cf. Ellegård 1953; Lightfoot 1979, 1991, 1999; Kemenade 1987; Kroch 1989; Roberts 1993a; Pintzuk 1999 among many others). . Within a generative setting, similar ideas have previously been proposed by Kiparsky (1995), who argues that the V2 property characteristic of Germanic developed as a consequence of the diachronic emergence of the C-system in combination with the requirement to lexicalize the head of this projection (in analogy to embedded clauses where C0 is overtly filled by a set of complementizers). In addition, the idea that synchronic properties of individual functional items such as determiners, tense markers etc. are determined by their historical development is common in traditional approaches to grammaticalization






phenomena (Hopper & Traugott 1993; Lehmann 2002). Recent generative approaches to grammaticalization include Abraham (2003), (2004), van Gelderen (2004) and Roberts and Roussou (2003). . Related ideas can be found in the work of John Hawkins and his colleagues (cf. e.g. Hawkins 1990; Cutler, Hawkins, & Gilligan 1985) who argue that language universals result from processing preferences that operate during language acquisition. The role of language change in the emergence of language universals is emphasized by Kirby (1999) whose approach combines processing considerations with the assumption that the range of possible typological variation is limited by UG. . Givón stresses that even in languages that tolerate indefinite subjects, subjects tend to be definite and referential. . Longobardi analyzes the construct state as the result of N-to-D movement, which is taken to account for the characteristic properties of this particular construction: (i) peculiar word order (head noun-first); (ii) a (Semitic: overt, Romance: overt or empty) phrase understood as a genitive argument follows the head noun; (iii) no article possible; (iv) strict adjacency between head noun and genitive DP; (v) the head noun is (often) deaccented and phonologically reduced. . Normally, Latin final-syllable /a/ is preserved in Old French (and later stages). However, what we find in Modern French is not the expected reflex (ib), but rather the irregular (ia). The regular reflex of CASA(M) existed as an allomorph of chies in Old French, but has been lost in later stages. (i)

a. b.

CASA(M) > OF chies > chez CASA(M) > OF chiese > *chèse

. According to Longobardi, this is a result of N-to-D movement of the head noun which led to a shift of the major phrasal stress to the genitive; cf. Roberts and Roussou 2003 for the idea that movement to a functional head goes hand in hand with phonological reduction typical of grammaticalization processes. . Another example of this line of research is a recent paper by Keenan (2003) who argues that certain puzzles for Binding theory can be explained if the diachronic development of reflexive pronouns in the history of English is taken into consideration.

References Abney, S. (1987). “The English Noun Phrase in Its Sentential Aspect.” Doctoral dissertation, MIT. Abraham, W. (2003). “Autonomous and non-autonomous components of ‘grammaticalization’: Economy criteria in the emergence of German negation.” Sprachtypologie und Universalienforschung, 56(4), 325–365. Abraham, W. (2004). “The grammaticalization of the infinitival preposition – toward a theory of ‘grammaticalizing reanalysis’.” Journal of Comparative Germanic Syntax, 7, 111–170.

Introduction

Alexiadou, A. & G. Fanselow (2001). “Laws of diachrony as a source for syntactic generalisations: The case of V to I.” GLOW Newsletter, 46, 57–58. Alexiadou, A. & G. Fanselow (2002). “On the correlation between morphology and syntax: The case of V-to-I.” In J.-W. Zwart & W. Abraham (Eds.), Studies in Comparative Germanic Syntax (pp. 219–242). Amsterdam/Philadelphia: John Benjamins. Allen, C. (1995). Case Marking and Reanalysis: Grammatical Relations from Old to Early Modern English. Oxford: Clarendon. Anderson, S. R. (1980). “On the development of morphology from syntax.” In Jacek Fisiak (Ed.), Historical Morphology (pp. 51–69). Berlin: Mouton de Gruyter. Aronoff, M. (1976). Word Formation in Generative Grammar. Cambridge, MA: MIT Press. Battye, A. & I. G. Roberts (Eds.). (1995). Clause Structure and Language Change. Oxford: Oxford University Press. Beckman, J. N. (Ed.). (1995). Proceedings of NELS 25. Vol. 2: Papers from the Workshops on Language Acquisition and Language change. University of Pennsylvania. Belletti, A. (1990). Generalized Verb Movement: Aspects of Verb Syntax. Turin: Rosenberg and Sellier. Berwick, R. C. (1985). The Acquisition of Syntactic Knowledge. Cambridge, MA: MIT Press. Bloomfield, L. (1933). Language. New York: Henry Holt. Bobaljik, J. (1995). “Morphosyntax: The syntax of verbal inflection.” Doctoral dissertation, MIT. Bobaljik, J. (2002). “Realizing Germanic inflection: Why morphology does not drive syntax.” Journal of Comparative Germanic Syntax, 6(2), 129–167. Bopp, F. (1816). Über das Conjugationssystem der Sanskritsprache in Vergleichung mit jenem der griechischen, lateinischen, persischen und germanischen Sprache. Frankfurt an Main: Andreäische Buchhandlung. Borer, H. (1984). Parametric Syntax. Dordrecht: Foris. Bybee, J. L. (1985). Morphology. Amsterdam: John Benjamins. Bybee, J. L., W. Pagliuca, & R. Perkins (1990). “On the asymmetries in the affixation of grammatical material.” In W. Croft, K. Denning, & S. Kemmer (Eds.), Studies in Typology and Diachrony (pp. 1–42). Amsterdam: John Benjamins. Cardinaletti, A. (1998). “On the deficient/strong opposition in possessive systems.” In A. Alexiadou & C. Wilder (Eds.), Possessors, Predicates and Movement in the Determiner Phrase (pp. 17–53). Amsterdam/Philadelphia: John Benjamins. Chomsky, N. (1965). Aspects of the Theory of Syntax. Cambridge, MA: The MIT Press. Chomsky, N. (1981). Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. (1986a). Knowledge of Language: Its Nature, Origin and Use. New York: Praeger. Chomsky, N. (1986b). Barriers. Cambridge, MA: MIT Press. Chomsky, N. (1991). “Some notes on economy of derivations and representations.” In R. Freidin (Ed.), Principles and Parameters in Comparative Grammar (pp. 417–454). Cambridge, MA: The MIT Press. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: The MIT Press. Chomsky, N. (2000). “Minimalist inquiries: The framework.” In R. Martin, D. Michaels, and J. Uriagereka (Eds.), Step by Step. Essays on Minimalist Syntax in Honor of Howard Lasnik (pp. 89–155). Cambridge, MA: The MIT Press.






Chomsky, N. (2001a). “Derivation by Phase.” In M. Kenstowicz (Ed.), Ken Hale. A Life in Language (pp. 1–52). Cambridge, MA: MIT Press. Chomsky, N. (2001b). “Beyond explanatory adequacy.” MIT Occasional Papers in Linguistics, 20, 1–28. Chomsky, N. & M. Halle (1968). The Sound Pattern of English. New York: Harper and Row. Cinque, G. (1999). Adverbs and Functional Heads: A Cross-linguistic Perspective. Oxford: Oxford University Press. Clark, R. & I. G. Roberts (1993). “A computational model of language learnability and language change.” Linguistic Inquiry, 24, 299–345. Cutler, A., J. Hawkins, & G. Gilligan (1985). “The suffixing preference: A processing explanation.” Linguistics, 23, 723–758. Dresher, E. (1999). “Charting the learning path: Cues to parameter setting.” Linguistic Inquiry, 30(1), 27–67. Dresher, E. & J. Kaye (1990). “A computational learning model for metrical phonology.” Cognition, 34, 137–195. Ellegård, A. (1953). The Auxiliary Do: The Establishment and Regulation of its Use in English. Stockholm: Almquist and Wiksell. Falk, C. (1993). “Non-referential subjects in the history of Swedish.” Doctoral dissertation, University of Lund. Fodor, J. D. (1998). “Unambiguous triggers.” Linguistic Inquiry, 29(1), 1–36. Fox, A. (1995). Linguistic Reconstruction: An Introduction to Theory and Method. Oxford: Oxford University Press. Fukui, N. (1986). “A theory of category projection and its application.” Doctoral dissertation, MIT. Fuß, E. & C. Trips (2002). “Variation and change in Old and Middle English. On the validity of the Double Base Hypothesis.” Journal of Comparative Germanic Linguistics, 4, 171– 224. Gelderen, E. van (2004). “Economy, innovation, and prescriptivism: From spec to head and head to head.” Journal of Comparative Germanic Linguistics, 7, 59–98. Gibson, E. & K. Wexler (1994). “Triggers.” Linguistic Inquiry, 25(3), 407–454. Givón, T. (1971). “Historical syntax and synchronic morphology. An archeologist’s field trip.” CLS, 7, 394–415. Givón, T. (1976). “Topic, pronouns, and grammatical agreement.” In C. N. Li (Ed.), Subject and Topic (pp. 149–188). New York: Academic Press. Greenberg, J. (Ed.). (1978). Universals of Human Language. Vol. I. Methods and Theory. Stanford: Stanford University Press. Haeberli, E. (1999). “Features, categories and the syntax of A-positions. Synchronic and diachronic variation in the Germanic languages.” Doctoral dissertation, University of Geneva. Haegeman, L. (1997). The New Comparative Syntax. London: Longman. Hale, M. (1996). “Theory and method in historical linguistics.” Ms., Concordia University, Montreal. Hale, M. (1998). “Diachronic syntax.” Syntax, 1(1), 1–18. Harris, A. & L. Campbell (1995). Historical Syntax in Cross-linguistic Perspective. Cambridge: Cambridge University Press.

Introduction

Hawkins, J. (1990). “A parsing theory of word order universals.” Linguistic Inquiry, 21, 223– 261. Holmberg, A. & C. Platzack (1991). “On the role of inflection in Scandinavian syntax”. In W. Abraham, W. Kosmeijer, & E. Reuland (Eds.), Issues in Germanic Syntax (pp. 93– 118). Berlin: Mouton de Gruyter. Hopper, P. & E. Traugott (1993). Grammaticalization. Cambridge: Cambridge University Press. Hyams, N. (1986). Language Acquisition and the Theory of Parameters. Dordrecht: Reidel. Jespersen, O. (1922). Language: Its Nature, Development and Origin. London: Allen and Unwin. Jones, W. (1807). The Works of Sir William Jones. London: John Stuckdale. Keenan, E. (2002). “Explaining the creation of reflexive pronouns in English.” In D. Minkova & R. Stockwell (Eds.), Studies in the History of English: A Millenial Perspective (pp. 325– 355). Berlin: Mouton de Gruyter. Keenan, E. (2003). “A historical explanation of some Binding theoretical facts in English.” In John Moore & Maria Polinsky (Eds.), The Nature of Explanation in Linguistic Theory (pp. 152–189). Stanford: CSLI Publications. Kemenade, A. van (1987). “Syntactic case and morphological case in the history of English.” Doctoral dissertation, Rijksuniversiteit te Utrecht. Kemenade, A. van & N. Vincent (Eds.). (1997). Parameters of Morphosyntactic Change. Cambridge: Cambridge University Press. King, R. (1969). Historical Linguistics and Generative Grammar. New York: Prentice-Hall. Kiparsky, P. (1965). Phonological Change. Cambridge, MA: MIT Press. Kiparsky, P. (1995). “Indo-European origins of Germanic syntax.” In A. Battye & I. Roberts (Eds.), Clause Structure and Language Change (pp. 140–169). Oxford: Oxford University Press. Kiparsky, P. (1997). “The rise of positional licensing.” In A. van Kemenade & N. Vincent (Eds.), Parameters of Morphosyntactic Change (pp. 460–493). Cambridge: Cambridge University Press. Kirby, S. (1999). Function, Selection, and Innateness: The Emergence of Language Universals. Oxford: Oxford University Press. Koopman, H. & D. Sportiche (1991). “The positions of subjects.” Lingua, 85, 211–258. Kosmeijer, W. (1986). “The status of the finite inflection in Icelandic and Swedish.” Working Papers in Scandinavian Syntax, 26, 1–41. Kroch, A. (1989). “Reflexes of grammar in patterns of language change.” Language Variation and Change, 1, 199–244. Kroch, A. (1994). “Morphosyntactic variation”. In K. Beals, J. Denton, B. Knippen, L. Melnar, H. Suzuki, & E. Zeinfeld (Eds.), Proceedings of the Thirthieth Annual Meeting of the Chicago Linguistics Society, Vol. II (pp. 180–201). Chicago: Chicago Linguistics Society. Kroch, A. (2001). “Syntactic change”. In M. Baltin & C. Collins (Eds.), The Handbook of Contemporary Syntactic Theory (pp. 699–730). Malden, MA: Blackwell. Kroch, A. & A. Taylor (1997). “Verb movement in Old and Middle English: Dialect variation and language contact.” In A. van Kemenade & N. Vincent (Eds.), Parameters of Morphosyntactic Change (pp. 297–325). Cambridge: Cambridge University Press.






Kroch, A., A. Taylor, & D. Ringe (2000). “The Middle English verb-second constraint: A case study in language contact and language change”. In S. Herring, P. van Reenen, & L. Schoesler (Eds.), Textual Parameters in Old Language (pp. 353–391). Amsterdam/Philadelphia: John Benjamins. Labov, W. (1994). Principles of Linguistic Change. Oxford: Blackwell. Larson, R. (1988). “On the double object construction.” Linguistic Inquiry, 19, 335–391. Lehmann, C. (2002). Thoughts on Grammaticalization. Second edition. Arbeitspapiere des Seminars für Sprachwissenschaft der Universität Erfurt Nr. 9. Lightfoot, D. (1979). Principles of Diachronic Syntax. Cambridge: Cambridge University Press. Lightfoot, D. (1991). How to Set Parameters: Arguments from Language Change. Cambridge, MA: MIT Press. Lightfoot, D. (1999). The Development of Language: Acquisition, Change and Evolution. Malden, MA: Blackwell. Longobardi, G. (2001). “Formal syntax, diachronic minimalism, and etymology: The history of French chez.” Linguistic Inquiry, 32(2), 275–302. Niyogi, P. & R. C. Berwick (1998). “The logical problem of language change: A case study of European Portuguese.” Syntax, 1(2), 192–205. O’Grady, W. (1997). Syntactic Development. Chicago: Chicago University Press. Ouhalla, J. (1991). Functional Projections and Parametric Variation. London: Routledge. Paul, H. (1880/1920 [5th edn.]). Prinzipien der Sprachgeschichte. [Konzepte der Sprach- und Literaturwissenschaft, 6.] Tübingen: Niemeyer. Pintzuk, S. (1999). Phrase Structures in Competition: Variation and Change in Old English Word Order. New York: Garland. Pintzuk, S., G. Tsoulas, & A. Warner (Eds.). (2000). Diachronic Syntax. Models and Mechanisms. Oxford: Oxford University Press. Platzack, C. (1988). “The emergence of a word order difference in Scandinavian subordinate clauses.” McGill Working Papers in Linguistics: Special Issue on Comparative Germanic Syntax, 215–238. Pollock, J.-Y. (1989). “Verb movement, UG and the structure of IP.” Linguistic Inquiry, 20(3), 365–424. Postal, P. (1968). Aspects of Phonological Theory. New York: Harper and Row. Rizzi, L. (1982). Issues in Italian Syntax. Dordrecht: Foris. Rizzi, L. (1997). “The fine structure of the left periphery.” In L. Haegeman (Ed.), Elements of Grammar: Handbook in Generative Syntax (pp. 281–337). Dordrecht: Kluwer. Roberts, I. (1993a). Verbs and Diachronic Syntax. A Comparative History of English and French. Dordrecht: Kluwer. Roberts, I. (1993b). “A formal account of grammaticalization in the history of Romance futures.” Folia Linguistica Historica, 13, 219–258. Roberts, I. (1995). “Object movement and verb movement in Early Modern English”. In H. Haider, S. Olsen, & S. Vikner (Eds.), Studies in Comparative Germanic Syntax (pp. 269–284). Dordrecht: Kluwer.

Introduction

Roberts, I. (1997). “Directionality and word order change in the history of English.” In A. van Kemenade & N. Vincent (Eds.), Parameters of Morphosyntactic Change (pp. 397–427). Cambridge: Cambridge University Press. Roberts, I. (1999). “Verb movement and markedness.” In M. deGraff (Ed.), Language Creation and Language Change: Creolization, Diachrony and Development (pp. 287– 329). Cambridge, MA: MIT Press. Roberts, I. & A. Roussou (2003). Syntactic Change. A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. Rohrbacher, B. (1994). “The Germanic VO languages and the full paradigm: A theory of V-to-I raising.” Doctoral dissertation, University of Massachusetts, Amherst. Sapir, E. (1921). Language. New York: Harcourt, Brace and World. de Saussure, F. (1967). Cours de linguistique générale. Wiesbaden: Harrassowitz. Schlegel, F. von (1808). Über die Sprache und Weisheit der Inder: ein Beitrag zur Begründung der Alterthumskunde. Heidelberg: Mohr and Zimmer. Simpson, A. & Z. Wu (2002). “Agreement, shells and focus.” Language, 78(2), 287–313. Szemerényi, O. (1989). Einführung in die vergleichende Sprachwissenschaft. Darmstadt: Wissenschaftliche Buchgesellschaft. Trips, C. (2002). From OV to VO in Early Middle English. Amsterdam/Philadelphia: John Benjamins. Vance, B. (1997). Syntactic Change in Medieval French: Verb Second and Null Subjects. Dordrecht: Kluwer. Vikner, S. (1994). “Finite verb movement in Scandinavian embedded clauses”. In N. Hornstein & D. Lightfoot (Eds.), Verb Movement (pp. 117–147). Cambridge: Cambridge University Press. Vikner, S. (1995). Verb Movement and Expletive Subjects in the Germanic Languages. Oxford: Oxford University Press. Vikner, S. (1997). “V-to-I movement and inflection for person in all tenses.” In L. Haegeman (Ed.), The New Comparative Syntax (pp. 189–213). London: Longman. Wexler, K. & P. Culicover (1980). Formal Principles of Language Acquisition. Cambridge, MA: The MIT Press. Wexler, K. & R. M. Manzini (1987). “Parameters, Binding theory, and learnability.” Linguistic Inquiry, 18, 413–444.



On the development of possessive determiners Consequences for DP structure* Artemis Alexiadou University of Stuttgart

In this paper I argue that an investigation of the diachronic development of possessive pronouns in English, German and French provides important insights into the structure of possessive DPs. More specifically, I claim that the grammaticalization path that possessive pronouns in e.g. English followed involves a change from possessive adjectives to possessive determiners, which is captured by a division of possessive pronouns into different sub-types along the lines of Cardinaletti (1998). In Cardinaletti’s system there is a structural difference between clitic possessors that move to D◦ , and weak/strong (adjectival) possessors that are XPs and either move to the specifier of a DP-internal AgrP (weak possessors) or may stay behind in their base position (Spec,nP, strong possessors). The diachronic data suggest that possessive determiners arise from weak adjectival possessors under two conditions: (a) the establishment of D◦ and (b) the loss or rather weakening of their agreement features.

.

Introduction

As is well known, in several languages the presence of a pre-nominal possessive pronoun excludes the presence of a determiner. This is illustrated in (1), from Schoorlemmer (1998: 58f.), with English, German and French examples respectively: (1) a. (*the) my book b. (*das) mein Buch c. (*le) mon livre

English German French



Artemis Alexiadou

However, in other languages pre-nominal possessive pronouns freely co-occur with determiners, as exemplified in (2) with an Italian example. Italian (2) il mio libro the my book

Possessors of the type in (1) are referred to as possessive determiners, while possessors of the type in (2) are referred to as possessive adjectives. As has been noted in the literature, the type of the pre-nominal possessive pronoun used in a language correlates with the definiteness of the whole noun phrase in this language. This correlation is illustrated in (3), from Schoorlemmer (1998: 60): the whole noun phrase is definite in languages which have possessive determiners (1), while it may be indefinite in languages which have possessive adjectives (2). Recent studies of this correlation include Cardinaletti (1998), Demske (1999), and Longobardi (1996) among others: (3) Pre-nominal possessors: Type 1 Type 2 article in the DP – + possessive construction may be – + indefinite in the presence of a possessor

An additional aspect of this correlation discussed in Schoorlemmer (1998) seems to be the fact that type 1 languages use a special form, different from the one that appears in pre-nominal position in ellipsis and predicative contexts. This is illustrated in (4) for English:1 (4) a. This book is mine. b. John will bring his car and I will bring mine.

In this paper I am concerned with a rather different aspect of this discussion. In particular, I investigate the diachronic development of the possessive elements. I argue that such an examination provides important insights into the structure of possessive DPs. More specifically, I show that the type 1 languages discussed here were type 2 languages in some period of their historical development. In other words, the data in (5) below are parallel to that in (2):2 Old English (5) a. se heora halga bisceop the their holy bishop

(Ælfrics Life of Saints 36.83)

On the development of possessive determiners

Old French b. le mien ami the my friend Old High German c. thia min-a frewida allo this my joy all

(Gamillscheg 1957: 169)

(Demske 1999: 122)

In particular I demonstrate that possessive pronouns in the three languages discussed here underwent a change in which possessive adjectives became possessive determiners. This historical development is a well known process of grammaticalization, since its result is the appearance of new functional material (Hopper & Traugott 1993; Roberts & Roussou 2003 and others). This suggests that the pattern in (3) is systematic and is relevant for processes of language change, in the sense that there is a historical relationship linking the two types with a clear direction: type 2 → type 1, i.e. adjective → determiner. In addition, this development coincides with the emergence of special forms which appear in ellipsis and predicative contexts, due to the breakdown of agreement morphology (at least in English, which will be discussed here in detail). The factors that relate to or could be considered as triggers of this re-arrangement will be extensively discussed here. The historical development as such fits well within the tripartite distinction of possessive elements and the general analysis of possessors presented in Cardinaletti (1998). This will be presented in detail in Section 2. Another aspect of the correlation in (3) that will be discussed here concerns the [±definite] features of possessive pronouns. It is often assumed that the variation in (3) is suggestive of a distinction between two types of possessors: Italian-like possessors, which are indefinite, and English-like possessors, which are definite. Here I pursue the hypothesis that definiteness in possessive constructions is related to the structural position the possessor occupies, rather than its inherent feature specification. It seems unlikely that the possessor acquires definiteness specification, although its other features (importantly agreement) get reduced. The paper is structured as follows. In the next section I introduce my theoretical assumptions concerning the structure of the DP, and Cardinaletti’s tripartition of possessive pronouns, so that I can refer back to it in the course of my discussion. In Section 3, I briefly summarize the evidence for the presence of a DP layer in Old English. In Sections 4 and 5 I am concerned with the possessive pronouns of Old and Middle English, and the various word order patterns that can be observed during these periods. In Section 6 I turn to





Artemis Alexiadou

some general remarks on the path of development presented for English and the feature specification of possessors. In Section 7 I briefly discuss the parallel changes of the possessive pronouns observed in German and French. In Section 8 I conclude my discussion.

. Theoretical background Following Carstens (2000), Radford (2000), Alexiadou (2003) and others, possessive DPs have the structure shown in (6):3 (6)

Building on a large amount of recent literature, I assume that only DPs can appear in argument positions, and the function of D is to render the noun phrase into an argument (see e.g. Stowell 1981; Longobardi 1996). Possessors are base generated in Spec,nP parallel to subjects in clause structure. FP1 corresponds to nominal Infl/Agr in e.g. Szabolcsi (1994). FP2 has been given several labels in the literature; I call it NumbP (Ritter 1991) here. It is not crucial for my discussion whether or not further projections are available between FP2 and nP. Arguably the syntactic projections in (6) contain semantic features and have distinct functions. DP is the level which encodes definiteness, and thus hosts elements that can check definiteness; NumbP hosts numerals and is responsible for number specification (i.e. plurality). nP introduces the possessor


much like vP introduces the agent in the clausal domain. Agr is a projection that is related to Case (genitive) (see Longobardi 1996; Siloni 1997) and in its specifier possessive adjectives of the Italian type are situated (see Cardinaletti for extensive discussion).4 Taking D to be the locus of definiteness, I assume that in order for a possessive DP to be (in)definite either Spec,DP or D must be filled. This relates to the general idea that features can be checked either by a head merged or moved to the head of a functional projection, or by an XP merged or moved to its specifier (see e.g. Alexiadou & Anagnostopoulou 1998 for such a proposal with respect to EPP-checking). Turning now to the sub-types of possessive pronouns seen in (1–2) above, the question is whether a distinction between adjectives and determiners is sufficient or whether further distinctions are necessary. Giorgi and Longobardi (1991) and Longobardi (1996) propose a bipartite division, in which some possessives are determiner-like, while others are adjectives. Concentrating mainly on Romance data, Cardinaletti (1998) proposes a tripartite division for possessive pronouns, which splits adjectival possessors into two sub-types.5 Cardinaletti crucially extends the tripartite clitic-weak-strong division proposed for personal pronouns in Cardinaletti and Starke (1999) to the possessive system. All three types of possessives are base generated in the lexical domain of the NP (or Spec,nP in (6)). The clitic possessive pronouns and the weak possessive pronouns move to a higher licensing position in the DP. Specifically, clitics incorporate into D. Because clitic possessives undergo head-movement to D, they cannot co-occur with articles. Possessive pronouns that do co-occur with articles must be analysed as either weak or strong. Weak possessors are maximal projections and move to a pre-nominal specifier position below D, Spec,AgrP in (6). Strong possessors can remain in their base position. If in addition N moves then strong possessive pronouns will be postnominal. (7a) summarizes the different types of structural positions in the DP in terms of functional structure, as elaborated in Cardinaletti’s work. (7b) contains a clitic possessor, (7c) a weak possessor, (7d) contains a strong possessor. (7) a. b. c. d.

[DP [D clitic] [AgrP weak [np strong [n [NP ]]]]] [DP [D mai ] [AgrP ti voiturek [np ti [n tk [NP tk ]]]]] [DP la [AgrP miai . . . [macchinak [np ti [n tk [NP tk ]]]]]] [DP la [AgrP . . . [macchinak [np mia [n tk [NP tk ]]]]]] car the my





Artemis Alexiadou

Table 1. Different types of possessors

post-nominal definite article Isolation Predicative Focalization Modification Coordination Ellipsis

Strong

Weak

Clitic

Examples

+ + + + + + + –

– + – – – – – +

– – – – – – – –

(7d) (7a, 2) (9) (10) (11) (12) (13) (14)

Evidence that the structure (7c) correctly represents the position of weak possessors comes from the fact that they precede all other adjectives in the noun phrase (from Cardinaletti 1998: 18, her (2a–b)): (8) a.

la sua bella casa the his/her nice house b. le sue due altre probabili goffe reazioni immediate alla tua lettera the his/her two other probable clumsy reactions immediate to-the your letter (Cinque 1994: 95)

As Cardinaletti (1998) convincingly shows, the three types of possessives are systematically distinguished from one another in a number of syntactic environments, summarized in Table 1. The following examples illustrate the properties listed above for Italian. As will be shown later on, these environments are relevant for my discussion of the historical development of possessive pronouns in English, German and French and provide important clues as far as their status is concerned (i.e. weak, clitic, strong).6 (9) A: Di quale ragazzo è questo libro? of which boy is this book? B: Suo. His

(Cardinaletti 1998: 20, her (13))

(10) Questo libro è suo. this book is his (11) a.

La casa MIA, non tua. the house my, not yours b. *La mia macchina, non tua (see Cardinaletti 1998: 44, Note 2, though)


(12) a.

la casa proprio sua the house really his/her b. *la proprio sua casa

(13) a.

La casa sua e tua the house her and your b. *La sua e tua casa

(14) Ho invitato i miei amici e Gianni i suoi have invited the my friend and Gianni the his

(Cardinaletti 1998: 36)

How does this description relate to the data in (1) and (2)? Cardinaletti (1998: 24ff.) proposes that the pronominal possessors in (1) qualify as cliticlike. The possessors in (2) are weak possessors. French and German differ from English in that they have access to weak possessor forms, which the languages use in ellipsis contexts, e.g. Mon ami m’a presenté le sien ‘my friend [to] me has introduced the his’. The idea is that “ellipsis requires possessives with adjectival agreement inflection, which means that they are necessarily pre-nominal”, i.e. weak.7 On the other hand, the possessives that appear in ellipsis contexts in English (4) are strong, as English lacks adjectival agreement altogether.8 Having identified the layers of functional structure within the DP and the properties of the different types of possessors, let us first examine the English case, and consider the behavior of English possessives in the history of the language.

. The DP layer in Old English The literature on Old English suggests that the language did not have a definite or indefinite article, making use of demonstratives and numerals instead. Hence a first issue that needs to be addressed here is whether Old English (OE) had a DP layer or not. The demonstratives used in OE were se ‘the/that’ and þes ‘this’, which sometimes have to be translated as that or this, sometimes (most often, in fact) as the. As (15) shows, in contrast to Modern English, the possessive can co-occur with a demonstrative in OE: (15) þam minum mynstre the-dat my-dat monastery-dat

(from Pintzuk 2002)

In her very detailed study of word order in the OE noun phrase, Wood (2003) very convincingly argues that since demonstratives, possessives and adjectives are strictly ordered, there must be some functional layers above NP in OE (see





Artemis Alexiadou

Table 2. (from Crisma 2000)

ælm. God God ælm.

+se

–se

127 0

22 32

also 4.2). Wood finds no evidence for the presence of a dedicated article, and assumes that OE only had demonstratives, which are analysed as specifiers of DP, hence specifying the whole DP as definite, in agreement with the remarks made in Section 2. A different argument in favor of the presence of a DP layer in OE, which, however, assumes the presence of a determiner-like element, comes from the work of Crisma (2000). Crisma observes that there is one construction involving the noun God and the OE adjective ‘almighty’, where there is a correlation between the presence vs. absence of a determiner and the position of the adjective; the interesting thing here is that only in this construction can the adjective either precede or follow the noun. Table 2 shows that the adjective uniformly precedes the noun when the determiner se is present, while it can either precede or follow the noun in the absence of se. Crisma takes this as evidence that OE has a particular case of N-to-D movement, though limited to a single lexical item, God, modified by the adjective ælmihtig ‘almighty’. The total absence of the co-occurrence of a determiner together with a post-nominal adjective is predicted by Longobardi’s (1994, 1996) N-to-D hypothesis. It is interesting to note here that the adjective appears in two different forms depending on the presence of the determiner: se ælmihtiga and ælmihtig. The former, the week adjectival form, is the one that was used in OE in modifying contexts, while the latter, the strong form, was used in predicative contexts (see also Section 4.1). This distribution could suggest that the adjective is base-generated in a position to the right of the head noun in the latter case, and hence the presence of the word order pattern N> Adj is irrelevant for the purposes of head-movement (Cinque 1994) and the presence of D in OE. In addition, this construction is very limited and occurs only in a particular text. Still, building on Wood’s results, I assume that there is a DP layer in OE, although the head of this DP is not occupied by a definite article. One could compare the situation found in OE to other languages which lack determiners, but have demonstratives marking the DP layer, e.g. Russian.9 Hence I conclude that the OE noun phrase contains a DP layer.


. Possessive pronouns in Old English The next question that must be dealt with concerns the behavior of Old English possessive pronouns. As will be shown below, these exhibit a rather different behavior than their modern counterparts. . Morphology and distribution As is well known, possessive pronouns carried relatively rich inflection in OE. In Table 3 at the bottom of this page, the counterpart of Modern English my is presented (Demske 1999; Presta 2003). As can be seen, the possessive pronoun carries case, gender and number information. The same applies for the other persons. There are several criteria that support an analysis of OE possessive pronouns as adjectives (see Demske (op.cit.), Wood (op.cit.), and Presta (op.cit)). The data in this section are all taken from Demske (1999: 162f.). First of all, much like other adjectives, the possessive pronoun generally precedes the noun (16a); however, one does find examples where it follows the noun as in (16b). (16) a.

ðe [hiora ðeninga] cuðen understondan on Englisc those who their mess texts could understand in English (GP10.13) b. þe her worhte to me [of liðum minum] he that here could save me from sufferings my (Genesis 282.816)

Secondly, the OE possessive pronoun can co-occur with definite and indefinite determiner-like elements; in (17) we see that the possessive co-occurs with a demonstrative pronoun, while in (18) it co-occurs with a numeral: (17) a.

Hie þa lærde [se heora halga bisceop] they there taught the their holy bishop (Ælfrics Life of Saints 36.83) b. [þes min gefea] is gefylled this my joy is extinct (John 3.29)

Table 3. Old English possessive pronoun 1st person mîn

Nom. Gen. Dat. Acc.

Singular masc.

fem.

neuter

Plural masc.

fem.

neuter

Ø -es -um -ne

-u, -o -re -re -e

Ø -es -um Ø

-e -ra -um -e

-a, -e -ra -um -a, -e

-u, -o -ra -um -u, -o





Artemis Alexiadou

(18) þær after (. . . ) wearþ se cyng Willelm on huntnoþe fram [his after that was King William on hunt by his anan men (. . . ) ofsceoten] one servant slain (Saxon Chronicle 1562/1100.6)

Thirdly, when possessive pronouns appear together with other adjectives, then they usually precede them, as (19a) shows. But there are also nominal phrases in which possessive pronouns follow the adjective (19b): (19) a.

& lædaþ [eowerne gyngstan broðor] to me, þæt ic and guide your youngest brother to me that I wite þæt ge sceaweras ne sind know that you are no spies b. of [inneweardre his heortan] from the more inner of his heart

(Genesis 42.34)

Fourthly, we observe several differences in the agreement patterns, which are identical to those found with other adjectives (see also Section 3). As example (20) shows, in predicative contexts the strong form of the possessive pronoun appears, as is the case with other adjectives used in predicative contexts (21). (20) & min-o alle ðin-o sint & ðin-o min-o sint and mine all yours are and yours mine are (21) Durh godes gife ge sind gehealden-e through God’s gift you are hold

(John 17.10) (Ælfrics Homilies I.7 78.114.5)

As with other Germanic languages, the appearance of weak and strong adjectival inflection correlates with the (in)definiteness of the respective determiner. The general pattern is illustrated in (22), from Demske (1999: 163). (22) a.

[on þone gren-an weald] into the green wood.masc.sg.acc b. [sum-ne adlig-ne mannan] some noble man.masc.sg.acc

(Genesis 287.835) (Ælfrics Homilies II.31 295.239)

In (22a), the adjective carries the ending -an which belongs to the weak adjectival declension, while in (22b) the adjective carries the ending -ne which belongs to the strong adjectival declension. Generally, when adjectives follow the possessive pronoun, the weak adjectival inflection appears in the majority of the cases (23a), but one finds instances of strong adjectival inflection as well (23b). Demske notes that while demonstratives control the adjectival inflection


in OE, this holds only partially for the possessive pronouns, unlike the situation in Old High German. (23) a.

Forhwan forwyrndest þu me [þæs min-es agen-an yrfes] why refuse you me this my own inheritance.neuter.sg.gen (HomS 35.75) b. [[min-es agen-es lifes] and [mines folces feores]] my own life and my peoples’ soul.neuter.sg.gen (Ælfrics Homilies M14 76.251)

From the above description, we can conclude that the possessive pronouns in Old English behave like adjectives. . Word order patterns in Old English Summarizing the word order patterns containing a possessive pronoun in OE discussed in the previous section, we find the following (see also Wood 2003, citing Mitchell 1985, and Pintzuk 2002): (24) a. b. c. d. e. f.

possessive > noun noun > possessive demonstrative > possessive > noun demonstrative > possessive > adjective > noun possessive > demonstrative > (adjective)> noun demonstrative > noun > possessive

Pattern (24e) has not been discussed yet, and is illustrated in (25). This pattern is interesting, as, according to Wood, this ordering is observed only when an adjective is present: (25) his sio gode modor his that good mother (Orosius 270.26, cited in Wood 2003, Chapter 4)

Concerning the order in (24f), both Wood and Pintzuk agree that it is very rare, and to the extent that it can be found in texts at all, it does not seem to be an OE construction. (24f) would be the only pattern containing a strong possessor in Cardinaletti’s sense and N-movement, given the discussion in Section 2. Its marginality, however, clearly suggests that N-movement does not seem to be a regular process within the OE DP, and that strong possessors must have disappeared very early in the history of the language as is the case for e.g. French (see Section 7.2). Taking this as evidence that OE lacks N-movement, naturally,





Artemis Alexiadou

Table 4. (from Pintzuk 2002): Positions of possessives in OE poetry

Arguments Predicates Vocatives

Total

Number inverted

Inversion not for metrical reasons

564 19 19

132/23.4% 8/42.1% 16/84.2%

6 0 4

Table 5. (from Pintzuk 2002): Positions of possessives in OE prose

Arguments Predicates Vocatives

Total

Number inverted

14224 465 138

7 1 20

something more needs to be said about pattern (24b), to which I turn below, see also Section 3. Consider then pattern (24b). Pintzuk’s quantitative study suggests this pattern is also very marginal. In fact she points out that possessives in OE can appear post-nominally, but only because of meter, since they are found in very small numbers in prose, and the numbers of inversion not for metrical reasons are indeed very limited (cf. Tables 4 and 5). These numbers, nevertheless, do suggest that this word order is marginally possible in OE and that it is not totally excluded from the grammar of the language.10 Again, as discussed in Section 3, the order noun>possessive could be argued to be the result of N-movement to D. But Pintzuk’s study shows that it is very limited. Hence we can conclude that N-movement is very marginal in the OE DP, to the extent that it takes place at all. The fact that it is a possible metrical alternation suggests that this process existed in stages of English prior to the texts we have at our disposal. Note, however, that the N-movement analysis of this pattern is only one possibility. If fact nothing prevents an analysis of pattern (24b) as involving XP movement of the noun to Spec,DP. Consider next the distribution of and the differences between (24a and c). The presence of (24a) would suggest that OE possessors are clitics in Cardinaletti’s sense, since they precede the noun and do not require a determiner-like element to precede them. On the other hand, the presence of (24c) would suggest that they behave like weak possessors, since they follow a determiner-like element. What is more intriguing is the fact that the two patterns seem to cooccur; in fact there is evidence that (24a) appears later than (24c). The evidence is discussed in Wood (2003). Wood observes that there is a difference between the earlier and the newer version of Gregory’s dialogue: the demonstrative is


missing in the later version, while it is present in the earlier version. This is illustrated in the examples below, taken from Wood’s work (Chapter 4; (26) is actually an example of (24e)): (26) min þæt ungesælige mod my that unfortunate spirit (27) min ungesælige mod

(GD:C 4.9) (GD:H, 4.9)

Wood further observes that the reviser did not remove the demonstrative on every occasion. In (28), the later H version has no se. But in (29) and (30), the determiner is retained in the newer version. (28) & þæt man ongæte hwæt se his wisdom wære and that one grasped, who that.masc.sg.nom his wisdom.masc were (GREGD4,35.12) (29) þa se ylca þegn þæt his hors þe he geseah acyrred fram when that same servant that his horse that he saw turned from his wedenheortnesse his madness (GREGD4,78.15) (30) þæt he þa gita on þære his wædle wæs swyðor geangsumod that he still on that his poverty was more vexed (GREGD3,57.10)

The above distribution suggests that the two patterns seem to be equivalent. Moreover, pattern (24c) is the old pattern, while pattern (24a) seems to be the new pattern. Pattern (24c) is immediately accounted for in terms of the structure in (6). This is also the conclusion drawn in Wood’s work. Hence the structure of e.g. (28), containing a weak possessor is as in (31): (31) [DP se [IP hisi [np [n ti [NP wisdom]]]]]

On this view, (31) is very much like its Italian counterpart in (2). The behavior of OE possessive pronouns of this type is similar to weak possessors: they follow demonstratives, they precede the head noun, they agree with it in all relevant features, and finally they precede all adjectives. It is not clear, however, whether at this stage the demonstrative could be analysed as being in the head of D. Rather it must be located in Spec,DP. In fact examples such as (25) and (26) seem to argue for a Spec analysis (see also the discussion round (32)). The fact that both (24c) and (24a) are possible suggests that the possessive is already undergoing syntactic change. Again it is far from clear whether already at this stage speakers analyse the possessor as a clitic. There are several reasons for this. First of all, the possessor is fully specified for





Artemis Alexiadou

all phi-features, and secondly, OE does not have a dedicated article at this stage, occupying D. Thus pattern (24a) could be an instance of a weak possessor in Spec,AgrP, or even in Spec,DP. Finally, consider pattern (24e). In this case, the possessive precedes the demonstrative. A potential analysis of this pattern is to suggest that the possessive pronoun has moved to a position higher than that of the demonstrative, an instance of XP-movement. Assuming that the demonstrative is in Spec,DP, this word order pattern could be taken to instantiate a topic-like construction. In fact the contrast is similar to what we find in Slavic languages e.g. Bulgarian (Vulchanova & Giusti 1998: 351, their (53)): (32) a.

tezi novi knigi na Ivan these new books to Ivan b. ?na Ivan tezi novi knigi to Ivan these new books

In (32b) the possessor is placed in Spec,TopicP, a projection to the left of DP, and the same could be argued to hold for the OE examples in (25) and (26). If Wood is correct in suggesting that (24e) in OE is regulated by the presence of an adjective, its existence could also imply that the possessive and the adjective compete for the same position, i.e. Spec,AgrP, within the OE DP. Once this is taken by the adjective, the possessor must move to a higher position. In the next section I turn to the possessive system of Middle English.

. Possessive pronouns in Middle English Table 6 below shows the possessive pronouns in Middle English. As is clear from the Table 6, in Middle English possessive pronouns have lost all inflectional agreement. Two important changes correlate with the loss of inflectional agreement in the possessive system. First, as reported in Jespersen (1961), an important novelty of Middle English is the possibility of dropping the -n in the possessive pronouns of the 1st and 2nd person singular, mîn and þîn, provided Table 6. Possessive pronouns in Middle English

singular plural

1. person

2. person

3. person masc./neuter fem.

min(e), mî ur(e), our(e)

þin(e), þi, thin(e) yûr(e), your(e), oûre

his hir(e), heore, her(e) here, heore, hore, þair, þar


that the following head noun begins with a consonant (except /h/). The forms mî, my, þî, þy are present when the next word begins with a consonant, data from Demske (1999: 164f.). (33) a.

thou toke [my lady] swete you took my lady sweet b. To [þi wif ] þu me take! to your wife you me take

(BD line 483) (KH line 536)

When the subsequent word begins with a vowel or an /h/, the forms mîn and þîn appear: (34) That wolde chaunge his youthe for [myn age]; that would change his youth for my age

(PT II. 724)

(35) The centre . . . hath but 24 holes, as [in myn instrument] the centre has but 24 holes as in my instrument (EPlanets 34/37)

The above examples suggest that the distribution of the allomorphs mî vs. mîn and þî vs. þîn is phonetically conditioned. What the above examples further show is that the possessive pronoun can appear without a determiner-like element, unlike what we saw in Section 4 for OE. However, one also finds examples where the possessive pronoun co-occurs with definite determiners (36). In fact co-occurrence of demonstratives and possessives can be found till Early Modern English (EMnE), see (37) cited in Rosenbach (2002: 225): (36) [þe myne dedis] the my dids

(Benedictine Rule 13.32)

(37) a. This his goodness b. that your Monastery

The presence of both patterns Determiner>Possessive>Noun and Possessive>Noun suggest that the ME possessives are ambiguous between being clitic and weak elements, a pattern that started already in OE, as discussed in Section 4. Similar to their OE counterparts, Middle English possessive pronouns can marginally follow the head noun (38): (38) [at tabel myne] at table mine

(KAlex. 4193)

A second novelty of Middle English is the development of new strengthened forms, which appear in predicative contexts (see the discussion in Allen 2002).





Artemis Alexiadou

Recall that in OE the strong form of the pronoun would appear in such contexts. But in Middle English new forms are created. Roughly there are two forms, the -n forms and the -s forms (39). According to Allen, the forms in -s are of Northern origin, spreading into the Midlands, while forms in -n are associated with the South and Midlands. (39) a.

and moren hit of hiren and increase it of theirs ‘and increase it of their own’ b. this gold is oures this gold is ours

(PT II. line 784-6)

Allen (2002: 204f.) summarizes the development of the strengthened forms as follows: The first strengthened forms of the independent possessive come from the first part of the 13th century. [...] It is not until the 14th century that we find clear evidence that strengthened forms were used in other areas of the country, but the old un-strengthened forms continued to be used in the South until the 17th century. There is no clear evidence that the -s forms were first used with a particular combination of person and number, but it seems clear that the new -n forms of the West Midlands began in the third person feminine and plural, spreading to the remaining person/number combinations by the early 15th century. With the rise of the standard based on the London dialect, which used the -s forms rather than the -n forms, the -n forms came to be considered substandard.

I agree with Allen that it is unlikely to be a coincidence that both the -n forms and the -s forms make their first appearance at the same time. It seems thus logical to conclude that the strengthened forms did not appear earlier, as before this period, English was a language with inflected possessives to agree with what they modified. With the loss of agreement morphology, however, it became possible for old suffixes like -s to be redeployed to convey different information or for new suffixes like -n, based on a phonological distinction, to arise. These developments led to a split of the pronoun system of the type in (40): (40) a. b. c. d.

This is my book. This is mine. *This is my. John will bring his car and I will bring mine/*my.

Let me summarize this development. In this section I first noted that Middle English used possessive pronouns without determiners and that they used new


strong forms in predicative contexts. However, I also mentioned that Middle English used possessive pronouns together with demonstratives. Moreover, as Allen states, other dialects managed perfectly well with the un-strengthened independent possessives for quite a long time. Thus we seem to have two patterns that co-existed for a long period of time. A pattern with clitic possessors and special forms for ellipsis/predicative contexts, and a pattern with weak possessors and no special forms for ellipsis/predicative contexts. This is a characteristic feature of language change, where new forms exist parallel to old forms and structures. This state of affairs seems to suggest that although the possessive pronouns of Middle English lack agreement morphology (and the language is characterized by a decline of case-marking in general), the other syntactic changes, i.e. new strong forms and change in the placement of the pronoun with respect to the article happened later. I take this to imply that loss of agreement might have been a necessary but not a sufficient cause for the advent of strengthened possessives and the re-analysis of possessive adjectives as possessive clitics/determiners. That is: new strengthened forms could only develop because of the loss of agreement, but the syntax of the possessive pronouns remained identical long after the loss of agreement. An inter-related factor seems to be the status and the position of determiners in general (definite and indefinite). As already mentioned, prior to the changes affecting the possessive system, there was no unique determiner occupying D◦ in English. Hence a string such as the my book, could be analysed as involving two adjective-like elements (41a). Naturally as long as agreement morphology was present this was in a sense the ‘default’ analysis. With the loss of morphology and the re-analysis that both these elements underwent, the only possible structure that corresponds to the placement of articles as well as to that of possessors is the one in (41b). In this structure the two compete for the same position: (41) a. [ the [ my [ book]]] b. [the/my [book ]]

old structure new structure

The re-assignment of syntactic positions was thus facilitated by the establishment of a unique determiner (i.e. the demonstrative which itself underwent grammaticalization and became a head) that makes D visible in English. In sum, the syntactic change that affected the possessive system of English was the result of loss of agreement in the pronominal system, the emergence/visibility of D◦ , as well as the change in status of the possessive pronouns themselves to which I turn in the next section.





Artemis Alexiadou

. From XP to clitic The change just described is a well known case of language change, and of grammaticalization in particular, as it relates to the development of new grammatical (functional) material out of ‘autonomous’ words. Initially autonomous words, namely the possessive pronouns, become determiner-like. In terms of phrase structure this means that they can no longer be situated in Spec positions, and they need to cliticise to D. As the overview in the previous sections suggests, the change in the possessive system is gradual, since it started some time during the OE period but it was completed only in the Early Modern English period. Second, the path it followed is: strong→weak→clitic. I assume here that prior to OE or in the early stages of OE strong possessors could be identified, and we have seen that such examples were marginal in OE. Third, it is characterized by the loss of agreement morphology and phonological reduction. However, loss of morphological properties is not a sufficient condition for this change. This is especially clear by the Middle English facts, where possessors lack agreement, but maintain syntactic properties of the OE system. That is, although agreement is reduced, the ‘older’ grammar is preserved for many years after the loss of morphology. This suggests that the re-analysis of the constructions is facilitated by other factors, e.g. the emergence of the determiner system. A related issue here concerns the internal structure of the DP in English. I have been assuming thus far that the presence of agreement within the Old English noun phrase is evidence for the presence of AgrP within the OE DP. One could assume that the loss of agreement is also reflected by the ‘loss’ of AgrP, leaving the English DP with only the category Number, as suggested in Munn and Schmitt (to appear). In other words, AgrP is no longer visible. Hence (41) could be viewed as containing only NumberP and not an additional AgrP, the properties of AgrP being ‘resumed’ by D and/or Number in English (see Haeberli, this volume). Does this change affect the [±definite] specification of the possessors? In other words, were OE possessors indefinites, while their Modern English counterparts are definite? In Section 2, I took D to be locus of definiteness determination. I proposed that in order for a possessive DP to be definite either Spec,DP or D must be filled. The presence of an overt determiner in D or a demonstrative in Spec,DP necessarily renders the DP definite. The possessor itself, if it is not situated in D, cannot have this function. Still, the question remains: why do possessors appear in DP? A standard assumption in the literature is that possessors are prototypically human and


animate. Possessive pronouns originate from the personal pronoun paradigm, hence one can assume that they carry person specification, or in the case of third person pronouns they carry definiteness specification (Ritter 1995). Such features are very similar to the features located in D (see for instance Chomsky 2001 who relates D to person features). Structurally then, D encodes features such as animacy and person and placement within the noun phrase is regulated by these features, in ways reminiscent of the clausal domain. On this view, D parallels the role of T in the verbal domain, which anchors/licenses person as well as animacy features in certain languages (Lecarme 1996 and references therein). This leads us to the following interpretation: as soon as the structural status of possessors changes in a language, as is the case in English, possessors necessarily appear in D. As a result, the D head is always filled by a possessive clitic and the whole DP is definite. In the next section I turn to some remarks on a similar development that took place in the history of German and French, drawing from the discussion in Demske (1999) and Arteaga (1995a–b) and references therein.

. Possessive pronouns in German and French Thus far I have been discussing the development of the English possessive system. A natural question to ask at this point is the following: since this development has a set of well-described properties and follows a clearly-defined path of grammaticalization, one would expect that this is found across languages. In fact this has been argued to be the case in German by Demske (1999), and could easily be shown to hold for French as well. . German As noted in Demske (op.cit.), in German, the possessive pronouns originate from a number of different pronouns, and thus the individual possessive pronouns differ in their behaviour. In the 1st and 2nd person singular and plural the possessive pronoun has developed from the genitive forms of the personal pronouns (42a), while in the 3rd person singular masculine and neuter the possessive has developed from the genitive form of the reflexive pronouns (42b). This is restricted to sîn. For all other 3rd person forms, that is the feminine singular and the plural, there is no possessive pronoun in Old High German. This function is taken over by the genitive of the personal pronoun of the 3rd





Artemis Alexiadou

Table 7. The possessive pronouns of Old High German

Nom. Gen. Dat. Acc.

Singular masc.

fem.

neuter

Plural masc.

fem.

neuter

Ø, -er -es -emu, -emo -an

Ø, -iu -era -eru, -ero -a

Ø, -az -es -emu, -emo Ø, -az

-e -ero -em, -en -e

o -ero -em, -en -o

-iu -ero -em, -en -iu

person singular feminine and the 3rd person plural (42c), which is similar to Latin (unless otherwise cited the examples come from Demske 1999). (42) a. mîn, dîn, unser, iuwer b. sîn- masc./neuter.sg c. ira- fem.sg, iro- 3.ps.pl

The inflection of the possessive pronoun follows the pronominal/strong adjectival inflection. Table 7 at the top of this page shows the paradigm of the possessive pronouns in Old High German using as an example mîn. In the nominative singular the non-inflected form appears; this generally corresponds to the genitives of the personal pronouns, see (43). In the plural (44a), as well as in the genitive (44b) and in the dative (44c), corresponding inflected forms are found; these agree with the noun in gender, number and case. (43) a.

ich bim alt, inti [mîn-ø quena] fram ist gigangan in ira tagun I am old, and my-fem.sg.nom wife far is gone in her days (T 8.1) b. [sîn-ø muos] uuas heuuiskrekco inti uuildi honag his-neut.sg.nom food were grasshoppers and wild honey (T 13.11)

(44) a.

wir birun [thine scalka] we are your fools (O II.24.21) b. thaz queme [mines truhtines muoter] zi mir? that would come my-masc.sg.gen lord’s mother to me? (T 4.3) c. min odouuan lon ni habet [mit iuuaremo fater] else (you) reward not have from your-masc.sg.dat father (T 33.1)

The forms of the feminine singular as well as those of the plural exist without inflection for all cases till the 14th century, as the following examples demonstrate. (45) a.

Want [ira anon] warun thanana gotes drutthegana because [her ancestors].nom were from there, God’s disciples (O I.11.27)


b. ougdun [iro wisduam], ougdun [iro cleini] in thes tihtonnes reini (they) showed [their-pl.gen wisdom].acc, showed [their-pl.gen artistic knowledge].acc in the perfection of their rhyming (O I.1.5)

Presta (op.cit.) and Demske (op.cit.) note that although the inflectional paradigm suggests that the possessive pronouns can be categorised as adjectives, it is rather problematic to categorise them as such in Old High German. On the one hand, this is due to the great frequency of non-inflected forms which correspond to the genitive form of the respective personal pronouns. On the other hand, the possessive pronouns inflect in the same manner as other pronouns do in Old High German. Moreover, they can appear between simple demonstrative pronouns and nouns. The examples in (46) show the inflected and non-inflected forms of the possessive pronouns. (46) a.

thia min-a frewida allo this my-fem.sg.acc. whole joy b. ther ira-ø sun guater this her son good ‘her good son’ c. Ih hiar giscribe follon [then thinan muatwillon] I here put down perfectly this your intention

(O II.13.16)

(O I.6.4) (O IV.1.41)

In Middle High German the ad-nominal possessive pronouns seem to have been established as an adjectival category. While the possessive pronouns in Old High German always have a strong inflection form, no matter in which syntactic environments they appear, the possessive pronouns in Middle High German surface both with strong and weak inflection after definite articles. In (47a) an example of the strong inflection (and non-inflection) pattern and in (47b) one of the weak inflection pattern are shown. Recall that the weak inflection pattern appears, like with other adjectives, after a definite article. This situation is reminiscent of what we have seen in Section 4 for OE. (47) a.

diu sin-iu keiserlich-en bein the his Kaiser(adj.) leg b. die iuwer-n scoen-en tohter the your beautiful daughter

(Tristan and Isolde, line 710) (Nibelungenlied, line 2188.4)

In that period the possessive pronoun typically appears following the determiner and preceding all other adjectives, see (48): (48) [ein myn lieber frunt] one my dear friend

(PL 162.9)





Artemis Alexiadou

As (49) shows, the possessive pronoun is not only able to precede other adjectives but can also follow them. Arguably these data suggest that the adjective has been moved to a position preceding the possessive, perhaps Spec,TopicP, see Section 4. (49) [Liebes myn kint] dear my child

(PL 180.23)

Demske’s discussion suggests that at this stage the German possessive pronoun exhibits properties of both weak and clitic possessors in Cardinaletti’s tripartition, since it appears both with and without a determiner. The status of the pronoun in Modern German is, however, controversial. According to Demske, the possessive pronoun has been re-analysed as a possessive determiner, much like what we have seen thus far for English. However, Löbel (1996) points out that the possessive can co-occur, although marginally, with demonstratives, as shown in (50): (50) diese meine Bücher these my books

In addition, Modern German possessive pronouns agree with the noun they modify in gender, number and case, which is not the case in English, but is what we see in French (for number and gender). It is interesting to point out here that certain dialects of Modern German, e.g. Franconian, have developed a system similar to that of Modern English, having possessive pronouns lacking (number and gender) agreement (51). These dialects use a different form for predicative contexts (52): Franconian11 (51) mei Buch dei Buch mei Bücha dei Bücha my book your book my books your books mei Tür dei Tür mei Türa dei Türa my door.fem your door.fem my doors your doors mei Lehrer dei Lehrer mei Lehrer dei Lehrer my teacher.masc your teacher.masc my teachers your teachers (52) des buch is meis the book is mine

The above pattern suggests that at least in this dialect of German the possessive pronoun has a clitic status, similar to its English counterpart. On the other hand, the standard German pronoun seems to not have completed its grammaticalization cycle yet. But the fact that these pronouns are having reduced


agreement features can be observed from the fact that they do not happily occur in predication contexts, although they do occur in ellipsis contexts. . French The situation in French is reminiscent of what we have seen for English and German. Arteaga (1995a–b) notes that Old French possessives were of two types, clitic-like and weak, in Cardinaletti’s sense. This is exemplified below: (53) a.

Mes pedre me desidret my father me desires b. ... que li miens cuers that the my heart

(Arteaga 1995a: 68)

According to Arteaga, both structures are found in the same period, and could be syntactically conjoined, suggesting that they had the same meaning. Gamillscheg (1957) notes that only rhythmical or metrical reasons determine the choice between the two types. The two types co-occur until the 17th century. Arteaga also notes that the pattern demonstrative>noun>possessive (e.g. *li amis miens) is not attested. According to her, Old French does not seem to have any strong possessives. We have seen that the evidence for strong possessive adjectives in OE is marginal as well. In Old French possessors were already weak, the strong forms being expressed via PPs or oblique noun phrases: (54) a.

Fiz a rei son to king b. les cornes del cerf the horns the deer c. la femme son voisin the wife the neighbour

(Arteaga 1995b: 80 )

Arteaga (1995b) notes that examples of the type in (54c) disappeared in the Middle French period. How do German and French resemble or differ from English? First of all, all three languages developed weak adjectival possessives. Arguably both French and (dialects of) German developed clitic possessors similar to those of English. Thus far the same principles apply. Where French and German differ from English is in the domain of agreement. First of all, both French and German are still characterized by adjective-noun agreement. Second, French and German have forms of the possessive pronouns different from the prenominal





Artemis Alexiadou

ones, which are used in ellipsis contexts. Assuming that ellipsis is licensed under some sort of recoverability of information, if a language has an agreeing prononinal possessor, that possessor will appear in ellipsis. In Cardinaletti’s system these are weak possessors. In fact the Old French data support this analysis: as we have just seen in older stages of French the forms that appear in ellipsis in Modern French could follow the article. Thus the language has a ‘relic’ of weak possessors used in ellipsis. English lacks agreement altogether and hence necessarily makes use of the strong possessive form in these contexts. Note here that German and French differ from one another. In particular, in predicative contexts, German uses special forms, similar to those found in ellipsis while French makes use of PP possessors, and cannot use the form used in ellipsis: (55) a.

mein Buch my book b. das Buch ist meins the book is mine

This seems to be related to the details of the development of the possessives in the individual languages. The German pattern here seems to be reminiscent of the English development, but it is not exactly identical to it, as German still retains agreement. French differs from both German and English, for reasons that presumably have to do with the early loss of weak possessors in the language and the ways possession is expressed in this language in particular. Thus the possessive system of French could not re-organize itself in a way similar to that of English or German.

. Summary In this paper I examined three languages that show a development path according to which possessive adjectives become possessive determiners, focussing on English. This historical development is a well known process of grammaticalization, since its result is the appearance of new functional material. I showed that in English this development coincides with the emergence of special forms which appear in ellipsis and predicative contexts, due to the breakdown of agreement morphology. The stages of the development of possessive pronouns provide further evidence for a tripartite distinction for possessive pronouns along the lines of Cardinaletti (1998). In Cardinaletti’s system there is a structural difference be-


tween clitic possessors that move to D◦ , and weak/strong possessors that are XPs and either move to the specifier of a DP-internal AgrP (weak possessors) or may stay behind in their base position (Spec,nP). The diachronic data suggest that possessive determiners arise from weak adjectival possessors under two conditions: (a) the establishment of D◦ and (b) the loss or rather weakening of their agreement features. It is important to note here once more that reduction of morphological and phonological features is crucial but not sufficient for syntactic re-analysis. For the case in point, English, German and French have access to a system of clitic possessors throughout their history, and irrespectively of the presence vs. absence of agreement. Speakers analyse ‘ambiguous’ structures in two different ways and in the end one of these options is maintained, corroborated by other factors as well, such as the emergence of D◦ , a result of the grammaticalization of the demonstrative. It should also be kept in mind that the three languages differ as to whether they make use of a residue of agreement in specific environments. As we have seen, the total loss of agreement within the English noun phrase results in the special forms used in ellipsis contexts in the language. On the other hand, German and French use forms available to them, which contain (a residue of) agreement.

Notes * I would like thank the participants to the workshop Diachronic Clues to Synchronic Syntax in Stuttgart in November 2002 for comments and discussion. Many thanks to the editors of this volume, one anonymous reviewer and Johanna Wood for helpful suggestions. . This holds for French as well. In (i) we see that French uses a special form of the pronoun in ellipsis contexts, while (ii) shows that another form is used for predicative contexts. I come back to the French pattern in 7.2. (i)

Mon ami m’a presenté le sien. my friend [to] me has introduced the his

(Cardinaletti 1998: 38, her (84a))

(ii) Le livre est à moi. the book is mine Note that the correlation between the absence of a determiner and a special form for both ellipsis and predication does not hold for German, where different forms are possible in predicative contexts (dependent on register/dialectal influence). Presumably this is related to the general dialectal differences concerning the stages of grammaticalization of the possessive pronoun in German. See also Section 7.1.





Artemis Alexiadou . It should be kept in mind that in the Old English and the Old High German examples we are rather dealing with a demonstrative and not with a determiner. I come back to this in Section 3. . Comparable structures are assumed in e.g. Cardinaletti (1998), Schoorlemmer (1998) among others, the difference being that these authors do not make use of an nP shell. (6) as well as comparable structures build on Cinque’s (1980) proposal that possessors are subjects in the noun phrase. . It has been argued that such a projection is present only in languages that show overt agreement in the DP e.g. the Romance languages (see Munn & Schmitt to appear for arguments that Agr and Numb are fused in Modern English). In these languages then case checking should take place in this fused category. . Cardinaletti’s typology is further refined in Ihsane (2000). Note here that ‘strong’ possessors are not to be viewed only as a sub-type of adjectival possessors, since PP possessors are necessarily strong. . As already mentioned in Note 5, PP possessors pattern like strong adjectival possessors. Cardinaletti’s data suggest that agreement, in the sense of concord, characterizes both weak and strong adjectival possessors. This implies that movement to Spec,AgrP is not what triggers overt agreement. . Cardinaletti notes that French seems to lack strong adjectival possessors. In all contexts in which a strong possessor is required a PP surfaces instead, e.g le livre est à moi ‘the book is mine’. See also 7.2. . With respect to the distribution of full DP possessors, English unlike the Romance languages has one of two options: prenominal genitive e.g. John’s book vs. post-nominal ofphrase. The former cannot co-occur with an overt determiner. The choice between the two patterns is regulated by animacy and presumably topichood. A question arises here with respect to the absence of a determiner in the former case: the assumption made is that in this case ‘s is located in D◦ , hence blocking the determiner from appearing there; see Longobardi (1996) and Alexiadou (to appear) for a recent discussion of these patterns. . In fact Schoorlemmer’s (1998) discussion of Russian and Bulgarian possessors suggests further similarities between OE and these languages. The reader is referred to her paper for details. . Note here that these tables relate to the pattern (24f) as well. . Data courtesy of Susann Fischer.

References Alexiadou, A. (2003). “Some notes on the structure of alienable and inalienable possessors.” In M. Coene & Y. D’hulst (Eds.), From NP to DP (pp. 167–188). Amsterdam: John Benjamins. Alexiadou, A. (to appear). “Possessors and (in)definiteness.” Lingua.


Alexiadou, A. & E. Anagnostopoulou (1998). “Parametrizing Agr: Word order, verbmovement and EPP-checking.” Natural Language and Linguistic Theory, 16(3), 491– 539. Allen, C. L. (2002). “The development of strengthened possessive pronouns in English.” Language Sciences, 24, 189–211. Arteaga, D. (1995a). “Strong and weak possessives in Old French.” Language Quarterly, 33, 67–80. Arteaga, D. (1995b). “On Old French genitive constructions.” In J. Amastac, G. Goodall, M. Montalbetti, & M. Phinney (Eds.), Contemporary Research in Romance Linguistics (pp. 79–90). Amsterdam: John Benjamins. Cardinaletti, A. (1998). “On the deficient/strong opposition in possessive systems.” In A. Alexiadou & C. Wilder (Eds.), Possessors, Predicates and Movement in the DP (pp. 17– 53). Amsterdam: John Benjamins. Cardinaletti, A. & M. Starke (1999). “The typology of structural deficiency: A case study of the three classes of pronouns.” In H. van Riemsdijk (Ed.), Clitics in the Languages of Europe [Vol. 8 of Language Typology] (pp. 145–233). Berlin: Mouton de Gruyter. Carstens, V. (2000). “Concord in minimalist theory.” Linguistic Inquiry, 31, 319–355. Cinque, G. (1980). “On extraction from NP in Italian.” Journal of Italian Linguistics, 5, 47– 100. Cinque, G. (1994). “On the evidence for partial N-movement in the Romance DP.” In G. Cinque, J. Koster, J.-Y. Pollock, L. Rizzi, & R. Zanuttini (Eds.), Paths Towards Universal Grammar. Studies in Honour of Richard Kayne (pp. 85–110). Washington, DC: Georgetown University Press. Chomsky, N. (2001). “Derivation by Phase.” In M. Kenstowicz (Ed.), Ken Hale: A Life in Language (pp. 1–52). Cambridge, MA: MIT Press. Crisma, P. (2000). “Sintassi formale e lingue medievali: l’articolo in inglese antico.” Archivio Glottologico Italiano, LXXXV(1), 38–84. Demske, U. (1999). Merkmale und Relationen: Diachrone Studien zur Nominalphrase des Deutschen. Habilitationschrift, Universität Jena. Gamillscheg, E. (1975). Historische Französische Syntax. Tübingen: Max Niemeyer Verlag. Giorgi, A. & G. Longobardi (1991). The Syntax of Noun Phrases. Cambridge: Cambridge University Press. Haeberli, E. (this volume). “Syntactic effects of inflectional morphology and competing grammars.” Hopper, P. & E. Traugott (1993). Grammaticalization. Cambridge: Cambridge University Press. Ihsane, T. (2000). “Three types of possessive modifiers.” Generative Grammar in Geneva, 1, 21–54. Jespersen, O. (1961). A Modern English Grammar on Historical Principles. London: George Allen and Unwin LTD. Lecarme, J. (1996). “Tense in the nominal system: the Somali DP.” In J. Lecarme, J. Lowenstamm, & U. Shlonsky (Eds.), Studies in Afroasiatic Syntax (pp. 159–178). The Hague: Holland Academic Graphics.





Artemis Alexiadou

Löbel, E. (1996). “Kategorisierung der Possessiva als Adjektiva in der NP/DP.” In T. Tappe & E. Löbel (Eds.), Die Struktur der Nominalphrase [Wuppertaler Arbeitspapiere zur Sprachwissenschaft 12] (pp. 58–94). Longobardi, G. (1994). “Reference and proper names: A theory of N-movement in syntax and logical form.” Linguistic Inquiry, 25, 609–665. Longobardi, G. (1996). “The syntax of N-raising: a minimalist theory.” Ms., University of Venice. Mitchell, B. (1985). Old English Syntax. Oxford: Oxford University Press. Munn, A. & C. Schmitt (to appear). “Number and definiteness.” Lingua. Pintzuk, S. (2002). “The syntax of Old English poetry: the position of heads in noun phrases and prepositional phrases.” Talk presented at the 8th Germanic Linguistics Annual Conference, Bloomington IN. Presta, E. (2003). “Possessive pronouns in Old and Middle English in comparison with possessive pronouns in Old High and Middle High German.” Term paper, University of Stuttgart. Radford, A. (2000). “NP-shells.” Essex Research Reports in Linguistics, 33, 2–20. Ritter, E. (1991). “Two functional categories in noun phrases: Evidence from Modern Hebrew.” In S. Rothstein (Ed.), Perspectives on Phrase Structure [Syntax and Semantics 26] (pp. 37–62). New York: Academic Press. Ritter, E. (1995). “On the syntactic category of pronouns and agreement.” Natural Language and Linguistic Theory, 13, 405–443. Roberts, I. & A. Roussou (2003). Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. Rosenbach, A. (2002). Genitive Variation in English. Conceptual Factors in Synchronic and Diachronic Studies. Berlin: Mouton de Gruyter. Schoorlemmer, M. (1998). “Possessors, articles and definiteness.” In A. Alexiadou & C. Wilder (Eds.), Possessors, Predicates and Movement in the DP (pp. 55–86). Amsterdam: John Benjamins. Siloni, T. (1997). Noun Phrases and Nominalizations. Dordrecht: Kluwer. Stowell, T. (1981). “Origins of phrase structure”. Doctoral dissertation, MIT. Szabolcsi, Anna (1994). “The noun phrase.” In F. Kiefer & K. E. Kiss (Eds.), The Syntactic Structure of Hungarian [Syntax and Semantics 27] (pp. 179–274). New York: Academic Press. Vulchanova, M. & G. Giust (1998). “Fragments of Balkan nominal structure.” In A. Alexiadou & C. Wilder (Eds.), Possessors, Predicates and Movement in the DP (pp. 33–360). Amsterdam: John Benjamins. Wood, J. (2003). “Definiteness and number: Determiner phrase and number phrase in the history of English.” Doctoral dissertation, Arizona State University.

Diachronic clues to pro-drop and complementizer agreement in Bavarian* Eric Fuß University of Frankfurt

This paper proposes a diachronic explanation of the fact that in Bavarian, complementizer agreement and pro-drop are limited to 2nd person contexts (along with 1pl in some dialects). It is argued that this restriction follows from a conspiracy of morphological and syntactic factors that guided the reanalysis of subject clitics as markers of verbal agreement in the history of Bavarian. More specifically, it is shown that this change could only take place in inversion contexts, leading to partial pro-drop and the presence of an agreement morpheme on C. The person/number restrictions are attributed to the workings of morphological blocking effects that confined the grammaticalization of new verbal agreement morphology to previously underspecified slots of the agreement paradigm. The diachronic analysis is based on a new approach to complementizer agreement that attributes the presence of inflection in the C-domain to the post-syntactic insertion of a ‘dissociated’ Agr-morpheme.

.

Introduction

In contrast to Standard German, Bavarian shows two series of pronominal elements: (i) a paradigm of full forms that may bear stress, and (ii) a set of reduced, atonic forms (cf. Table 1 at the top of the following page). The latter are C-oriented clitics that attach to the right of elements located in the C-system. The enclitic forms of the 2nd person nominative pronouns (2sg -st, 2pl -ts) exhibit special properties that set them apart from the other pronominal clitics. It is commonly assumed that the special behavior of the 2nd person ‘clitics’ should be taken to indicate that they are actually not pronominal elements. Instead, they are analyzed as pieces of inflectional morphology, namely agreement markers that attach to elements in the C-domain (cf. Bayer 1984;



Eric Fuß

Table 1. Nominative pronominal forms of Bavarian (Bayer 1984: 230) Full form 1sg 2sg 3sg, masc 3sg, fem 3sg, neut 1pl 2pl 3pl

I du er, der (demonst.) sie, die (demonst.) es, des (demonst.) mir ihr, es sie, die (demonst.)

Enclitic /i/ /du/ /er/, /der/ /si/, /di/ /es/, /des/ /mir/ /ir/, /ås/ /si/, /di/

-a, -e -(s)t -a -s -s -ma -(t)s -s

/a/, /e/ /sd/ /a/ /s/ /s/ /mer/ /ts/ /s/

Altmann 1984; Weiß 1998). The particular properties of the 2nd person ‘clitics’ are illustrated in the following. First, there are a number of syntactic (i.e. distributional) differences between 2nd person forms and the other pronominal clitics. In contrast to the latter, the 2nd person forms are obligatory in all contexts, that is, they cannot be simply replaced by the relevant full forms: (1) a.

ob-st noch Minga kumm-st whether-2sg to Munich come-2sg b. *ob du noch Minga kumm-st whether you.sg to Munich come-2sg ‘whether you come to Munich’

(2) a.

ob-ts noch Minga kumm-ts whether-2pl to Munich come-2pl b. *ob es/ihr noch Minga kumm-ts whether you.pl to Munich come-2pl ‘whether you come to Munich’

In addition, a full 2nd person subject pronoun is only acceptable if it co-occurs with the relevant clitic form, giving rise to obligatory subject clitic doubling. In these contexts, the full pronoun normally bears focal stress. (3) a.

ob-st DU noch Minga kumm-st whether-2sg you.sg to Munich come-2sg ‘whether you come to Munich’ b. ob-ts ES/IHR noch Minga kumm-ts whether-2pl you.pl to Munich come-2pl ‘whether you come to Munich’

This contrasts with the behavior of the other subject enclitics which are not obligatory and which cannot be doubled by full pronouns:1

Complementizer agreement and pro-drop in Bavarian

(4) a.

ob’e (*I) noch Minga kumm whether-1sg I to Munich come-1sg b. ob i noch Minga kumm whether I to Munich come-1sg ‘whether I come to Munich’

Apart from these syntactic differences, we can also observe some crucial morpho-phonological differences that underline the special status of 2nd person forms. First, in contrast to the other clitics, the 2nd person forms cannot be derived from the relevant full pronouns via phonological reduction processes (cf. Table 1). Second, the 2nd person ‘clitics’ are identical with the relevant verbal agreement suffixes, cf. Table 2. Note that normally (but not necessarily), clitics are derived from the full pronominal forms via phonological reduction processes. This is not the case with the 2nd person forms in Bavarian. Instead, the 2nd person forms show a clear similarity with verbal agreement endings, which already suggests that they are in fact inflectional elements. Accordingly, Pfalz (1918), Bayer (1984), Altmann (1984), Zwart (1993) and Weiß (1998), (2002) (among others) propose that the special properties of the 2nd person subject clitics can be explained on the assumption that these elements are in fact some form of inflection located in C. In this paper, I follow Bayer (1984) and analyze the 2nd person ‘enclitics’ as agreement morphemes that are attached to C (henceforth Agron-C). Note, however, that this assumption leaves an open question, namely why complementizer agreement is restricted to 2sg, 2pl forms. Another consequence of the above analysis is that Bavarian is regarded a ‘partial pro-drop’ language (Bayer 1984; Weiß 1998): sentences which contain only a 2nd person clitic have to be analyzed as instances of pro-drop. According to Bayer (1984), the overt manifestation of agreement in C (with 2sg, 2pl) serves to license referential pro in present-day Bavarian: (5) a.

Kumm-st pro noch Minga, dann muaß-t pro me b’suacha. come-2sg to Munich then must-2sg me visit ‘If you come to Munich you must visit me.’

Table 2. 2nd person full/reduced pronominal forms and verbal agreement

2sg 2pl

Full pronoun

Enclitic pronoun

Verbal agreement

du es/ihr

-st -ts

-st -ts





Eric Fuß

b. Kumm-ts pro noch Minga, dann müaß-ts pro me b’suacha. come-2pl to Munich then must-2pl me visit ‘If you come to Munich you must visit me.’ (6) a. *Kumm pro noch Minga . . . come-1sg to Munich ‘If I come to Munich, . . . ’ b. *Kumm-t pro noch Minga? come-3sg to Munich ‘Will he/she/it come to Munich?’

Again, the restriction to 2nd person is somewhat mysterious. For example, it cannot be attributed to some special morphological property of the 2nd person agreement suffixes, in the sense that 2nd person forms are ‘more distinctive’ than e.g. 1sg or 3sg, enabling them to license pro, in contrast to the other endings. Table 3 shows that the only non-distinctive (i.e. homophonous) agreement endings are 1pl and 3pl, whereas all other endings serve to unambiguously signal person and number of the subject. Thus, it appears that the restrictions on complementizer agreement and pro-drop observed in Bavarian escape a deeper synchronic account and can only be captured by a stipulation that simply rules out the non-existing cases (cf. Bayer 1984). In this paper, it is argued that the restrictions in question can be more fully understood if diachronic evidence is taken into account. More specifically, it is claimed that the restrictions on pro-drop and complementizer agreement follow from a set of syntactic and morphological factors that determined the reanalysis of subject clitics as verbal Agr-morphemes in the history of Bavarian. First, it is shown that this reanalysis could only take place in contexts with V-to-C movement, forcing the learner to assume the existence of (i) pro-drop and (ii) an agreement morpheme on C, leading to complementizer agreement. In addition, it is argued that the grammaticalization of new verbal agreement suffixes is shaped by blocking effects that favor the use of more specified forms Table 3. Verbal agreement paradigm (pres. indic.) of Bavarian Verbal agreement 1sg 2sg 3sg 1pl 2pl 3pl

-∅ -st -t -an -ts -an


over less specified forms (Kiparsky 1973, 1982; Anderson 1992; Halle 1997). The person/number restrictions observed above are shown to result from the fact that the change affected only defective/underspecified slots of the verbal agreement paradigm. Thus, the restrictions in question are analyzed as a residue of the independent development of new verbal agreement markers that affected only 2nd person forms and could only take place in contexts where the finite verb moved into the C-system. The paper is organized as follows. Section 2 briefly introduces the relevant diachronic facts concerning the development of new verbal agreement markers in the history of Bavarian. More specifically, it is shown that most varieties acquired new 2nd person markers and that a small number of Lower Bavarian dialects show a similar development for 1pl. In Section 3, an analysis of these facts is proposed. The diachronic explanation argued for in this paper is based on a new synchronic analysis of complementizer agreement couched in the framework of Distributed Morphology which is presented in Section 3.1. Syntactic aspects of the change in question are dealt with in Section 3.2. Section 3.3 discusses morphological factors involved in the rise of new verbal agreement in Bavarian. Section 4 summarizes the empirical and theoretical findings of this paper and discusses a set of predictions for the grammaticalization of agreement morphology across languages.

. The diachronic development of Agr-on-C in Bavarian In all varieties of Bavarian, C-oriented enclitics were reanalyzed as enlargements of existing agreement markers, giving rise to new agreement suffixes for 2nd person (cf. Bayer 1984; Wiesinger 1989). Table 4 below lists the inherited verbal agreement endings and the forms resulting from the reanalysis of the pronominal clitics for 2sg t(hu) 2pl, (¯e)s (originally a dual that developed into a 2pl pronoun in Bavarian). Interestingly, it can be shown that this grammaticalization process was initially confined to certain syntactic environments. The following data suggest Table 4. Old and new agreement suffixes for 2nd person in Bavarian

2sg 2pl

‘Old’ inherited ending

‘New’ enlarged ending

Lexical source

-s -t

-s+t -t+s

t(hu) (¯e)s





Eric Fuß

that the historical development of the new 2nd person agreement morphemes affected first finite verbs in C and spread later to other verbal positions. . 2sg -st The development of 2sg -st began already in early Old High German (OHG) presumably during the 9th century AD. The resulting new ending is found in all modern German varieties. It is commonly assumed (cf. Brinkmann 1931; Braune 1950: 252) that the ending 2sg -s+t resulted from a reanalysis of the combination verb + clitic pronoun t(hu) in inversion contexts, possibly on the analogy of the preterite-presents which already showed -st for the 2sg present indicative (kanst ‘can’, tarst ‘dare’, muost ‘must’, weist ‘know’ etc.) and the 2sg of ‘be’ bist which resulted from an independent and earlier development (cf. Lühr 1984).2 Examples such as (7) from the OHG Tatian seem to suggest that the new agreement ending was initially confined to verbs that have undergone V-to-C movement: (7) Ih forahta, [CP uuanta thu grim man bist, [C’ nimi-st [IP thaz thu since you grim man are take-2sg that you I feared ni sázto-s]]] inti [CP [C’ arno-st [IP thaz thu ni sáto-s. ]]] earn-2sg that you not sow-2sg neg plant-2sg and ‘Since you are a grim man, I feared that you take what you haven’t planted and earn what you haven’t sowed.’ (Tatian ζ 151,7; Sievers 1961: 228)

In (7), we can see that the new ending shows up on verbs that occupy the C position in V2 contexts, whereas finite verbs in clause-final position still exhibit the older ending -s.3 However, a second look at the relevant data reveals that the picture is actually more complex. To settle this matter requires a detailed quantitative analysis of the OHG texts in question (which is a topic for future research; see Section 3.3.2 below for morphological aspects of the development of 2sg -st in OHG/Bavarian). . 2pl -ts The grammaticalization of 2pl -ts (-t+-s < ¯es, originally a dual, as noted above) is a later development that began in the 13th century and is confined to Bavarian (i.e. it cannot be observed in other varieties of German). Again, it can be shown that the new ending appeared first on finite verbs that underwent movement to C0 . In some northern Bavarian dialects (spoken in Lauterbach &


Sangerberg), the new ending for 2pl -ts still attaches only to conjunctions and verbs in C, but not to verbs in clause-final position (Pfalz 1918: 232): (8) wei-ts iw6 t’pruk khumt-∅ sea-ts s’wi6tshaus when-2pl over the-bridge come see-2pl the-tavern ‘When you cross the bridge, you will see the tavern.’

These facts suggest that the new verbal agreement morphology developed via a transitional stage where the new ending was confined to the C-position. This claim is supported by data from Lower Bavarian dialects which exhibit a related grammaticalization process affecting 1pl forms. . 1pl -ma in Lower Bavarian In a set of Lower Bavarian dialects, the 1st person plural subject enclitic -ma developed in a similar way as the 2nd person enclitics (cf. Pfalz 1918; Bayer 1984; Altmann 1984; Kollmer 1987; Wiesinger 1989; Abraham 1995; Weiß 1998, 2002). That is, the pronominal enclitic 1pl -ma shows a similar behavior as the 2nd person forms: it is obligatory in all contexts, it can be doubled by full forms for emphatic reasons, and 1pl contexts license pro-drop in these dialects (Bayer 1984: 252): (9) a.

wem-ma aaf Minga fon when-1pl to Munich drive b. wem-ma MIA aaf Minga fon when-1pl we to Munich drive c. *wem mia aaf Minga fon when we to Munich drive ‘when we drive to Munich’

MIA fom-ma hoam. we drive-1pl home ‘We go home.’ b. *Mia fon hoam. we drive home ‘We go home.’

(Weiß 2002: 9)

(10) a.

(Weiß 2002: 9)

(Helmut Weiß, p.c.)

(11) Fom-ma pro noch Minga? drive-1pl to Munich ‘Will we go to Munich?’

Therefore, it is plausible to assume that in these dialects, -ma developed into an additional instance of Agr-on-C (cf. Bayer 1984; Weiß 1998, 2002). This





Eric Fuß

claim is supported by the fact that in inversion contexts, -ma replaces the original agreement ending in bisyllabic verbs such as laffa ‘to run’, gengan ‘to go’, soucha(n) ‘to seek’ etc. (cf. Weiß 2002; Kollmer 1987): (12) a.

Mia laff-ma/*laff-a hoam. we ran-1pl/ran-1pl home ‘We are running home.’ b. Mia gem-ma/*geng-an hoam. we go-1pl/go-1pl home ‘We are going home.’

Crucially, no such replacement is possible when the finite verb shows up in sentence-final position: (13) wa-ma hoam laff-a/*laff-ma. because-1pl home go-1pl ‘because we are going home’

In other words, the dialects in question show a complementary distribution of the new suffix -ma and the old ending for 1pl, -a(n) (cf. Kollmer 1987: I, 357): -ma systematically appears on verbs in V2 clauses (main and embedded), cf. (12), whereas verbs in sentence-final position maintain the old ending -a(n), cf. (13). In a subset of these Lower Bavarian dialects,4 -ma has spread to frequently used (light) verbs such as ‘have’ and ‘do’ in clause-final position as well (Kollmer 1987: I, 357; Wiesinger 1989: 38; Weiß 2002: 9). Note that -ma must be analyzed as an agreement marker in the following examples, since enclitics cannot attach to clause-final verbs in Bavarian. (14) dass-ma (mia) koã geid ned hã-ma [instead of 1pl hã-n] that-1pl we no money not have-1pl ‘that we have no money’ (Kollmer 1987: I, 362) (15) we-ma (mia) des ned dou-ma. . . [instead of 1pl dou-n] if-1pl we that not do-1pl ‘if we don’t do that...’ (Kollmer 1987: I, 358)

The above discussion has shown that in Bavarian, new verbal agreement morphology developed from enclitic subject pronouns in inversion contexts. With respect to 2nd person forms, the available evidence suggests that the new 2nd person agreement markers developed first on verbs located in C0 and spread to other verbal positions in a subsequent development. The latter claim is supported by data from Lower Bavarian dialects, where the new agreement marker


for 1pl, -ma, is obligatory on C0 (i.e. on verbs and complementizers), but still impossible on most clause-final verbs.

. Towards an analysis This section presents an account of the historical development of Agr-on-C in Bavarian that links the rise of complementizer agreement and partial prodrop with the grammaticalization of new verbal agreement morphology. The analysis is framed in a realizational model of grammar where word building operations are distributed over several components of grammar (Distributed Morphology, Halle & Marantz 1993, 1994). It is assumed that the morphological component (called Morphological Structure, henceforth MS) operates postsyntactically, interpreting the output of the syntactic derivation. Accordingly, the structural design of the grammar looks like (16). (16) Lexicon (morphosyntactic/semantic features) Syntactic derivation Spell-out MS

LF

PF

In this model of grammar, the syntactic operations Merge and Move operate on bundles of morpho-syntactic features that constitute syntactic terminal nodes (i.e. heads). The syntactic terminal nodes are referred to as morphemes. At MS, the terminal nodes are realized by phonological exponents in a process called Vocabulary insertion. The idea that phonological content is added after syntax is also known as Late insertion. The information that links phonological exponents with morphosyntactic features (i.e. insertion contexts) is stored in individual Vocabulary items. For example, the English verbal inflection 3sg.pres.indic. /-z/ is associated with the following Vocabulary item (which can be read as an insertion rule): (17) [3, sg, pres.] ↔ /-z/

The insertion procedure requires that the feature specification of the Vocabulary item is nondistinct from the features of the insertion site (i.e. a certain





Eric Fuß

morpheme). Usually, this requirement is met by several items, which then enter into a competition where the item that realizes the greatest subset of features is chosen for insertion. In the example on hand, the availability of the Vocabulary item in (17) blocks the insertion of the less specified exponent /-∅/, which is found in all other contexts, representing the ‘elsewhere’ case (cf. Kiparsky 1973; Aronoff 1976; Anderson 1986 for the use of elsewhere principles in phonology/morphology). This implies that Vocabulary items may be underspecified for the feature complexes they realize (see below Section 3.3 for details). Before we turn to the diachronic analysis in Sections 3.2 and 3.3, the next section develops a new synchronic account of complementizer agreement in Germanic which makes use of the idea that agreement morphemes may be inserted post-syntactically (Marantz 1992; Halle & Marantz 1993), another key feature of Distributed Morphology. . A synchronic account of complementizer agreement in Germanic A number of non-standard varieties of Germanic is characterized by the presence of inflection in the C-domain, usually referred to as complementizer agreement (cf. Bayer 1984; Altmann 1984; Weiß 1998, 2002 on Bavarian; Bennis & Haegeman 1984; Haegeman 1992 for West Flemish; Hoekstra & Marácz 1989 on Frisian; Zwart 1993; Shlonsky 1994; Poletto 1999 for a number of Northern Italian dialects). The phenomenon of complementizer agreement is commonly attributed to a structural relation that holds between C and a lower inflectional head. This relation is often modeled in terms of syntactic head movement which provides C with inflectional features (e.g. Hoekstra & Marácz 1989: I-to-C; Zwart 1993: Agrs-to-C).5 Alternatively, it has been proposed that complementizer agreement heads its own projection in the C-domain (AgrCP) and is licensed in a specifier-head relation (Shlonsky 1994). Both assumptions can be shown to face a number of serious problems (cf. Carstens 2003 for some discussion). For example, an analysis in terms of head movement raises the question of how the verb can be inflected when the relevant inflectional head moves to C0 to license complementizer agreement. An analysis in terms of spechead agreement faces the problem that the relevant spec-head relation between the subject and AgrC0 is never overtly realized, either because the subject appears to the right of the complementizer or because there is simply no subject overtly present (as in the Bavarian data above). In addition, both analyses can be shown to make wrong empirical predictions. This will be pointed out when we come to the relevant data below.


Recently, Carstens (2003) has proposed an analysis of complementizer agreement in Germanic that makes use of the probe/goal mechanism developed in Chomsky (2000, 2001a, 2001b). Carstens assumes that the agreement morpheme on C0 (in her analysis: Fin0 ) is valued under closest c-command (i.e. via the operation Agree) by the interpretable phi-features of the subject which occupies SpecTP.6 Before this Agree operation takes place, the subject’s phi-set has already been accessed by the uninterpretable phi-features of T0 (in its base position SpecνP, prior to movement to SpecTP), case-licensing the subject and ensuring that the finite verb carries agreement morphology as well. Note that this instance of ‘multiple Agree’ represents a deviation from the assumption that a category with an inactive (i.e. deletion-marked) case feature should no longer be accessible to syntactic operations (Chomsky 2000, 2001a). That is, it should not be possible for the subject to enter into an additional probe-goal relation with the non-interpretable phi-features located in C0 after its case feature has been deletion-marked by T0 . Accordingly, Carstens slightly relaxes Chomsky’s original proposal and claims that a case-feature that has been marked for deletion remains syntactically active until the next strong phase is reached (cf. Pesetsky & Torrego 2001 for a similar assumption), which enables the subject to participate in multiple Agree relations (in the case on hand first with T0 and then with C0 ). However, it can be shown that this analysis does not account for a number of restrictions on complementizer agreement in a satisfactory way. The first set of problematic data has to do with the fact that the availability of complementizer agreement is subject to an adjacency requirement (cf. Haegeman 1992 for West Flemish, Carstens 2003; Ackema & Neeleman 2003 on Hellendoorn. Günther Grewendorf pointed out to me that similar facts can be observed in Bavarian). For example, in the East Netherlandic dialect Hellendoorn, the presence of an adjunct that intervenes between C0 and the subject blocks the availability of complementizer agreement. This restriction holds for embedded clauses as well as inversion contexts (Carstens 2003: 397f., citing Ackema & Neeleman 2001).7 (18) a.

dat/*datt-e op den wärmsten dag van’t joar wiej tegen that/that-1pl on the warmest day of-the year we against oonze wil ewärkt heb-t. our will worked have-1pl ‘that on the warmest day of the year we have worked against our will’





Eric Fuß

b. Volgens miej loop-t/*lop-e op den wärmsten dag according-to me walk-1pl/walk-1pl on the warmest day van’t joar ook wiej noar’t park. of-the year also we to-the park ‘According to me we are also walking to the park on the warmest day of the year.’

Note that these facts are completely unexpected if an analysis in terms of head movement is adopted (the intervening PP should not block head movement). Carstens attributes the absence of complementizer agreement in (18) to an intervention effect in the sense of Chomsky (2000, 2001a). More precisely, she claims that the adjoined adverbials in (18) bear an abstract Case feature that identifies the adverbial as a possible goal for the phi-set in C0 . As a consequence, the adverbial “disrupts closest c-command of the subject by C0 ” (p. 398), thereby blocking the valuing and realization of complementizer agreement. However, apart from the fact that it is quite unusual to assume that PP adverbials like op den wärmsten dag van’t joar carry a Case feature, this analysis leads to a serious problem with respect to the Agree relation that serves to value the phi-set of T0 . Recall that Carstens assumes that the phi-set of T0 initiates an Agree relation with the subject in SpecνP. Now, under her analysis, we should expect that adverbials that intervene between T0 and the base position of the subject should give rise to the same kind of intervention effects that are taken to block Agree between C0 and the subject in SpecTP. Of course, this is not the case. Therefore, we can conclude that the analysis of Carstens (2003) does not provide an adequate account of the Hellendoorn data, in contrast to the author’s claims. The fact that complementizer agreement is sensitive to an adjacency requirement suggests that this form of agreement is not a purely syntactic phenomenon, but at least partially determined by properties of PF (or MS) (see Ackema & Neeleman 2003 for a similar conclusion). This hypothesis is corroborated by data from Bavarian comparatives. The assumption that C0 carries its own set of non-interpretable phi-features (which initiate an Agree operation) should lead us to expect that the establishment of complementizer agreement is independent of the realization of verbal agreement. At least in Bavarian, however, this expectation is not borne out by the facts. Consider the following data from Bavarian, involving comparative clauses (Bayer 1984: 269):8 (19) a.

D’Resl is gresser [als wia-st du bist] the-Resl is taller than what-2sg you are ‘Resl is taller than you are.’


b. *D’Resl the-Resl c. D’Resl the-Resl

is is is is

gresser taller gresser taller

[als than [als than

wia-st du] what-2sg you wia du] what you

In comparatives, overt agreement on C leads to ungrammaticality if the finite verb is absent from the structure, cf. (19b). The sentence becomes acceptable when the complementizer bears no inflection, cf. (19c), an apparent violation of the generalization that complementizer agreement is obligatory in all contexts (for 2nd person, see above). This suggests that the possibility of complementizer agreement depends on the presence of a finite verb within the same clause. In other words, it seems that Agr-on-C is parasitic on the presence of another already valued agreement node, Agr-on-T. Crucially, these examples suggest that agreement between the complementizer and the subject cannot be implemented in terms of Agree between the set of phi-features on C0 (acting as a probe) and the lower subject (the goal). Otherwise one would expect examples such as (19b) to be grammatical. Moreover, comparatives such as (19b, c) provide further evidence for the above suggestion that complementizer agreement must operate post-syntactically. It is standardly assumed that comparatives are to be analyzed as the result of post-syntactic PF-operations that delete the inflected verb in the second clause, as shown in (20) (cf. Bresnan 1973; Lechner 2001). (20) D’Resl is gresser [als wia (*-st) du bist] the-Resl is taller than what-2sg you (are) ‘Resl is taller than you are.’

Then, however, we would not expect any interaction with complementizer agreement if insertion and valuing of agreement morphemes were to take place in the syntax, since the finite verb would be present throughout the whole syntactic derivation, being subject to deletion only after the structure has been transmitted to Morphological Structure (and, finally, PF). In other words, it would remain a mystery why post-syntactic deletion of the finite verb affects the realization of Agr-on-C in these contexts in the way it does. In contrast, this interaction comes out much more naturally in a Late insertion model where the insertion and valuing of Agr-morphemes on C0 takes place after syntax as well, in the sense that the post-spell out operations that bring about complementizer agreement may be sensitive to other post-spell out operations such as deletion of the finite verb in examples like (19b, c). We can therefore conclude that certain properties of complementizer agreement in Germanic cannot be





Eric Fuß

accounted for by a purely syntactic analysis in terms of Agree. Instead, the sensitivity to morpho-phonological properties such as linear adjacency or PFdeletion seems to indicate that the phenomenon of complementizer agreement is to be located in the post-syntactic components of grammar, that is, at MS or, more generally, the mapping to PF. Thus, I propose a hybrid model of agreement, where ‘canonical’ subject agreement is the result of agreement morphemes attached to T which are inserted into the syntactic derivations together with T and valued by the operation Agree.9 At MS, these syntactically valued/licensed morphemes are then subject to insertion of phonological exponents in the usual way (see above). (21) [CP ... [TP T+Agr ... [νP subject ... ]]] | ↑ Agree

In contrast, complementizer agreement is analyzed as resulting from a morphological operation, the post-syntactic insertion of a dissociated Agr-morpheme at the level of MS.10 Hence, I assume that the constituent structure of morphemes derived in the syntax can be modified by the post-syntactic insertion of (functional) morphemes which adjoin to syntactic terminal nodes (cf. Marantz 1992; Halle & Marantz 1993; Embick 1997). Following Embick (1997), these morphemes are called ‘dissociated’, since they are not present in the syntactic derivation and merely reflect properties expressed by structural configurations in the syntax proper. This process is illustrated by the following pair of phrase markers, where (22a) represents the output of the syntactic component and (22b) shows the structure resulting from the insertion of a dissociated Agr-morpheme to the C head, which is subsequently subject to Vocabulary insertion (note that the Agr-morpheme on T is already present and valued in the syntax).


(22) a.

b.

CP C’

spec C0

CP C’

spec C0

TP C0

T’

subj.

TP A subj.

T’

(inserted at MS)

T0

VP T0 V0

T0

VP T0

A T0

V0

A T0

Let’s now turn to the question of how the feature content of Agr-on-C can be identified/valued. The Bavarian comparatives in (19) above suggest that the licensing of Agr-on-C depends on the presence of Agr-on-T. To capture this intuition, I assume that the feature content of the dissociated Agr-morpheme on C is determined by the feature content of an Agr-morpheme that has already been valued in the syntax, that is, Agr-on-T. More generally, I propose that dissociated morphemes that express agreement with an argument (here: the subject) are merely copies of a relevant Agr-node that has been valued in a syntactic Agree relation (here: Agr-on-T). This mechanism ensures feature identity between these different types of Agr-morphemes (which both reflect the phi-feature content of the same argument). Thus, we can formulate the following condition on the insertion of dissociated Agr-morphemes.11 (23) The insertion of a dissociated Agr-morpheme is bound to the presence of an Agr-morpheme that has been valued in the syntax.

In other words, complementizer agreement is primarily a morphological process that is nevertheless sensitive to the syntactic environment. This account explains the restriction on complementizer agreement observed in Bavarian comparatives if we assume that the insertion of dissociated Agr-morphemes is a morphological operation that applies after the deletion of the inflected verb in comparatives.12 In Distributed Morphology, rules operating at MS are usually subject to locality constraints. More specifically, operations such as Impoverishment or Morphological Merger may target only structurally adjacent morphemes, where structural adjacency is defined in terms of head government (cf. Halle & Marantz 1993 and Note 18 below). With respect to the insertion of dissociated





Eric Fuß

Agr-morphemes, the relevant condition must capture the intuition that the relation between the syntactically valued Agr-morpheme and its late-inserted copy is sufficiently local. This can be achieved by the following condition on the insertion of dissociated Agr-morphemes (and the definition of structural adjacency in (25)).13 (24) Insertion of dissociated Agr-morphemes A dissociated Agr-morpheme can attach to a functional head X only if X is structurally adjacent to another functional head Y hosting an Agrmorpheme that has been valued in the syntax. (25) Structural adjacency A terminal node X and the closest terminal node Y c-commanded by X are structurally adjacent.

In other words, (25) guarantees that a head X is structurally adjacent to the head Y of its complement. Hence, dissociated Agr-on-C can only be inserted if C0 is structurally adjacent to T0 hosting a valued Agr-morpheme. Interestingly, it is now possible to attribute the adjacency effects observed in Hellendoorn to the locality condition (24), if we assume that (scrambled) XPs intervening between C0 and TP are not adjoined to TP but occupy the specifier of a projection (Haeberli 2002: AgrsP, Grewendorf 2004: TopP/FocP) that is only projected if necessary.14 In the structure (26), the (optionally projected) TopP hosting the PP op den wärmsten dag van’t joar disrupts structural adjacency between C0 and T0 , thereby blocking the insertion of a dissociated Agr-morpheme on C0 . As a consequence, complementizer agreement leads to ungrammaticality in contexts such as (26). (26) *[CP datt-e [TopP op den wärmsten dag van’t joar [TP wiej that-1pl on the warmest day of-the year we tegen oonze wil ewärkt heb-t.]]] against our will worked have-1pl ‘that on the warmest day of the year we have worked against our will’

Interestingly, not all elements that intervene between C0 and the subject block the realization of complementizer agreement. In Bavarian which exhibits a similar adjacency requirement as Hellendoorn (as shown in (27)), modal particles such as aber, halt, ja and clitic object pronouns may intervene between Agr-on-C and TP/the subject:15


(27) *obwoi-st gesdan du ins Kino ganga bist. although-2sg yesterday you to-the movies gone are ‘although you went to the movies yesterday’ (Günther Grewendorf, p.c.) (28) dass-st oaba du ibaroi dabei bis-st. that-2sg prt you everywhere with-it are ‘that you really are involved everywhere’ (29) wia-sd-n du gseng hoasd when-2sg-3sg.clit you seen have ‘when you saw him’

(Altmann 1984: 205)

(Pfalz 1918: 231)

Note that these examples do not represent counterexamples to the analysis proposed above if we assume that the structural positions of clitics and modal particles differ from that of XPs that are moved into a specifier position of a TopP or FocP intervening between C0 and TP. With respect to modal particles, it is commonly assumed that they are base generated as adjuncts (here: TP-adjuncts) (cf. e.g. Abraham 1995).16 Accordingly, they do not require the projection of a separate TopP or FocP and do not disrupt structural adjacency between C0 and TP. Concerning the placement of object clitics, I assume that their ultimate surface position is determined by late MS-processes such as Prosodic Inversion that apply at the mapping to PF (cf. Bonet 1991; Halpern 1992; Schütze 1994). Therefore, they reach their surface position after the insertion and valuation of dissociated Agr-morphemes has been completed. Again, no interaction between these two processes is expected. Summing up, the correct generalization seems to be that only XPs that have undergone syntactic movement to a topic or focus position between C0 and TP block the realization of complementizer agreement, whereas base generated adjuncts (i.e. modal particles) and object clitics that have undergone PF-movement do not disrupt the structural adjacency between C0 and T0 . . The diachrony of Agr-on-C: syntactic aspects This section argues that the reanalysis of pronominal elements as (verbal) agreement markers is only possible if it meets a number of preconditions imposed by the syntax. Here, I will focus on a selection of syntactic restrictions on the categorial reanalysis in question including linear adjacency, conditions imposed by Theta-theory, economy, and the syntactic licensing of agreement markers. The restrictions in question are based on the assumption that agreement morphemes do not head their own projection in the syntax. Rather, they are parasitic on other functional heads (Iatridou 1990; Speas 1991; Halle &





Eric Fuß

Marantz 1993; Mitchell 1994; Chomsky 1995, 2000; Halle 1997; Julien 2002). Following the analysis developed in the previous section, agreement morphemes can combine with core functional categories such as C, T, ν or D in one of the following ways. Either the agreement morpheme attaches to its functional host prior to the insertion of that host into the syntactic derivation (‘canonical agreement’), or it is inserted as a dissociated Agr-morpheme that attaches to its host post-syntactically. ‘Canonical’ subject-verb agreement is then the result of an agreement morpheme attached to T (henceforth Agron-T). That is, the following complex head consisting of an Agr-morpheme adjoined to T is merged with νP in the syntax: (30)

T T

A

By assumption, any of the core functional categories is in principle capable of hosting agreement morphemes. The actual presence of an agreement morpheme is dependent on whether the learner can detect the presence of an Agrmorpheme on a certain functional head on the basis of the evidence available to him/her. Under the assumption (see above) that words are built in the syntactic component (i.e. bundles of morpho-syntactic features are combined via Merge and Move) and later realized by the insertion of phonological exponents, an inflected verb can only be spelled out if it is combined with its inflectional affixes (e.g. T, Agr, Asp, Mood etc.) prior to Vocabulary insertion.17 This can be accomplished either by overt verb movement to a functional head or by late morphological operations such as Morphological Merger or Fusion that combine the in-situ verb with a higher functional head under (structural) adjacency (‘affix hopping’ as in the case of finite main verbs in English, cf. Halle and Marantz 1993, Bobaljik 1994, 1995, Embick and Noyer 1999).18 That is, the requirement that an inflectional affix is attached to a lexical host is treated as a morpho-phonological requirement which must be satisfied prior to PF. The latter considerations correlate with the perhaps most obvious syntactic restriction on the reanalysis of pronouns as verbal agreement affixes: a pronoun can only be re-interpreted as an inflectional affix if it is adjacent to the verb.19 (31) Adjacency requirement A pronoun can become an agreement affix attached to a host X only if it is string-adjacent to X.


The intuition captured by (31) can be put down in more formal terms by stating that a pronominal clitic can only be reanalyzed as an agreement morpheme on a functional head X if the verb moves to X in the overt syntax or if X and the verb are linearly adjacent and can combine via operations such as Morphological Merger at MS. Since this requirement concerns the process of word building (i.e. the creation of complex morpho-syntactic units either in the syntax or at MS), I refer to it as the Word building constraint: (32) Word building constraint The reanalysis of a pronoun as a (bound) verbal agreement marker on a functional head X requires that X combines with the verb prior to Vocabulary insertion.

In other words, one might say that the presence of the finite verb in a functional head X (as a result of overt movement) signals the learner that X is capable of hosting verbal agreement morphemes/features.20 In addition, the feature content of the newly created agreement morpheme must be licensed (or, ‘valued’ using the terminology of Chomsky 2000, 2001a) in a local structural configuration. If UG allows for two kinds of Agr-morphemes (i.e. ‘syntactic’ and ‘dissociated’ Agr-morphemes) which are subject to different licensing mechanisms (see above), then the reanalysis of pronominal elements is restricted by the following condition: (33) Identification of feature content The reanalysis of a pronoun is licit if the resulting agreement morpheme is licensed (i) in the syntax by a local Agree relation with a matching bundle of interpretable phi-features, or (ii) at MS as a dissociated morpheme that represents a copy of a syntactically licensed Agr-morpheme.

The reanalysis in question must presumably conform to general economy principles. In generative approaches to language acquisition and change, it is often assumed that the learner assigns a given input string the most simple or economic structure/derivation that is compatible with the evidence (cf. e.g. Clark & Roberts 1993; Roberts & Roussou 2003). The preference for simplicity may trigger a reanalysis of a given structure if the trigger for a more complex target grammar has been obscured by phonological erosion or independent change processes. In other words, if the evidence available to the learner is ambiguous, he/she will opt for the most economic structural option. This intuition is captured by the following principle.





Eric Fuß

(34) Least Effort Of the parametric choices compatible with the evidence available to the learner, he/she will acquire the most economical structural description for a given input string (in the sense of computational/derivational complexity).

Hence, the development of agreement markers from pronominal elements is expected to be only possible if the resulting structure is at least equally economical as a competing structural alternative where the clitic is perceived as a pronominal element. However, the impact of economy principles on language change raises a number of intricate questions that go well beyond the scope of this paper (for discussion cf. Lightfoot 1999; Roberts & Roussou 2003). Note for example that it is usually assumed (cf. e.g. Chomsky 1995: 348) that the reference sets that are compared for economy include only those derivational alternatives that start out from identical numerations. On this assumption, it is quite unlikely that the learner can decide on the structure of a given input string by comparing a derivation where a certain element is a pronoun with a derivation where it is an agreement marker. Therefore, I will not pay much attention to the role of economy principles in the remainder of this paper (see Fuß in prep. for some discussion). Finally, a rather obvious condition on the categorial reanalysis of pronouns is imposed by Theta-theory. Since a pronoun usually carries a thematic role, it must be ensured that the role in question can still be assigned when the pronoun is reanalyzed as an agreement marker and disappears from the set of arguments in a given clause: (35) Preservation of argument structure The reanalysis of a pronoun as an agreement marker must preserve the predicate’s argument structure.

In other words, a categorial reanalyis of a subject pronoun as an agreement marker can only take place if the relevant thematic role can be assigned to another element, either another overt DP (e.g. a former topic in a clitic left dislocation structure, cf. Givón 1976) or a covert pronominal element, that is, pro. This leads to some interesting questions concerning the interaction between the development of verbal agreement and the pro-drop property. More specifically, it seems that pro-drop may be either a precondition for or the result of the development of (new) verbal agreement marking. On these assumptions, the basic proposal for the rise of new verbal agreement markers in Bavarian is illustrated by the phrase markers in (36) and (37).


Thus, it is proposed that structures such as (36) were reanalyzed as (37), where the reanalysis of the former subject clitic as an agreement morpheme forced the learner to assume the presence of a referential pro in the subject position, giving rise to the limited pro-drop properties of present-day Bavarian (cf. Bayer 1984 on pro-drop in Bavarian; Weiß 2002 for a related proposal): CP

(36)

C’

Topic

TP

C+Vfin DPi

T’

Dclit. ti

(37)

CP C’

Topic C

TP T’

C+Vfin A proi ti

Let’s now address the question of how this change meets the set of restrictions introduced above. In Note 19, it was observed that the adjacency requirement is trivially fulfilled in Bavarian since subject enclitics are always adjacent to the verb in V2 contexts. Here, I will argue that the V2 property in fact provides the only pathway to new verbal agreement morphology in languages with C-oriented clitics and basic SOV order like Bavarian if the reanalysis in question has to satisfy the Word building constraint stated in (32).21 Note that in a SOV+V2 grammar, a C-oriented clitic cannot be reanalyzed as an agreement morpheme on C which combines via Morphological Merger with a verb that occupies a lower head position, since these heads are usually not adjacent at MS (e.g. the subject regularly intervenes between C and T etc.). Hence, overt movement of the finite verb to C0 is a necessary precondition for a reanalysis that leads to the existence of verbal agreement features on C0 .22 The V2 property arguably plays also a role in extending the domain of the new endings to other verbal positions. Recall that the change in question did





Eric Fuß

not immediately lead to a complete replacement of the old ending (i.e. the phonological exponent of Agr-on-T) by the new one. Rather, it proceeded via a stage where the new agreement ending was confined to C (as the phonological realization of Agr-on-C), presumably due to the fact that embedded clauses (without V-to-C) still provided enough evidence for the old ending. Thus, initially, the exponent of Agr-on-C replaced the old agreement ending only in V2 contexts.23 The new inflection may then gradually spread to other verbal positions when the learner reanalyzes the exponent of Agr-on-C as the ‘canonical’ agreement ending (i.e. as ‘Agr-on-T’ instead of ‘Agr-on-C’). The latter reanalysis depends on the V2 property as well: only cases where the original agreement ending is replaced by the exponent of Agr-on-C can feed a possible misinterpretation of the latter as Agr-on-T.24 By assumption, the reanalysis leading to (37) satisfies the restriction imposed by Theta-theory (Preservation of argument structure) by assigning the subject/agent theta-role to an empty pronominal element that is inserted into SpecνP and subsequently undergoes movement to SpecTP. As already noted above, the rise of verbal agreement from former pronominal elements may in this way give rise to new pro-drop properties formerly absent in the target grammar. In addition, the rise of referential pro-drop was presumably facilitated by the fact that Bavarian is not a strict non-pro-drop language. Instead, it generally licenses expletive pro, for example in impersonal passives (in contrast to e.g. English, but similar to Standard German, cf. Grewendorf 1989): (38) Heit wird pro geoabeid. today is-pass worked

(Günther Grewendorf, p.c.)

So far, we have not addressed the question of how the newly created agreement morpheme on C0 is syntactically licensed, that is, how the reanalysis in question can satisfy condition (33). In line with the analysis of complementizer agreement presented in the previous section, I assume that the subject clitic was reanalyzed as a dissociated Agr-morpheme that is post-syntactically adjoined to C0 . This reanalysis fulfills (33) since the new agreement morpheme is licensed under structural adjacency with Agr-on-T which has been valued by an Agree relation established in the syntax. At this point, one might speculate that the reanalysis of clitics as dissociated agreement morphemes constitutes a necessary intermediate stage prior to the development of ‘canonical’ Agrmorphemes (such as Agr-on-T) that are part of the syntax. Initial support for this hypothesis comes from the diachronic facts observed in Bavarian, where the new agreement endings started on C as (the exponents of) dissociated morphemes, replacing (the exponents of) Agr-on-T in a subsequent development.


In addition, it can be argued that the difference between clitics and dissociated morphemes is rather subtle and therefore easily missed during language acquisition. Note that dissociated Agr-morphemes share basic properties with clitics such as post-syntactic positioning (for post-syntactic accounts of clitic placement cf. Bonet 1991; Halpern 1992; Schütze 1994 among others). In Bavarian, the clitic-like behavior of Agr-on-C is also evident in cases where the agreement morphology does not attach to an element in C0 , but rather to an XP located in SpecCP. The following examples illustrate that in the absence of a filled C-head, the inflection can attach to any element that occurs in the left periphery of the clause such as DPs (39a), adjectives (39b), or adverbs (39c): (39) a.

Du soll-st song [CP [an wäichan Schuah]-st [IP du you should-2sg say the which-one shoe-2sg you wui-st]]]. want-2sg ‘You should say which one of the shoes you want.’ wurscht. b. [CP [Wia oit]-ts [IP ihr/es sei-ts]] is mir how old-2pl you are-2pl is for-me not-important ‘How old you are makes no difference to me.’ c. [CP [Wia schnäi]-ts [IP ihr/es fahr-ts]]! how fast-2pl you drive-2pl ‘How fast you drive!’ (Bayer 1984: 235)

Note that this behavior of Agr-on-C in Bavarian contrasts with the properties of well-behaved inflectional affixes which usually select for a unique host/syntactic category they attach to. The systematic differences between dissociated and ‘syntactic’ Agr-morphemes (adjacency effects, ‘weak’ morphological selection, interaction with PF-processes such as deletion or clitic placement) can be used as diagnostic criteria for telling apart the two types of agreement. The similarities between clitics and dissociated morphemes presumably abet a reanalysis of the former as the latter since they blur the boundary between these elements. Thus, it seems that at least in Bavarian, the development of new agreement endings proceeded via an initial reanalysis of subject clitics as dissociated Agr-morphemes. Whether this change represents a necessary stage on the diachronic pathway towards canonical argument-predicate agreement across languages is a question I leave open for future research. .. Interim summary This section has argued that there is a diachronic link between complementizer agreement, pro-drop and the development of new verbal agreement markers in





Eric Fuß

Bavarian. More precisely, it has been shown that the development of new verbal agreement markers from clitic pronouns could only take place in inversion contexts, forcing the learner to assume that the subject position is occupied by referential pro. In addition, we have seen that the change in question was shaped by a number of (syntactic) conditions that had to be satisfied for the grammaticalization of new agreement markers to take place. The diachronic analysis proposed in this section is based on a synchronic account of complementizer agreement in terms of a dissociated Agr-morpheme that is inserted post-syntactically and represents the copy of an Agr-morpheme that has been valued in the syntax (Agr-on-T). On these assumptions, the diachronic development in Bavarian has been shown to proceed via a stage where the learner assumes the existence of a dissociated Agr-morpheme which is initially confined to C (Agr-on-C). In a subsequent change, the phonological exponent of Agr-on-C spreads to other verbal positions. Present-day Bavarian still shows a residue of these diachronic processes, namely a dissociated Agr-morpheme on C, giving rise to the phenomenon of complementizer agreement. Thus, prodrop and complementizer agreement are reflexes of the historical development of new verbal agreement morphology in the history of Bavarian which had to proceed via C0 in V2/inversion contexts. However, although we have managed to establish a diachronic connection between complementizer agreement and pro-drop in Bavarian, the investigation of syntactic aspects has not provided an answer to the question of why complementizer agreement and pro-drop are restricted to 2nd person contexts in present-day Bavarian. In the next section, it is demonstrated that this restriction is a result of morphological conditions that shaped the development of new verbal agreement morphology in the history of Bavarian. . Morphological aspects This section aims at providing a diachronic explanation for the person/number restrictions observed for complementizer agreement and pro-drop in presentday Bavarian. The analysis presented here makes crucial use of the theoretical notions of morphological blocking and underspecification. The mechanism of morphological blocking is based on the intuition that the availability of a more specified morphological form blocks the use/insertion of a less specified form (for discussion of blocking effects cf. Kiparsky 1973, 1982; Aronoff 1976; Anderson 1986, 1992; Halle & Marantz 1993; Kroch 1994; Sauerland 1996; Halle 1997). Thus, if more than one Vocabulary item is compatible with the feature content of a certain terminal node, the most specified Vocabulary item


must be inserted, that is, the item that realizes the greatest subset of the features contained in the node (cf. Halle’s 1997 Subset Principle). This trait of the insertion procedure implies that phonological exponents may be underspecified with respect to the morphosyntactic features contained in the morpheme they realize. That is, for insertion to take place, a phonological exponent need not be specified for all features present in a terminal node. On these assumptions, it is shown that the person/number restrictions follow from the workings of morphological blocking effects that confined the historical development of new verbal agreement morphology (via the grammaticalization of Agr-on-C) to underspecified slots of the agreement paradigm. .. The Blocking Principle A closer look at morphological aspects of the rise of new agreement suffixes in Bavarian reveals that the individual diachronic developments share an interesting characteristic: the grammaticalization of 2pl -ts and 1pl -ma resolved existing homophony in the verbal agreement paradigm. The grammaticalization of 2pl -ts (< clit. -(¯e)s) began in the 13th century (in Northern and Middle Bavarian, cf. Wiesinger 1989: 72f.). Prior to this development, the verbal agreement paradigm exhibited homophonous forms for 3sg and 2pl. Thus, the development of a new distinctive ending for 2pl can be said to have repaired this ‘defect’ of the paradigm. The relevant facts are shown in Table 5 below (the 3sg, 2pl forms are printed in boldface). Interestingly, it can be shown that a similar situation existed in those varieties that developed a new agreement ending for 1pl, -ma. Due to phonological reduction, final -t was lost in the 3pl, leading to homophony of 3pl and 1pl forms in most Bavarian dialects in the 18th century. In some dialects, this form of syncretism was resolved by the development of 1pl -ma (from the relevant subject clitic) as a new agreement ending (initially confined to C, see above), cf. Table 6. Table 5. Verbal agreement paradigms (pres. indic.), 13th century Bavarian

1sg 2sg 3sg 1pl 2pl 3pl

Old paradigm

New paradigm

-∅ -st -t -an -t -ant

-∅ -st -t -an -ts -ant





Eric Fuß

Table 6. Verbal agreement paradigms (pres. indic.), late 18th century Bavarian


Old paradigm

New paradigm

-∅ -st -t -an -ts -an

-∅ -st -t -ma -ts -an

We can thus conclude that both changes apparently repaired a previously defective agreement paradigm. Interestingly, it seems that this observation is a general characteristic of the grammaticalization of agreement markers across languages. For example, Gerlach (2002) shows that the development of new prefixal agreement markers in Non-standard French and the Northern Italian dialect Piattino affects only those slots of the agreement paradigm where the existing verbal inflection is not distinctive. Let’s therefore assume that (40) represents a valid descriptive generalization on the grammaticalization of new verbal agreement morphology (see Fuß in prep. for more examples and discussion):25 (40) New verbal agreement morphology arises historically only for those slots of the agreement paradigm where the existing verbal inflection is nondistinctive.

In other words, it seems that the reanalysis of clitics as agreement markers is triggered if the change leads to the elimination of syncretism in a defective agreement paradigm. Note that (40) is merely a description, but no explanation. Therefore, it would be desirable if (40) could be shown to follow from deeper principles of grammar. In what follows, it is argued that the generalization in (40) is the outcome of blocking effects that operate during language acquisition and block the acquisition of a less specified form if a more specific form is attested in the Primary Linguistic Data.26 In a Late insertion model such as Distributed Morphology, this general idea can be formalized as in (41): (41) Blocking Principle If several appropriate PF-realizations of a given morpheme are attested in the Primary Linguistic Data, the form matching the greatest subset of the morpho-syntactic features included in the morpheme must be chosen for storage in the lexicon.


The Blocking Principle is to be understood as a cognitive economy principle that applies during language acquisition and guarantees an optimal and non-redundant lexicon (and paradigm) structure. Note that this kind of morphological (or lexical) economy differs significantly from syntactic/structural economy principles that are part of the computational system and shape the course of syntactic derivations (cf. e.g. Chomsky 1995; Collins 1997; Kitahara 1997). The latter usually exhibit a ‘least effort’ character, in the sense that they favor less complex, that is, less marked syntactic derivations/representations (cf. Clark & Roberts 1993; Roberts & Roussou 2003 on the role of ‘least effort’ principles in language change). In contrast, the Blocking Principle prefers more specific, that is, more marked lexical entries over less marked ones. Still, this can be seen as a form of economy as well, as noted above. Coming back to the historical developments observed in Bavarian, it can be shown that the new agreement suffixes 2pl -ts, 1pl -ma satisfy the Blocking Principle due to the fact that they are more specified than their respective predecessors. Turning first to the change that took place in 13th century Bavarian, a closer look at Table 5 shows that the formative /-t/ occurs in 3sg and 2pl contexts, that is, the relevant Vocabulary item must be underspecified for [person] as well as [number]. In other words, it represents the elsewhere case that is inserted as the default agreement ending: (42) elsewhere ↔ /-t/

Hence, the introduction of 2pl /-ts/ was licensed by the Blocking Principle since the new form is specified for [person] and [number], resolving the existing homophony between 3sg and 2pl:27 (43) [2, pl] ↔ /-ts/

Similar facts can be observed in 18th century Bavarian. There are only two different plural forms in the agreement paradigm: /-ts/ is inserted in the context [2, pl], whereas /-an/ realizes the feature combinations [1, pl] and [3, pl]. Thus, the lexical entry for /-an/ must be underspecified for the feature [person]. In other words, /-an/ is simply the elsewhere case among the plural forms, cf. the following insertion rules: (44) [2, pl] ↔ /-ts/ [pl] ↔ /-an/

Again, the potential ‘new’ form for 1pl (-ma) is more specified than the existing Vocabulary item /-an/, since it is in addition specified for [person]. This state of





Eric Fuß

affairs facilitates the grammaticalization process in question, leading to a fully distinctive set of plural agreement markers:28 (45) [1, pl] ↔ /-ma/ [2, pl] ↔ /-ts/ [pl] ↔ /-an/

It now becomes clear that the person/number restrictions on complementizer agreement (and pro-drop) observed in present-day Bavarian reflect the workings of the Blocking Principle in the grammaticalization of new verbal agreement morphology in the history of Bavarian. As noted above, the Blocking Principle licensed the development of new agreement formatives only for those slots of the agreement paradigm where the existing phonological exponents were less specified than their newly emerging competitors, that is, in 2nd person and 1pl contexts. At this point, some clarifying remarks on the nature and the workings of the Blocking Principle (henceforth BP) are in order. It is important to understand that the BP restricts, but does not trigger grammaticalization processes (otherwise one would perhaps expect that all dialects developed 1pl -ma). The BP merely ensures that the development of new inflectional formatives can affect only weak/underspecified slots of the paradigm, replacing Vocabulary items that are not distinctive. This complies with the view that language change is contingent and non-deterministic (cf. Lightfoot 1999).29 On this view, properties of UG draw up the boundaries of possible changes, but crucially, they do not trigger changes. For example, it is quite obvious that languages with a defective agreement paradigm may exist happily over centuries (alone the fact that such a grammar can be acquired shows that it is line with UG and that change is not necessary). In fact, we can observe that in most languages, the effects of phonological reduction win out over the processes that create new inflectional morphology (the Germanic languages, in particular English, are a typical example).30 Thus, the BP is to be seen as a morphological condition on the development of new inflectional morphology: in syntactic contexts where the categorial reanalyses is possible (in the case on hand, V2/inversion), the grammaticalization of new forms may affect only Vocabulary items that are less specified than their competitors. Furthermore, recall that the old forms are not replaced instantly. The change in question introduces a more specified inflectional formative that replaces the older, less specified first in a certain context before it gradually gains a wider distribution, eventually driving the old form out of the grammar (cf. Kroch 1994 on this form of morpho-syntactic competition).


.. The diachrony of 2sg -st As noted above, it is commonly assumed that the form -st resulted from a reanalysis of -s+thu sequences, where the onset of the pronoun (or the enclitic -t(u)) was mistakenly analyzed as part of the verbal agreement ending (in analogy with the 2sg of the preterite-presents). This process could only take place in inversion contexts (i.e. V2/V1) where the finite verb preceded the subject clitic, which is in line with the syntactic analysis presented in Section 3.2 above. However, the development of 2sg -st represents a problem for an account in terms of the BP. Consider the forms listed in Table 7 below.31 A look at Table 7 suggests that the change from 2sg -s to -st apparently did not involve the creation of a Vocabulary item that is more specific than its predecessor. Both items seem to realize the same set of morpho-syntactic features, cf. (46) [2,sg, pres.] ↔ /-s/ [2,sg, pres] ↔ /-st/

Thus, it appears that the creation of the new ending -st conflicts with the BP, since it apparently does not lead to a more specified form. In what follows, however, I will argue that the change in question proceeded in accordance with the BP, since it helped to create an inflectional formative for 2sg that was additionally specified for verbal mood, in contrast to its predecessor. Thus, the feature specifications of the relevant Vocabulary items are actually those in (47). (47) [2,sg, pres.] ↔ /-s/ [2,sg, pres., indic.] ↔ /-st/

In early OHG, the 2sg endings of many verbs were identical in the present indicative and the present subjunctive, whereas verbal mood was clearly distinguished in other person/number combinations (apart from 2pl). This is Table 7. Agreement paradigms (pres. indic.) for nëmen ‘take’, early OHG


Old paradigm

New paradigm

nim-u nim-is nim-it nëm-emês (-êm, -ên) nëm-êt nëm-ant

nim-u nim-ist nim-it nëm-emês (-êm, -ên) nëm-êt nëm-ant





Eric Fuß

Table 8. Conjugation of salbôn ‘anoint’ (class 2, present tense), early OHG


Present indicative

Present subjunctive

salbôm salbôs salbôt salbômês salbôt salbônt

salbo salbôs salbo salbôm salbôt salbôn

Table 9. Conjugation of habên ‘have’ (class 3, present tense), early OHG


Present indicative

Present subjunctive

habêm habês habêt habêmês habêt habênt

habe habês habe habêm habêt habên

illustrated is Tables 8 and 9 for the weak verbs salbôn ‘anoint’ (conjugation class 2) and the very frequent habên ‘have’ (conjugation class 3), which show the characteristic inflections of their respective verb classes.32 A closer look at the tables above reveals that identical verb forms show up in the slots for 2sg present indicative and 2sg present subjunctive. We can therefore conclude that the forms for 2sg are underspecified with respect to verbal mood. Interestingly, the development of the new formative -st began in the present indicative (cf. Brinkmann 1931). This suggests that the development in question was licensed by the fact that the new ending was unambiguously specified for verbal mood (i.e. indicative) in contrast to the earlier formative -s. Thus, despite appearances, the development of 2sg -st does not represent a counterexample to the BP. Instead, it can be shown to provide further support for the analysis proposed in this paper if a larger portion of the inflectional paradigm (i.e. the distinction of verbal mood) is taken into account.

. Conclusions In this paper, it has been claimed that the puzzling person/number restrictions on pro-drop and complementizer agreement found in present-day Bavarian can only be fully understood if diachronic data is taken into consideration.


More specifically, I have argued that the relevant synchronic properties are reflexes of a conspiracy of syntactic and morphological factors that guided the development of new verbal agreement morphology (for 2nd person and 1pl forms) in the history of Bavarian. With respect to syntactic factors involved in this change, it has been shown that new verbal agreement morphology could only develop in inversion contexts, where enclitic pronouns were reanalyzed as (dissociated) agreement morphemes on C. This change forced the learner to assume that the subject position is occupied by pro, giving rise to partial pro-drop which is restricted to the contexts where the reanalysis of clitic pronouns took place (i.e. 2nd person, 1pl). After this change, new doubling structures could emerge where a (stressed) full pronoun is inserted instead of pro (similar to e.g. Italian). In addition, the development of Agr-on-C led to the phenomenon of complementizer agreement which is a characteristic of present-day Bavarian. The diachronic analysis proposed in this paper is based on a new approach to complementizer agreement in Germanic where the inflection found in the C-domain is analyzed as resulting from the post-syntactic insertion of a dissociated Agr-morpheme which adjoins to C0 and is licensed in a local structural configuration with an Agr-morpheme that has been valued in the syntax (Agr-on-T). Accordingly, agreement phenomena may result either from Agrmorphemes that are present and valued in the syntax or from late-inserted dissociated Agr-morphemes. It has been demonstrated that an analysis in terms of a late-inserted Agr-morpheme is superior to purely syntactic approaches since it accounts for a number of restrictions on complementizer agreement (adjacency effects, necessary presence of an inflected verb) that have a ‘morpho-phonological flavor’ and are left unexplained by previous (syntactic) analyses. On these assumptions, I have speculated that the development of new agreement markers generally involves an initial step where clitics are reanalyzed as dissociated agreement morphemes before they develop further into ‘canonical’ Agr-morphemes (such as Agr-on-T) that are part of the syntax. This hypothesis has been shown to be supported by the diachronic facts observed in Bavarian where the new agreement endings were first confined to C0 (as dissociated morphemes) and spread later to other positions, eventually replacing the existing verbal agreement morphology (i.e. the exponents of Agr-on-T). In the second part of the analysis, the person/number restrictions on prodrop and complementizer agreement are attributed to morphological factors. Here, it has been shown that the grammaticalization of 2pl -ts and 1pl -ma affected only defective/underspecified slots of the verbal agreement paradigm.





Eric Fuß

This observation is accounted for by the assumption that the acquisition of inflectional morphology is guided by blocking effects which prefer ‘new’ verbal agreement morphology to be more specific than existing morphology. The relevant theoretical mechanism, the so-called Blocking Principle, is considered a cognitive economy principle that applies during language acquisition and guarantees an optimal and non-redundant lexicon (and paradigm) structure. Interestingly, we have seen that the Blocking Principle accounts not only for the cases 2pl -ts/1pl -ma, but also for the much earlier (and at first sight problematic) development of 2sg -st. Concerning the latter development, it has been demonstrated that the change led to a 2sg present form that was unambiguously specified for verbal mood (i.e. indicative), in contrast to its predecessor -s, in line with the Blocking Principle. At this point, I want to add another piece of evidence indicating that the ideas put forward in this paper (i.e. the Blocking Principle) are on the right track. Across the world’s languages, we can observe that verbal agreement markers for 1st and 2nd person subjects are much more common than for 3rd person subjects (cf. Bybee 1985; Ariel 2000; Cysouw 2001). For example, Bybee (1985) shows that 54% of the languages (in her sample) which manifest agreement do not mark third person on the verb. Cases in point include Basque (Arregi 1999), a number of Mongolian languages (Poppe 1960; Comrie 1980), Turkish (no verbal agreement for 3sg, Kornfilt 1990), and many native languages of North America (Mithun 1991). Similar person restrictions can also be observed in languages that are currently developing new agreement markers from clitic pronouns such as Non-standard French and Northern Italian dialects (Gerlach 2002). Note that related constraints are reflected by the historical developments in Bavarian discussed above. In other words, it appears that 1st and 2nd person elements play a pioneering role in the development of verbal agreement marking across languages. The special role of 1st and 2nd person in grammaticalization processes has inspired numerous functionalist explanations which mostly rely on the fact that speaker and hearer are the most salient participants in a speech event (cf. Mithun 1991; Ariel 2000), that is, they exhibit a high degree of ‘givenness’, ‘discourse accessibility’ or ‘discourse prominence’ etc. Interestingly, these facts can be explained in purely formal terms if we combine the Blocking Principle with the widely held assumption that ‘3rd person’ actually constitutes no separate person feature at all. Instead, ‘3rd person’ is analyzed as the result of the absence of (positive values for) the features 1st and 2nd person (cf. Benveniste 1972; Bayer 1984; Halle 1997; Grimshaw 1997; Poletto 1999; Ariel 2000; Cysouw 2001; Harley & Ritter 2002). The fact that


cross-linguistically, 3rd person agreement forms arise later (if at all) than forms for 1st and 2nd person can then be attributed to the workings of the Blocking Principle in the following way: if UG requires that newly grammaticalized markers realize a greater subset of agreement features than existing markers, the development of 3rd person forms is considerably hindered, due to the inherent underspecification of ‘3rd person’ forms with respect to the feature [person]. As a consequence, even in a grammar that previously lacked agreement markers completely, a new 3rd person marker can only develop if it is specified for some other inflectional feature like [gender], [number] etc. We can therefore conclude that the workings of the Blocking Principle ensure that the grammaticalization of new 3rd person forms is less likely to be triggered than the development of forms specified for a separate [person] feature, that is, 1st and 2nd person markers. This explains the typological tendency in question, lending further support to the proposal that the acquisition of inflectional morphology is guided by blocking effects that favor more specified over less specified forms.

Notes * For helpful comments and discussion I thank Werner Abraham, Patrick Brandt, Günther Grewendorf, Shin-Sook Kim, Cécile Meier, Helmut Weiß, and in particular Carola Trips. Various portions of this material were presented to audiences at the Universities of Frankfurt and Stuttgart, GLOW 26 (Lund) and the annual GGS meeting in Cologne (2003). Funding for this research was provided by the DFG project #559/6-1, Die Rolle funktionaler Kategorien in Grammatikalisierungs- und Sprachwandelprozessen. . In a number of dialects, the 1pl clitic -ma has developed similar properties as the 2nd person forms, cf. Section 2.3. . The early OHG manuscripts written in the monastery of Fulda show this change in the process of its development, cf. the Hildebrandslied (preserved in an early 9th century copy of the original text dating from the late 8th century), the Basel Recipes (around 800), or the Tatian (translated around 830–840. This translation was then copied in the second half of the 9th century); see Moulton (1944), Sievers (1961), Sommer (1994) for details. . Note that verbs such as ‘fear’ may optionally take V2 complement clauses in German. Therefore the verbs nimist ‘take-2sg’ and arnost ‘earn-2sg’ can be taken to occupy the C-head of an embedded V2 clause. . These dialects are spoken in the Bavarian Forest, in an area the boundaries of which are (roughly) marked by Cham in the west, Lam in the east, Furth i. W. in the north and Kötzting in the south, cf. Kollmer 1987, I. Similar phenomena can be observed in a small number of Carinthian dialects (cf. Wiesinger 1989):





Eric Fuß

(i)

wos m#r wöl-m#r what we want-1pl

(Wiesinger 1989: 38)

. See den Besten (1982) for an early account of complementizer agreement in terms of a rule Move Tense and Bennis and Haegeman (1984) on West Flemish data. In an early analysis of complementizer agreement in Bavarian, Bayer (1984) develops an account that is based on the idea that in V2 languages, there is an abstract agreement relation between Comp, V/Infl and the subject leading to co-indexation of all three elements. In the case of 2nd person subjects in Bavarian, this form of agreement is overtly realized due to a linking rule that copies the phi-features located in Infl to Comp. The overt manifestation of agreement serves to identify the referential content of the subject DP, thereby licensing referential pro in the subject position. . Note that this mechanism is somewhat reminiscent of the idea of Bayer (1984) that there exists an abstract agreement relation between between Comp, V/Infl and the subject. . Note that in Hellendoorn, the morphological realization of complementizer agreement differs from canonical verbal agreement, similar to the Lower Bavarian data discussed above. This trait is a characteristic of many East Netherlandic dialects and is also found in Brabants cf. Zwart (1993). . Note that similar facts hold for 1pl in the Lower Bavarian dialects discussed above, cf. (i)

De san g’scheider [(als) wia-ma mir san] they are more-intelligent than what-1pl we are ‘They are more intelligent than we are.’

(ii) *De san g’scheider [(als) wia-ma mir] they are more-intelligent than what-1pl we (iii) De san g’scheider [(als) wia mir] they are more-intelligent than what we

(Bayer 1984: 271)

This can be taken as further evidence that the formative 1pl -ma has evolved into an agreement morpheme in these varieties of Bavarian. . That is, it is assumed that agreement morphemes do not head their own projection in the syntax. Instead, they are parasitic on other functional heads (see Section 3.2 below for some discussion). Following Chomsky (2000, 2001a), an agreement morpheme with noninterpretable phi-features (the probe) intiates an Agree operation, looking for an element with matching interpretable phi-features (the goal). The search space of the Agree operation is confined to elements dominated by the sister node of the probe, with locality reduced to closest c-command. . Within DM, this mechanism is commonly used to account for case and agreement phenomena. For example, Marantz (1992), Halle and Marantz (1993), Halle (1997) analyze subject-verb agreement in terms of the post-syntactic adjunction of an [Agr] morpheme to T. However, I hesitate to adopt this mechanism for all kinds of agreement phenomena for the following reasons. First, an analysis of canonical cases of agreement in terms of the insertion of dissociated morphemes requires that MS has powerful syntax-like mechanisms at its disposal, for example in order to detect the correct agreement controller, to value Agr-


morphemes etc. That is, this analysis seems to establish a syntax after the real syntax, which is conceptually not very attractive. Second, it is well-known that agreement phenomena such as long-distance agreement are sensitive to intricate syntactic restrictions such as intervention effects created by topicalized constituents, where it is rather doubtful that these restrictions can be handled by morphological processes alone (cf. e.g. Polinsky & Potsdam 2001 on Tsez, Bruening 2001 on Passamaquoddy, Bobaljik, & Wurmbrand 2002 on Itelmen). . The idea that complementizer agreement is parasitic on verbal agreement is further supported by the observation that across Germanic, there are no languages with complementizer agreement but without verbal agreement, while there are many languages that exhibit verbal agreement in the absence of complementizer agreement (Hoekstra & Smits 1999). Thus, it seems that cross-linguistically, the availability of complementizer agreement is dependent on the overt realization of verbal agreement morphology. . At first sight, it seems that this analysis is contradicted by those varieties (like Hellendoorn, or 1pl in Lower Bavarian) where the shape of Agr-on-C differs from that of Agr-on-T. However, these phonological differences can be explained as the result of conditioned allomorphy that is sensitive to the syntactic environment, that is, the adjunction site of the Agr-morpheme (C or T). Presumably, this form of allomorphy is introduced by the historical reanalysis of a subject clitic as a dissociated Agr-morpheme where the phonological exponent of latter will in most cases differ from the existing verbal morphology due to its lexical source (the subject clitic), see below. . A similar notion of ‘structural adjacency’ is proposed in Zeller (2001: 36). However, in contrast to the analysis proposed here, Zeller assumes that ‘structural adjacency’ constitutes a local domain between two heads that is only syntactically relevant. . Cf. Rizzi (1997) for a similar approach to the presence of TopP/FocP in the left periphery. . Note that I adapted the examples taken from Altmann and Pfalz to the spelling conventions used in this paper. . At first sight, the shape of many modal particles seems to suggest an analysis as X0 elements. However, the fact that these elements do not block V-to-C movement in the Germanic V2 languages shows that they cannot be analyzed as syntactic heads. In addition, similar locality problems arise for XP-movement into the C-domain if they are analyzed as specifiers (cf. van Gelderen 2001 for discussion and further diachronic arguments against an analysis as specifiers from the history of English). These facts can be taken as further support for the idea that modal particles are base generated in adjoined positions. . Thus, in contrast to proposals by Chomsky (1993), the option that the inflectional features of verbs are checked via covert operations is not available. . Marantz (1988: 261) defines the operation Morphological Merger as follows: (i)

“At any level of syntactic analysis (d-structure, s-structure, phonological structure), a relation between X and Y may be replaced by (expressed by) the affixation of the lexical head of X to the lexical head of Y.”

It is generally assumed that (structural) adjacency at MS is a necessary condition for Merger to take place (cf. Bobaljik 1994). Thus, in present-day English, T is affixed to VP-internal V at MS if nothing intervenes between these head positions; the presence of negation triggers





Eric Fuß

do-support to pick up the inflectional morphemes in T (cf. Chomsky 1957; Halle & Marantz 1993). Note that this analysis requires that VP-adjoined material such as adverbs does not count for the evaluation of adjacency at MS (cf. Bobaljik 1994, 1995, 2002 for a concrete proposal). . Here, the question arises of whether adjacency must hold without exceptions for a reanalysis to be possible or whether it is sufficient when adjacency is given in the majority of cases. Note that this does not play a role in the Bavarian data on hand, since adjacency between a filled C0 head and the subject clitic is obligatory. However, a related problem might arise in cases where C0 is empty and the clitic/agreement marker attaches to other elements in the C-domain (see below). . Presumably, (32) has to be reformulated in more general terms to capture cases where the agreement marker does not attach to the verb, but rather to other (free) inflectional morphemes such as tense. Relevant examples come from Australian languages such as Wambaya (showing a clitic Agrs/Agro/T marker, Nordlinger 1995), the Mayan language Jacaltec (Craig 1977) where Agrs attaches to a clause-initial completive marker and Finnish where Agrs may attach to the negation. . Thus, the rise of new agreement formatives (in C) appears to be intimately linked to the V2 property in languages such as Bavarian. This corresponds to the observation that crosslinguistically, complementizer agreement/Agr-on-C is a marked phenomenon which goes hand in hand with another marked syntactic property, V2. . In the case on hand, therefore, the presence of the finite verb in C0 can be said to signal that C0 is capable of hosting an Agr-morpheme. . This raises the question of why Agr-on-C cannot be realized in addition to Agr-onT, that is, what rules out verbs with ‘double agreement’ (i.e. *V+AgrT+AgrC). Note that this question relates to all Germanic varieties with complementizer agreement (cf. e.g. Carstens 2003). Tentatively, I assume that the impossibility of doubly inflected finite verbs results from a morphological condition ensuring that only the hierarchically highest Agrmorpheme is spelled out in a given head complex (cf. Kinyalolo 1991; Carstens 2003 for similar proposals). . Note that it is misleading to say that “Agr-on-C is misinterpreted as Agr-on-T”. Rather, it is the phonological exponent of Agr-on-C that is misinterpreted as the phonological exponent of Agr-on-T. Note that the feature content of the two Agr-morphemes is identical. Thus, it would not make much of a difference if Agr-on-C were reanalyzed as Agr-onT. This is in line with a general theoretical approach which regards grammaticalization as a process which mainly changes the phonological exponents of underlying, universally present functional morphemes (cf. von Fintel 1995; Roberts & Roussou 2003). . Note that (40) is trivially fulfilled if agreement morphology develops in a language that previously showed no agreement at all. . See Section 4 for further arguments from language typology that support the idea that the acquisition of agreement morphology is shaped by a principle such as (41). . For 13th century Bavarian, the sets of Vocabulary items that competed for insertion into the Agr-morpheme (present tense indicative) are thus as follows. (i) lists the set of items


prior to the development of 2pl -ts, (ii) shows the situation after the change in question has taken place. [1, +pl] [1, -pl] [2, -pl] [pl] elsewhere

↔ ↔ ↔ ↔ ↔

/-an/ ∅ /-st/ /-ant/ /-t/

(ii) [1, +pl] [2, +pl] [1, -pl] [2, -pl] [pl] elsewhere

↔ ↔ ↔ ↔ ↔ ↔

/-an/ /-ts/ ∅ /-st/ /-ant/ /-t/

(i)

. The relevant sets of Vocabulary items for 18th century Bavarian are listed below. (i) lists the items prior to the rise of 1pl -ma, while (ii) shows the paradigm resulting from the change in question. [2, +pl] [1, -pl] [2, -pl] [pl] elsewhere

↔ ↔ ↔ ↔ ↔

/-ts/ ∅ /-st/ /-an/ /-t/

(ii) [1, +pl] [2, +pl] [1, -pl] [2, -pl] [pl] elsewhere

↔ ↔ ↔ ↔ ↔ ↔

/-ma/ /-ts/ ∅ /-st/ /-an/ /-t/

(i)

. In contrast to views widely held in typological work where it is often claimed that grammaticalization processes are highly deterministic, in the sense that an item that once entered a grammaticalization cline must continue along the different stages of the cline. . In addition, we have to acknowledge that there is a difference between the change as such and its diffusion in the speaker community. Thus, even if a given change has taken place (in the grammar of some speakers), it does not necessarily prevail, due to socio-linguistic factors. In the case of 1pl -ma, for example, the change is restricted to some Lower Bavarian varieties which became marginalized due to the influence of urban (i.e. Munich) Bavarian and are nearly extinct today (cf. Altmann 1984; Kollmer 1987). . Note that the initial vowel in formatives such as -emês is actually not part of the agreement suffix, but rather a so-called ‘theme vowel’ that originally served to derive verb stems from roots. . Strong verbs and the weak verbs of conjugation class 1 exhibit -is and -ês for 2sg present indicative and 2sg present subjunctive, respectively. Here, the difference in vowel quality





Eric Fuß

was perhaps not salient enough to differentiate the forms. Furthermore, the difference was presumably further weakened by phonological erosion that affected non-stressed final syllables. Alternatively, one might assume that the change first affected the weak verbs of the conjugation classes 2 and 3 and spread later to other verb classes by analogy.

References Abraham, W. (1995). Deutsche Syntax im Sprachenvergleich. Tübingen: Gunter Narr. Ackema, P. & Neeleman, A. (2001). “Context-sensitive spell-out and adjacency.” Ms., Utrecht University and University College London. Ackema, P. & Neeleman, A. (2003). “Agreement checking at PF.” Paper presented at the Comparative Germanic Syntax Workshop 18, University of Durham. Altmann, H. (1984). “Das System der enklitischen Personalpronomina in einer mittelbairischen Mundart.” Zeitschrift für Dialektologie und Linguistik, 51(2), 191–211. Anderson, S. (1986). “Disjunctive ordering in inflectional morphology.” Natural Language and Linguistic Theory, 4, 1–31. Anderson, S. (1992). A-morphous Morphology. Cambridge: Cambridge University Press. Ariel, M. (2000). “The development of person agreement markers: From pronoun to higher accessibility markers.” In S. Kemmer & M. Barlow (Eds.), Usage-based Models of Language (pp. 197–261). Stanford: CSLI Publications. Aronoff, M. (1976). Word Formation in Generative Grammar. Cambridge, MA: MIT Press. Arregi, K. (1999). “Person and number inflection in Basque.” In V. Lin, C. Krause, B. Bruening, & K. Arregi (Eds.), MIT Working Papers in Linguistics 34 (pp. 229–264). Cambridge, MA: MIT. Bayer, J. (1984). “COMP in Bavarian syntax.” The Linguistic Review, 3, 209–274. Benveniste, E. (1972). Problèmes de linguistique générale. Paris: Editions Gallimard. Bennis, H. & Haegeman, L. (1984). “On the status of agreement and relative clauses in West Flemish.” In W. de Geest & Y. Putseys (Eds.), Sentential Complementation (pp. 33–55). Dordecht: Foris. Besten, H. den (1982). “On the interaction of root transformations and lexical deletive rules.” In W. Abraham (Ed.), On the Formal Syntax of the Westgermania (pp. 47–132). Amsterdam: John Benjamins. Bobaljik, J. (1994). “What does adjacency do?” In H. Harley & Colin Phillips (Eds.), The Morphology-Syntax Connection, MIT Working Papers in Linguistics 22 (pp. 1–32). Cambridge, MA: MIT. Bobaljik, J. (1995). “Morphosyntax: The syntax of verbal inflection.” Doctoral dissertation, MIT. Bobaljik, J. (2002). “A-chains at the PF interface.” Natural Language and Linguistic Theory, 20, 197–267. Bobaljik, J. & Wurmbrand, S. (2002). “Notes on agreement in Itelmen.” Linguistic Discovery, 1(1), 1–27 (http://linguistic-discovery.dartmouth.edu/). Bonet, E. (1991). “Morphology after syntax. Pronominal clitics in Romance.” Doctoral dissertation, MIT.


Braune, W. (1950). Althochdeutsche Grammatik. 7. Auflage. Halle: Niemeyer. Bresnan, J. (1973). “The Syntax of the comparative clause construction in English.” Linguistic Inquiry, 4, 275–343. Brinkmann, H. (1931). Sprachwandel und Sprachbewegungen in althochdeutscher Zeit. Jena: Frommann. Bruening, B. (2001). “Syntax at the edge: Cross-clausal phenomena and the syntax of Passamaquoddy.” Doctoral dissertation, MIT. Bybee, J. L. (1985). Morphology. Amsterdam: John Benjamins. Bybee, J. L., Pagliuca, W., & R. Perkins (1990). “On the asymmetries in the affixation of grammatical material.” In W. Croft, K. Denning, & S. Kemmer (Eds.), Studies in Typology and Diachrony (pp. 1–42). Amsterdam: John Benjamins. Carstens, V. (2003). “Rethinking complementizer agreement: Agree with a Case-checked goal.” Linguistic Inquiry, 34(3), 393–412. Chomsky, N. (1957). Syntactic Structures. The Hague: Mouton. Chomsky, N. (1993). “A minimalist program for linguistic theory.” In K. Hale & S. J. Keyser (Eds.), The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger (pp. 1–52). Cambridge, MA: MIT Press. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: The MIT Press. Chomsky, N. (2000). “Minimalist inquiries: the framework.” In R. Martin, D. Michaels, & J. Uriagereka (Eds.), Step by Step. Essays on Minimalist Syntax in Honor of Howard Lasnik (pp. 89–155). Cambridge, MA: MIT Press. Chomsky, N. (2001a). “Derivation by Phase.” In M. Kenstowicz (Ed.), Ken Hale: A Life in Language (pp. 1–52). Cambridge, MA: MIT Press. Chomsky, N. (2001b). “Beyond explanatory adequacy.” MIT Occasional Papers in Linguistics, 20, 1–28. Collins, C. (1997). Local Economy. Cambridge, MA: MIT Press. Comrie, B. (1980). “Morphology and word order reconstruction: Problems and prospects.” In J. Fisiak (Ed.), Historical Morphology (pp. 83–96). The Hague: Mouton. Clark, R. & Roberts, I. (1993). “A computational model of language learnability and language change.” Linguisitic Inquiry, 24, 299–345. Craig, C. G. (1977). The Structure of Jacaltec. Austin: University of Texas Press. Cysouw, M. (2001). “The paradigmatic structure of person marking.” Doctoral dissertation, Katholieke Universiteit Nijmegen. Embick, D. (1997). “Voice and the interfaces of syntax.” Doctoral dissertation, University of Pennsylvania. Embick, D. & Noyer, R. (1999). “Locality in post-syntactic operations.” In V. Lin, C. Krause, B. Bruening, & K. Arregi (Eds.), MIT Working Papers in Linguistics 34 (pp. 265–316). Cambridge, MA: MIT. Fintel, K. von (1995). “The formal semantics of grammaticalization.” In J. Beckman (Ed.), Proceedings of NELS 25: Workshop on Language Change (pp. 175–189). Amherst: GLSA, University of Massachusetts. Fuß, E. (in prep.). “The rise of agreement. Grammaticalization and its syntactic effects.” Doctoral dissertation, University of Frankfurt. Givón, T. (1976). “Topic, pronoun, and grammatical agreement.” In C. N. Li (Ed.), Subject and Topic (pp. 149–188). New York: Academic Press.





Eric Fuß

Gelderen, E. van (2001). “Modal particles in Germanic.” Folia Linguistica Historica, 22(1–2), 310–333. Gerlach, B. (2002). Clitics between Syntax and Lexicon. Amsterdam: John Benjamins. Grewendorf, G. (1989). Ergativity in German. Dordrecht: Foris. Grewendorf, G. (2004). “The discourse configurationality of scrambling.” Ms., University of Frankfurt. Grimshaw, J. (1997). “The best clitic: constraint conflict in morphosyntax.” In L. Haegeman (Ed.), Elements of Grammar (pp. 169–196). Dordrecht: Kluwer. Haeberli, E. (2002). Features, Categories and the Syntax of A-Positions. Cross-Linguistic Variation in the Germanic Languages. Dordrecht: Kluwer. Haegeman, L. (1992). Theory and Description in Generative Syntax: A Case Study in WestFlemish. Cambridge: Cambridge University Press. Halle, M. (1997). “Distributed Morphology: Impoverishment and Fission.” In PF: Papers at the Interface. MIT Working Papers in Linguistics, 30, 425–450. Cambridge, MA: MIT. Halle, M. & Marantz, A. (1993). “Distributed Morphology and the pieces of inflection.” In S. J. Keyser & K. Hale (Eds.), The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger (pp. 111–176). Cambridge, MA: MIT Press. Halle, M. & Marantz, A. (1994). Some key features of Distributed Morphology. Papers on Phonology and Morphology, MIT Working Papers in Linguistics, 21, 275–288. Cambridge, MA: MIT. Halpern, A. (1992). “Topics in the placement and morphology of clitics.” Doctoral dissertation, Stanford University. Harley, H. & Ritter, E. (2002). “Structuring the bundle: A universal morphosyntactic feature geometry.” In H. J. Simon & H. Wiese (Eds.), Pronouns – Grammar and Representation (pp. 23–39). Amsterdam: John Benjamins. Hoekstra, E. & Smits, C. (1999). “Everything you always wanted to know about complementizer agreement.” In E. van Gelderen & V. Samiian (Eds.), Proceedings of WECOL 1998. Fresno, CA: California State University Press. Hoekstra, J. & Marácz, L. (1989). “On the position of inflection in West Germanic. Working Papers in Scandinavian Syntax, 44, 75–88. Iatridou, S. (1990). “About Agr(P).” Linguistic Inquiry, 21, 551–577. Julien, M. (2002). Syntactic Heads and Word Formation. Oxford: Oxford University Press. Kinyalolo, K. (1991). “Syntactic dependencies and the spec-head agreement hypothesis in Kilega.” Doctoral dissertation, UCLA. Kiparsky, P. (1973). “‘Elsewhere’ in phonology.” In S. Anderson & P. Kiparsky (Eds.), A Festschrift for Morris Halle (pp. 93–106). New York: Holt, Rinehart and Winston. Kiparsky, P. (1982). “Word-formation and the lexicon.” In F. Ingemann (Ed.), Proceedings of the 1982 Mid-America Linguistics Conference (pp. 3–29). University of Kansas. Kitahara, H. (1997). Elementary Operations and Optimal Derivations. Cambridge, MA: MIT Press. Kollmer, M. (1987). Die schöne Waldlersprach. Bd. I–III. Prackenbach: Kollmer (Eigenverlag). Kornfilt, J. (1990). “Turkish and the Turkic languages.” In B. Comrie (Ed.), The Major Languages of Eastern Europe (pp. 227–252). London: Routledge.


Kroch, A. (1994). “Morphosyntactic change.” In K. Beals (Ed.), Proceedings of the 30th Annual Meeting of the Chicago Linguistics Society, Vol. 2 (pp. 180–201). Chicago: Chicago Linguistics Society. Lambrecht, K. (1981). Topic, Antitopic and Verb Agreement in Non-standard French. Pragmatics and beyond, Vol. II: 6. Amsterdam: John Benjamins. Lechner, W. (2001). “Reduced and phrasal comparatives.” Natural Language and Linguistic Theory, 19, 683–735. Lightfoot, D. (1991). How to Set Parameters. Cambridge, MA: MIT Press. Lightfoot, D. (1999). The Development of Language. Acquisition, Change, and Evolution. Oxford: Blackwell. Lühr, R. (1984). “Reste der athematischen Konjugation in den germanischen Sprachen.” In J. Untermann & B. Brogyanyi (Eds.), Das Germanische und die Rekonstruktion der Indogermanischen Ursprache (pp. 25–90). Amsterdam: John Benjamins. Marantz, A. (1988). “Clitics, morphological merger, and the mapping to phonological structure.” In M. Hammond & M. Noonan (Eds.), Theoretical Morphology: Approaches in Modern Linguistics (pp. 253–270). San Diego: Academic Press. Marantz, A. (1992). “Case and licensing.” In G. Westphal, B. Ao, & H.-R. Chae (Eds.), Proceedings of ESCOL 1991 (pp. 142–159). Columbus, Ohio: Department of Linguistics, Ohio State University. Mitchell, E. (1994). “When Agro is fused to Agrs: What morphology can tell us about the functional categories.” In H. Harley and C. Phillips (Eds.), The Morphology-Syntax Connection. MIT Working Papers in Linguistics 22 (pp. 111–130). Cambridge, MA: MIT. Mithun, M. (1991). “The development of bound pronominal paradigms.” In W. P. Lehmann & H.-J. Jakusz Hewitt (Eds.), Language Typology 1988: Typological Models in Reconstruction (pp. 85–104). Amsterdam: John Benjamins. Moulton, W. (1944). “Scribe γ of the Old High German Tatian translation.” PMLA, 59(2), 307–334. Nordlinger, R. (1995). “Split tense and mood inflection in Wambaya.” Berkeley Linguistics Society, 21, 226–236. Pesetsky, D. & Torrego, E. (2001). “T-to-C movement: causes and consequences.” In M. Kenstowicz (Ed.), Ken Hale: A Life in Language (pp. 355–426). Cambridge, MA: MIT Press. Pfalz, A. (1918). “Suffigierung der Personalpronomina im Donaubairischen.” Republished in 1983 in P. Wiesinger (Ed.), Die Wiener dialektologische Schule. Grundsätzliche Studien aus 70 Jahren Forschung. Wiener Arbeiten zur germanischen Altertumskunde und Philologie 23 (pp. 217–235). Wien: Karl M. Halosar. Poletto, C. (1999). “The internal structure of Agrs and subject clitics in the Northern Italian dialects.” In H. van Riemsdijk (Ed.), Clitics in the Languages of Europe (pp. 581–620). Berlin: Mouton de Gruyter. Polinsky, M. & Potsdam, E. (2001). “Long distance agreement and topic in Tsez.” Natural Language and Linguistic Theory, 19, 583–646. Poppe, N. (1960). Buriat Grammar. Bloomington: Indiana University Publications. Rizzi, L. (1997). “The fine structure of the left periphery.” In L. Haegeman (Ed.), Elements of Grammar: Handbook in Generative Syntax (pp. 281–337). Dordrecht: Kluwer.



 Eric Fuß

Roberts, I. & Roussou, A. (2003). Syntactic Change. A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. Sauerland, U. (1996). “The late insertion of Germanic inflection.” Ms., MIT. Schütze, C. (1994). “Serbo-Croatian second position clitic placement and the phonologysyntax interface.” In A. Carnie, H. Harley, & T. Bures (Eds.), Papers on phonology and morphology. MIT Working Papers in Linguistics 21 (pp. 373–473). Cambridge, MA: MIT. Shlonsky, U. (1994). “Agreement in Comp.” The Linguistic Review, 11, 351–375. Sievers, E. (Ed.). (1961). Tatian. Darmstadt: Wissenschaftliche Buchgesellschaft. Sommer, T. (1994). Flexionsmorphologie des Verbs im althochdeutschen Tatian. München: Tuduv. Speas, M. (1991). “Functional heads and inflectional morphology.” The Linguistic Review, 8, 389–417. Weiß, H. (1998). Die Syntax des Bairischen. Studien zur Grammatik einer natürlichen Sprache. Tübingen: Niemeyer. Weiß, H. (2002). “Agr-in Comp and partial pro-drop.” Ms., Universität Regensburg. Wiesinger, P. (1989). Die Flexionsmorphologie des Verbums im Bairischen. Wien: Verlag der Österreichischen Akademie der Wissenschaften. Zeller, J. (2001). Particle Verbs and Local Domains. Amsterdam: John Benjamins. Zwart, J. W. (1993). “Clues from dialect syntax: Complementizer agreement.” In W. Abraham & J. Bayer (Eds.), Dialektsyntax. Linguistische Berichte Sonderheft 5 (pp. 246– 270). Opladen: Westdeutscher Verlag.

Syntactic effects of inflectional morphology and competing grammars* Eric Haeberli University of Geneva

It is a long-standing observation that cross-linguistic syntactic variation sometimes seems to correlate with variation in the inflectional morphology. An attractive consequence of this observation is that it may provide the basis for genuine explanations of certain aspects of syntactic variation and thereby replace accounts in which the variation is simply due to random differences in parameter settings. However, in many instances in which such a relation between the syntax and the morphology has been postulated, occasional counterexamples to the generalization can be found and they shed doubts on its validity. In this paper, it is argued that, once we consider such counterexamples not just from a synchronic but also from a diachronic perspective, we are not necessarily forced to abandon the basic intuition behind the generalization. Taking data concerning the order of arguments (‘free’ vs. fixed) and the distribution of subjects and adjuncts in subject-verb inversion contexts as illustrations, it will be shown that apparently problematic cases can often be dealt with if we adopt a specific conception of how the syntax is related to inflectional morphology and of how grammars change.

.

Introduction

An important feature of generative syntax over the last three decades has been its comparative dimension. The goal of the comparative approach is to identify common properties and areas of variation across languages in order to determine the nature of Universal Grammar and, more specifically, its principles and parameters. Although most comparative work in generative syntax focuses on living languages, the comparative perspective has also contributed to a resurging interest in diachronic evidence, i.e. the comparison of a sequence of historically related synchronic grammars.

 Eric Haeberli

Considering the status of diachronic syntax within the comparative approach, the question might be raised whether a diachronic comparison is to a large extent equivalent to a comparison of any other set of grammars of living or not directly related dead languages, or whether the diachronic perspective can provide insights that could not be obtained from other types of comparative studies. The following point made by Kroch (1989: 200) suggests that the latter may be true. [. . . ] with information about the time course of language change . . . we may hope to learn how the grammars of languages change from one state to another over time and, from an understanding of the process by which they change, to learn more about their principles of organization. After all, perturbing a complex system and observing its subsequent evolution is often an excellent way of inferring internal structure. (Kroch 1989: 200)

Thus, if we can observe for example that the grammatical system is perturbed by a change A and that A co-occurs with or is followed by a change B in another domain of the grammar and if the changes are not clearly related to some external factor such as language contact, it is plausible to conclude that the phenomena involved in A and B are underlyingly related. Diachronic change may therefore provide clues as to whether and how some aspects of the grammar are related to certain other aspects of the grammar. It could be argued now that connections between different components of the grammar can also be identified on the basis of non-diachronic comparative evidence. If the grammatical properties before change A and B and those after change A and B systematically co-occur across languages, it would also be plausible to infer an underlying relationship between these properties. However, I will argue in this paper that synchronic snapshots can sometimes be insufficient when trying to determine the connections among different aspects of the grammar and that diachronic developments have to be taken into account in order to obtain a complete picture. Thus, diachronic considerations can provide us with clues for the analysis of the grammatical system which are not available on the basis of purely synchronic evidence. The area I will be focusing on is the relationship between syntactic phenomena and morphological properties. More precisely, I will consider two cases for which a connection between syntax and morphology has been proposed, namely the relationship between the distribution of subjects and agreement morphology in the Germanic languages and the correlation between word order freedom and case morphology. The main point arising from the discussion of these phenomena is that certain apparent counterexamples to

Syntactic effects of inflectional morphology 

proposed relations between syntactic and morphological properties can be explained as intermediate stages of diachronic developments if we adopt the position that periods of linguistic change are characterized by what has been referred to as competing grammars (cf. e.g. Kroch 1989). Thus, correlations that may look problematic from a purely synchronic point of view can be strengthened if diachronic considerations are taken into account. Before turning to these issues in more detail, let us briefly review the basic aspects of the analysis of syntactic variation within generative grammar and some of the proposals that have been made on the relationship between syntactic variation and morphological variation.

. Syntactic variation and morphological variation Within the Principles and Parameters framework (cf. e.g. Chomsky 1981, 1986) and within the Minimalist Program (cf. e.g. Chomsky 1995), the innate language faculty (Universal Grammar) is assumed to consist of principles, i.e. grammatical properties holding across languages, and parameters, i.e. grammatical properties whose content is not determined universally but which allow variation among languages. A parameter is generally assumed to provide a choice between two options, and the task of the language learner is thus simply to determine the parameter setting that corresponds to the language she or he is exposed to. If we look at some of the parameters that have been proposed to account for cross-linguistic variation in the literature, we can observe that for many of them it seems fairly arbitrary whether the parameter is set one way or the other in a given language. Consider for example one of the textbook cases of a parameter, the subjacency parameter. As observed by Rizzi (1982), Italian and English differ with respect to the movement options of wh-elements. This is illustrated in (1). Italian (1) a. tuo fratello, [a cui]i mi domando che storie abbiano raccontato ti your brother to whom myself I-ask which stories they-have told b. *your brother [to whom]i I wonder which stories they told ti

The structures in (1a) and (1b) seem to be identical in the two languages, but whereas Italian allows extraction of a wh-element out of the embedded interrogative clause in the context of relativization, the same process leads to an ungrammatical result in English. This contrast has been accounted for in terms

 Eric Haeberli

of bounding theory. The central principle of bounding theory is the subjacency condition which postulates that movement cannot cross more than one bounding node. As a principle of Universal Grammar, the subjacency condition holds across languages. What gives rise to cross-linguistic variation is the notion of bounding node. It is proposed (cf. again Rizzi 1982) that while NP is a bounding node in both Italian and English, the two languages differ as to which projection counts as a bounding node at the clausal level. In Italian it is CP while in English it is IP. This explains the contrast in (1). In (1), [Spec, CP] of the embedded interrogative is occupied by a wh-constituent and the higher wh-element therefore has to move directly to the [Spec, CP] of the higher clause. Doing so, it crosses two IPs (the lower one and the higher one) but only one CP (the lower one). Thus, in a language like English in which IP is a bounding node the subjacency condition is violated because two bounding nodes are crossed and the structure in (1) is ungrammatical. But in a language like Italian where CP is a bounding node only one bounding node is crossed and subjacency is respected. Thus, the contrast in (1) has been accounted for in terms of a parameter determining the categorial status of bounding nodes: IP or CP. However, no explanation seems to be available as to why a language chooses IP or CP. It could just as well be English that selects CP and Italian that chooses IP, or the parameter could be set identically in both languages. Cross-linguistic variation arising from the subjacency parameter therefore seems to be to a large extent random. Another aspect of wh-questions provides a further illustration of this apparent arbitrariness with respect to parameter setting. As often observed, we can distinguish three main systems of wh-question formation: (i) one or more wh-constituents can be fronted (e.g. Bulgarian, cf. Rudin 1988); (ii) a single wh-constituent can be fronted (e.g. English); (iii) wh-constituents do not undergo overt movement (e.g. Japanese). This suggests that there is a parameter determining how many wh-constituents can be fronted overtly. The parameter setting in Bulgarian allows multiple fronting, the setting in English licenses only one movement process, and the setting in Japanese licenses no overt movement. Although proposals have been made to relate some aspects of this variation to morphological properties of wh-elements (cf. e.g. Cheng 1991; Grewendorf 2001), the choice of certain parametric options remains unexplained. For example, nothing seems to prevent Japanese from behaving like Bulgarian, or English from having the Japanese parameter setting. To a large extent, the parametric variation with respect to overt wh-movement thus appears to be arbitrary.

Syntactic effects of inflectional morphology 

In terms of current syntactic theory, it is difficult to see how the crosslinguistic situation discussed above could be explained in a principled way. It may therefore very well be that, for certain parameters, no deeper reasons can be identified as to why the setting is one way or the other in a given language. However, this does not mean that all aspects of syntactic variation must be related to parameters whose setting seems to be random. Instead, there are areas of variation where plausible accounts can be given for the behavior of individual languages. A good illustration for this observation is again provided by one of the early parameters discussed in the generative literature. The pro-drop parameter (cf. e.g. Rizzi 1982) has two attractive properties. First, it relates several aspects of cross-linguistic syntactic variation (pro-drop, that-trace effects, postverbal subjects) to a single source (the pro-drop parameter). Thus, we are not just looking at individual phenomena but at an entire cluster of properties, and the crosslinguistic variation becomes slightly less random in the sense that there is not a different parameter for each area of variation. But the pro-drop parameter has a second attractive property which makes the cross-linguistic variation even less arbitrary. It is assumed that the setting of the pro-drop parameter is related to a different, non-syntactic property, namely the richness of the verbal agreement morphology. Thus, in languages with a sufficiently rich agreement morphology the pro-drop parameter is set positively (hence pro-drop, no that-trace effects and postverbal subjects) whereas in other languages it is set negatively (hence no pro-drop, that-trace effects and no postverbal subjects). More detailed research on pro-drop and the related phenomena has shown that the situation is more complex than just described, but for our purposes it is not essential to pursue this issue any further. The main point here is the spirit rather than the details of the pro-drop parameter, i.e. the hypothesis that parametric variation is sometimes by no means random but determined by other factors and more particularly by inflectional morphology. Another illustration of this type of approach is the variation found with respect to the distribution of finite verbs. As often observed, certain types of adverbs can precede the finite verb in some languages (e.g. English) whereas they have to follow it in some other languages (e.g. French): French (2) a. John (often) eats (*often) chocolate b. Jean (*souvent) mange (souvent) du chocolat John (often) eats (often) chocolate

 Eric Haeberli

The way this contrast has been analyzed (cf. e.g. Emonds 1978; Pollock 1989) is that the verb stays in V in (2a) whereas it moves past the adverb into the inflectional domain in (2b). Thus, there seems to be a parameter determining whether a verb moves to I or not. As in the case of the pro-drop parameter, it has been proposed that variation in the setting of the V-to-I parameter is not entirely random but that it is the consequence of variation in the domain of verbal agreement morphology (cf. e.g. Holmberg & Platzack 1995; Pollock 1989; Roberts 1985; Rohrbacher 1994; Vikner 1995). Languages with a certain richness of agreement morphology have a parameter setting that triggers verb movement to I, whereas languages with an impoverished agreement system lack verb movement. The two parameters just discussed (pro-drop and verb movement) are in principle very attractive because an attempt is made to provide genuine explanations for syntactic variation. Language X has the syntactic property A because it has the morphological property C, and language Y has the syntactic property B because it has the morphological property D. However, the occurrence of counterexamples sometimes sheds doubts on these types of correlations. For example, it has long been observed that there is a dialect of Swedish that seems to have verb movement despite an impoverished verbal agreement paradigm (cf. e.g. Platzack & Holmberg 1989: 73–74). The aim of this paper is to show that such counterexamples may not always be fatal to generalizations concerning the relation between syntax and morphology. This conclusion will be based on an examination of two areas of syntactic variation other than pro-drop and verb movement, namely the distribution of subjects and adjuncts in subject-verb inversion contexts and the order of arguments (‘free’ vs. fixed).1 Both of these areas of syntactic variation can be argued to be related to variation in the inflectional morphology, and in both cases there seem to be potential counterexamples to the correlation. But once we consider such counterexamples not just from a synchronic but also from a diachronic perspective, we are not necessarily forced to abandon the basic intuition behind the generalization. Apparently problematic cases can be dealt with if we adopt a specific view of how the syntax is related to inflectional morphology and of how grammars change.

. ‘XP-subject’ orders in Germanic The first phenomenon I would like to consider is the distribution of postverbal subjects in the modern Germanic Verb Second (V2) languages. As shown

Syntactic effects of inflectional morphology 

in Haeberli (1999, 2002a) and Vikner (1995), there is considerable variation as to whether a postverbal definite full DP subject has to be adjacent to the finite verb in C or whether a constituent can occur between the finite verb and a definite subject (‘XP-subject’ orders). Some languages accept ‘XP-subject’ orders, others do not. I will start my discussion by focusing on the West Germanic languages, and I will then consider some issues that the Mainland Scandinavian languages raise for the analysis of this phenomenon. . ‘XP-subject’ in West Germanic The following data show the variation with respect to ‘XP-subject’ orders in the West Germanic languages.2 German (3) a. Wahrscheinlich wird (später) Hans dieselbe Uhr kaufen. West Flemish b. Misschien goa (*loater) Jan tzelfste orloge kuopen. Dutch c.

Waarschijnlijk zal (%later) Jan hetzelfde horloge gaan kopen.

Frisian d. Wierskynlik wol (letter) Jan itselde horloazje keapje. Afrikaans e. Waarskynlik sal (*later) Jan dieselfde oorlosie gaan koop. probably will (later) John the-same watch (go) buy ‘Probably, John will buy the same watch (later).’ Yiddish f. Minastam vet (shpeter) Moyshe koyfn dem zelbikn zeyger. probably will (later) Moyshe buy the same watch

Whereas an adjunct can intervene between the finite verb and a full DP subject in German, Frisian, Yiddish and (with some restrictions) Dutch, strict adjacency is required in West Flemish and Afrikaans. Given the traditional assumption that in asymmetric V2 languages (i.e. all languages in (3) apart from Yiddish) the complementizer in subordinate clauses is in the structural position occupied by the finite verb in main clauses, the pattern in (3) is reproduced in subordinate clauses. In other words, definite full DP subjects have to be ad-

 Eric Haeberli

jacent to the complementizer in West Flemish and Afrikaans, while adjuncts can intervene between the complementizer and the subject in German, Frisian and Dutch. A similar kind of restriction as in West Flemish and Afrikaans (3b–e) can also be found in Modern English residual V2 contexts. This is illustrated in (4) where the fronted auxiliary cannot be separated from the subject by another constituent. (4) a. Under no circumstances would (*later) John buy the same watch. b. Why did (*yesterday) Mary buy the same watch?

The question that arises now is how the variation found in (3) and (4) can be accounted for. One possibility, proposed by Holmberg (1993) and Vikner (1995), would be to say that the different languages vary with respect to whether they license adjunction to the highest inflectional projection, say AgrP. Hence, in languages like German, Dutch, Frisian and Yiddish, AgrP is an adjunction site for adjuncts whereas in West Flemish, Afrikaans and English AgrP-adjunction is ruled out. This kind of parametric variation would be of the type discussed for subjacency in Section 2. In the same way that it is not clear why CP is a bounding node in some languages but IP in others, there is no obvious reason why adjunction should be restricted in some languages but not in others. The variation with respect to the occurrence of ‘XP-subject’ orders would therefore remain entirely random. From the point of view of an adjunction analysis, we could not be surprised if some languages behaved differently from what is shown in (3) and (4) or even if the division among the West Germanic languages were exactly the opposite, with German, Dutch, Frisian and Yiddish requiring adjacency and West Flemish, Afrikaans and English allowing non-adjacency. In Haeberli (1999, 2002a: Chapter 4), I propose an alternative analysis of the data in (3) and (4), the main goal of which is to account for the variation in a less random fashion by relating it to other grammatical factors. The basic assumptions made for this analysis are the following:3 i. Adjunction to inflectional projections is ruled out. ii. The occurrence of an adjunct in pre-subject position is evidence for additional structure above TP. iii. Additional structure above TP is related to AgrP. iv. The presence of verbal Agr features and, hence, of AgrP is related to verbal agreement morphology.4 Agr is syntactically represented if a language has rich agreement morphology, where ‘rich agreement’ is defined as the

Syntactic effects of inflectional morphology 

co-occurrence of an agreement and a tense morpheme (cf. Bobaljik & Thráinsson 1998; Thráinsson 1996) or as consisting of more than a twoway (default/non-default) distinction.5 Given these basic assumptions, the variation in the West Germanic languages in (3)–(4) can be analyzed as follows. First, with respect to ‘XP-subject’ orders, the languages that allow this word order all have rich agreement (as defined above) and they therefore have an AgrP-TP structure in the inflectional domain. Concerning the placement of the subject, we can assume that it moves to TP to be licensed (Case). Given that all the West Germanic languages with ‘XP-subject’ orders license empty expletives, the second step is then to propose that a subject can remain in TP because [Spec, AgrP] can be occupied by an empty expletive. And the final ingredient for the analysis of ‘XP-subject’ orders is to assume that the XP is merged ‘parasitically’ on the presence of AgrP, either through Agr’-adjunction or insertion in the specifier of an independent adjunct projection whose presence depends on the presence of AgrP. Thus, we obtain the following structure for the ‘XP-subject’ orders in (3). (5) [CP C [AgrsP e XP [TP SU . . . ]]]

For verb-subject adjacency in (3) and (4), we have to distinguish two scenarios. The first scenario accounts for the absence of ‘XP-subject’ orders in Afrikaans and English. Both of these languages do not have rich agreement. Given point (iv) above, we may therefore assume that they also do not have Agr-features and an AgrP-level in the inflectional domain. Hence, once a subject has moved to be licensed in [Spec, TP], it is in a position adjacent to a verb or an auxiliary in C. The structural representation of this analysis is shown in (6). (6) [CP C (*XP) [TP SU . . . ]]

Finally, there is a second scenario giving rise to verb-subject adjacency. This scenario applies to West Flemish. West Flemish has rich agreement and therefore, in terms of (iv) above, an AgrP-TP structure. But what distinguishes West Flemish from the other West Germanic languages with rich agreement is that it does not license empty expletives.6 We may therefore assume that subjects in West Flemish cannot stay in TP but have to move to AgrP so that the relevant features in AgrP can be checked. The absence of ‘XP-subject’ orders can thus be accounted for in the following way. (7) [CP C (*XP) [AgrsP SUi [TP ti . . . ]]]

 Eric Haeberli

The result we have obtained in terms of the above analysis is that the variation in (3) and (4) is not simply the result of the apparently random setting of a parameter. Instead it is the consequence of other properties of the different languages, in particular the richness of verbal agreement morphology and the licensing of empty expletives. The ‘XP-subject’ variation can therefore be considered as being of the type discussed for pro-drop and verb movement in Section 2: A syntactic property is crucially linked to a morphological property. . ‘XP-subject’ in Mainland Scandinavian If the conclusions reached in the previous subsection are correct, then we would obviously expect this analysis to account for phenomena in other languages as well. In Haeberli (2000), it is shown that we indeed obtain the right results for Old English and for what seems to be a case of dialect variation in Middle English. However, once we consider the Mainland Scandinavian languages, the correlation between rich agreement morphology and ‘XP-subject’ orders appears to encounter some difficulties. In Swedish and Norwegian, we can find adjuncts occurring between a finite verb and a subject. This is shown in (8). Swedish (8) a. Den här klockan hade (senare) min gamle far köpt Norwegian b. Denne klokka hadde (seinere) min gamle far kjøpt this watch had (later) my old dad bought

In terms of the analysis of (3)–(4) outlined above, the grammaticality of (8) is surprising at first sight. If ‘XP-subject’ orders are related to the presence of AgrP above TP and if the presence of AgrP is related to rich agreement morphology, we would not expect ‘XP-subject’ orders in Swedish and Norwegian because both languages lack verbal agreement. A possible explanation for the situation in Swedish and Norwegian can be found if we consider the status of ‘XP-subject’ orders in Danish and in particular its diachronic development. Modern Danish behaves like West Flemish, Afrikaans and English in that postverbal subjects have to be adjacent to the finite verb. Thus, the equivalent of (8) with an intervening adjunct is ungrammatical in Danish.

Syntactic effects of inflectional morphology

Danish (9) Dette ur vil (*senere) min far købe this watch will (later) my dad buy

The ungrammaticality of non-adjacency as in (9) follows from the analysis presented above for the West Germanic languages. Just like Afrikaans and English, Danish lacks rich subject-verb agreement. The adjacency requirement in Danish can therefore be analyzed as in (6), i.e. in terms of the absence of AgrP. What is interesting for our purposes now is the observation made by Vikner (1995: 128) that there is “evidence which indicates that the difference between Danish and Swedish . . . is a rather recent difference.” Vikner provides references to Mikkelsen (1911) and Diderichsen (1962) for examples showing adjuncts in a position between the finite verb and a postverbal subject in Danish. The change to a grammar which does not license ‘XP-subject’ orders thus does not seem to have been completed before the 20th century. So what about verbal agreement morphology? As shown by Vikner (1997: 206), the verbal agreement paradigm is already extremely impoverished in the 14th century. There is no evidence for the co-occurrence of tense and agreement morphology at this time any more. So from this point of view, there would be no need to postulate verbal Agr features and AgrP. However, if, as suggested under point (iv) above, an agreement distinction which goes beyond a simple two-way (default/non-default) distinction is sufficient to trigger the presence of Agr features, then the paradigm given by Vikner could still be argued to entail the presence of AgrP in the clause structure of 14th century Danish because we can distinguish three different morphemes which seem to have no other function than to mark agreement (-ær present singular and imperative plural, -æ present plural, and zero for imperative singular). I have to leave it for further research to determine exactly when these distinctions were further reduced to the point where Danish did not meet the requirements for rich agreement defined in (iv) above any more, but it is very likely that this happened well before the 20th century. Returning to the correlation between ‘XP-subject’ orders and rich verbal agreement, the conclusion with respect to the history of Danish thus is that the loss of the syntactic phenomenon occurred very recently, whereas the impoverishment of the morphology seems to have occurred considerably earlier, possibly several centuries before the syntactic change. At first sight, this, like the modern Swedish and Norwegian data, may appear to undermine the hypothesized link between syntax and morphology in the context of verb-subject





Eric Haeberli

non-adjacency. This conclusion can be rejected, however, if we take a closer look at the mechanisms of diachronic change. . Morphology, syntax and diachronic change In this subsection, I will show that a delay between a morphological change and a related syntactic change as discussed for Danish is not entirely unexpected if we adopt a certain conception of the relation between morphology and syntax and a certain conception of how syntax changes. First, with respect to the relation between morphology and syntax, it is important to point out that the analysis of the variation in (3)–(4) outlined above does not suggest that there is a direct relation between a morphological property and a syntactic phenomenon. Instead, the relation is mediated by a syntactic feature and the syntactic structure this feature projects, an assumption which is also at the heart of the approach discussed in Bobaljik (2002), Bobaljik & Thráinsson (1998) and Thráinsson (1996). The basic idea is thus that there is a close link between morphology and a syntactic feature/structure, and a close link between this syntactic feature/structure and a syntactic phenomenon. Represented schematically, we obtain a relation like (10b) rather than one like (10a).7 (10) a. *morphology → syntactic phenomenon b. morphology ↔ syntactic feature/syntactic structure ↔ syntactic phenomenon

From the point of view of language change, the distinction in (10) has the following consequences. In terms of (10a), the loss of a morphological property would have to entail the immediate loss of the syntactic phenomenon. If the morphology is not available, the syntactic property cannot exist, either. In terms of (10b), however, the loss of the relevant morphology does not necessarily lead to the loss of the related syntactic phenomenon because the mediating syntactic structure can be maintained on the basis of the evidence provided by the relevant syntactic phenomenon. Instead of (10b), we then get (11). (11) morphology ↔ syntactic feature/syntactic structure ↔ syntactic phenomenon

(11) explains why a morphological change does not have to be immediately followed by a change in the syntactic property that seems to be related to the morphology. However, a different issue arises now. The question is not any more why there is a delay between the morphological change and the syntactic


change but rather why the change would ever occur given that the syntactic phenomenon could be sufficient to maintain the syntactic structure for any generation to come. As discussed by Bobaljik (2002), a strong interpretation of a scenario like (10b) indeed means that the lack of morphology does not necessarily make any predictions with respect to the future of the related syntactic structure.8 If there is one related syntactic phenomenon, the morphology is one among two pieces of evidence suggesting the presence of the relevant syntactic structure to the language learner, i.e. the morphology is one among two of what Lightfoot (1999) calls cues. If there is more than one related phenomenon, the morphology is one among several cues. The loss of the morphology therefore simply means that there is one fewer cue to the next generation of learners, but the grammar with the relevant syntactic structure may remain perfectly learnable on the basis of the syntactic evidence. The diachronic developments that have been observed in contexts of a hypothesized relation between morphology and syntax and, more particularly, the loss of a syntactic phenomenon after the loss of the related morphological property thus cannot be immediately explained on the basis of the scenario represented in (10b)–(11). One way to deal with this problem might be to question the assumption that all types of cues are of equal importance. Instead, it could be argued that the morphology is the more relevant cue in the acquisition process, and that once this cue is lost the related syntactic structure is too weakly motivated to be maintained. However, such an assumption may be problematic. In his discussion of V-to-I movement, Bobaljik (2002: 162) is very skeptical as regards the relevance of verbal morphology as a cue for the acquisition of verb movement. He cites evidence suggesting that children acquire verb movement in French before they fully master the verbal morphology that has been related to the presence of verb movement. Although, as Bobaljik points out, various open questions concerning the interpretation of the acquisition data remain, these data nevertheless seem to shed doubts on an approach which would want to attribute to the morphology the crucial role in (10b). But the opposite conclusion, i.e. that the syntactic cues are more important than the morphological ones, does not seem to be warranted, either. In the case of V-to-I movement in French, the relevant syntactic phenomenon is very salient in the primary linguistic data the child is exposed to. Given that negation is generally used as a diagnostics for V-to-I movement, each negative declarative clause with a verb preceding the negator provides evidence for the occurrence of verb movement, and so do numerous other declarative clauses with certain adverbs. In other cases, however, the syntactic phenomenon might not be as salient. For example the phenomenon discussed in this section, ‘XP-





Eric Haeberli

subject’ orders in Germanic, is no doubt much rarer in the language learner’s input. In main clauses, this phenomenon is restricted to cases with subjectverb inversion, a word order pattern which is already considerably less frequent than subject-initial V2. Furthermore, adjunct XPs can generally also occur to the right of the subject and, although I am not aware of any statistical data on this at the moment, there is no doubt that the ‘subject-XP’ order is more frequent than the ‘XP-subject’ order. So in this case where we have a fairly marked construction, it is less likely that children start using the relevant syntactic phenomenon before they have acquired the morphological property that I have related to it above (i.e. rich agreement).9 What this example thus shows is that the salience of syntactic cues can vary considerably to the point that the morphological cue may turn out to be more salient. In other words, the issue in (10b) may not be whether morphological cues are more, equally or less important than syntactic ones to determine the presence of a syntactic feature/structure, the issue may simply be one of relative salience of different cues for the language learner, and the relative salience may vary from one context to another. Sometimes the syntax may be crucial in the acquisition process, sometimes the morphology may be more important. To return to our diachronic issue of the connection between morphological and syntactic change, the above observations may lead us to the following hypothesis: A morphological change leads to a related syntactic change whenever the morphological cue is the most salient one for the language learner. Below, I will argue that this hypothesis is not correct, but let us suppose for the moment that it is. Reconsider now the status of ‘XP-subject’ orders in Germanic. If, as suggested above, ‘XP-subject’ is a relatively infrequent construction, the morphological property (rich agreement) that is related to the syntactic feature/structure licensing ‘XP-subject’ (Agr) may be more salient than the syntactic phenomenon. Therefore, when rich agreement is lost, the status of Agr is considerably weakened. But how can we express this weakening formally? I propose that this can be done by adopting the view, put forward by Kroch (1989) and further developed in much subsequent work (cf. Kroch 2001 for an overview), that syntactic change involves a transitional phase where different grammars are in competition. The loss of rich agreement means that a grammar without Agr would become a possibility. Assuming that language learners postulate the most economical grammar and that a grammar with fewer features and less structure is more economical, they start introducing an Agr-less grammar once rich agreement is lost. However, even though ‘XP-subject’ orders may not be very frequent, their occurrence would still require a grammar in which Agr is postulated. Hence, a situation arises in which two grammars


compete: one grammar with Agr producing ‘XP-subject’ orders and one grammar without Agr not being able to derive ‘XP-subject’ orders. Schematically, this situation can be represented as shown in (12). (12) Grammar A: morphology ↔ syntactic feature/syntactic structure ↔ syntactic phenomenon Grammar B: morphology ↔ syntactic feature/syntactic structure ↔ syntactic phenomenon

The two grammars remain in competition for some time, and the old grammar A is finally driven out. As for the reasons for the loss of grammar A, it would be plausible to assume that grammar B wins out because it is more economical and eliminates an apparent conflict between the morphology and the syntax. An important question that arises in this context is what the expected timescale for the elimination of grammar A is. It seems impossible to give a precise figure here, but for our purpose it is sufficient to point out that long periods of competition are attested. For example, Ellegård’s (1953) curve, which represents the rise of do-support and which Kroch (1989) reinterprets in terms of competing grammars, covers about 300 years and at the end of the curve the change is not actually completed yet. Similarly, Santorini’s (1993) discussion of the rise of Infl-initial clause structure in the history of Yiddish suggests that the Infl-initial grammar and the old Infl-final grammar were in competition for around 400 years. Before reconsidering Mainland Scandinavian ‘XP-subject’ orders in the light of the above observations, let me just briefly point out that the scenario outlined before can also account for cases where the syntactic phenomenon is arguably a more salient cue for the syntactic structure than the morphological property. As discussed above, V-to-I movement may be relevant here. Suppose that, as suggested by Bobaljik (2002) children acquire V-to-I movement in French mainly on the basis of syntactic evidence rather than on the basis of the relevant verbal morphology. This may very well be true, but it does not mean that the morphology plays no role at all. Instead, I propose that the role of the morphology is to confirm the syntactic structure postulated on the basis of the syntactic evidence or to disconfirm it. The latter case can be argued to give rise to grammar competition. As Kroch (2001: 722f.) points out, grammar competition can be viewed as a kind of bilingualism with the learner acquiring two grammatical systems. In the case of bilingualism, the acquisition of the two grammars obviously does not have to coincide. One grammar can





Eric Haeberli

be acquired earlier than the other. The same thing could be argued to happen in cases where grammar competition emerges. The child might acquire a certain syntactic structure first on the basis of the syntactic evidence. But once the relevant morphological distinctions are acquired, the syntactic structure can be reassessed, and if the syntactic structure contains elements that are not required by the morphology a second, more economical grammar is postulated. Thus, I propose that even if the morphological cues and the syntactic cues differ in terms of salience to the language learner, they are both ultimately taken into account. If the two types of evidence match, a single grammar is acquired. If there is a mismatch, grammar competition emerges which, over time, leads to the elimination of the mismatch. In short, we obtain the result we were looking for at the beginning of this subsection: An account of how a morphological change gives rise to a delayed syntactic change. . ‘XP-subject’ in Mainland Scandinavian revisited Let us now return to the issue of ‘XP-subject’ orders in Mainland Scandinavian. Given the observations made in the previous subsection, a possible scenario for the loss of ‘XP-subject’ in Danish looks as follows: Stage (i): Rich agreement and Grammar A in which Agr is syntactically represented. ‘XP-subject’ orders are licensed due to the presence of AgrP. Stage (ii): Loss of rich agreement. Stage (iii): Competition between the old Grammar A and a new Grammar B in which Agr is absent. Grammar A can still derive ‘XP-subject’ orders if we assume that abstract (i.e. morphologically redundant) Agr may remain syntactically weak in the sense of not overtly attracting the subject to its specifier (cf. Haeberli 2002a: 239).10 Grammar B, however, rules ‘XPsubject’ orders out, as shown in (6) above. Possible evidence for grammar competition would be a gradual decrease in the frequency of ‘XP-subject’ orders, a decrease in the frequency of potentially related phenomena,11 and variation among speakers (see discussion below). Stage (iv): Loss of Grammar A, possibly after several centuries of competition. Given this diachronic analysis of Danish, we can now reconsider the problem raised by modern Swedish and Norwegian, i.e. the occurrence of ‘XP-subject’ orders despite the absence of rich agreement (see example 8). What I propose is that Swedish and Norwegian are still at stage (iii) of the changes that Danish has undergone. The grammar with syntactically represented Agr and ‘XP-subject’ orders is thus in a state of decline, but, contrary to Danish, Swedish


and Norwegian have not reached stage (iv) of the diachronic development yet. As discussed in Haeberli (2002a: 242–245), the conclusion that Swedish and Norwegian are in a diachronically unstable state may be supported by the fact that there is considerable variation among speakers with respect to the acceptability of ‘XP-subject’ orders. In Swedish, at least three types of speakers can be identified. The first one accepts ‘XP-subject’ orders in main and subordinate clauses. The second one has the same judgements, but requires some degree of contrastive focus on the subject (Holmberg 1993). Finally, for the third type of speakers (as a matter of fact the large majority consulted), ‘XP-subject’ is possible in main clauses, but virtually impossible in subordinate clauses. In Norwegian, we find similar variation. Some speakers accept ‘XP-subject’ in both main clauses and subordinate clauses, some others reject this order in subordinate clauses, and others reject it in subordinate clauses and find it highly marginal even in main clauses. In terms of the proposals made here, the variation among speakers of Swedish and Norwegian can be interpreted as evidence for the decline of the grammar in which Agr is syntactically represented, with some speakers licensing additional structure related to Agr only in main clauses but not in subordinate clauses. . Summary Let us now sum up the main points of this section. The variation among the West Germanic languages with respect to ‘XP-subject’ orders suggests that the richness of verbal agreement morphology plays a role in the licensing of this word order. However, the Mainland Scandinavian languages seem to raise a problem for this generalization because Swedish and Norwegian show ‘XPsubject’ orders despite the absence of agreement morphology. So from a purely synchronic perspective, the Mainland Scandinavian languages would look fatal for the generalization relating ‘XP-subject’ to rich agreement. However, once we assume that the relation between the syntactic phenomenon and the morphological property is not a direct one but mediated by a syntactic feature and the structure it projects, and once we take diachronic issues into account, the generalization does not have to be abandoned. A morphological loss does not immediately give rise to the loss of a related syntactic feature/structure because the syntactic evidence for this feature/structure remains available. Instead, the morphological change introduces a competing grammar which then drives the old grammar out, possibly after a considerable time span of grammar competition. On the basis of this observation and on the basis of the diachronic developments in Danish, I have argued that the situation in Swedish and Nor-





Eric Haeberli

wegian with respect to ‘XP-subject’ orders actually reflects such a stage of grammar competition during which there may be a mismatch between the morphology and the relevant syntactic phenomenon. Swedish and Norwegian therefore do not provide conclusive counterevidence against the intuition behind the analysis presented for the West Germanic languages in Section 3.1. From a diachronic point of view, apparently problematic cases for a synchronic generalization concerning the relationship between syntax and morphology are thus not unexpected.12

. Word order freedom and case morphology in the history of English In this section, I will briefly consider another syntactic domain where a correlation between inflectional morphology and a syntactic property has been proposed in the literature, and I will argue that an examination of the diachronic developments in English in this domain provides some support for the conclusions reached in the previous section. The relevant domain is the order of arguments. According to a long-standing observation, rich case morphology on nominal constituents tends to go together with freedom of argument order. Furthermore, this link between morphology and syntax has also been observed in the context of diachronic change. Sapir (1921: 168) for example refers to “the drift toward abolition of most case distinctions and the correlative drift toward position as an all-important grammatical method”. As proposed by Haeberli (2001, 2002a) and Weerman (1997), the relation between case morphology and argument order can be obtained along the lines of the schema in (10b) above (i.e.: morphology ↔ syntactic feature/syntactic structure ↔ syntactic phenomenon). The link between the two phenomena is not a direct one, but it is mediated by a syntactic feature or the syntactic structure. This result can be obtained by assuming, in contrast to standard generative syntactic theory, that case is not a universal property. Instead, case has to be syntactically represented in languages with a rich case morphology, but it can remain absent in other languages. In Haeberli (2002a), this approach is implemented in terms of the following basic proposals.13 i.

Phenomena traditionally related to the concept of abstract Case are captured in terms of the interaction of categorial feature matrices. ii. Case features are syntactically represented in languages with rich case morphology, where richness of case morphology is defined as a case system with more than a two-way (default/non-default) distinction.14


iii. Case features on verbal heads (V, T) are uninterpretable at LF and PF and therefore have to be checked. They project independent projections in order to establish a checking relation and the order of these projections is variable. This variation gives rise to argument order freedom. Within this framework, the presence of a rich morphological case system generally entails argument order variation.15 But as in Section 3, the question arises as to what the predictions of this approach are for languages that lack rich case morphology. Would a language exhibiting argument order variation despite the absence of rich case morphology undermine the generalization linking syntax to morphology? As we will see, a stage in the history of English may indeed have had the properties of such a language. But, by analogy to the argument presented in Section 3, I will propose that such a stage is not problematic for the traditional correlation between morphology and syntax. It only looks problematic from a purely synchronic perspective, but can be explained once we consider diachronic issues. As is well known, Old English allows the two DP objects of a ditransitive verb to occur in both possible orders (example from Koopman 1994: 104/5) (13) a.

þæt he [ðam adligan menn] [his hæle] forgeafe that he the sick man (DAT) his health (ACC) granted ‘that he would grant the sick man his health’ (ÆLS (Swithun) 120) b. he ageaf [þone cnapan] cucenne [his meder] he returned the boy (ACC) alive his mother (DAT) ‘he returned the boy alive to his mother.’ (ÆLS (Martin) 1027)

In (13a) the indirect object precedes the direct object whereas in (13b) the order of the two objects is inverted. Koopman’s (1994) statistical evidence suggests that the two orders are almost equally frequent in Old English. In his corpus consisting of nearly 2000 clauses with two full DP objects, Koopman found frequencies of 54% for the order ‘indirect object-direct object’ and of 46% for the inverted order. Given the hypothesis that argument order freedom is related to the presence of rich case morphology, Koopman’s findings are not surprising. Old English had a rich morphological case system distinguishing between four (or possibly five) cases. In terms of (ii) above, case therefore has to be represented syntactically, and case feature checking then gives rise to argument order variation. But the rich case system starts eroding towards the end of the Old English period. Furthermore, within the 300 years after the Old English period, the inverted order ‘direct object-indirect object’ seems to be lost as well. Referring to the inverted order, Allen (1995: 419) observes: “By the late four-



 Eric Haeberli

teenth century, the construction seems to have disappeared completely; I have found no convincing examples with two nominals in texts written after 1340.” Given these general diachronic developments in early English, let us consider the loss of object order variation and its connection to the loss of case morphology a bit more closely. Unfortunately, the amount of data available for the crucial period is very limited, but some observations can nevertheless be made. Polo (2002) examines the First and the Second Continuation of the Peterborough Chronicle. For the First Continuation (1070–1121), she finds that the case morphology on full DPs is already lost to a large extent but ‘direct object-indirect object’ orders with full DP objects still occur. However, there may still be some evidence for a rich morphological case system as defined under (ii). 3rd person pronouns in the First Continuation show a three-way distinction Nominative/Accusative/Dative.16 Case features may therefore still be required to make these distinctions at this stage. However, in the Second Continuation, Accusative and Dative have become identical even with 3rd person pronouns. In terms of the criterion in (ii) above, case would therefore not have to be represented syntactically any more because what is left is a simple default/non-default distinction. Nevertheless, the order ‘direct object-indirect object’ can still be found in the Second Continuation. Yet, the quantitative evidence in support of this observation is very slim unfortunately. In the entire Second Continuation, there is only one example involving two DP objects in a double object construction and the order of the arguments in this example is ‘direct object-indirect object’ (Polo 2002: 138). Even though we have to treat this conclusion with some caution in view of the small amount of data, Polo’s finding nevertheless suggests that the Second Continuation of the Peterborough Chronicle may represent a grammar in which object inversion occurs despite the absence of rich case morphology. Some further indications concerning the development of the loss of object inversion and the loss of case morphology can be found in Allen (1995). With respect to some 13th century Early Middle English texts, Allen makes the following observations. First, in the Southeast Midlands dialect (represented by Vices and Virtues in Allen’s corpus) the change has not made much progress yet because “the old case marking system is fairly well preserved” (1995: 185) and both orders of objects can still be found in double object constructions. A bit more interesting is the situation in the West Midlands dialect (represented by Ancrene Wisse and the Katherine Group in Allen’s study). The two objects in a double object construction are not formally distinguished any more, but their order is not fixed yet (Allen 1995: 182–183). However, there are some suggestions that the case system is not impoverished enough to make syntactically


represented case features entirely redundant from a morphological point of view. For example, the Accusative/Dative distinction in the pronominal system is still occasionally made, but only with 3rd person singular masculine pronouns and only very rarely (1995: 183). Allen suggests that these remaining instances may simply be due to a scribe who copied the work of someone whose grammar still did make the pronominal Accusative/Dative distinction productively.17 Even if we discard this evidence, there still remain two additional traces of a morphological case system in the West Midlands dialect (Allen 1995: 182/4). First, the old Dative ending -e still occurs fairly frequently in the West Midland texts. But its use has been restricted considerably and nouns inflected with -e can mainly be found as objects of prepositions but also, more rarely, as complements of adjectives. And secondly, although such constructions seem to be rare, verbal complements still can bear Genitive sometimes. The occasional occurrence of the Dative and Genitive inflection on nouns together with the pronominal Nominative/Accusative distinction with subjects and objects would require the presence of syntactically represented case features in terms of condition (ii) above. Variation in the order of objects would then be a consequence of this case feature system. Of course, if an earlier manuscript may have had an influence on the use of case distinctions with 3rd person singular masculine pronouns, as suggested by Allen, then similar interference may also have played a role for the occurrence of the -e Dative and of verbal complements in the Genitive. However, on the basis of the textual evidence available, we cannot safely conclude that a syntactically represented case feature system is not required by the case morphology in the West Midlands dialect even though the case system is certainly considerably weakened. The next set of data Allen (1995) considers is from the early 14th century. At this stage, the morphological case system seems to have disappeared entirely, and we are left with the pronominal Nominative/Accusative distinction familiar from modern English. Hence, in terms of condition (ii) above, the morphology does not require syntactically represented case features any more because the case system has been reduced to a two-way (default/non-default) distinction. Variation in the order of two object DPs can nevertheless still be found in Allen’s data. However, a potential problem that arises here is the type of data that is considered. Given the scarcity of sources from the late 13th/early 14th century, Allen includes verse texts in her sample, and verse data are potentially less reliable sources for determining syntactic properties. Nevertheless, some of Allen’s data are quite suggestive. Thus, she observes (1995: 418) that in The Metrical Chronicle of Robert Gloucester 9 ‘direct object-indirect object’ orders occur and 33 ‘indirect object-direct object’ orders (21.4% vs. 78.6%). A



 Eric Haeberli

frequency of more than 20% for object inversion seems to be rather high if this word order were simply a stylistic poetic device. Thus, it may not be implausible to conclude that object inversion was possible in early 14th century Middle English despite the loss of the morphological case system. However, the total decline of this word order was imminent since, as pointed out earlier, the last attested example of the order ‘direct object-indirect object’ that Allen has been able to identify dates from 1340. In summary, it is relatively difficult to obtain a complete and precise picture of how object order variation and rich case morphology were lost in the history of English. This is to a large extent due to the very limited textual evidence that is available from the crucial period. An additional complicating factor is the fact that there is dialect variation and that we do not have an uninterrupted sequence of texts from a single dialect. Nevertheless, Polo’s (2002) work on the Second Continuation of the Peterborough Chronicle and Allen’s study of early 14th century texts suggest that there may have been a period (or different periods in different dialects) in the history of English when variable argument order was still available even though the relevant morphological case distinctions had been lost. If we looked at such a stage from a purely synchronic point of view, we would have to conclude that the generalization relating the syntactic phenomenon to a morphological property cannot be maintained because variable argument order occurs independently of rich case morphology. However, once diachronic aspects are taken into account, we can explain the occurrence of such a stage without abandoning the generalization. Object inversion after the loss of rich morphological case is the manifestation of a transitional period. This transitional period can again be analyzed in terms of competing grammars along the lines discussed in the previous section. The competition is schematically represented in (12) above, repeated here as (14).18 (14) Grammar A: morphology ↔ syntactic feature/syntactic structure ↔ syntactic phenomenon Grammar B: morphology ↔ syntactic feature/syntactic structure ↔ syntactic phenomenon

In the context discussed in this section, Grammar A has case features that are syntactically represented. The presence of these case features is not motivated by morphological evidence any more but by syntactic phenomena, and more particularly by argument order variation. But the loss of case morphology gives

Syntactic effects of inflectional morphology 

rise to an alternative, more economical Grammar B in which case is not syntactically represented any more. Over time, the case-less grammar wins out, and argument order becomes entirely fixed. In summary, the points made above provide some additional support to the conclusions reached in the previous section. The observation that rich case morphology and argument order variation are related phenomena is a traditional one in the literature. Although this correlation generally seems to be correct, an examination of the developments in the history of English suggests that there may be a transitional period during which argument order variation is maintained even after the loss of a rich morphological case system. A synchronic study of this transitional period would shed doubts on the adequacy of the generalization. But once we take diachronic issues into account, the apparent counterexample to the generalization can be explained.

. Conclusion Work within generative grammar has identified various parameters that account for syntactic variation across languages. For a large number of these parameters, the cross-linguistic variation seems to be fairly random in the sense that it remains unexplained why one language sets the parameter in one way and why another language sets it the other way. However, for some areas of the grammar, attempts have been made to provide systematic explanations for syntactic variation by relating syntactic properties to aspects of the inflectional morphology. More precisely, proposals have been of the type that if a language has a certain morphological property M it also has the syntactic phenomenon S. The natural extension of this is then to say that if a language does not have M it also does not have S. Although such correlations seem attractive from the point of view of systematically accounting for syntactic variation, certain counterexamples seem to undermine their validity. In particular, it has often been observed that the absence of M does not always entail the absence of S. On the basis of two case studies, the variation with respect to ‘XP-subject’ orders in the Germanic languages and argument order variation in the history of English, I have argued in this paper that apparent counterexamples to generalizations relating morphology and syntax do not necessarily force us to abandon the basic intuition behind the generalization. To obtain this result, a first important assumption to be made is that the link between M and S is not a direct one but that it is mediated by syntactic features or syntactic structure that license S. This assumption is also an important aspect of the analyses

 Eric Haeberli

developed in Bobaljik (2002), Bobaljik & Thráinsson (1998) and Thráinsson (1996). In terms of such an approach, the center stage is taken by the syntactic feature/structure, and both M and S motivate the presence of this syntactic feature/structure. Once M is lost, S can still provide evidence to the language learner for the presence of the syntactic feature/structure, and the absence of M therefore does not necessarily entail the absence of S. Bobaljik (2002) concludes from this that relations between morphology and syntax are one-way entailments. If M is present, it entails the presence of syntactic structure which then licenses the occurrence of S. As for the case when M is absent, no synchronic predictions are made. S could be attested or not. Nevertheless, there seems to be a tendency for the related entailment (i.e. if M is absent, S is absent) to hold and for the loss of M to be followed by the loss of S. Thus, a generalization might be lost if the relation between morphology and syntax were limited to an entailment involving the presence of M and if nothing was said about the absence of M. In this paper, I have proposed that once we take a closer look at the mechanisms of diachronic change we can indeed include the absence of M in our picture of the correlation between morphology and syntax. More precisely, what I have proposed is that the loss of M does have a syntactic consequence in that it introduces a new, more economical grammar in which the syntactic feature/structure related to M is eliminated. This grammar enters into competition with the old grammar in which the syntactic feature/structure is still maintained due to the evidence provided by S. Thus, the change caused by the loss of M is not a dramatic one from the point of view of the surface syntax. But after a period of competition which can be of considerable duration, the new grammar drives the old grammar out, as in other cases of grammar competition discussed in the literature. The result is the complete loss of S. In the two case studies discussed, I have related one syntactic phenomenon (‘XP-subject’ orders) to rich verbal agreement morphology and a second syntactic phenomenon (object order variation in double object constructions) to rich case morphology. In both cases, there is evidence that the syntactic phenomenon can occur despite the absence of the relevant rich inflectional morphology. These counterexamples look problematic from a purely synchronic point of view. However, once we consider them from a diachronic point of view, we can show that they do not invalidate the basic generalizations. Mismatches between the syntax and the morphology can be analyzed as periods of transition in which two grammars are in competition. The generalizations discussed here which relate a syntactic phenomenon to a morphological property can therefore be maintained, even in a rather strong form which makes

Syntactic effects of inflectional morphology 

reference to both the presence and the absence of the relevant morphological property.

Notes * I would like to thank the participants at the workshop on Diachronic Clues to Synchronic Syntax and the editors of this volume for helpful comments on earlier versions of this paper. . For recent discussions of the status of counterexamples to the correlation between verb movement and verbal agreement morphology, see Alexiadou and Fanselow (2002) and Bobaljik (2002). See also Warner (1997: 381–383) for an analysis of the diachrony of verb movement along very similar general lines as proposed below for other phenomena. . The intervening element in this example is an adjunct. German and Yiddish differ from Dutch and Frisian in that they also productively allow arguments in this position. However, this variation is not essential for the points made in this section and it will therefore be left aside. But see Haeberli (2002a: Chapters 3 and 4) for a discussion of how argument order variation can be integrated into the analysis presented in this section. . What follows is meant to provide a very brief overview of the main aspects of the analysis. For a more detailed discussion of these points, I refer the reader to the sources cited (in particular Haeberli 2002a: Chapter 4 for the most recent version of this approach). . The distinction between Agr as a feature and AgrP is made here (and below) since in Haeberli (2002a) Agr is analyzed as a feature on T which then projects an independent projection above TP to establish a checking relation. . The condition of co-occurrence of inflectional morphemes is based on the assumption that inflectional morphemes correspond to inflectional heads in the syntax. If a tense and an agreement morpheme co-occur, two inflectional heads must be available for these morphemes to be able to be inserted in the morphology. As for the second condition, which refers to distinctions going beyond a two-way agreement distinction, it has the effect of triggering the presence of AgrP even when there are no forms in which agreement and tense morphology co-occur. This proposal is based on the assumption that the syntax has to provide the morphology with the relevant information for spelling out the different inflectional heads. If agreement is a simple two-way distinction as in English (Ø vs. -s), T can simply be marked as non-default in a specific structural context (i.e. with a 3sg subject in its specifier if T is [+present]) and then spelled out as non-default -s in the morphology. However, if there is a third form, a simple default/non-default distinction is not sufficient any more. Instead, I propose that the relevant distinctions are made in terms of agreement features which have to be checked in the syntax (giving rise to an AgrP) and which then determine the spell-out form of the verb at PF. For a more detailed discussion of these points cf. Haeberli (2002a: 209ff.). . The licensing of empty expletives has generally also been related to richness of agreement morphology (cf. e.g. Platzack 1987). The behavior of West Flemish therefore suggests that richness of agreement for the purposes discussed in the text (i.e. triggering the presence of AgrP) cannot be defined in the same way as for the purposes of the licensing of empty

 Eric Haeberli

expletives. For some speculations on what determines the absence of empty expletives in West Flemish and on what the relevant morphological properties for the licensing of empty expletives are, see Haeberli (2002b: 99ff.). . As suggested by the bi-directional arrows in (10b), the links between the different properties are mutual. The syntactic feature and the structure projected by this feature license a certain syntactic phenomenon, and this syntactic phenomenon then is a manifestation of the presence of the syntactic feature/structure. Similarly, the morphology triggers the presence of a certain syntactic feature/structure, and the syntactic structure also has an influence on the morphology in the sense that for example a simple IP-structure makes the co-occurrence of two inflectional morphemes impossible in the morphology. . As Bobaljik (2002) points out, the only strong prediction that something like (10b) makes concerns the case when the relevant morphological property is present. If it is (e.g. rich agreement), a syntactic structure which does not reflect the morphology (e.g. an unsplit IP) is impossible. But the absence of morphology does not necessarily entail the absence of the related syntactic structure. . Note however that things are a bit more complex than suggested in the text. When looking at relations of the type shown in (10b), we may not be dealing with a single syntactic phenomenon that is related to a morphological property and a certain syntactic structure, but with two or more, and each of them would contribute some weight to the syntactic evidence. For example, V-to-I movement may be related to ‘XP-subject’ orders in Germanic (cf. Haeberli 2002a: 284). Yet, V-to-I is not as salient in Germanic as in French. If we assume that all V2 clauses involve V-to-C, including subject-initial ones (cf. e.g. Vikner 1995), there is actually no evidence for independently triggered V-to-I in Germanic main clauses. And even in subordinate clauses, the evidence can be extremely limited (only embedded questions in Icelandic) or unavailable (in Yiddish) (cf. e.g. Bobaljik & Thráinsson 1998: 45, 48ff. for discussion). So V-to-I may not (or at least not always) contribute a frequent cue for the presence of AgrP in Germanic. Bobaljik and Thráinsson (1998) also relate object shift and transitive expletive constructions to the presence of AgrP. However, in terms of the approach developed in this section, where the presence of AgrP derives ‘XP-subject’ orders, object shift and transitive expletive constructions can only be indirectly related to AgrP (cf. Haeberli 2002a: 280ff. for discussion). Nevertheless, there may be other syntactic properties that are related to Agr in this framework, as discussed in Haeberli (2002a). For example, Agr plays an important role for deriving Nominative-Dative inversion in Dutch (2002a: 225ff.) and it allows the promotion of both objects of a ditransitive verb to the status of subject in Swedish and Norwegian (2002a: 239ff.). However, I will not pursue these issues any further here. As we will see below, an exact measure to determine the importance of the syntactic and the morphological cues may not be required for a satisfactory analysis of the issues raised in this section. . Following e.g. Bobaljik (1995) and Groat and O’Neil (1996), I assume here that this option involves movement of the subject to AgrP with its phonological features staying behind in [Spec, TP]. The analysis proposed for German, Dutch, Frisian and Yiddish in Section 3.1 (i.e. ‘XP-subject’ as the result of the insertion of an empty expletive in [Spec, AgrP]), would presumably not be available at this stage given that, as pointed out in Note 6, the licensing of empty expletives has also been related to richness of agreement morphology.

Syntactic effects of inflectional morphology  . Based on evidence from modern Swedish and Norwegian, Haeberli (2002a: 239ff.) proposes that additional phenomena that depend on the presence of AgrP are the promotion of both objects to subjecthood in passivized ditransitives and long object shift. . It should be pointed out that Icelandic raises an additional potential problem for the analysis of the ‘XP-subject’ variation presented in this section. The problem is the opposite of what we have seen for Swedish/Norwegian. Icelandic has rich verbal agreement morphology but no ‘XP-subject’ orders. At first sight, this seems problematic for the correlation between agreement morphology and ‘XP-subject’ orders. According to (10b) and assumption (iv) in Section 3.1, Agr would have to be represented in Icelandic. Given furthermore that empty expletives are licensed in Icelandic, we would expect that ‘XP-subject’ orders can be derived as in German, Dutch, Frisian and Yiddish (see the discussion of representation 5 in Section 3.1). An analysis of this unexpected situation in Icelandic is presented in Haeberli (2002a: Chapter 6). The main proposal there is that Agr is indeed syntactically represented in Icelandic, but in a more complex way than in West Germanic. Whereas West Germanic has a single subject agreement feature, Icelandic distinguishes between person and number agreement. It is proposed that this distinction interferes with the derivation of ‘XP-subject orders’ and also contributes to the occurrence of other unexpected syntactic properties of Icelandic. . For a more detailed discussion of these points see Haeberli (2002a), in particular Chapter 2 for point (i) and Chapter 3 for points (ii) and (iii). . The motivation for this requirement is as given above for the syntactic representation of agreement (see Note 5). In the case of a two-way distinction, one form can simply be marked as non-default and a structural context can be defined in which the non-default form is spelled out. If more than two forms are distinguished, a simple default/non-default distinction is not sufficient, and case features are introduced to make the relevant distinctions (see Haeberli 2002a: 144ff., 181ff. for discussion). . As in the case of ‘XP-subject’ orders (cf. Note 12), Icelandic raises a potential problem for this generalization in that it has rich case morphology and therefore licenses case features, but argument order variation is restricted to object inversion. An analysis of this property of Icelandic can be found in Haeberli (2002a: Chapter 6) where it is argued that the restrictions on free argument order are part of a cluster of properties including oblique subjecthood, the absence of ‘XP-subject’ orders and a strict definiteness effect and that these properties are all underlyingly related. . I am leaving possessives aside in the discussion of the decline of the case system in English. Possessive pronouns may best be analyzed as determiners rather than case-bearing elements. As for full DP possessors, I assume that they also do not provide evidence for maintaining a syntactically represented case system. For example, modern English possessive ‘s could be argued to be the manifestation of a licensing element rather than an instance of a case on the possessor in the theoretical system outlined in (i) to (iii) above (cf. Haeberli 2002a: 74, Fn. 49). . This point obviously raises an important issue here, namely the difficulty of reliably dating the loss of morphological distinctions on the basis of written sources. Apart from scribes copying the work of other scribes, there is the more general possibility that distinctions made in the spelling simply reflect spelling conventions rather than the actual spoken

 Eric Haeberli

language of that time. But of course, the question whether written sources are good representations of the spoken language at a given point in time also arises when we consider syntactic constructions. So if there is a delay until a completed morphological change appears clearly in the written records, there may very well also be a delay in the related syntactic change. But whether the delays are of equal length would obviously remain uncertain. Any conclusion concerning the chronology of a morphological change and a related syntactic change therefore often has to remain rather tentative. . If the West Midlands text discussed above indeed reflects the grammar of a speaker of Early Middle English, it might add another potential aspect of competition to this picture. In a transitional period, speakers might vary as to whether they use case morphemes in certain contexts or not. In particular, the use of the -e Dative, which gives rise to a third distinction apart from the pronominal Nominative/Accusative distinction, does not seem to have been systematic in the Early Middle English West Midlands dialect. Thus, while in (14) the relevant morphological property is represented as lost, there could be a slightly earlier stage in which there is also competition in Grammar A with respect to the presence of the morphological property, i.e. a competition leading to inconsistent use of inflectional endings.

References Alexiadou, A. & G. Fanselow. (2002). “On the correlation between morphology and syntax: The case of V-to-I.” In J.-W. Zwart & W. Abraham (Eds.), Studies in Comparative Germanic Syntax (pp. 219–242). Amsterdam: John Benjamins. Allen, C. (1995). Case Marking and Reanalysis. Grammatical Relations from Old to Early Modern English. Oxford: Clarendon Press. Bobaljik, J. (1995). “Morphosyntax: The syntax of verbal inflection.” Doctoral dissertation, MIT. Bobaljik, J. (2002). “Realizing Germanic inflection: Why morphology does not drive syntax.” The Journal of Comparative Germanic Linguistics, 6, 129–167. Bobaljik, J. & H. Thráinsson (1998). “Two heads aren’t always better than one.” Syntax, 1, 37–71. Cheng, L. (1991). On the Typology of Wh-Questions. Doctoral dissertation, MIT. Chomsky, N. (1981). Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. (1986). Knowledge of Language: Its Nature, Origin and Use. New York: Praeger. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT Press. Diderichsen, P. (1962). Elementær Dansk Grammatik. 3rd edition. Copenhagen: Gyldendal. Ellegård, A. (1953). The Auxiliary Do: The Establishment and Regulation of its Use in English. Stockholm: Almqvist and Wiksell. Emonds, J. (1978). “The complex V-V’ in French.” Linguistic Inquiry, 9, 151–175. Grewendorf, G. (2001). “Multiple wh-fronting.” Linguistic Inquiry, 32, 87–122. Groat, E. & J. O’Neil (1996). “Spell-out at the LF interface.” In W. Abraham, S. D. Epstein, H. Thráinsson, & J.-W. Zwart (Eds.), Minimal Ideas. Syntactic Studies in the Minimalist Framework (pp. 113–139). Amsterdam/Philadelphia: John Benjamins.

Syntactic effects of inflectional morphology 

Haeberli, E. (1999). “On the word order ‘XP-subject’ in the Germanic languages.” The Journal of Comparative Germanic Linguistics, 3, 1–36. Haeberli, E. (2000). “Adjuncts and the syntax of subjects in Old and Middle English.” In S. Pintzuk, G. Tsoulas, & A. Warner (Eds.), Diachronic Syntax: Models and Mechanisms (pp. 109–131). Oxford: Oxford University Press. Haeberli, E. (2001). “Deriving syntactic effects of morphological case by eliminating abstract case.” Lingua, 111, 279–313. Haeberli, E. (2002a). Features, Categories and the Syntax of A-Positions. Cross-Linguistic Variation in the Germanic Languages. Dordrecht: Kluwer. Haeberli, E. (2002b). “Inflectional morphology and the loss of verb second in English.” In D. Lightfoot (Ed.), Syntactic Effects of Morphological Change (pp. 88–106). Oxford: Oxford University Press. Holmberg, A. (1993). “Two subject positions in IP in Mainland Scandinavian.” Working Papers in Scandinavian Syntax, 52, 29–41. Holmberg, A. & C. Platzack (1995). The Role of Inflection in Scandinavian Syntax. Oxford: Oxford University Press. Koopman, W. (1994). “The order of dative and accusative objects in Old English.” Ms., University of Amsterdam. Kroch, A. (1989). “Reflexes of grammar in patterns of language change.” Language Variation and Change, 1, 199–244. Kroch, A. (2001). “Syntactic change.” In M. Baltin & C. Collins (Eds.), The Handbook of Contemporary Syntactic Theory (pp. 699–729). Oxford: Blackwell. Lightfoot, D. (1999). The Development of Language. Acquisition, Change and Evolution. Oxford: Blackwell. Mikkelsen, K. (1911). Dansk Ordföjningslære. Copenhagen: Lehmann and Stage. Platzack, C. (1987). “The Scandinavian languages and the Null-Subject Parameter.” Natural Language and Linguistic Theory, 5, 377–401. Platzack, C. & A. Holmberg (1989). “The role of AGR and finiteness.” Working Papers in Scandinavian Syntax, 43, 51–76. Polo, C. (2002). “Double objects and morphological triggers for syntactic case.” In D. Lightfoot (Ed.), Syntactic Effects of Morphological Change (pp. 124–142). Oxford: Oxford University Press. Pollock, J.-Y. (1989). “Verb movement, UG and the structure of IP.” Linguistic Inquiry, 20, 365–424. Rizzi, L. (1982). Issues in Italian Syntax. Dordrecht: Foris. Roberts, I. (1985). “Agreement parameters and the development of English modal auxiliaries.” Natural Language and Linguistic Theory, 3, 21–58. Rohrbacher, B. (1994). “The Germanic languages and the full paradigm: A theory of V-to-I raising.” Doctoral dissertation, University of Massachusetts, Amherst. Rudin, C. (1988). “On multiple questions and multiple wh-fronting.” Natural Language and Linguistic Theory, 6, 445–501. Santorini, B. (1993). “The rate of phrase structure change in the history of Yiddish.” Language Variation and Change, 5, 257–283. Sapir, E. (1921). Language: An Introduction to the Study of Speech. New York: Harcourt, Brace and World.

 Eric Haeberli

Thráinsson, H. (1996). “On the (non-)universality of functional categories.” In W. Abraham, S. D. Epstein, H. Thráinsson, & J.-W. Zwart (Eds.), Minimal Ideas. Syntactic Studies in the Minimalist Framework (pp. 253–281). Amsterdam: John Benjamins. Vikner, S. (1995). Verb Movement and Expletive Subjects in the Germanic Languages. Oxford: Oxford University Press. Vikner, S. (1997). “V-to-I movement and inflection for person in all tenses.” In L. Haegeman (Ed.), The New Comparative Syntax (pp. 189–213). London: Longman. Warner, A. (1997). “The structure of parametric change, and V-movement in the history of English.” In A. van Kemenade & N. Vincent (Eds.), Parameters of Morphosyntactic Change (pp. 380–393). Cambridge: Cambridge University Press. Weerman, F. (1997). “On the relation between morphological and syntactic case.” In A. van Kemenade & N. Vincent (Eds.), Parameters of Morphosyntactic Change (pp. 427–459). Cambridge: Cambridge University Press.

Language change versus grammar change What diachronic data reveal about the distinction between core grammar and periphery* Roland Hinterhölzl Humboldt-Universität zu Berlin

The paper focuses on two questions. A) How can we explain that language change often proceeds in a slow, gradual fashion, extending over long periods of time in which the old and the new pattern coexist side by side? This fact conflicts with the idea that grammar change via parameter resettings should give rise to abrupt changes. Alternatively, this paper proposes that the gradualness results from a change in language use which eventually may lead up to a change in grammar. More precisely, it is proposed that the grammar provides a limited amount of optionality in the form of stylistic or peripheral rules that can be exploited by speakers for their communicative purposes. These rules may affect word order and prosodic phrasing to derive information-structurally marked forms, which, over time, may loose their stylistic force and become reanalysed as (obligatory) rules of the core grammar. B) What is the nature of word order change? In a framework subscribing to the Universal Base Hypothesis (UBH), changes in word order cannot be relegated to changes in the head complement parameter. Based on the above conception of language change, this paper proposes a novel account of the well known OV-VO change in the history of English, in which word orders resulting from stylistic Light Predicate Raising are reanalysed as the result of obligatory VP-intraposition. Furthermore, it is claimed that a similar change (from OV+VO to OV) took place in the history in German, in which a stylistic rule that moved focussed DPs into a preverbal position led to a change in the syntax/phonology interface, affecting the possible unmarked word orders in the language.

 Roland Hinterhölzl

.

Introduction

Diachronic processes are often characterized by a slow gradual change that eventually is followed by an abrupt change that marks the new parameter settings. Typically, before the innovative pattern replaces the old pattern, there can be a relatively long period of time in which both patterns are used in the language. Examples of diachronic processes that proceed in this way actually abound. Take, for instance, the slow rise of VO-orders in the history of Icelandic. As Hróarsdóttir (1998) has shown, OV and VO orders coexisted in Icelandic for many centuries. A similar case is the slow decline of V-XP orders in embedded clauses in the history of German. This process starts in the 9th century (cf. Näf 1979 who diagnoses a major decline of postverbal accusative objects in the writings of Notker), but the declining pattern is still available in the presentday language in restricted contexts. Another example that does not come from the field of word order is the slow rise of expletive subjects in the history of Swedish as it has been described by Falk (1993). Under the standard assumption that reanalysis and parameter resettings give rise to abrupt change (cf. Lightfoot 1991, 1999), the period of gradual change must be viewed as a period of language change without a change in grammar which sets the stage for the abrupt change instantiating the grammar change proper. This scenario, plausible from an empirical point of view, raises the question of how language change is possible without a change in grammar. One approach that pays attention to the gradualness of change and the parallel availability of two (or more) alternative forms is the concept of grammar competition (cf. Kroch 1989; Pintzuk 1996). This approach was proposed to account for the development of English from an OV- to a VO-language, but clearly has applications beyond this particular historical process and beyond word order changes in general. An alternative would be to assume change in language use within one grammar. In order for adult speakers to be innovative in their language use within the limits of the grammar they acquired, we need to assume that a grammar defines a certain space of limited options that can be exploited by innovative speakers to achieve certain communicative effects in specific discourse situations. I propose that the area of limited optionality pertains to the field of stylistic form, more particularly to what has been called information packaging (cf. Vallduvi 1992), that is, to the arrangement of information according to features like new/old or prominent/non-prominent and the like. Stylistic or

Language change versus grammar change

information-structural rules (in short IS-rules) involve mainly alternation of word order and alternatives in prosodic phrasing. In the following sections, I will provide concrete examples from German and English of the type of options that are available in languages. To delimit the space of restricted options in grammar, I reinvoke the old distinction between the core grammar and the periphery and the notion of markedness as the distinguishing criterion (cf. Chomsky 1965). What is derived in the core grammar is unmarked and stylistically or information-structurally marked forms are derived by peripheral or stylistic rules. Having introduced the distinction between core rules and peripheral/stylistic rules in an intuitive way, the question arises whether the distinction can be made in more general terms. The standard notion of a stylistic rule so far has been that they apply postcyclically, after the core derivation (hence also the name of a peripheral rule), that is, at PF. As PF-rules, stylistic rules may change the word order and alter the prosodic form of the syntactic output. Thus, it seems that relegating these rules to the PF-component derives the relevant properties that we would like to ascribe to stylistic rules. However, a closer inspection reveals that this conception of “peripheral” rules is untenable. The case of stylistic fronting in Icelandic reveals that this rule cannot apply at PF, since it is subject to a syntactic restriction, namely that the subject position be empty (cf. Maling & Zaenen 1990). Since syntactic information cannot be taken to be available at the PF-interface, the effect can only be obtained if the rule applies in narrow syntax. This implies that socalled peripheral rules, a misnomer as it turns out, apply in the core derivation as well and that the distinction between core rules and stylistic rules should possibly be made on grounds of the type of features that they check respectively. In the remainder of the paper, I will argue on the basis of historical data that a peripheral/stylistic rule can be defined as a rule that leads to a marked prosodic output. The goal of this paper is to evaluate the two accounts, the one using grammar competition and the one employing the distinction between core and periphery, by investigating the change from OV to VO in English as well as the development of the German sentence bracket. The major claims of the paper are that a) we have to distinguish between the base order and the unmarked order in a language, b) language change can be caused by the rise of certain stylistic rules which when crossing a certain threshold in their use are reinterpreted as rules of the core grammar (cf. Lightfoot 1999) and c) the inspection of diachronic data provides important insights into the structure of synchronic grammars with respect to the division



 Roland Hinterhölzl

of labour between core and peripheral rules and the realization of word order variations.

. The change from OV to VO in English Let us first look at the change of word order that occurred in the transition between the Old English period (OE) and the Middle English period (ME). The standard answer to the question of what the pertinent change consists in is that there has been a change in the head complement parameter. This analysis is based on two assumptions: A) OE was an OV-language akin to modern Dutch or German and B) the base order changed from OV to VO. The first assumption can be shown to be plainly wrong (cf. 2.2) and the second assumption can be shown to be insufficient. To evaluate the second assumption, let us take a look at the differences between a so-called typical VO-language like (Modern) English and a typical OV-language like (Modern) German. . Differences between German and English In this section, I will discuss certain differences between German and English that cannot possibly be subsumed under the head complement parameter. A property of adverbs in VO languages that is in need of an explanation is the fact that adjuncts that can occur between the subject and the VP in VOlanguages are subject to restrictions absent in OV-languages. (1) a. John (more) often (* than Peter) read the book. b. Hans hat öfter (als der Peter) das Buch gelesen.

Descriptively speaking, the head of the adjunct must not have material to its right. This is only possible if the adjunct appears in sentence final position. An option, on the other hand, that is not available in OV-languages, as the contrast illustrated in (2) shows. (2) a. John read the book more often than Peter. b. *Hans hat das Buch gelesen öfter (als Peter).

From (1) and (2), I conclude that only light adverbs (of a restricted class that have been called adverbs of indefinite time by Ellegård 1953) can appear in the middle field in English. This property can of course not be derived by the head complement parameter. This distinction between English and German may seem to be just one (of many) stylistic difference between the two lan-


guages, but I will argue that it is part of a larger development that also affected the order of object and verb and can be given a uniform historical explanation. A property that correlates with the position of the object with respect to the verb and that has received little attention in this connection so far, is the position of event-related adverbs, that is, Time, Place and Manner adverbs. These adverbs occur preverbally in the order T>P>M in OV-languages but postverbally in the exact mirror image in VO-languages (cf. Haider 2000; Hinterhölzl 2002). This is illustrated in (3). (3) a. C T P M-V b. C V-M P T

OV-languages VO-languages

The properties of event-related adjuncts raise various interesting questions. First, their distribution within OV- and VO-languages raises the question of what makes exactly these adjunct types special such that their positioning, but not the positioning of, say, higher adverbs seems to be correlated with the head complement parameter. Secondly, given Cinque’s seminal work on adverbs, these adverbials are peculiar in several respects (cf. Cinque 1999). A) They appear to differ from adverb phrases (AdvPs) proper in not being rigidly ordered. B) In contrast to AdvPs proper, they can be interchangeably in the scope of each other, as is illustrated in (4). C) They differ from AdvPs proper in being typically realised in the form of PPs or bare NPs. And D) Scope may go from right to left (cf. (5a)), but binding only from left to right as is illustrated by the contrast in (5bc). In Hinterhölzl (2002), I argue that properties A, B and D follow from property C and the assumption that English – contrary to the standard view – has preserved scrambling of the Dutch type. In particular, I have argued that they are base-generated in an unmarked order but that, due to their nature of being typically realized as PPs, they can scramble to satisfy their scopal requirements and their binding properties. (4) a. They met students everyday of the week in a different university. b. They met students in each university on a different day. (5) a. John met Mary in a (different) park every Sunday. b. *Sue met Mary in his house on everybody’s birthday. c. Sue met Mary on everybody’s birthday in his house.

Thirdly, event-related adverbs give rise to Pesetsky’s paradox. The standard account of postverbal adverbs in VO-languages was given in terms of layered adjunction to the VP on the right, as is illustrated in (6). Right-adjunction structures, either base-generated or derived by movement, are incompatible



 Roland Hinterhölzl

with the Universal Base Hypothesis (Kayne 1994). Independently of the Universal Base Hypothesis, Larson (1988), Stroik (1990) and Pesetsky (1995) have argued that the standard approach to the syntax of adverbs is mistaken, since it fails to account for basic c-command relations between them and the complements of the verb. Typical c-command diagnostics, as NPI-licensing (7a) and quantifier-bound pronouns (7b), indicate that postverbal adjuncts are in the c-command domain of postverbal complements. (6) [IP SU [VP [VP V DO] Adjunct]] (7) a. John saw no student in any classroom. b. John met every girl on her birthday.

Since in the representation in (6) the direct object fails to c-command the postverbal adjunct, Larson (1988) proposed that event-related adverb(ial)s are part of a (multi-) layered VP-shell in which these elements are deeper embedded than the complements of the verb as is indicated in (8). (8) [VP SU V [VP DO tV Adjunct]]

In the Larsonian approach, event-related adverbs are analysed as a sort of (optional) complements in the VP. While this analysis neatly accounts for the c-command relations between postverbal complements and adjuncts, it fails to account for standard constituency tests such as VP-preposing and VP-ellipsis which show that verb and object form a constituent excluding postverbal adjuncts. This state of affairs is called Pesetsky’s paradox and led him into proposing a dual structure: a cascading (Larsonian) structure to account for the binding facts and a layered structure (parallel to the traditional analysis given in (6)) to account for the constituency facts. This state of affairs is highly unsatisfactory. It would be advantageous to settle for one basic underlying structure and derive the effects of the other structure via movement. In Hinterhölzl (2002), I have shown that the Larsonian approach is untenable. Here, I will briefly outline only the two decisive arguments. The first argument involves the semantic interpretation of event-related adverbs and the second argument pertains to the comparative dimension of accounting for the different distribution of these adverbs in OV- and VO-languages. The Larsonian approach to the syntax of event-related adjuncts raises questions about the proper interpretation of these elements. In a Larsonian shell, temporal adverbs are deeper embedded than manner adverbs, as is shown in (9). (9) a.

John wrote the letter carefully today.

Language change versus grammar change 

b. [VP John wrote [VP the letter tV [VP carefully tV today]]]

Following Ernst (1998), Haider (2000) and others, I assume that the attachment of adverbs is determined by their scopal properties. The scopal requirements of an adverb include selection for a clausal argument of a particular type. Ernst (1998) specifies a schema of abstract clausal entities relevant for the interpretation of adverb(ial)s. (10) Speech Act > Fact > Proposition > Event > Specified Event

From (10) it follows, for instance, that evaluative adverbs like unfortunately selecting for a fact cannot attach lower to the clausal skeleton than modal adverbs like probably selecting for a proposition, though they can otherwise occupy various positions in the clause, as is illustrated in (11). (11) a. (Unfortunately) Eddie (unfortunately) has (?unfortunately) left. b. *Probably Eddie unfortunately has left.

From a semantic point of view, manner adverbs specify an aspect of only a part of the event, namely the process component of the event, while temporal adverbs situate the entire event with respect to the speaking time.1 Thus, standard assumptions about the interaction of syntactic structure and semantic interpretation predict that temporal adverbs should attach to the clause higher, not lower as in the Larsonian approach, than manner adverbs. Secondly, here is what I call the comparative argument. If the English order is basic, then it is not clear how the German order is to be derived. A roll-up structure that moves a constituent containing the temporal adverb in front of the manner adverb and subsequently moves that larger constituent in front of the final position of the verb fails to account for the scopal properties of these adverbs in the middle field. In the German middle field an adverb always scopes over the adverb to its right. However, if a roll up analysis of this type is adopted, the temporal adverb still fails to c-command the manner adverb and therefore falls short of meeting the crucial condition for scope taking. A derivation in terms of movement of the adverbs by themselves raises several questions. First, the question arises why the hierarchy of adverbs in German is different from the hierarchy of adverbs in English, as is illustrated in (12). (12) 2.

1.

Secondly, the question arises what the motivation of adverb movement in German is. It is not clear why these adverbs are licensed in situ in English but

 Roland Hinterhölzl

have to undergo licensing movement in German. To the extent that we cannot find a satisfactory answer to this question, the Larsonian approach is rendered unattractive. Given the above considerations, in particular taking serious the semantic argument, I concluded in Hinterhölzl (2002) that the order of event-related adverbs observed in German, namely T>P>M, is closer to the base than the English order and proposed that the English order is derived from the German order via successive cyclic intraposition of verbal projections. In the following, I assume that manner adverbs are base-generated in the VP while Time and Place adverbs are base-generated above VP as is indicated in (13). (13) [Temp . . . [Loc . . . [ v [ Manner [ V DO ]]]]]

Under these assumptions, the English sentence in (14a) is derived from the base structure in (14b) via successive intraposition. In this derivation, first the VP containing the verb and the direct object moves in front of the locative PP and then the resulting structure is moved in front of the temporal PP, as is indicated in (14c). (14) a. John visited them in Vienna on Friday. b. [IP Johni [on Friday [in Vienna [VP ti visited them]]]] c. [IP John [[[visited them]i in Vienna ti ]j on Friday tj ]]

This account derives the VP-constituency facts from a base common to German and English. To account for the c-command effects, I have argued that English has preserved scrambling of the Dutch type, that is, movement of arguments across adjuncts, but spells out the lower copy, as is indicated by underlining in (7’). The derivation of (7b) is illustrated in (7’), where Pesetsky’s paradox is resolved in that LF interprets the higher copy in the middle field, while PF interprets the lower copy in the VP. (7’) a. John met every girl on her birthday. b. John [on her birthday [met every girl]] => base structure c. John [every girl [on her birthday [met every girl]] => scrambling (+ binding) d. John [[met every girl] [every girl [on her birthday [ tVP ]]]] => VP-intraposition

If this account of Modern English is correct, then the question arises how the order with postverbal adjuncts came about and why it is exactly event-related

Language change versus grammar change 

adverbs and not other adverbs that occur postverbally in English. It is this question that we turn to in the following section. . Word order and peripheral rules in the older stages of Germanic In this section, I will scrutinize the assumption that OE was an OV-language of the modern German type, which is basic to the description that there was a change in the head complement parameter. I will evaluate the evidence from OE against the broader picture provided by its sister languages Old High German (OHG) and Old Icelandic (OI). In the traditional literature, descriptive generalisations abound according to which the distribution of arguments and adjuncts is determined crucially by stylistic factors in the older stages in Germanic. One such statement is given in (15). (15) Light elements precede heavy elements in OE, OI and OHG. (Behaghel 1932: Das Gesetz der wachsenden Glieder)

In contrast to the traditional view, word order variation in OE and ME is recently often modelled in terms of competing parametric choices that give rise to the variety of word order patterns observed in these historical stages of English. One such approach is the so-called double base hypothesis (Pintzuk 1999) according to which word order variation follows from the co-existence of competing grammars that differ with respect to the head parameter of VP and IP. One of the central assumptions of the double base hypothesis which is based on the idea of grammar competition is that VO orders are an EMEinnovation that was brought about by language contact between Anglo-Saxons and the Scandinavian settlers in the 10th century. This thesis itself involves two assumptions which are questionable: A) OE was an OV-language and B) the grammar of the Scandinavian settlers was (already) VO. In the following section, I demonstrate that OE already contained a substantial amount of VO-structures. As to the second claim about Scandinavian, there is simply no evidence that the grammar of the Scandinavian settlers in England was predominantly VO. The oldest written Old Nordic documents are Icelandic texts from the 12th century. These documents show a preponderance of OV-orders.2 A recent corpus study of OI confirms the observations that have been made by traditional grammarians (cf. Hróarsdóttir 2002). A) There is variation between OV and VO structures and B) the distribution of syntactic elements is determined by stylistic/information-structural factors. Hróarsdóttir (2002) shows that the preverbal or postverbal occurrence of DP-arguments is determined by the in-

 Roland Hinterhölzl

teraction of two factors: the heaviness of the argument and its discourse status (whether it represents old or new information). A similar picture arises for OHG. My own studies of OHG texts warrant a strengthening of Behaghel’s observation. Given that background elements are usually light and focussed elements, by virtue of carrying stress, count as (prosodically) heavy, the hypothesis in (16) is worth investigating. (16) C background V focus

(16) expresses the hypothesis that in the older stages of English, German and Icelandic, which did not have a determiner system or were about to grammaticalize one, the discourse status of an argument was signalled by its positioning with respect to the verb. To conclude, it is unwarranted to assume that VO-orders came about in the history of English by language contact with Scandinavians, since we find both OV and VO orders in the older stages of German and Icelandic as well, suggesting that the variation in argument positioning in OE, OHG and OI is simply part of the common Germanic heritage. .. VO-features in Old English There is consensus among researchers that the predominant word order in the clause is verb-final in subordinate clauses and verb-second in main clauses. Thus OE has been analyzed as an OV-language akin to modern German or modern Dutch (cf. Kemenade 1987). In fact, OE does display a couple of other features typical of OV-languages, according to typological criteria (cf. Greenberg 1963; Hawkins 1983). Next to the finite verb appearing in final position in embedded clauses, it is a typical feature of OV-languages that verbal particles precede the verb and non-finite verbs precede the auxiliary. All three properties are found in high numbers in OE and are illustrated in (17). (17) a.

þæt ic þas boc of Ledenum gereorde to Engliscre spræce awende that I this book from Latin language to English tongue translate ‘that I translated this book from Latin to English’ (van Kemenade 1987: 16) b. þæt he his stefne up ahof that he his voice up raised ‘that he spoke up’ (Pintzuk 1991: 77)


c.

forþon of Breotone nædran on scippe lædde wæron because from Britain adders on ships brought were ‘because snakes were brought on the ships from Britain’ (Pintzuk 1991: 117)

However, OE also displays a number of VO-features. First, we find a considerable number of instances where the complement follows its selecting head in cases of Verb Raising (18a, b) and Verb Projection Raising (18c). These facts are not too surprising since we find the same kind of order in an undisputable OV-language like Dutch and its dialects, especially in West Flemish. (18) a.

þe æfre on gefeohte his hande wolde afylan who-ever in battle his hands wanted defile ‘whoever wanted to defile his hands in the battle’ (Pintzuk 1991: 102) b. & from Offan kyninge Hygebryht wæs gecoren and by King O. Henry was chosen ‘and Henry was chosen by king O.’ (Pintzuk 1991: 102) c. þæt nan man ne mihte a meniu geniman that no man could the multitude count ‘that no one could count the multidude’ (Pintzuk 1991: 33)

Secondly, we find a considerable number of extraposed PPs, CPs and DPs. While extraposed CPs and PPs are unproblematic, extraposed DPs are rather rare and marked in Modern Dutch and Modern German, though they did occur more frequently in older varieties of German.3 Pintzuk and Kroch (1989) show on the basis of a metrical analysis of Beowulf that these DPs receive stress. A case in point is given in (19a). (19b) shows the parallel structure in modern German, which is ungrammatical. (19) a.

þæt ænig mon atellan mæge [ealne one demm] that any man relate can all the misery ‘that no one can tell all the misery’ ? b. * dass jeder Mann erzählen kann all dieses Elend that every man relate can all this misery

(Pintzuk 1991: 36)

However, there is further evidence that points against a pure OV-character of OE. A) Verbal particles, though they cannot occur after a non-finite verb, can be moved along with the verb in V2-contexts, be stranded somewhere in the middle field and as such be followed by a DP (cf. (24) below). B) Small clauses can follow the finite verb in embedded clauses and can also follow the particle in V2 clauses, as is shown in (20a) and (20b), respectively. (20a’) and (20b’) show that the parallel structure is ungrammatical in modern German.



 Roland Hinterhölzl

(20) a.

forðam ðe he licettað hie unscyldige because that they pretended themselves innocent ‘therefore they pretended to be innocent’ (van Kemenade 1987: 35) a.’ *darum gaben sie aus sich selbst als unschuldig therefore gave they out themselves as innocent b. he ahof þæt cild up geedcucod and ansund he raised the child up quickened and healthy ‘he lifted the child up which was refreshed and healthy’ (van Kemenade 1987: 36) b.’ *er hob das Kind auf erquickt und gesund he lifted the child up refreshed and healthy

C) OE allows postverbal adverbs in embedded clauses, as is illustrated for manner adverbs in (21a). In Dutch and German, manner adverbs may not follow the finite verb in embedded clauses (cf. (21b) for German). (21) a.

þæt hie þæt unaliefede dod aliefedlice that he the unlawful did lawfully ‘that he correctly performed an unjust deed’ (van Kemenade 1987: 36) b. *dass er das Unrechtmässige ausführte rechtmässig that he an unjust deed performed correctly

The conclusion to be drawn from these observations is that there is strong evidence that the finite verb moves leftward even in embedded clauses (cf. Pintzuk 1996, 1999). Note, however, that this rule will not account for all cases of VOorder. (22) shows a statistics by W. Koopman (1994). His corpus contains in addition to 340 OV-sentences all sorts of VO-structures involving non-finite verbs (total number = 84). (22) a. b. c. d. e.

C. . . Vf DO IO C. . . Vf IO DO C. . . DO Vf IO C. . . IO Vf DO C. . . Vf DO V IO

70 86 42 29 11

f. g. h. i. j.

C. . . Vf IO V DO C. . . Vf V DO IO C. . . Vf V IO DO C. . . DO V Vf IO C. . . IO V Vf DO

6 22 38 2 5

In order to account for these and similar facts, in her dissertation Pintzuk (1991) (published as Pintzuk 1999; the idea is also adopted by Kroch & Taylor 1997) proposes, what I call, a doubly double base. She assumes that not only the VP can be head-initial or head final, but also that the IP can be head-initial or head-final in OE. Not only is this proposal cumbersome, it also allows for a grammar that overgenerates, as is acknowledged by Pintzuk (1999). We do not find sentences of the form V XP Aux, which would result from combining

Language change versus grammar change 

a head-initial VP with a head-final IP (see Fuß and Trips 2002 for discussion and an alternative proposal in terms of grammar competition that excludes the order V-O-Aux). Alternatively, I assume a VO-based grammar plus licensing movement of arguments and VP-internal predicates to designated positions in the middle field, as has been shown for Dutch by Zwart (1993) and for German by Hinterhölzl (1999). This will derive all the OV-properties of OE (cf. Roberts 1997 for a VO based account of OE) and is illustrated in (23). (23a) is derived from the VO-base structure (23b) in a cyclic fashion by XP-movement of the particle into a licensing position for predicates (23c) and by movement of the arguments into Case-licensing positions (23d). (23) a.

þæt he his stefne up ahof that he his voice up raised ‘that he spoke up’ b. [CP þæt [VP he ahof his stefne up]] c. [CP þæt [ up [VP he ahof his stefne]]] d. [CP þæt [he [his stefne [ up [VP ahof ]]]]]

To account for the apparent VO-properties of OE, I will assume optional movement of an XP containing the verb to a medial position (light predicate raising, LPR). Assuming this rule has the following advantages: a) it can apply to finite as well as to non-finite verbs and b) it can pied-pipe verbal particles. Thus, the OE sentence (24a) is derived within a grammar that yields basic OV-structures (24b) by successive application of light predicate raising and V2, as illustrated in (24c). (24) a.

þa ahof Paulus up his heafod then raised Paulus up his head ‘then Paulus lifted his head (up)’ (van Kemenade 1987: 33) b. [CP þa [IP Paulus his heafod up ahof ]] c. [CP þa ahof i [IP Paulus [VP up ti ] his heafod]]

If OE had a rule of LPR then the distribution of event-related adverbs in Modern English, especially their postverbal occurrence which distinguishes them from all other adverb classes is no surprise anymore. Note that event-related adjuncts are primarily realized as PPs or NPs and that (heavy) NPs and PPs are the prime candidates for appearing in clause-final position, as will be shown in the following section.

 Roland Hinterhölzl

.. The trigger of VP-adverb inversion In this section, I report on the results of a corpus analysis. I studied the distribution of DPs, PPs and adverbs with respect to non-finite verbs, that is, infinitives and participles, in main clauses in two OE-texts and three EMEtexts. The results are given in (25). The data give us a pretty good idea of how VP-adverb inversion in the history of English came about. Secondly, I address a problem that is raised by the double base hypothesis but which is hardly dealt with by its proponents. How does a speaker that possesses two grammars, an OV-grammar and a VO-grammar, decide in which situations he uses which grammar? We either expect the choice to be arbitrary or, more likely, to be determined by sociolinguistic factors (the formal or informal character of the discourse situation, prestige and ethnicity of the interlocutors, etc.). But we do not expect the choice to be determined by language internal factors. However, this is exactly what we find when we take a closer look at the data.4 (25) Old English Texts: Boeth = King Alfreds OE version of Boethius’ De Consolatione Philosophiae (850–950) postverbal adverbs 15/150 10% postverbal objects (DPs only) 44/164 27% postverbal PPs 69/113 60% Aelive= Aelfric’s Lives of Saints (950–1050) postverbal adverbs 22/85 26% postverbal objects 52/121 43% postverbal PPs 89/137 64% Middle English Texts: Sawles Warde: West Midlands (ca. 1200) postverbal adverbs 5/9 55% postverbal objects 17/22 77% St. Katherine: West Midlands (ca 1200–1220) postverbal adverbs 24/47 50% postverbal objects 58/89 62% postverbal PPs 57/86 66%

Language change versus grammar change 

Vices and Virtues: East Midlands (ca. 1200) postverbal adverbs postverbal objects postverbal PPs postverbal Adjunct-NPs

11/129 89/363 123/240 8/27

7% 25% 55% 29%

The two OE-texts are taken from the middle and the later period, just to see whether any development already happened in the OE-period. The ME-texts are all from the same early period. I discarded the results from Sawles Warde since the data are too small in numbers to warrant any statistically sound conclusions. Two things can be read off these numbers. In both OE-texts as well as in Vices and Virtues we see a clear difference in postverbal occurrences according to syntactic category which can be explained most coherently as a length effect. It seems to me that this state of affairs is not easy to explain within the double base hypothesis. To explain the data in Boethius, the proponents of competing grammars would either have to claim that speakers use a VO grammar in 60% of the cases in sentences with PPs, but only in 27% and 10% of the cases with DPs and adverbs respectively. To assume this is certainly absurd. One way out for the proponents of the double base hypothesis would be to assume that there is a type of metagrammar which embodies statements like “Heavy and stressed elements are preferably placed at the right edge”, according to which OV or VOstructures are selected. The alternative for the DBH is to accept the existence of optional, stylistic rules. For instance, Pintzuk (1999) explicitly assumes that OE had a rule of extraposition which optionally adjoined heavy material to the right of the clause. This opens the question when the placement of a given constituent is subject to stylistic rules or the result of the choice of the head complement parameter. Note that these facts can be handled very elegantly by the peripheral rule of VP-intraposition that I proposed above if we assume that the rule is conditioned by the heaviness of the category that ends up at the right boundary of the clause. If the threshold is just around three words, then most of the PPs, around 60%, will end up in the postverbal domain, while much less DPs, namely only the longer ones and very few adverbs, very likely only those that are focussed, may trigger light predicate raising. The data also show that the frequency of postverbal adverbs rises in the transition from OE to ME. When postverbal adverbs reached a certain threshold, the stylistic rule of LPR was reanalysed as an obligatory rule that applies

 Roland Hinterhölzl

in the core grammar (cf. Hinterhölzl 2002 for the details). After this reanalysis had taken place, preverbal adverbs were analysed as the result of a (new) stylistic rule that placed light adverbs in a preverbal position. This is the situation that we find in Modern English. At this point some remarks are in order as to the general assumption about the interaction between core rules and peripheral rules. Based on a quantitative analysis, Pintzuk (1999) shows that the rate/frequency of extraposition did not change over the centuries from OE to (Early) Modern English. This is one important argument of hers against Stockwell (1977) and the hypothesis that the outcome of (an increased use of) stylistic rules was reanalysed in terms of VO base structures. However, if the analysis of postverbal adjuncts given in Section 2.1 is correct, then postverbal adjuncts have indeed undergone a process of reanalysis: what initially was a derived stylistic order became an unmarked base order. The important point here is not extraposition and its quantitative development per se. It is the category and the information-structural role of the postponed constituent that are crucial in this context. Extraposed CPs must and extraposed PPs can form an intonational phrase of their own. What matters is the prosodic status of postverbal arguments, adverbs and focussed PPs that are the result of LPR. These elements are typically integrated into the intonational phrase containing the verb and can thus alter the prosodic structure of the core IP. In the following section, I will outline how an alternation in the prosodic makeup of a language can lead to a change of what counts as marked or unmarked word order in that language. Another question that arises at this point is whether LPR is triggered and, if so, by which (type of) feature. One option is to assume that LPR is untriggered and applies as a last resort to fulfil a PF-requirement that evaluates sequences of phonological phrases according to non-decreasing weight. The other option is to assume that LPR is triggered by pragmatic (information-structural) features. A good candidate would be the feature [+ background] that can be ascribed to a predicate that is either presupposed, prementioned or implied in a given context. Such a trigger that moves ‘light’ elements leftwards is in line with traditional accounts of other developments in Germanic (cf. Wackernagel’s 1892 law about the placement of unstressed pronouns, Delbrück’s 1911 characterization of the development of V2 from a prosodic rule to a categorical rule). At this point, I leave it as an open question whether ‘light’ in this case is to be interpreted as prosodically light or semantically light, that is to say, whether LPR is due to a PF- or an LF-interface requirement.5

Language change versus grammar change 

To summarize, I have provided evidence that explains why it is exactly event-related adverbs and not other adverbs that are placed postverbally in English. The reason is that they were realized in their majority as NPs and PPs. As such they were placed in the postverbal field by the stylistic rule of LPR in OE and EME. Furthermore, I have shown that the nature of the variation between OV and VO orders that we (already) find in OE and in the transition to ME cannot be explained well by the double base hypothesis.

. Unmarked word order and the Universal Base Hypothesis If LPR occurred in the core derivation in OE, it yields, when combined with scrambling, a postverbal occurrence of a destressed (light) background element, as is illustrated in (26). (26) a. [C . . . [Manner [VP V DO]]] base structure scrambling b. [C . . . DOi [Manner [VP V ti ]]] c. [C . . . [VP V ti ] DOi [Manner tVP ]] VP-intraposition

Elements that undergo scrambling in German and Dutch represent old information and are destressed. To obey the extension condition, the VP must move to a position above the position of the scrambled object in (26c). Thus, if LPR is more and more extensively used by speakers of early English, it will, over time, have a profound effect on the prosodic repertoire of the language and thereby on what counts as marked or unmarked word order in this language, as I will argue below. We have come to identify as marked or unmarked word order in a language what goes against or conforms to the particular setting of the head complement parameter in this language. When we dispense with the head complement parameter and adopt the Universal Base Hypothesis (UBH) (Kayne 1994), then what counts as an unmarked order in a language is not a basic property anymore, but has to be derived from other properties in the language. We may assume that the unmarked word order is defined by the amount of obligatory movement of the different constituents in a language. For instance, one can say that German has unmarked OV order because objects obligatorily move out of the VP and the verb stays low, presumably in the VP (except in cases of V2). This is a coherent notion which involves dissociating the unmarked order and the base order. Dissociating base order and unmarked order may be problematic when it comes to the acquisition of the unmarked order in OV-languages. German

 Roland Hinterhölzl

children virtually make no word order mistakes and use the correct word orders from very early on. If the child needs to learn all or the majority of movement operations in its language before it can produce unmarked word order patterns, one should expect that in early utterances the German child produces VOstructures, which conform to the base order. But this simply does not happen. Note, however, that the standard account to the acquisition of the head complement parameter faces a similar problem as well. The common assumption is that in order for the child to set this parameter, it needs to be able to segment the speech stream into words. Thus, the head complement parameter cannot be set before the child has attained a substantial part of the lexicon. Alternatively, Nespor, Guasti and Christophe (1996) propose that the headcomplement parameter is set on the basis of prosodic information (the rhythmic activation principle). The advantage of this proposal is that the unmarked word order can be determined very early indeed, that is, before any lexical knowledge is attained. More specifically, they propose that the decisive information for the child is the placement of main prominence within the phonological phrase, a constituent of the prosodic hierarchy. I will adopt this account and propose that the unmarked word order in the phrases of a language is determined, independently of obligatory movement operations, by the predominant, that is, unmarked prosodic patterns in a language.6 If a language displays only one prosodic pattern, this pattern will define the unmarked word order. If a language displays several prosodic patterns, the predominant pattern or the predominant patterns will define the unmarked word order(s),7 which have to be derived in the core grammar by obligatory movements, while less frequent patterns will be relegated as marked word orders to the periphery, to be derived by stylistic rules. The frequency component within the definition of a marked prosodic pattern is important for explaining language change. If a peripheral rule like LPR is extended and more and more widely used (when it becomes popular), it will be less and less marked and may, when crossing a certain (statistical) threshold give rise to a new or an additional unmarked word order. When this happens a peripheral rule will be reanalysed as a rule of the core grammar.

. Options in grammar and the role of prosodic constraints In German, a question like the one in (27) can be answered either with (27a) or with (27b) with capitalized letters indicating the stressed (= focussed) con-

Language change versus grammar change 

stituent. In both versions, das Buch is destressed. Native speakers report that the answer in (27a) is preferred over or more natural than (27b). (27) Q: Wem hat Otto das Buch gegeben? Who has Otto the book given? A: a. Otto hat das Buch dem PETER gegeben. b. Otto hat dem PETER das Buch gebeben. Otto has (the book) to Peter (the book) given

So we are dealing with a real option in grammar here. I have proposed in Hinterhölzl (2002) that one area of optionality in grammar may involve the spell-out of copies (the relevant condition seems to be that the feature that is checked by the pertinent movement operation is interpretable at both interfaces). More specifically I propose that scrambling that is triggered by the discourse feature familiarity may spell out either the higher or the lower copy. Thus, both in (27a) and (27b) the direct object has scrambled across the indirect object, with the difference in word order following from the difference in Spell-out. The choice is determined by prosodic constraints, as is illustrated in (28). (28) shows the prosodic structure of both answers, where round brackets indicate phonological phrases (φs) and iP indicates an intonational phrase. We see that (28a) optimally fulfils the prosodic condition in (29), while (28b) violates this condition. I propose that this is the reason why (27a) is preferred over (27b). (27b) is repaired by being assigned a stronger pitch accent, while in (27a) the assignment of the normal sentence accent suffices to mark the focussed constituent. Thus (27b) is prosodically more marked than (27a), but speakers are free to use the more marked forms for their communicative purposes, whatever they are. (28) a. [iP (φ Otto hat) (φ das Buch) (φ dem PEter gegeben) ] b. [iP (φ Otto hat ) (φ dem PEter) (φ das Buch gegeben)] (29) Interface Condition: The phonological phrase containing the focus (main accent) must be rightmost within its intonational phrase. (cf. Chierchia 1986; Hayes & Lahiri 1991)

Grammar change is dependent on a certain amount of variation. One source of variation, as I have argued above, is options in the grammar that can be exploited by (adult) speakers for stylistic purposes. Alternative phrasings and linear orderings are typically prosodically marked, which qualifies them as stylistic means for conveying different communicative effects. For instance,

 Roland Hinterhölzl

the postponing of PPs in modern German allows the speaker to exbraciate information which forms the background for an assertion in a given discourse situation, as is illustrated in (30). (30) weil der Hans die MAria getroffen hat | gestern in Wien since Hans the Maria met has yesterday in Vienna

However, if a certain stylistic means is used regularly, it loses its stylistic force and an originally marked prosodic pattern is integrated into the core grammar, sometimes leading to the reanalysis of a derived word order pattern as a basic one.

. Stylistic change? The development of the German sentence bracket As already mentioned in the introduction, postverbal occurrences of arguments and adjuncts were slowly diminishing in the history of German. The process started in the middle of the 9th century, where scholars have diagnosed a sharp decline of postverbal accusative objects in the later texts of Notker (cf. Bolli 1975; Näf 1979; Borter 1982) and has not been fully completed till today. To illustrate the development, let us look at the statistics of Kavanagh (1979), cited by Lenerz (1984), who studied the development of the verb final position, traditionally called the sentence bracket, in German. His counts show that the verb final pattern has risen from 50% occurrence in OHG, via an average number of 80% in the period of Middle High German (MHG) to more than 90% in Modern German. If this statistics is correct, it shows two things. First, as we have argued that OE cannot be considered an OV-language, 50% verb final patterns indicate that German in its oldest accessible stage cannot be considered an OV-language either. Secondly, the change was rather slow and gradual and we may assume that starting from the MHG period, V XP-orders represented marked options in the grammar. However, the evaluation of word order regularities in OHG is a quite intricate and difficult task since the majority of texts are translations from Latin. Thus many researchers consider only those examples which deviate from the Latin word order as revealing real properties of the grammar of OHG. Based on this rationale, it was argued that already the earliest OHG records display much stronger OV properties than comparable OE texts (cf. Robinson 1997; Dittmer & Dittmer 1998; Fuß 2002).


It is interesting to note but compatible with the above rationale that while the change from OE to Modern English word order has been described as a grammatical change (a change in the head complement parameter), the inverse change in German has been characterized as mere stylistic in the German traditional literature. Also Lenerz (1984) states that with respect to the development of verb positioning from OHG to ModG, no syntactic change in the strict sense has taken place: What has changed is the functional relation of certain verb orders to specific pragmatic types. He claims that the changes that you find should be treated as the variation of parameters which are not part of the core grammar (p. 126). He backs up his position with the claim that in general the relative proportions of extraposable constituents have remained constant over the whole period. Citing Kavanagh (1979), he states that a) adjuncts are most frequently postponed and arguments least frequently and b) longer and heavier constituents are more easily postponible than short and light ones. Furthermore, he argues that the change cannot be a categorial one, since even arguments can still be postponed in Modern German, as is illustrated in (31). (31) Auf Gleis 5 fährt ein | der IR nach Straubing. at platform 5 comes in the Interregio to Straubing

The question only is whether this (traditional) picture of the development of German is correct. I think it is at least incomplete. Note also that the study of Robinson (1997) reveals that certain embedded clauses display only 50% verb final order: “in consecutive and purposive sentences it [non-final word order] occurs 12 times out of 24” (p. 83). If OHG was a simple OV-language, this result is rather surprising. Secondly, Robinson (1997) also points out that “predicative elements (predicate adjectives, nouns, prepositional phrases, etc.) regularly follow any non-finite element at the end of the clause.” Postverbal predicates in Modern German are completely ungrammatical in any context and style. Thus, if the development of the sentence bracket in German was simply a matter of stylistic change, this fact cannot be explained. The latter finding of Robinson (1997) is in line with the observations of Dittmer and Dittmer (1998) about reorderings in Tatian with respect to the original Latin text. With some writers reorderings are quite principled and systematic. Within DPs, postnominal adjectives are preposed regularly as is standard in Modern German. Within the sentence, pronouns are almost always preposed and the Latin verb complex is broken up to form the typical modern German sentence bracket by preposing light verbs. Dittmer and Dittmer (1998)





Roland Hinterhölzl

take these specific reorderings as indicative of elements of grammar that are typical of modern German. Two other conclusions, I think, are equally warranted. First, Tatian cannot be considered a simple word to word translation of a Latin text. Secondly and more importantly, it may be warranted to consider also as systematic and principled the orders which remain unchanged with respect to the Latin origin. In other words, the systematic non-reordering of postverbal non-light DPs may also be indicative of a grammatical element typical of OHG, especially in a context where verbs, whose positioning is anyway more fixed in the languages of the world, are rearranged quite freely. The rationale being that if writers felt free to rearrange verbs (and pronouns) they would also have rearranged DPs if the grammar of OHG had required it. Thus the word orders found in the earliest translation texts may all be indicative of grammatical rules of OHG. According to my own readings, Tatian shows an abundant number of verb-complement orders and it may turn out that the statistics of Kavanagh (1979) is actually correct. Let us now return to Lenerz’s argument that exbraciated arguments in Modern German are indicative of a mere change in style rather than of a change in grammar proper. Sentence (31), which involves a postponed subject is okay, but it is very rare and highly marked in modern German, as the corresponding prosodic structure in (32) indicates. The sentence with a postponed subject cannot be phrased with just one intonational phrase. Hence it is marked since DPs normally do not form intonational phrases on their own. As a result, the focussed subject may not be contained in the intonational phrase that contains the verb and thus violates the prosodic condition in (33). (32) [iP ( Auf Gleis 5 ) ( fährt ein ) ] [iP (der Interregio) (nach Straubing)] (33) Focus constituents are mapped into the intonational phrase which contains the verb. (Nespor & Vogel 1986)

Based on the above discussion, I conclude, contrary to Lenerz and the tradition, that there was a change in grammar, since it seems unreasonable to assume that there was a language stage, namely OHG, in which texts displayed such a high number of marked structures in the form of verb complement orders. . The role of information-structure According to traditional grammarians, word order in OHG was determined to a high degree by stylistic factors. Behaghel (1932) observes that pronouns (definite and indefinite ones) and single nouns (remember that a determiner


system has not yet developed in OHG) occur mostly preverbally, while nouns modified by an adjective or a genitive complement occur mostly postverbally. This suggests that the positioning of arguments (and adjuncts) with respect to the verb was determined by the factor weight. I have assumed above (cf. (16)) that the weight factor is just a corollary of another factor, namely the information-structural role of the respective constituent. According to (16), focussed elements occur postverbally and background elements occur preverbally. In this context, it is interesting to note that we find quite a number of postverbal occurrences of light arguments in OHG texts, some of which are given in (34) (examples from Behaghel 1932). (34) a.

tar. . . dero goto rat festenoe dia hitat there advice of these gods hardened the act of fathering ‘in this place the advice of the gods confirmed the act of fathering’ (Notker 726, 1) b. so aber die sorgun gruozent tiu herzen so but the sorrows make-beat the hearts ‘in this way the sorrows make the heart beat’ (Notker 788,3) c. daz tu mit tages liehte irbarost tie nahtsculde that you with day light disclosed the sins of the night ‘that you disclosed the sins of the night at the break of day’ (Notker 843, 32)

As I also noted above a determiner system was not established yet, and I assume that the positioning of an argument with respect to the verb signalled its discourse status, with elements representing old information preceding the verb and elements representing new information following it. But the grammatical system was in a state of transition. In the late OHG period, the grammaticalization of definite determiners from demonstrative pronouns, which originally only reinforced the referential interpretation of an argument, started. Behaghel notes that determiners first show up with NPs in preverbal position: ‘Substantiva mit Pronomen stehen auf der Seite der einfachen Wörter; zum Teil mag das daher rühren, dass ihnen der Artikel früher fehlte’ (Behaghel 1932: 79). So the introduction of an obligatory definite determiner could have contributed to the breakdown of a grammatical system in which the positioning of an argument (wrt. the verb) signalled its discourse status. The question, though, arises whether this could have been the decisive factor that led to the gradual loss of V-XP orders. One wonders why the grammaticalization of definite determiners did not yield the same effect in the history of English.8 In the



 Roland Hinterhölzl

following section, I propose that another factor played a decisive role in this development. . The grammar of focus The question arises whether the picture that (16) draws of a language that has variation between OV and VO orders is complete. When we take a look at a modern Germanic language that has preserved OV and VO orders, namely Yiddish, the picture is a bit more complex. Diesing (1997) claims that definite und specific indefinite NPs in Yiddish precede the verb while indefinites follow it. Insofar as definites and specific indefinites can be taken to be background elements and indefinites to be focussed, this confirms our hypothesis. However, there is a third group of elements which disturb this picture. Diesing claims that arguments that are contrastively focussed also precede the verb. Three notions of focus are relevant for our discussion here: broad and narrow presentational focus and contrastive focus. These are illustrated in (35). In (35), brackets mark the focussed constituents and capital letters mark a high pitch accent, which is typical for contrastively focussed elements. The picture that arises for Yiddish is given in (36). (35) a.

What did John do? (broad presentational focus) John [gave a book to Mary]. b. What did John give to Mary? (narrow presentational focus) John gave [a book] to Mary. c. John gave Mary [a BOOK], not a pen. (contrastive focus)

(36) C background contrastive focus V presentational focus

This distribution can be implemented into the following phrase structure representation. The verb moves into the head position of FocP. Contrastively focussed constituents move into [Spec,FocP]. Constituents that are part of the presentational focus remain in the scope of the head of FocP. Background elements move out of the scope of Foc into licensing positions in the higher middle field. (37) [C background [FocP ContrastF V [AgrP PresentationF [VP]]]]

It is interesting to see that the grammar of Yiddish implements a syntactic distinction between contrastive and non-contrastive, that is, presentational focus. Note in this respect that prosody makes a distinction between narrow and broad focus. Constituents that are broadly focussed do not restructure with the verb. They are mapped into prosodic constituents according to the syn-


tactic structure alone, while narrowly focussed constituents integrate into the phonological phrase of the verb (cf. Nespor & Vogel 1986; Frascarelli 2000). Let us assume for the sake of discussion that the same situation as is exemplified by Yiddish held in OHG. Then we can envisage the following development. If speakers of late OHG, for whatever stylistic purposes, moved more and more narrowly focussed constituents into [Spec,FocP], then there will be a greater number of phonological phrases that, due to focus restructuring, instantiate the pattern (s w), with s being the focussed element and w being the verb. Since according to the rhythmic activation principle, unmarked order is determined on the basis of the metric structure of phonological phrases (cf. Nespor, Guasti, & Christophe 1996), such a stylistic rule, if it catches on and reaches a certain threshold, will have a profound effect on what counts as unmarked or marked word order in the language. More specifically, we may assume that if the application of the stylistic rule crosses a certain threshold, it will cause the IS-prosodic parameter in (38) to be set to the value (F>V), that is, Focus restructures to the right. If this happens, postverbal focussed constituents will be literally exbraciated prosodically: They are excluded from the intonational phrase containing the verb and will thus be highly marked and be useable only for certain stylistic effects as was shown in (32). (38) IS-requirements on the prosodic mapping of syntactic structures: A focussed constituent introduces an iP-boundary on one side and restructures with the adjacent verb on its other side. (generalized from Frascarelli 2000)

Harris and Campbell (1995) in their comparative survey of types of word order changes in the languages of the world, point out that it frequently happens that an originally pragmatically defined position provides the basis for the fixing of word order through reanalysis (p. 235). For example, Comrie (1988) argues that an earlier stage of Armenian placed focussed constituents in preverbal position and that in Modern East Armenian this word order has been fixed as a base order. Thus, there is cross-linguistic evidence for the process of the neutralization of a focus position that I have proposed for German. Of course, the empirical investigations have to follow. But I have outlined the grammatical basis for such a development: namely prosodic bootstrapping in the acquisition of unmarked word order and focus restructuring. To conclude, traditional grammarians were right in that the development of the German sentence bracket was brought about by stylistic rules. How-



 Roland Hinterhölzl

ever, these stylistic rules as they crucially affected the prosodic make-up of the language eventually led to a change in grammar. This happened when the IS-Prosodic parameter was set to or restricted to one out of three options, namely F>V.9 Furthermore, I have argued that the head-complement parameter is dispensable. I have shown that German and English underwent changes that cannot be subsumed under the headcomplement parameter. Instead, I have sketched an account in terms of the distinction between core grammar and peripheral rules embodying a dynamic concept of (prosodic) markedness and outlined the conditions and factors that led German and English to develop from a common source into opposite directions. Along these lines, German can be claimed to be a VO-language, whose unmarked order happens to be OV, because the IS-prosodic parameter was fixed to F>V. Finally, I have argued that in the acquisition process, prosody helps the child to decide which structures belong to the core grammar and which ones are peripheral. This entails that the child acquires the prosodic patterns of its language first and ranks them in their markedness according to their frequency.

Notes * A special thanks goes to the editors of this volume, Carola Trips and Eric Fuß, for extensive discussions and helpful and valuable comments on earlier versions of this paper. . More specifically, temporal adverbs specify the reference time (in a Reichenbachian sense) in relation to the speaking time. In non-complex tenses the reference time is identical with the event time. . An anonymous reviewer points out that the presence of OV orders as such (in OI) does not necessarily constitute a problem for the language contact scenario of OV-to-VO in English, the rationale being that these orders are derivable from a VO base by leftward movement operations. If the learner would be able to detect this VO base, this might still influence his grammar (that develops in a bilingual environment), so the argument goes. But this begs the question. In this line of reasoning, also the grammar of OE would have contained evidence for a VO base. Without any further argument as to why the contact situation in general would favour VO orders (rather than OV orders) or as to which feature of OI would favour a VO analysis, the contact scenario seems mute to me. . An anonymous reviewer points out that examples with right dislocation of arguments are very natural in Modern German (cf. (ia)). (ia) is indeed quite natural but immaterial to the point made above, since the core sentence contains an argument, namely the pronoun, in a non-extraposed position, while the DP ‘den Peter’ can be considered to be


an afterthought. The crucial point here is that without resumption, extraposition of a DP argument is extremely marked (cf. (ib)). (i)

a.

weil ich ihn gestern getroffen habe, den Peter since I him yesterday met have, the-acc Peter b. *?weil ich gestern getroffen habe, den Peter since I yesterday met have, the-acc Peter

. Already Pintzuk (1999) points out that the choice of OV versus VO order is sensitive to various grammar internal factors such as clause type (cf. Fuß and Trips 2002 for an attempt to explain the impact of these factors). What is largely missing, however, is a framework which explains how these internal factors can be integrated into a theory that accounts for word order variation by competing grammars. . My intuition is that the conditioning factor in OE was semantic lightness and that the grammar contained a requirement that semantic light elements are treated as prosodically light elements. But more empirical work is to be done to evaluate these options. . To get prosodic bootstrapping off the ground, we have to assume that the child has access to the following universal principle that if a head and a non-head are combined in a single phrase it is the head (or the functional element) that is weak. So if the child encounters the pattern (weak strong), it knows that the head precedes the complement or that a functional element precedes a lexical category. . Note that a language may have several unmarked prosodic patterns. In German, the unmarked pattern with nouns and prepositions is weak-strong (w s), but with verbs it is (s w). . The question needs further investigation. At this point I can only offer some speculation: if determiners are just added to the nominal in the position where the nominal was placed originally we derive that the majority of DPs will occur preverbally (German). If, however, determiners are introduced and the corresponding phrase is positioned according to its weight, we derive that the majority of DPs will occur postverbally (this could be what happened in English). Ideally, the choice should depend on how reduced the determiner already was when it became obligatory (and in which positions it was introduced first). . We have to assume that in Yiddish, OHG and OE both options, V>F and F>V were available.

References Behaghel, O. (1932). Deutsche Syntax. Band 4. Heidelberg: Carl Winters Universitätsbuchhandlung. Bolli, E. (1975). Die verbale Klammer bei Notker. Untersuchungen zur Wortstellung in der Boethius-Übersetzung. Berlin/New York: Mouton de Gruyter. Borter, A. (1982). Syntaktische Klammerbildung in Notkers Psalter. Berlin/New York: Mouton de Gruyter.



 Roland Hinterhölzl

Chierchia, G. (1986). “Length, syllabification and the phonological cycle in Italian.” Journal of Italian Linguistics, 8, 5–34. Cinque, G. (1999). Adverbs and Functional Heads. A Crosslinguistic Perspective. Oxford: Oxford University Press. Comrie, B. (1988). “Topics, grammaticalized topics and subjects.” In S. Axmaker, A. Jaisser, & H. Singmaster (Eds.), General Session and Parasession on Grammaticalization (pp. 265–279). Berkeley Linguistic Society. Chomsky, N. (1965). Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Delbrück, B. (1911). Zur Stellung des Verbums [Germanische Syntax 2]. Leipzig: Teubner. Diesing, M. (1997). “Yiddish VP order and the typology of object movement in Germanic.” Natural Language and Linguistic Theory, 15, 369–427. Dittmer, A. & E. Dittmer (1998). Studien zur Wortstellung-Satzgliedstellung in der althochdeutschen Tatianübersetzung. Göttingen: Vandenhoeck and Ruprecht. Ellegård, A. (1953). The Auxilliary DO: The Establishment and Regulation of its Use in English. Stockholm: Almquist and Wiksell. Ernst, T. (1998). “The scopal basis of adverb licensing.” In P. Tamanji & K. Kusumoto (Eds.), Proceedings of NELS 28 (pp. 127–142). Amherst: GLSA. Falk, C. (1993). “Non-referential subjects in the history of Swedish.” Doctoral dissertation, University of Lund. Frascarelli, M. (2000). The Syntax-Phonology Interface in Focus and Topic Constructions in Italian. Dordrecht: Kluwer. Fuß, E. (2002). “Wortstellungsvariation in den älteren germanischen Sprachen.” Ms., University of Frankfurt. Fuß, E. & C. Trips (2002). “Variation and change in Old and Middle English. On the validity of the Double Base Hypothesis.” Journal of Comparative Germanic Linguistics, 4, 177– 224. Greenberg, J. (1963). “Some universals of grammar with particular reference to the order of meaningful elements.” In J. Greenberg (Ed.), Universals in Language (pp. 33–59). Cambridge, MA: MIT Press. Haider, H. (2000). “Adverb placement – convergence of structure and licensing.” Theoretical Linguistics, 26, 95–134. Harris, A. & L. Campbell (1995). Historical Syntax in Cross-linguistic Perspective. Cambridge: Cambridge University Press. Hayes, B. & A. Lahiri (1991). “Bengali Intonational Phonology.” Natural Language and Linguistic Theory, 9, 47–96. Hawkins, J. A. (1983). Word Order Universals. New York: Academic Press. Hinterhölzl, R. (1999). “Restructuring infinitives and the theory of complementation.” Doctoral dissertation, University of Southern California. Hinterhölzl, R. (2001). “Event-related adjuncts and the OV/VO Distinction.” In K. Megerdoomian & L. Bar-el (Eds.), WCCFL 20 Proceedings (pp. 276–289). Somerville: Cascadilla Press. Hinterhölzl, R. (2002). “Parametric variation and scrambling in English.” In W. Abraham & J. W. Zwart (Eds.), Studies in Comparative Germanic Syntax (pp. 131–150). Amsterdam & Philadelphia: John Benjamins.

Language change versus grammar change 

Hróarsdóttir, T. (1998). “Verb phrase syntax in the history of Icelandic.” Doctoral dissertation, University of Tromsø. Hróarsdóttir, T. (2002). “Explaining language change: A three step process.” LIP, 19, 103– 141 [Papers from the workshop “Language change from a Generative Perspective”. Thessaloniki, February 2002]. Kavanagh, J. (1979). “The verb final rule in German: A diachronic analysis of surface structure constraints.” Doctoral dissertation, University of Michigan. Kayne, R. (1994). The Antisymmetry of Syntax. Linguistic Inquiry Monograph 25. Cambridge, MA: MIT Press. Kemenade, A. van (1987). Syntactic Case and Morphological Case in the History of English. Dordrecht: Foris. Kemenade, A. van & N. Vincent (1997). Parameters of Morphosyntactic Change. Cambridge: Cambridge University Press. Koopman, W. (1994). “The order of dative and accusative objects in Old English.” Ms., University of Amsterdam. Kroch, A. (1989). “Reflexes of grammar in patterns of language change.” Language Variation and Change, 1, 199–244. Kroch, A. & A. Taylor (1997). “Verb movement in Old and Middle English: Dialect variation and language contact.” In A. van Kemenade & N. Vincent (Eds.), Parameters of Morphosyntactic Change (pp. 297–325). Cambridge: Cambridge University Press. Larson, R. (1988). “On the double object construction.” Linguistic Inquiry, 19, 335–391. Lenerz, J. (1984). Syntaktischer Wandel und Grammatiktheorie. Tübingen: Niemeyer. Lightfoot, D. (1991). How to Set Parameters. Cambridge, MA: MIT Press. Lightfoot, D. (1999). The Development of Language. Acquisition, Change and Evolution. Oxford: Blackwell. Maling, J. & A. Zaenen (1990). Modern Icelandic Syntax. Syntax and Semantics 24. San Diego/New York: Academic Press. Näf, A. (1979). Die Wortstellung in Notkers Consolatio. Berlin/New York: Mouton de Gruyter. Nespor, M. & I. Vogel (1986). Prosodic Phonology. Dordrecht: Foris. Nespor, M., M. T. Guasti, & A. Christophe (1996). “Selecting word order: The Rhythmic Activation Principle.” In Ursula Kleinhenz (Ed.), Interfaces in Phonology (pp. 1–26). Berlin: Akademie Verlag. Pesetsky, D. (1995). Zero Syntax. Cambridge, MA: MIT Press. Pintzuk, S. & A. Kroch. (1989). “The rightward movement of complements and adjuncts in the Old English of Beowulf.” Language Variation and Change, 1, 115–143. Pintzuk, S. (1991). “Phrase structures in competition: Variation and change in Old English word order.” Doctoral dissertation, University of Pennsylvania. Pintzuk, S. (1996). “Old English verb-complement word order and the change from OV to VO.” York Papers in Linguistics, 17, 241–264. Pintzuk, S. (1999). Phrase Structures in Competition: Variation and Change in Old English Word Order. New York: Garland. Roberts, I. (1997). “Directionality and word order change in the history of English.” In A. van Kemende & N. Vincent (Eds.), Parameters of Morphosyntactic Change (pp. 397– 426). Cambridge: Cambridge University Press.

 Roland Hinterhölzl

Robinson, O. W. (1997). Clause Subordination and Verb Placement in the Old High German Isidor Translation. Heidelberg: C. Winter. Stockwell, R. (1977). “Motivations for exbraciation in Old English.” In C. Li (Ed.), Mechanisms of Syntactic Change (pp. 291–314). Austin: University of Texas Press. Stroik, T. (1990). “Adverbs as V-sisters.” Linguistic Inquiry, 21, 654–661. Vallduvi, E. (1992). The Informational Component. New York: Garland. Wackernagel, J. (1892). “Über ein Gesetz der indogermanischen Wortstellung.” Indogermanische Forschungen, 1, 333–435. Zwart, J. W. (1993). “Dutch syntax: a minimalist approach.” Doctoral dissertation, University of Groningen.

The EPP, fossilized movement and reanalysis Andrew Simpson SOAS London

This paper addresses the question of whether all operations of movement necessarily have a clear morphological, semantic or pragmatic motivation, and how the EPP is to be understood as a trigger/motivation for syntactic movement. Considering patterns in Thai, Taiwanese and other languages of east and southeast Asia, it is argued that certain applications of movement have no genuinely understandable motivation when considered from a purely synchronic point of view, and that the use of EPP features to trigger movement is a formal mechanism made available by the grammar for the legitimization of a movement whose genuine semantic, pragmatic or morphological trigger has become lost over time.

.

Introduction

This paper sets out to provide answers to the questions below concerning movement and its motivations: (A) Do all operations of movement necessarily have a motivation which can be identified and clearly understood in morphological, semantic or pragmatic terms, or are there also real occurrences of movement without such an identifiable trigger? (B) How is the EPP to be understood as a trigger/motivation for syntactic movement? Considering patterns found in Thai, Taiwanese and other languages of east and southeast Asia, it will be argued that certain applications of movement really have no genuinely understandable motivation when considered from a purely synchronic point of view, and that the use of EPP features to trigger movement is a formal mechanism made available by the grammar for the (continued) legitimization of a movement whose genuine semantic, pragmatic or morphological trigger has become lost over time. In Sections 2–4 evidence is presented

 Andrew Simpson

suggesting that when the original motivation for a movement disappears from a construction, this may generally not cause any simple discontinuation/loss of the movement and an automatic retreat to an earlier pre-/no-movement patterning as might be expected, but instead results in a significant reanalysis of the movement structure in two possible ways. One potential form of the reanalysis is suggested to be the reinterpretation of an original movement structure as new morphology, with an element previously moved to a second, higher position in the syntax being re-interpreted as a new affix base-generated in the higher position. Elsewhere however, when such reanalysis is not permitted, a second outcome is suggested to be the simple continuation of the original movement via the introduction of EPP features on an appropriate functional head as a blind/semantically meaningless replacement for the original (lost) trigger of movement. The broad consequences which follow on from this conclusion are that there may in fact be many cases of movement which have become “fossilized” in this way and have lost any original understandable (i.e. non-EPP) motivation, and secondly that once initiated in a regular way, applications of movement may possibly never be truly lost from a structure, unless they are converted into new morphology. The structure of the paper is essentially as follows. Section 2 first reviews data and arguments in Simpson (2001) for a focus-based analysis of a particular modal construction found in Thai and other southeast Asian languages. Section 3 then shows how such an analysis is further supported by complex tone sandhi patterns in Taiwanese present with the same modal element, but where focus significantly does not occur as part of the interpretation. Comparing Thai with Taiwanese and other varieties of Chinese, Section 4 then argues for the main conclusion of the paper that movement may indeed survive the loss of its original morpho-syntactic/semantic motivation and continue to occur synchronically in a construction for no clear reason as an occurrence of EPP-driven “fossilized” movement. Finally in Section 5, the paper examines other consequences of the possibility that movement is never diachronically lost from a construction even when its original motivation disappears and attempts to show how certain cases previously categorized as instances of loss of movement might in fact synchronically still be (disguised) movement structures.

The EPP, fossilized movement and reanalysis 

. Post-verbal modals in southeast Asia Standard Thai together with a variety of other southeast Asian languages (e.g Vietnamese, Hmong, Khmer) has been noted in Simpson (2001) and other works to show an unusual modal paradigm in which a single modal meaning ‘to be able’ consistently appears in a post-verbal position, as in (1). This post-verbal positioning is unexpected as these languages are all head-initial SVO languages with modal verbs that otherwise occur in a regular pre-verbal position, as shown in Thai (2): (1) khaw khian dai. he write can ‘He can write.’ (2) Daeng aat-ca/doong/khong maa. Daeng may/must/is-sure-to come ‘Daeng may/must/is sure to come.’

As plausible arguments can be given which relate the post-verbal potential modal in Thai to that in other neighbouring languages (Huffman 1973; Simpson 2001), the widespread occurrence of such an unexpected pattern of postverbal modals is assumed to most naturally be the result of borrowing and transfer among Thai and the other languages of the region. Attempting to probe the underlying structure of the modal construction, Simpson (2001) then identifies a number of key properties of the patterning in Thai. First of all it is pointed out that the post-verbal modal is not simply a morphological suffix as it occurs following not only lexical verbs but also other VP-internal elements such as objects and PPs, as shown in (3). The correct generalization of the pattern in Thai would therefore seem to be that the modal dai ‘to be able’ is VP-final rather than simply “post-verbal”: (3) khun pai kap khaw phrung-nii dai. you go with him tomorrow can ‘You can go with him tomorrow.’

A second relevant observation is that there is evidence indicating that the modal is (somehow) in a structurally higher position than the lexical verb and the VP. Yes-no questions in Thai are commonly answered in the affirmative by repetition of the highest verbal element present in a head position in the main projection line of a clause, and it is found that it is indeed the modal element dai which is used to answer yes-no questions, not the lexical verb which precedes it, as seen in (4):

 Andrew Simpson

(4) khaw phuut phasaa thai dai mai? he speak language thai can Q ‘Can he speak Thai?’ A1: dai A2: *phuut can speak ‘Yes’

In addition to this patterning, the positioning and scope of constituent negation provides further evidence for the high structural position of the modal dai. As illustrated in (5), if the negative element mai ‘not’ precedes the lexical verb it may only take scope over the verb and/or other VP-internal elements, and importantly can not take scope over dai. This suggests that the negation in (5) c-commands the verb and other VP-internal elements but not dai, so that dai itself must be assumed to occur in a position which is hierarchically higher than the VP: (5) khun mai pai kap khaw dai. you neg go with him can ‘You can (choose) not (to) go with him.’

One possible way of reconciling the above patterns is potentially to suggest that the sequence of elements which precedes the modal dai actually comprizes a sentential subject as in (6a), with a meaning similar to the English sentential subject form in (6b). In such a structure constituent negation will not scope over the modal and the modal will be the structurally highest verbal element in the clause: IP

(6) a. IP

I’

khun mai pai kap khaw you  go with him

I dai can

b. [That you do not go with him] is permissible/possible.

However, there is also evidence that dai-modal forms do not have the underlying structure in (6a) either. First of all, sentential subject structures in Thai show regular CED/island restrictions on the extraction of an element from within sentential subjects in topicalization and relative clause formation, yet it is perfectly acceptable to topicalize or relativize an element from the sequence

The EPP, fossilized movement and reanalysis 

which precedes dai, as shown in (7) and (8) below. This suggests that there is a significant difference between sentential subject structures and dai-sentences: (7) *phuu-chaai Oi thii [loon khop ti ] mai dii ko khuu. . . man rel she see neg good then be ‘The man who that she associates with is bad is. . . (e.g. John)’ (8) phuu-chaai Oi thii [loon khop ti ] mai dai ko khuu. . . man rel she see neg can then be ‘The man who she may not date/see is . . . (John)’

Secondly, where there is more than a single modal in a dai-sentence, it is possible for an epistemic modal element such as naa-ca ‘should’ which precedes the VP and dai to actually take scope over dai as shown in (9). If it is necessary for such an element to c-command dai in order to take scope over dai, this indicates that the linearly first modal cannot be contained within a sentential subject structure such as (6a): (9) khaw naa-ca pen pheuan kan dai. they should be friend together can ‘They should be able to be friends together.’

The analysis ultimately suggested in Simpson (2001) to account for the above and various other properties of dai-structures is that the VP in dai-sentences actually originates as a regular rightward complement to dai and then undergoes movement to a surface position to the left of dai, as indicated in (10). In such a derivation the subject is taken to be selected by dai and to also undergo leftward raising:1

 Andrew Simpson

(10)

Such an analysis is shown to be able to capture all of the important properties of dai-sentences. In the structure in (10), dai is the highest verbal element in a head position in the main projection line of the clause resulting in the yes/no question answer patterns (4), and dai in the underlying structure will c-command (but not be c-commanded by) the VP resulting in the restricted scope of negation attached to a fronted VP (5). The underlying regular complement position of the VP will allow for an account of the differences in extraction possibilities from pre-dai sequences compared with regular sentential structures, and the scope of higher modals over dai can be captured in hierarchical structures such as (11) below (using English words to represent sentences such as (9)). The higher modal naa-ca ‘should’ can be taken to head an epistemic Mood projection in (11) located higher than the XP in (10) and so will c-command dai ‘be able to’ in its lower root Mood position. (11) theyi should [be friends]k ti be able to tk

The EPP, fossilized movement and reanalysis 

Turning now to the motivation for the movements suggested to underlie daisentences, it can be noted that dai-sentences might seem to be regularly associated with some kind of focus interpretation, and that this may explain the special syntax of such constructions. Focus interpretations essentially show up in two forms with dai. First of all under certain circumstances it is found that the object of the lexical verb in dai-sentences can actually occur clause-finally following dai, as in (12). This is however only possible if the object is itself clearly focused: (12) khaw phuut dai laai phasaa. he speak can many languages ‘He can speak MANY languages.’

Secondly, if no focused object follows dai and dai is final in the clause then dai itself naturally carries a focal stress (e.g. in examples (1), (5), (9) etc.). Simpson (2001) therefore suggests that the motivation and function of the proposed VP-raising in dai-sentences is critically to de-focus the VP predicate by moving it away from the final focus position in dai-structures, allowing for either dai itself or alternatively an object following dai to receive the clause-final focus intonation and interpretation. Dai-sentences are then suggested to emphasize the possibility, ability or permission of carrying out a certain action (with stress on dai itself) or to emphasize a particular element relating to this possible action (with stress on a final object as in (12)), and the VP predicate is taken to represent pre-supposed/old information contrasting with the final focused modal/object. Syntactically, when a focused object occurs following dai, dai in ROOT MoodP is suggested to select for a Focus projection to whose specifier the object raises prior to VP preposing, as in the remnant movement sequence in (13–14), using English words for the derivation in (12) for ease of exposition:2

 Andrew Simpson

(13)

MoodP Spec

He

Mood’ FocP

Mood can Spec

Foc’

[many languages]k

Foc

VP V’

Spec pro V

NP

speak

(14)

tk

TP T’

Spec Hei

XP

T Spec [ speak tk]m

X’ MoodP

X

Mood’

Spec ti

Mood can

FocP Foc’

Spec

[many languages]k

Foc

VP tm

Ultimately then a principled account of the syntax of the Thai modal construction is argued to be possible as well as a clear functional explanation of why the hypothetical syntactic derivation occurs. Having now established the basic

The EPP, fossilized movement and reanalysis 

properties of such an account for Thai as suggested in Simpson (2001), Section 3 will now move on to consider a related modal patterning in Taiwanese and show how the analysis suggested to underlie object-final modal-focus sentences such (12) is in fact well supported by quite independent data from Taiwanese. Section 4 then compares the patterns in Thai and Taiwanese and also Cantonese and early Chinese and shows how the global paradigm associated with this exceptional modal element leads to interesting conclusions concerning the triggers for movement in the various languages considered.

. Taiwanese, tone sandhi and post-verbal potential modals Chinese is another set of languages in the east Asian area where a post-verbal modal with the same potential ‘can/be able to’ interpretation is found as in Thai, Vietnamese etc., and there are also arguments (presented in Simpson 2001) that the post-verbal modal construction found in the latter southeast Asian languages may indeed have originated in (early and middle) Chinese. In this section I would like to examine the patterning found with the modal as it occurs in the Taiwanese variety of Chinese, previously unexamined in earlier accounts, and show how Taiwanese adds further interesting evidence for the basic derivation outlined in (13) and (14) above. The critical evidence from Taiwanese relates to the odd and rather puzzling patterns of tone sandhi found in modal sentences such as (15) formed with the post-verbal modal element tit ‘can’, and requires a certain initial explanation about the phenomenon of tone sandhi and its general relation to syntactic structure in Taiwanese. (15) Li e ki tit goa e mia be? you will remember can I gen face Q ‘Do you remember my face?’

Taiwanese is a variety of Chinese commonly described as having eight tones. In addition to these there are also syllables which do not carry any tone, this sometimes being referred to as “neutral tone” NT. In the phenomenon of tone sandhi, the lexically-listed “citation” tone of a syllable undergoes modification according to fully regular rules when preceding some other tone-bearing syllable in the same tone sandhi domain. For example, if a syllable with tone 3 precedes another tone-carrying syllable in the same tone sandhi domain, the tone 3 will change into a tone 2, as illustrated in (16):

 Andrew Simpson

(16) khi3 pak8kiang1 → khi2 pak8kiang1 go Beijing ‘go to Beijing’

Tone sandhi does not occur in a syllable if it precedes a syllable which has only neutral tone/no tone, hence zau in (17) does not change its tone-2 when occurring before the toneless element a: (17) zau2 a-NT → zau2 a-NT run already ‘already ran’

Similarly, a syllable generally does not undergo tone sandhi if it occurs sentence-finally, hence in (18), the sentence-final ho does not change its citation tone-2 even though followed by a syllable (Kia) which does carry tone because the latter syllable occurs in a separate sentence.3 Note that from this point on for simplicity of representation tone sandhi change is indicated by means of a simple bolded dot following the relevant syllable. Thus if a syllable is followed by a bolded dot, this indicates that it undergoes tone sandhi change, and if a dot is absent, no tone sandhi change is possible. In (18) sentence-final ho is therefore not followed by a bolded dot as no tone sandhi change occurs in the sentence-final position: (18) Anke chin• ho. Kia ma• chin• ho. Anke very good Kia also very fine ‘Anke is fine, and Kia is also fine.’

Sentence-internally there would also seem to be other tone sandhi/TS domains relevant for the operation of tonal change, and broadly-speaking every syllable in such a domain will change its tone unless it is the last tone-bearing syllable in the domain. Significantly also, tone sandhi change in Taiwanese appears to relate to and reveal the underlying syntactic structure in a way which is not found in tone sandhi phenomena in Mandarin, Shanghainese and certain other varieties of Chinese. Here three generalizations noted in Simpson and Wu (2001) can be pointed out: (19) A head and its complement occur in the same TS domain The presence of an overt complement consistently triggers tone sandhi on its selecting head, indicating that a head and its complement are in a single TS domain. (20) A head and its specifier do not occur in the same TS domain A head does not trigger tone sandhi on the final syllable of its spec-

The EPP, fossilized movement and reanalysis

ifier. Consequently the specifier of a head constitutes an independent TS domain. (21) Adjuncts are self-contained TS domains The final syllable of an adjunct does not undergo tone sandhi even when followed by other tone-bearing syllables.

These general properties of tone sandhi in Taiwanese now give rise to an interesting puzzle when one considers sentences involving the modal element tit. In their text book on colloquial Taiwanese, Ko and Tan (1960) carefully describe the tone sandhi patterns which occur in spoken Taiwanese and observe the following patterns in sentences with tit. When the modal element tit occurs following a lexical verb in clause-final position as in (25) it is noted that: (a) the lexical verb does not undergo tone sandhi, and (b) the modal element tit does undergo tone sandhi, hence: (25) Anke be• ki tit• Anke neg remember can ‘Anke does not remember.’

This pattern is actually the opposite to what one might expect, and puzzling for the following reasons. First of all one might expect that the sentence-final element tit would not undergo any kind of tonal change as it occurs in final position in its tone sandhi domain. Secondly, it might be expected that the tone-bearing element tit following the lexical verb ki would indeed trigger tone sandhi on the latter in the same tone sandhi domain, yet Ko and Tan note that this does not happen (whereas in other cases where there are sequences of two verbal elements, the second verbal element in a series does regularly trigger tone sandhi on the element which precedes it). This patterning with tit is made all the more interesting by the observation in Ko and Tan that when an object occurs following the verb-modal sequence, this is noted to cause tone sandhi to occur on the lexical verb as in (26), although tone sandhi does not occur in the lexical verb in (25): (26) Anke be• ki• tit• Kia. Anke neg remember can Kia ‘Anke does not remember Kia.’

Again this is not anticipated and it is unclear why the introduction of an element in object position should result in tone sandhi occurring in an element that is non-adjacent to the object (i.e. the lexical verb ki), as tone sandhi is nor-



 Andrew Simpson

mally triggered by a tone-bearing element on the syllable which immediately precedes it. Such patterns can however arguably be explained if it is assumed (a) that tone sandhi may apply during the course of a syntactic derivation, and (b) sentences with the modal tit such as (25) and (26) have an underlying syntax essentially parallel to that of Thai dai structures. The first assumption is justified at length in Simpson and Wu (2001) on the basis of a study of sentencefinal particle constructions in Taiwanese, with tone sandhi being argued to occur mid-derivationally as part of cyclic Spell-Out (Chomsky 2000–2001). The second hypothesis that Taiwanese tit and Thai dai might share a common underlying syntax will be mostly justified by its ability to explain the tone sandhi patterns here, but is also supported by other indications tit and dai belong to the same wider paradigm of east and southeast Asian post-verbal potential modals: tit is the only post-verbal modal which occurs in Taiwanese, paralleling dai in Thai, tit and dai share the same potential meaning ‘to be able to’, and diachronically both tit and dai may well relate to the same modal de(i) found in early/middle Chinese. Supposing now that sentences such as (25)–(26) do relate to a derivation such as that in (10) and (13)–(14), it can be shown that the tone sandhi patterns found can be straightforwardly explained as follows. In the underlying structure, the VP containing the lexical verb ki may be assumed to be selected by the modal tit in the canonical rightward direction, as with all other modals in Taiwanese, and ki will therefore occur base-generated to the right of tit as in (27), modeling first the derivation of (25): (27) Anke be tit [VP ki] Anke neg can remember

Supposing the tone sandhi rules were to apply to the structure at this premovement point, the occurrences of tone sandhi noted in (25) would be correctly produced: the modal tit would be followed by a tone-bearing element ki and so tit would undergo tone sandhi, whereas the lexical verb ki would be in sentence-final position and so not be eligible to undergo tone sandhi. Following pre-movement application of the tone sandhi rules, the VP containing just the verb would raise leftwards to produce the surface attested form in (25).4 Considering (26) where an overt object is present in the structure, the underlying structure may be assumed to be (28), with both the lexical verb and the object in situ in the VP positioned to the right of the modal tit:

The EPP, fossilized movement and reanalysis 

(28) Anke be tit [VP ki Kia] Anke neg can remember Kia

Again, if the tone sandhi rules are applied at this point they will correctly output the patterns of tone sandhi noted in the surface form: (i) tit will be followed by a tone-bearing syllable and so undergo tone sandhi, (ii) the object will be in sentence-final position and so its final syllable will not undergo tone sandhi, and (iii) the lexical verb will here be followed by a tone-bearing syllable in its tone sandhi domain and so in this case itself undergo tone sandhi (unlike in (27)). If the derivation essentially follows the sequence of syntactic movements in Thai dai object-final sentences, there will be two further post-tone sandhi applications of movement transforming (28) into (26): first the object Kia will raise to a position higher than the VP but below the modal tit as in (29), and then the VP-remnant will raise higher to its surface position preceding the modal as in (30): (29) Anke be tit [Kia]i [VP ki ti ] Anke neg can Kia remember (30) Anke be [VP ki ti ]k tit [Kia]i tk Anke neg remember can Kia

The account outlined above is clearly able to make principled sense of the puzzling patterns of tone sandhi noted in (25) and (26), and it is not at all obvious how these output forms could otherwise be accounted for assuming that tone sandhi applies in a regular way. If this is indeed the correct way to interpret the patterns noted here, then we now have a second rather different set of data pointing to the same two-step derivation of post-verbal modal sentences that has been hypothesized for Thai, hence adding potential support to the analysis of Thai in Section 2. In Section 5 we will shortly see how it is useful to compare the Thai and Taiwanese modal patterns further still and how this leads to certain interesting general conclusions. First, however, I would like to add a third set of related patterns from Cantonese to the set of data to be considered.

. Cantonese Cantonese is another variety of Chinese that shares the striking property of many languages of the region of having a single post-verbal modal verb which turns out to have the meaning ‘can/be able to’ – the element dak. Because dak is the only modal element to occur in post-verbal position in an otherwise dom-

 Andrew Simpson

inantly head-initial SAuxVO language, it poses the same kinds of problems in its analysis as dai does in Thai, occurring in a quite unusual position given other properties of the language. Similar to Thai dai Cantonese dak can also not be analyzed away as a verbal suffix as dak can occur alone and unsupported in answers to questions: (31) ngo tai dak nei-bun-syu maa? I read can your-cl-book Q ‘Can I read your book?’ A: dak. can ‘Yes.’

Consequently, if dak is not a suffix and is assumed to be located in a functional head position (ROOT Mood), it might seem that one needs to assume that the lexical verb itself undergoes some kind of repositioning in order to consistently occur in a position adjacent to and preceding the modal. If one furthermore considers complex sequences of lexical verb, dak and an aspectual verb such as jyun as in (32), one finds that the latter aspectual verb is positioned to the right of dak. Such a left-right Mood>Aspect ordering of the modal verb dak and the aspectual verb jyun is precisely what one would expect in a head-initial language, according to Cinque (1999), if both these elements are in their base-generated functional head positions, with Mood cross-linguistically being higher in the functional structure than Aspect: (32) ngo m tai dak jyun bun-syu. I neg read can asp cl-book ‘I can’t finish reading the book.’

Making the common assumption that a projection of Aspect dominates and selects the VP predicate of a clause, the VP is therefore expected to be projected to the right of the aspectual verb jyun in Asp as diagrammed in (33): (33) [Mood dak [Asp jyun [VP . . . ]]]

However, as the lexical verb instead surfaces to the left of the modal dak, this suggests that the verb occurs in its surface location as the result of some leftward repositioning, and there is no way that a VP complement to Aspect could be simply base-generated to the left of dak if Mood is higher in the functional structure than Aspect as in other languages. This conclusion that the verb reaches its surface position from a VP basegenerated to the right of jyun/Aspect now seems to face a serious problem


however. If it is assumed that the verb simply raises out of the VP to the right of Aspect up to its position to the left of dak/Mood, such long head movement over two intervening, filled head positions is clearly expected to violate the Head Movement Constraint and be illicit, yet sentences such as (32) are perfectly well-formed. One possible way to avoid the conclusion that daksentences should consistently cause HMC violations is now to suggest that what undergoes movement to the left of dak is actually not just the V0 alone, but a larger XP-constituent which will not be subject to the HMC – the entire VP containing the verb. It can be suggested that the object of the verb first raises leftwards out of the VP and then the VP remnant constituent containing just the verb raises higher to its surface position to the left of dak, landing in either a higher specifier or XP-adjoined position, as schematized in (34): (34) a. [Mood dak [Asp jyun [OBi [VP V ti ]]]] b. [VP V ti ]k [Mood dak [Asp jyun [OBi ] tk ]]

Such a solution to the HMC problems posed by such structures may also account for the fact that more than just the verb/V0 is sometimes carried along to the pre-dak position, and one also finds cases where constituent VP negation/a NegP occurs raised with the verb as in (35), suggesting that this movement to the pre-dak position is again raising of a larger XP-type unit rather than simply X0 -movement: (35) nei [m lai] m dak. you neg come neg can ‘You can’t not come’ = ‘You must come.’

Consequently, to the extent that these different problems raised by the postverbal modal in Cantonese are again solved by assuming basically the same two-step derivation as posited in Thai and Taiwanese, we now have a third reason for assuming that such a derivation may perhaps be a general property of the post-verbal modal which has spread through various east and southeast Asian languages.

. Triggers for movement, fossilization and the EPP In Sections 2–4 we have examined a range of evidence in different languages all converging on the conclusion that a single basic two-step derivation underlies the post-verbal potential modal construction in these languages. The general syntax of this construction and the way it appears to have spread throughout



 Andrew Simpson

China and southeast Asia is very interesting and worthy of fuller description, but here I would like to focus on one particular aspect of the construction and see how it can be argued to lead to conclusions which directly address the issue of movement and its synchronic motivations. Specifically, we will now consider just a little more closely what triggers the two-step movement operation in the languages considered. In introducing the modal construction with an examination of Thai, it was argued that it is possible to identify a clear functional/pragmatic trigger behind the movements which occur with dai. Observing that focus is commonly part of the semantic force of the modal construction it was argued that VP-preposing in Thai is carried out in order to de-focus the predicate and cast the modal dai into focus in prominent sentence-final position. Such VP-preposing can therefore be understood to be a case of the wider phenomenon of p-movement discussed in Zubizarreta (1998), where various elements are raised out of a prominent focus position in order to cast a second element into focus and allow for this element to receive a natural focus-stress. When an object occurs focused in the modal construction in Thai, it was argued that this triggers raising of the object out of the VP to a dedicated FocusP selected and induced by the modal before (defocusing) p-movement of the VP-remnant. Both movements involved in such structures were therefore straightforwardly explained in terms of their contribution to the realization of focus in the modal construction. Following our consideration of Thai, we then continued on to look at the post-verbal modal patterns in Taiwanese and Cantonese, and in both cases found evidence of different types pointing towards the same kind of explanation for its odd post-verbal position and the conclusion that the verb in such cases reaches its pre-modal position via an operation of VP-remnant movement occurring after raising of the object out of the VP. What was however not mentioned in Sections 3 and 4 is the important fact that focus is actually not a critical part of the meaning of the modal construction in Cantonese and Taiwanese and there need be no interpretation of focus present in the structure. Furthermore, considerations of focus do not govern the positioning of the object in Cantonese and Taiwanese and objects of all types are always positioned following the modal, whether focused or not, unlike in Thai where only heavily focused objects may occur in this position. The question now is therefore how one should attempt to make sense of this potentially important difference between Thai and Cantonese/Taiwanese. Is focus something which is isolated to just the Thai construction, and if it is indeed only present in Thai, how can the

The EPP, fossilized movement and reanalysis 

movement of the VP and the object in Taiwanese and Cantonese be motivated in the syntax? If we briefly now look back a little in the history of Chinese, there might seem to be interesting evidence that a focus patterning similar to that in Thai was in fact also present in earlier forms of Chinese. In middle Chinese, for example, one finds that objects which are not focalized do occur raised to the left of the post-verbal modal de exactly as in modern Thai, as shown in example (36), and objects which are clearly focused occur to the right of the modal, again as in modern Thai, as shown in (37): (36) shi qie [yao shou] bu de. cause wife wave hand neg can ‘It caused the wife not to be able to wave her hand.’ (37) cheng de ge shenme-bian shi? succeed can cl what matter ‘What can one accomplish?’

(Hanshu)

(Zutangji 3/105/7)

One also finds that the VP complement of the modal de in early Chinese occurred consistently on its right, as is significantly assumed to be the underlying base-generated position of the VP in modern Thai/Taiwanese/Cantonese, and then went through a period in middle Chinese when it was optionally raised to the left of de as a stylistic variant of the in situ form (see Sun 1996). Given this evidence for an underlying base position of the VP to the right of the modal and the positioning of focused objects after the modal and non-focused objects before, raised leftwards within a defocused VP, it can be suggested that middle Chinese indeed had a critically focus-based modal construction parallel to that in modern Thai, but that this role of focus has since disappeared from the modal construction in modern varieties of Chinese such as Cantonese and Taiwanese. If this is so, then we now have the case of a construction which used to have a clear stylistic force and function, but where this force has arguably been lost over time. What is highly significant for our present interests is that the syntactic movements associated with this construction appear to have significantly remained present in the modal construction through to modern Cantonese/Taiwanese despite the loss of stylistic force. This is a particularly important observation which can be argued to be potentially very revealing. If we reconsider the two movements commonly assumed to be involved in the modal construction, namely the movement of the object and the preposing of the VP, supposing there was indeed loss of the focus interpretation from the structure in Chinese it might be possible to suggest that the object movement

 Andrew Simpson

attested could have perhaps been reinterpreted by speakers in some way as occurring for some other non-focus reason such as Case, and been reanalyzed as overt raising to a low Agreement Phrase for licensing of accusative casefeatures. However, once all necessary focus interpretation has been lost from the modal construction, it seems impossible to see how the original defocusing p-movement of the VP could be similarly reinterpreted in any plausible nonfocus way. As there is however synchronic evidence and arguments that such VP-preposing nevertheless continues to occur in modern Chinese, it seems that we have here now identified an operation of movement which can genuinely be described as having no clearly understandable semantic, pragmatic or morphological motivation when viewed from a purely synchronic point of view.5 Viewed from a privileged diachronic or cross-linguistic perspective, a certain clear insight into the situation is indeed available, but from the isolated language learner’s position, it seems that a speaker will be confronted with evidence for a movement operation in the modern Chinese modal construction whose motivation he/she will never be able to understand, as its original, genuine motivation has simply been lost in time. This now importantly leads us to an initial answer to the original question (A) that we started the paper with and the significant conclusion that certain operations of movement may continue to occur in a language’s syntax long after their genuine motivations have disappeared, remaining on as “fossils” of an earlier period of the language when a genuinely understandable trigger did exist for such movement. The conclusion that cases of “fossilized movement” do indeed exist also immediately raises two other related and simple questions, namely (i) why is it the case that movement may fossilize instead of disappearing with its original motivation, and (ii) how is such movement formally licensed within a synchronic grammar? Concerning the first broad question, a natural expectation might be that when the semantic/pragmatic/morphological trigger behind a particular movement becomes lost over time, the movement itself would also cease to take place and there would simply be a return to a state in which no movement occurs in the relevant environment. Here one can certainly only speculate a little, but the conclusion that there is fossilization of the type described here may be interpreted as indicating that speakers may prefer to blindly mimic an aspect of language/movement present in their neighbours’ speech which they do not fully understand rather than choose the more radical alternative of changing their speech to a non-movement form for no stylistic gain. Quite possibly imitation/mimicry is simply perceived as an easier and less risky option than innovation to a new form which involves the

The EPP, fossilized movement and reanalysis 

drastic, negative step of undoing a regularized movement operation, and it may be that languages do not tolerate such purely backward steps, favouring instead innovation in more positive ways via reanalysis wherever possible. Mimesis/mimicry and “matching behaviour” has indeed been recently noted to be a cognitive trait which is particularly highly developed in humans and early learning processes displayed by children (Stern et al. 1985; Meltzoff & Moore 1993), and has also been implicated as being the critical distinctive quality leading to the evolution of language in humans rather than other species (Vihman & Depaolis 2000). If there is indeed some in-built tendency for mimesis and matching, and fossilization essentially results because of this, addressing question (ii) above, it can now be suggested that the syntax actually makes available a formal mechanism to respond to the desire for mimesis, and this is precisely what the much-discussed but poorly understood EPP/use of generalized EPP features ultimately is – a purely formal mechanism provided by the grammar for the continued legitimization of movements whose original characteristics have undergone change and loss, synchronically imposed as a feature-checking requirement on a relevant functional head, triggering “blind” movement (i.e. movement with no genuinely understandable motivation) to the specifier or (possibly) head position of the functional projection. Exploring the Thai/Cantonese/Taiwanese patterns thus leads to a set of possible answers to the original question (A) whether all operations of movement necessarily have a clearly understandable motivation, and question (B) what the role of the EPP may be as a trigger for syntactic movement. Such questions are essentially quite simple but nevertheless important as the answers to them significantly affect the way we engage in syntactic analysis. Whenever syntactic evidence is presented for an analysis of movement, the question is commonly asked why the hypothesized movement takes place, and if no obvious motivation can be identified, the analysis of movement may often be considered implausible and unlikely to be correct, despite the accompanying syntactic evidence in its favour. The contention of the current paper is however that maybe it is simply not possible to find a genuine, understandable synchronic motivation for certain real syntactic movements because the original semantic/pragmatic triggers for such movements have been lost in time. In cases where there is good syntactic evidence or theoretical reason for assuming the existence of an occurrence of movement, the lack of a clearly identifiable trigger for the movement should therefore NOT be seen as sufficient reason to reject the likelihood that movement does occur. The conclusions reached above also bring with them a final question concerning the direction of linguistic change. If the grammar makes available a

 Andrew Simpson

mechanism such as the EPP, and if there is indeed a conservative preference amongst speakers for reanalysis of a fossilized movement structure via the introduction of EPP features so that movement continues to be (blindly) made rather than discontinued, is it the case that language ever simply goes “into retreat” and reverts back to an earlier state (i.e. pre-/no-movement), or is it the case that all linguistic development must necessarily be in a forward direction, building on and reanalyzing existing structure rather than reducing it via the direct loss of features such as movement operations? Section 6 now closes the paper with a brief consideration of certain consequences of the possibility that language might in fact not license or tolerate purely negative, reversal-type changes.

. Reversal or reanalysis? If it is supposed that language does not readily go into simple retreat, then there may indeed be many instances where applications of movement once performed for stylistic reasons have undergone fossilization and continue to be triggered by semantically-empty EPP-features. Before we consider this issue further, it can be suggested that there may also be a second alternative outcome to the fossilization of movement, and that movement structures which have lost their original semantic/pragmatic triggers may perhaps also allow for reanalysis and conversion into morphology if the linguistic environment permits this. Reflecting on the case of the modal construction in Cantonese, the element dak now regularly occurs right-adjacent to the lexical verb but is syntactically distinguished as a free morpheme in a functional head position in virtue of (a) occurring as an unsupported answer form as in (31), (b) occurring separated from the lexical verb in double negation structures such as (35), and (c) due to the sequencing of morphemes in structures combining dak with a lexical verb and an aspectual element as in (32). Because of such data, it was argued that one needs to assume an analysis in which the object of the verb first raises out of the VP and then the VP-remnant raises higher beyond dak. However, if the patterns in (a–c) became rarer and disappeared, there would be nothing in the patterning of dak to block a reanalysis of this element as being base-generated as a suffix on the lexical verb. Were this to happen, there would no longer be any need to assume either object-raising or VP-remnant movement and the semantically/pragmatically redundant movement operation(s) licensed by EPP features could be fully eliminated from Cantonese, re-absorbed into the language via the creation of new verbal morphology. In Taiwanese were the tone


sandhi patterns to undergo certain changes then a similar result could be effected here too, and many modern speakers of Taiwanese may in fact seem to have made such a change. Reanalysis of movement as new morphology in this way would however not be the simple negative reversal of an earlier movement form to a state of non-movement (which would produce sequences such as [Subject Modal V Object] rather than [Subject V-Modal Object]), but instead result in the positive creation of a new morphological form and hence constitute a significantly forward development and evolution of the input form (here the modal paradigm). Where such reanalysis of movement as new morphology is however not possible, the view that linguistic change should always involve some kind of positive development leads one to expect that fossilized, EPP-driven movement should continue on until it can eventually be reanalyzed in some positive way, and importantly there should not be any simple loss/discontinuation of movement. Such a view, if plausible, now suggests that an interesting avenue of future research may be to carefully reconsider those instances where there does seem to be evidence for a direct loss of movement and to ask whether the relevant evidence may have possibly been mis-analyzed and might perhaps be open to other interpretations. Here I will briefly mention just two prominent sets of cases that have commonly been referred to as involving movement loss/reversal – (a) the “loss” of verb-movement in certain west European languages, and (b) “loss” of wh-movement. The first phenomenon is in fact considered in some depth in Haeberli (2002) and also Alexiadou and Fanselow (2002), which re-examine the development of verb-second structures in Old English (OE) into SVO forms during the Middle English period (ME). Haeberli points out that OE actually shows two distinct patterns in the clause-initial placement of subjects and verbs when some other (non-operator) XP occurs in first position in the sentence. When a non-pronominal subject is present, this most commonly follows the verb resulting in a V2 pattern: [XP V Sub. . . ]. However, when the subject is a pronoun, this actually precedes the verb, resulting in a V3 type pattern: [XP Subpronoun V. . . ]. Haeberli proposes that this difference in distribution of pronominal and non-pronominal subjects is due to the former raising to a higher subject position than the latter: pronoun subjects in OE would therefore occur in SU1 in (38), and non-pronominal subjects in SU2 . The verb itself would be raised to the Agr position between these two subject positions: (38) [CP XP C [AgrP SU1 Agr SU2 . . . ]]



 Andrew Simpson

Haeberli makes the interesting suggestion that non-pronominal subjects are able to stay lower in SU2 in OE because SU1 can be filled by an expletive pro which satisfies the EPP in SU1 , and provides convincing argumentation for the existence of such an element in OE. The critical change from verb-second patterns to SVO forms is then argued to have resulted from a simple loss of the expletive pro from English, and the necessity for not only pronoun subjects but also non-pronominal subjects to occur raised into SU1 to satisfy the EPP. Such a change from verb-second to SVO sequences has elsewhere been described as a loss of verb-movement to the position preceding (non-pronominal) subjects, as it appears that a V2 pattern [XP Vk Sub..tk . . . ] has developed into sequences where the verb occurs in a lower position: [XP Sub V..] (considering only the occurrence of non-pronominal subjects). However, reanalyzed in the way that Haeberli suggests, the change from verb-second to fully regular SVO forms would not in fact have resulted from any loss of verb-movement but from the introduction of movement of non-pronominal subjects to a higher position, SU1 (hence: [XP Subi Vk ti .. tk .. ]), to compensate for the loss of expletive pro in this position (and the verb would continue to raise to the Agr position as in OE). In such an analysis there is consequently a very plausible account of the development of “V2” into SVO patterns which allows one to significantly maintain a “no-reversal” view of movement, and that movement is not simply discontinued in language change causing structures to revert to some earlier pre-movement state.6 Concerning the second set of cases mentioned above and the loss of whmovement, this has been assumed to be a recent (optional) development in colloquial French which now allows a wh in situ strategy in direct questions whereas only wh-movement was possible in the past, and has also been hypothesized as a full development in Japanese (Watanabe 2002).7 Although cases such as French and Japanese might therefore be suggested to show the development of wh in situ out of wh-movement forms and hence a simple loss of movement, against this one can suggest that certain new perspectives on wh in situ and movement in general may also allow for plausible alternative interpretations of the patterning found. Following work in Bošković (2002), Simpson and Bhattacharya (2003), and Fujimoto (2001), at least four different perspectives of the “loss” of wh-movement might be argued to be available. In Simpson and Bhattacharya (2003) it is argued at length that a language previously categorized as wh in situ (Bangla/Bengali) is actually a language with regular overt wh-movement. Failure to notice the occurrence of such movement in the past is suggested to be due to the fact that wh-movement in Bangla is heavily disguised by the occurrence of other, secondary movements and the

The EPP, fossilized movement and reanalysis 

topicalization of all non-wh arguments and adjuncts to clause-initial positions preceding the landing-site of wh-movement. The existence of such movement is however clearly revealed in restrictions on wh scope and long-distance whclausal pied piping. If languages may therefore sometimes appear to be wh in situ on the surface but in fact have regular overt wh-movement disguised by other movement operations, it is possible that French and Japanese might actually not have developed new wh in situ possibilities, but instead have developed question-forms like Bangla where regular overt wh-movement has come to be hidden by other topicalization operations. In such a view, wh-movement would therefore not be “lost” but simply come to be disguised by other new movement operations. A second, potential view of the “loss” of wh-movement which maintains that movement is actually not lost from wh-questions is to consider the possibility that what undergoes change in a language is perhaps the size of constituents that undergo wh-movement. Assuming that the pied piping of non-wh material in wh-questions is licensed by wh-feature percolation, if wh-features came to be able to percolate to higher nodes in a syntactic structure than in previous stages of the language and identify larger constituents as wh-phrases eligible for wh-movement, a language might come to have whmovement of CP clauses, IPs, and other major constituents containing whphrases where prior to this only smaller DP-like phrases might have undergone wh-movement. In languages such as Japanese with apparently loose pre-verbal word order, it could therefore be that the occurrence of clausal wh-movement (i.e. movement of a CP containing a wh-phrase which has percolated its whfeatures to the CP-node from a position embedded within the CP) could go unnoticed as wh-movement and give the language the appearance of having developed wh in situ questions (as in Bangla, cf. Simpson & Bhattacharya 2003). Again as with the first alternative view of wh in situ considered, here wh-movement would not be lost but simply disguised by an independent development, here a potential increase in the size of XP undergoing wh-movement, so that wh-movement would no longer be so obviously noticeable as when restricted to smaller constituents. A third possible account of changes within a language which appear to lead to a loss of wh-movement might perhaps attempt to attribute new “wh in situ” patterns to phonological factors. Bošković (2002) offers persuasive evidence that in certain Slavic languages wh in situ patterns sometimes occur for purely phonological reasons, and that in such cases overt wh-movement actually takes place, but it is the lower copy of the movement chain which is phonetically spelt-out giving the appearance of wh in situ. If the spell-out of

 Andrew Simpson

wh-phrases in low chain internal “in situ” positions is possible in certain instances and caused by phonological factors, it is also possible that phonological changes in a language could result in such a low spell-out strategy becoming standardized so that the language might seem to develop wh in situ from whmovement. However, if there actually is wh-movement preceding this spell-out phenomena as suggested by the Slavic patterns, then such a change would again not constitute a loss of movement, just a difference in the phonetic realization of the resulting movement chain.8 Finally, there is also the possibility that claims of loss of wh-movement in a language may not be correct for the simple reason that a particular language actually may have never had genuine wh-movement in the first place. Such a proposal is made in Fujimoto (2001) for Japanese in an interesting reconsideration of patterns examined in Watanabe (2002). Whereas Watanabe (2002) suggests that classical Japanese had wh-movement to a clause-initial position and this has been lost in modern Japanese, Fujimoto (2001) argues that the relevant wh-structures in classical Japanese were actually reduced biclausal wh-cleft structures in which there was no movement of the wh-phrase, and that the change to modern Japanese has been simply a reanalysis of the bi-clausal cleft as a mono-clausal structure. Wh-movement is therefore argued not to have occurred in classical Japanese and also not to be present in modern Japanese. Consequently, in such a reanalysis of the wh-patterns, Japanese would again not be a case where wh-movement would have simply been lost. Given the availability of the above alternative possibilities of analysis to explain the “loss” of wh-movement in French, Japanese and other languages, and the fact that such explanations may in principle be available to explain other non-wh cases of movement “loss”, it can perhaps not so easily and automatically be concluded that there really are cases where movement has been lost from an earlier structure. To this it can be added that the most commonly cited and well-accepted examples of loss of movement are indeed the verbmovement and wh-movement cases discussed, hence cases where there are now plausible reasons to be skeptical of a simple movement loss analysis. Furthermore, if one attempts to catalogue the number of cases of loss of movement reported cross-linguistically, one finds that the general number of such cases is in fact surprisingly and suspiciously few, and a minute fraction of the set of cases where movement has evolved from an earlier non-movement structure. If movement can apparently develop so commonly in languages, and if movement can also be lost from a structure, it is not at all clear why there should be such a tremendous difference in the number of cases of development of movement vs. loss of movement. All of this may therefore instead seem to support

The EPP, fossilized movement and reanalysis 

the interesting alternative possibility that languages may only tolerate positive type changes, as suggested, and indicate that simple reversals and loss of movement perhaps may not occur. If this is indeed so, and if future research is able to offer good alternative explanations of cases of apparent loss of movement, such a conclusion will clearly lead to a helpful and significant narrowing of the type of diachronic change that may consequently be assumed to be possible in natural language.

. Concluding remarks This paper set out to examine the two questions (A) and (B) relating to the potential motivations for movement: (A) Do all operations of movement necessarily have a motivation which can be identified and clearly understood in morphological, semantic or pragmatic terms, or are there also real occurrences of movement without such an identifiable trigger? (B) How is the EPP to be understood as a trigger/motivation for syntactic movement? Investigating differences in the manifestation of a modal paradigm present in various east and southeast Asian languages it was argued that it is indeed possible to identify movements in syntax which do not have any clear synchronic trigger in obvious semantic, pragmatic or morphological terms. Broadly it was suggested that when an initial semantic, pragmatic or morphological trigger for a movement disappears over time, speakers may in fact continue to effect the original movement with a different purely formal trigger in the form of an EPP feature-checking requirement. The introduction of such an EPP featurechecking requirement on a relevant attracting functional head was therefore suggested to be a purely formal mechanism which grammars allow themselves for the continued legitimization of movements whose original characteristics have undergone change and loss, movements which can synchronically be viewed as “automatic fossilized reflexes” of earlier stages of the language and hence described as “fossilized movement”. Such conclusions from the study of diachronic patterns are potentially significant for the study of synchronic syntax because they suggest that it will simply not be possible to identify clearly understandable motivations for certain genuine synchronic movement in syntax and hence if syntactic evidence supports the assumption of an occurrence of movement but no deep explanation for the movement can be found, this

 Andrew Simpson

should not necessarily cause one to doubt the hypothesis that movement does indeed take place. Certain syntactic movements may become embedded and fossilized in a language as the result of diachronic change and synchronically occur for no reason other than a purely formal and semantically vacuous EPP trigger. The paper also suggested that the introduction and use of EPP features in a construction may correspond to a preference in speakers to blindly mimic a syntactic movement present in the speech of others whose motivation is poorly/not understood rather than to innovate with a discontinuation of such movement, and that possibly languages may not tolerate fully negative reversal-type changes such as the loss/discontinuation of movement. A consequence of this latter conjecture was noted to be that it may now be appropriate and interesting to re-examine cases which have been commonly assumed to be instances of loss of movement to see if synchronically they may in fact still contain movement disguised in certain ways.

Notes . Note that the landing-site of the movement is essentially left unspecified in (10). Movement of the VP is assumed to be to the specifier of some higher clausal functional projection located lower than TP (or perhaps alternatively be adjunction to MoodP or some other functional projection higher than MoodP). Note also that a pro subject is assumed to occur in SpecVP, controlled by the overt subject generated in SpecMoodP and raised to SpecTP. . Note that, generally, certain sequences of extraction and remnant movement have been argued to give rise to Proper Binding Condition violations (e.g. when an element XP is scrambled out of a constituent YP, and then YP is scrambled higher), whereas other sequences of extraction and remnant movement are commonly well-formed (e.g. scrambling of XP out of YP followed by topicalization of YP, as in the case of German VP remnant movement). The basic typological observation/claim made in Müller (1996) and Tsujioka (2001) relating to such cases is that if movement of the extracted XP and the remnant YP is of the same type (e.g. both are instances of scrambling), then the sequence of extraction and remnant movement is ill-formed. If, however, the initial extraction of the XP and the subsequent movement of the remnant YP are essentially different types of movement (e.g. scrambling vs. topicalization), then extraction of XP followed by remnant movement of YP is (frequently) permitted. In the case under consideration here, it can be suggested that the focus-movement of the object is sufficiently different from p-movement/de-focusing/scrambling of the VP that no Proper Binding Condition violation occurs in the sequencing in (13) and (14). . One exception to this is the case of the sentence-final particle kong whose special syntactic properties are examined at length in Simpson and Wu (2001). . Note that the range of patterns discussed in Simpson and Wu (2002) are argued to lead to the conclusions that (a) Cyclic Spell-Out applies at the clausal-phase level when a TP is

The EPP, fossilized movement and reanalysis 

built together with the C position which selects it, and not at the sub-clausal vP-phase level, and (b) that movement can apply to elements which have already been given a phonetic interpretation by Cyclic Spell-Out. In (25), (26) and other cases here, the occurrences of movement described are therefore assumed to take place following application of Cyclic Spell-Out at the clausal level. . What is meant by an “understandable morphological motivation” is a trigger for a movement which results in the licensing/checking of some clearly identifiable morphological specification such as Case, tense, number etc. . Alexiadou and Fanselow (2002) similarly argue that verb-movement is not discontinued in the switch from V2 to SVO patterns, and that all indications of verb-movement such as the position of adverbs relative to the verb commonly remain the same following the change to such an SVO order. See also Julien (2002: 263–273) who re-examines the distribution of finite verbs, adverbs and objects in Modern English and argues that Modern English in fact continues to show evidence of overt verb-movement to the tense position. . A reviewer of the paper asks about the change of OV word order to VO as a third possible case of movement loss, perhaps requiring the assumption that object shift is lost in the switch to a regularized VO ordering. Unfortunately, space requirements do not allow for a proper discussion or investigation of OV/VO shifts here. However, an obvious potential alternative analysis to attempt to explore, which would avoid the conclusion that movement is lost, would be the possibility that new VO sequences are produced not by any loss of object shift/movement, but rather by an innovation of verb-movement to a higher position over the object. As the development of verb-movement to a higher clausal head is a fairly frequently attested change in language, it is not unlikely that it could result in VO sequences developing out of earlier OV forms. In addition to this, it can be noted that the occurrence of object-shift/movement in OV sequences has in fact been seriously questioned in Pintzuk (2002) (for Old English), and OV/VO forms are argued to be related to each other by a mechanism which does not involve any kind of movement. Following the presentation of a wide range of evidence, Pintzuk proposes an analysis based on competition between headinitial and head-final structures which similarly avoids the conclusion that movement is lost in the emergence of VO forms from OV sequences. . A reviewer asks what the difference might be between “real” wh in situ and “apparent/phonological” wh in situ of the type described in Bošković (2002). Here one can note that data in Bošković (2002) show that there are clear syntactic effects indicating movement of the wh constituent to SpecCP in the cases of “phonological” wh in situ described, for example the licensing of parasitic gaps. Such effects are not anticipated to be present if there is no movement of the wh-phrase, and would be expected to be absent in instances of “real” wh in situ. For all purposes of interpretation and the licensing of syntactic relations, “real” wh in situ is therefore anticipated to pattern as if no movement to a higher position has taken place (as discussed in Simpson 2000, Chapter 1), whereas “phonological” wh in situ is expected to show signs that the wh-phrase can license syntactic relations/interpretations from a higher Comp position.

 Andrew Simpson

References Alexiadou, A. & Fanselow, G. (2002). “On the correlation between morphology and syntax: The case of V-to I.” In W. Abraham & J. W. Zwart (Eds.), Proceedings from the 15th CGSW (pp. 219–242). Amsterdam: John Benjamins. Bošković, Z. (2002). “Multiple wh-fronting.” Linguistic Inquiry, 33(3), 351–384. Chomsky, N. (2000). “Minimalist inquiries: The framework.” In R. Martin, D. Michaels, & J. Uriagereka (Eds.), Step by Step: Essays on Minimalism in Honor of Howard Lasnik (pp. 89–155). Cambridge, MA: MIT Press. Chomsky, N. (2001). “Derivation by phase.” In Michael Kenstowicz (Ed.), Ken Hale: A life in language (pp. 1–52). Cambridge, MA: MIT Press. Cinque, G. (1999). Adverbs and Functional Heads: A Cross-linguistic Perspective. Oxford: Oxford University Press. Fujimoto, S. (2001). “The syntax of Kakarimusubi construction and its historical change.” In M. Irie & H. Ono (Eds.), UCI Working Papers in Linguistics, Vol. 7 (pp. 20–38). Irvine Linguistics Students Association, University of California, Irvine. Haeberli, E. (2002). “Inflectional morphology and the loss of verb-second in English.” In D. Lightfoot (Ed.), Syntactic Effects of Morphological Change (pp. 88–106). Oxford: Oxford University Press. Huffman, F. (1973). “Thai and Cambodian – a case of syntactic borrowing?” Journal of the American Oriental Society, 93, 488–509. Julien, M. (2002). Syntactic Heads and Word Formation. Oxford: Oxford University Press. Ko, C. H. & P. T. Tan (1960). An Introduction to Taiwanese Colloquial. Taichung, Taiwan: Kupfer. Meltzoff, A. & Moore, M. (1993). “Why faces are special to infants: on connecting the attraction of faces and infants’ ability for imitation and cross-modal processing.” In B. de Boysson-Bardies, S. de Schonen, P. Jusczyk, P. MacNeilage, & J. Morton (Eds.), Developmental Neuro-cognition: Speech and Face Processing in the First Year of Life (pp. 211–225). Dordrecht: Kluwer. Müller, G. (1996). “A constraint on remnant movement.” Natural Language and Linguistic Theory, 14, 355–407. Pintzuk, S. (2002). “Verb-object order in Old English: Variation as grammatical competition.” In D. Lightfoot (Ed.), Syntactic Effects of Morphological Change (pp. 276–299). Oxford: Oxford University Press. Simpson, A. (2000). Wh-Movement and the Theory of Feature-Checking. Amsterdam: John Benjamins. Simpson, A. (2001). “Focus, presupposition and light predicate raising in East and Southeast Asian Languages.” Journal of East Asian Linguistics, 10, 89–128. Simpson, A. & Bhattacharya, T. (2003). “Obligatory overt wh-movement in a wh in situ language.” Linguistic Inquiry, 34(1), 127–142. Simpson, A. & Wu, Z. (2001). “Tone sandhi and the creation of sentence-final particles: Evidence for cyclic Spell-out.” Journal of East Asian Linguistics, 11(1), 67–99. Simpson, A. & Wu, Z. (2002). “Understanding cyclic Spell-out.” In M. Hirotani (Ed.), Proceedings of NELS 32 (pp. 499–518). GLSA, University of Massachusetts, Amherst.

The EPP, fossilized movement and reanalysis 

Stern, D., Hofer, L., Haft, W., & Dore, J. (1985). “Affect attunement: The sharing of feeling states between mother and infant by means of intermodal fluency.” In T. Filed & N. Fox (Eds.), Social Perception in Infants (pp. 249–268). Norwood, NJ: Ablex. Sun, C.-F. (1996). Word Order Change and Grammaticalization in the History of Chinese. Palo Alto, California: Stanford University Press. Tsujioka, T. (2001). “Improper remnant A-Movement.” In M. Kim & U. Strauss (Eds.), Proceedings of NELS 31 (pp. 483–500). GLSA, University of Massachusetts. Amherst. Vihman, M. & Depaolis, R. (2001). “The role of mimesis in infant language development: Evidence for phylogeny?” In C. Knight, M. Studdert-Kennedy, & J. Hurford (Eds.), The Evolutionary Emergence of Language (pp. 130–145). Cambridge: Cambridge University Press. Watanabe, A. (2002). “Loss of overt wh-movement in Old Japanese.” In D. Lightfoot (Ed.), Syntactic Effects of Morphological Change (pp. 179–195). Oxford: Oxford University Press. Zubizarreta, M. L. (1998). Prosody, Focus, and Word Order. Cambridge, MA: MIT Press.

Restructuring and the development of functional categories Zoe Wu University of Southern California

This paper attempts to show how a study of the diachronic development of resultative constructions in Chinese is able to shed better light on the current morpho-syntactic structure and status of verbal clusters in Chinese. Arguing that the secondary verbal elements in ‘read-finish’ and ‘wash-clean’-type resultative sequences have now formally developed into a new class of aspectual suffix, the paper suggests that functional projections may sometimes come into existence as the result of a revaluation process in which a grammatical property associated with the use of certain lexical items in a particular syntactic environment becomes more highly valued than the actual descriptive content of the lexical items. Considering certain critical word order changes which have occurred in the Chinese resultative verb construction over time, the paper also suggests that reasons of directionality may sometimes play an important role in the creation of suffixation in head-initial languages, providing a means to functionally re-align disharmonic head-final structures.

.

Introduction

This paper examines resultative verb constructions (RVCs) in Chinese and argues that a close consideration of the diachronic development of RVCs is able to shed better light on the current morpho-syntactic structure and status of verbal clusters in the language. It is argued that the secondary verbal elements in ‘read-finish’ and ‘wash-clean’-type resultative sequences have now formally developed into a new class of aspectual suffix, and that such a development resulted from two significant stages of reanalysis. First of all, it is suggested that the interpretation of telicity in RVCs created by verbs encoding result states came to be more highly valued, over time, than the actual descriptive content provided by such verbs. This resulted in a clear change in the type of verb which

 Zoe Wu

could participate in RVCs as the second verb. Secondly, and considerably later in the development of Chinese RVCs, the initial [V1 Object V2 ] sequencing of RVCs underwent a change to the modern [V1 -V2 Object] pattern. This second change in the development of the resultative construction is suggested to be the result of a reanalysis of the V2 elements as new aspectual suffixes/morphology. The paper shows how the creation of new morphology in such a way provides a means in which a head-initial language such as Chinese may be able to functionally re-align potentially ‘disharmonic’ head-final structures which are present in the language. The paper therefore makes two central proposals. First, it is suggested that functional projections may sometimes arise due to a revaluation process in which a grammatical property associated with the use of lexical items in a particular syntactic environment becomes more highly valued than the actual descriptive content of the lexical items. Second, the occurrence of suffixation in head-initial languages, which is often noted to be unexpected (e.g. Givón 1976), may sometimes arise due to reasons of directionality and provide a way of harmonizing a language in a uniform head-initial manner. The paper also shows that it is possible to arrive at a clear understanding of the synchronic structure underlying modern-day Chinese RVCs only once the various historical developments of the construction have been carefully considered and taken into account.

. Resultative Verb Constructions/RVCs Resultative Verb Constructions in Chinese are sequences of two adjacent verbal elements, V1 and V2 , where the second verb V2 signals the result or the termination of the action of V1 as schematized below: RVCs: Subject V1 V2 (Object) V2 signals the result or the termination of the action of V1

There are two main types of RVCs. In “Simple RVCs” the V2 provides a literal description of the state resulting from the action of V1 as in (1) and (2) where the clothes become clean as a result of the washing act and the horse is tired as a result of Zhangsan’s riding:

Simple RVCs (1) ta xi-ganjing yifu le. she wash-clean clothes le ‘She washed the clothes clean.’

Restructuring and the development of functional categories 

(2) Zhangsan qi-lei-le na-pi-ma. Zhangsan ride-tired-le that-cl-horse ‘Zhangsan rode that horse and that horse got tired.’1

In “Phase RVCs” (Li and Thompson 1981), the V2 simply signals completion of the action of V1 , as in (3) and (4):

Phase RVCs (3) ta nian-wan shu le. he study-end book le ‘He finished studying.’ (4) ta mai-dao-le nei-ben-shu. he buy-arrive-le that-cl-book ‘He managed to buy that book.’

In attempting to arrive at a synchronic analysis of RVCs in this paper, I would like to highlight and concentrate on three important aspects of RVCs past and present. The first of these is the property well noted in the literature, e.g. in Li and Thompson (1981), that the V2 in an RVC provides an endpoint to the activity described by the V1 . The V2 thus closes off and terminates the action depicted by the V1 and makes a dynamic event telic and aspectually bounded. For example, in (1) and (2) above, the state of the clothes becoming clean and the horse becoming tired marks the endpoint of the respective washing and riding actions. (5) a.

(A) Telic contribution of V2 The V2 in an RVC makes a dynamic event telic and aspectually bounded.

A second important issue in RVCs concerns the existence of a predication relation between the NP ‘object’ following V2 and the V2 itself. In examples such as (1) and (2) it might appear that the V2 predicates the properties of being clean and tired directly onto the NPs which follow the V2 . b. (B) Predication between V2 and the NP ‘object’? Is there a predication relation between the V2 and the NP ‘object’ following V2 ?

The third property I would like to connect to properties (A) and (B) is the fact that historically there was a significant change in the word order in RVCs, and although the modern-day Chinese RVC order is [V1 V2 object], earlier it was [V1 object V2 ], as schematized below:

 Zoe Wu

c.

(C) Historical re-positioning of V2 Historically, sequences of [V1 object V2 ] changed to [V1 -V2 object] as the V2 re-positioned itself right-adjacent to V1 : V1 object V2 → V1 – V2 object

During the course of the paper, it will be argued that a consideration of the interaction of these three properties of RVCs leads to an analysis of the V2 elements as synchronically being aspectual suffixes diachronically derived from independent verbs via a critical process of restructuring. In this two-step process, it will be suggested that the V2 elements first became re-analyzed as the heads of a sentence-final completive aspect projection and then later became re-analyzed as suffixes for reasons relating to directionality.

. V2 -object predication in RVCs Considering first property (B) above, the assumption that there is a predication relation between V2 and the NP object which follows it is a central component of recent syntactic analyses of RVCs such as Zhou (1994) and Sybesma (1999). Sybesma (1999) suggests that underlying modern RVCs there is a small clause structure in which the V2 predicates directly of the NP as in (7) representing the surface string in (6): (6) xi-ganjing yifu wash-clean clothes (7)

VP V’ V1

VP (=Small Clause/SC)

xi NP

V’

yifu

V2 ganjing

Movement of the V2 to the V1 is suggested to result in the surface word order [V1 V2 NP] as shown. Note that the NP yifu ‘clothes’ in (6/7) is actually the underlying subject of the V2 ‘be clean’. Its interpretation as the object of the V1 is

Restructuring and the development of functional categories 

suggested by Sybesma to be largely pragmatically derived. Specifically, structure (7) is argued to give rise to the interpretation that the clothes became clean as the result of a washing event, and simply imply the conclusion that the clothes were directly washed during this event. A large number of RVCs do appear to be naturally analyzable in terms of such a predication structure and resemble English resultatives such as (8) below, in which Hoekstra (1992) has analyzed the adjective ‘red’ as predicating of the NP ‘the door’ in a small clause structure: (8) John painted [SC the door red].

However, synchronically it can be pointed out that significantly there are also many other instances of RVCs where there is no obvious predication relation between V2 and the following NP. This is true for both English and Chinese. For example, consider (9)–(11) below: (9) Mary beat [SC [NP Peter] stupid ]. (10) The detective checked [SC [the information] out]. (11) John looked [SC [NP the reference] up].

The small clause analysis essentially predicts the interpretations in (12)–(13) for these sentences: (12) expected interpretation of (9): Peter is stupid because of a beating event carried out by Mary. (13) expected interpretation of (10): The information is out because of a checking event carried out by the detective. (14) expected interpretation of (11): The reference is up as a result of a looking event carried out by John.

However, in (9) it is not the case that Peter is actually stupid after the beating event, in (10) it would be incorrect to conclude that the information is out (i.e. made known) as the result of the detective’s investigation, and relating to (11) it is meaningless to attempt to state that ‘the reference is up’. Essentially the elements ‘stupid’, ‘out’ and and ‘up’ only convey the intended meanings in (9)–(11) through combination with the V1 ‘beat’, ‘check’ and ‘look’, and there is no isolated genuine predication relation possible between ‘stupid’/‘out’/‘up’ and the object NP. The small clause analysis clearly leads one to expect that the ‘object’ NP and the following predicate should be able to stand in a regular predication without the need for a higher predicate, yet this is not so. The same

 Zoe Wu

is true of many examples of RVCs in Chinese. For example, (14) would not seem to result from combining the parts in (15): (14) wo liu-zhu-le ta. I retain-live-le he ‘I stopped him.’ (15) [ta zhu-le = he lived] + [this was a result of a retaining event] → small clause structure predicted interpretation: ‘His living terminated a holding event I carried out.’

Similarly example (16) does not mean: [we talked] + [the state that [Xmas arrived] terminated this talking event], that is, it does not mean we talked until Xmas. However, such an interpretation is what is expected from a small clause underlying structure such as (17): (16) women tan-dao-le shengdan-jie. we talk-arrive-le Xmas ‘We talked about Xmas.’ (17) women tan [-le shengdan-jie dao] we talk le Xmas arrive → small clause structure predicted interpretation: [We talked] + [the state that [Xmas arrived] terminated this talking event] = ‘We talked until Xmas.’

With other V2 elements if one attempts to isolate the object NP and the V2 into a single sentence predication, the result is unacceptable in any reading as in (18) and (19). (18) a.

ta mai-zhao-le nei-ben-shu. he buy-catch-le that-cl-book ‘He managed to buy the book.’ b. *nei-ben-shu zhao-le. that-cl-book catch-le ‘???’

(19) a.

ta xianzai zhu-jin-le xin-fangzi. he now live-enter-le new-house ‘He has now moved into the new house.’ b. *xin-fangzi jin-le. new-house enter-le ‘???’

Restructuring and the development of functional categories 

With intransitive V1 elements, the small clause analysis of RVCs suggests that the surface subject originates as the subject of the small clause and is predicated of by the V2 , shown in (20). (20) pao [sc ta lei-le] → ta pao-lei-le. run he tired-le he run-tired-le ‘He got tired as a result of running.’

However, isolation of the V2 and such a subject NP again frequently results in quite unacceptable forms, as shown in (21): (21) a.

ta zou-kai-le. he walk-open-le ‘He walked away.’

*ta kai-le. he open-le (!He became open.)

The compositional account of resultatives in which the meaning of a small clause encoding a predication relation is added to the meaning of a higher verb can therefore be argued to be inappropriate. Either the small clause often has no meaning on its own or its meaning may be quite different from the interpretation found in complex V1 -V2 sequences. Consequently, if one reaches the conclusion that synchronically there is in fact no necessary predication relation between V2 and the following NP, this raises the important question of how the V2 is syntactically licensed in RVCs and what the correct synchronic analysis of RVCs should be. Here a consideration of the historical development of the resultative construction is critically able to provide important clues helping us understand the underlying structure of modern day RVCs.

. The historical development of RVCs Three important observations relating to the development of RVCs can now be noted and grouped together in (22) below. Although synchronically not all V2 ’s are in a necessary predication relation with the object NP in RVCs, historically it has been observed that in earlier forms of Chinese the V2 elements which occurred in RVCs all did seem to have genuine predicative-like relations with the NP ‘object’. Specifically, Wu (1999) notes that phase-type V2 elements which do not appear to have any predication relation with the NP object only appeared from the Wei, Jin and Northern and Southern dynasty periods onwards, whereas fully predicative-type V2 elements regularly occurred in RVCs from the Qin and Han dynasty time.

 Zoe Wu

(22) a.

In modern Chinese V2 elements do not have any necessary predication relation with the following NP ‘object’. b. In the Qin and Han dynasties when RVCs first appeared, V2 elements always did seem to predicate of the NP ‘object’. Non-predicative (Wu 1999) phase-type V2 elements did not occur. c. Non-predicative phase-type V2 elements occur only from the Wei, Jin and Northern and Southern dynasty periods onwards. (Wu 1999)

Furthermore it should be noted that during these periods of classical Chinese, the NP ‘object’ visibly occurred in a position between the V1 and the V2 , hence in a position where the NP might be argued to be the subject of the V2 . The critical generalisation from (22) can now be summed up as in (23): (23) Important historical change in Chinese RVCs Originally the V2 and the NP ‘object’ showed clear signs of being in a predication relation (Qin/Han). Later on (Wei/Jin/Northern/Southern dynasties and through to modern Chinese) there are signs that the V2 and the ‘object’ NP are no longer in a necessary predication relation.

I would therefore like to conclude that a predication relation between the V2 and the ‘object’ NP did indeed exist in the earliest RVCs in the form of a small clause structure such as (24), with the predication possibly mediated via a small pro as shown. Later on however, the generalisations in (22) and (23) suggest that the underlying structure of RVCs must have undergone some change reflecting the fact that a V2 no longer occurred in a necessary predication relation with the object NP. (24) Earliest V1 -NP-V2 with ‘literal’ necessarily predicative V2 ’s

If (24) corresponds to the earliest fully predication-based RVCs, the question now is how such structures evolved into later RVCs without any necessary predication relation between the V2 and the object NP. Furthermore, if the V2

Restructuring and the development of functional categories 

element is licensed in early RVCs due to being in a predication relation with the NP, and if such a relation no longer exists in later RVCs, one needs to ask how V2 elements are legitimized in the structure of later and modern-day RVCs. Essentially it would seem that some other means of legitimization must have become available to allow non-predicational phase-type V2 elements to occur in the basic RVC structure. Here I would like to suggest that the emergence of phase-type verbs in the V2 position in RVCs is the result of a first important stage of re-analysis which re-oriented the resultative construction in a particular way, subsequently allowing in certain V2 elements not instantiating any predication relation with the object NP. I would like to suggest that what was critical here is property (A), the observation that all V2 elements provide an aspectual bound to the action depicted by V1 and the object. Prior to the occurrence of phase verbs as V2 ’s, all V2 elements can be said to have contributed the two discrete properties listed in (25): (25) a.

V2 ’s lexically encoded the type of result which terminated the event (e.g., the clothes come to be in a clean state after being washed). b. V2 ’s functionally provided a telic bound to the event of the V1 .

Intuitively, the (first) important historical change which occurred in RVCs can be suggested to be that after some time V2 ’s could be licensed in resultative constructions by providing simply property (b) and not necessarily property (a) as well. Phase-type V2 ’s which do not predicate of the object NP and which do not describe the result state of the object can therefore be said to have become licensed in the V2 position in RVCs in virtue of providing property (b) alone, an aspectual bounding of the predicate. Cross-linguistically, it might seem that both properties (a) and (b) above are regularly present in resultative constructions. Reflecting on their frequent co-occurrence one might initially assume that they are equally weighted in terms of relative importance and that the lexical information provided describing the result state of an object in a resultative construction is necessarily as important and highly-valued as the bounding effect which occurs. However, it is also quite possible that there actually is an imbalance in the relationship between the two properties and that one of the two properties is or becomes more highly valued than the other. For example, as will shortly be discussed further, it is likely that the lexical descriptive property (a) is of more importance in English ‘so that’ resultative forms, and that the focus of interest largely falls on the characterization of the result-state, with the bounding effect (property (b)) being more of a secondary pragmatic by-product:

 Zoe Wu

(26) Mary laughed at John so much that he became embarrassed.

In (26) the emphasis is felt to be on describing the type of result state (‘embarrassed’) rather than focusing the completion of an activity (i.e., the termination of laughing). In ‘so that’ type resultatives it would therefore seem that there is indeed a clear imbalance between property (a) and property (b), with the former being of greater importance. In other cases it is not unlikely that an imbalance might swing in the opposite direction and the telic bounding property might become of greater importance than the actual description of the type of result-state. This is precisely what I would like to suggest has occurred in Chinese resultatives. From an earlier stage where the descriptive property (a) was clearly important and necessary for an element to occur as a V2 , later on the emphasis in resultatives switched to the bounding property (b) and simple completion of an event. In this regard, it has often been noted that the completive function signalled by V2 elements in RVC’s does indeed carry a special focus in the construction, for example Li and Thompson (1981) state that: . . . the primary communicative function of an RVC is to comment on whether the result of an action did or did not, can or cannot, take place. (p. 57)

The emphasis is therefore on whether an action is completed, and not so strongly on the characterization of the type of completion. Furthermore, (and this point will also be returned to shortly), with a V1 such as xi ‘wash’ there is almost no choice in what one could use or expect to hear as a V2 element, and ganjing ‘clean’ (or the phase V2 hao ‘well’) is neutrally anticipated as the result state V2 rather than other elements. The characterization of the result state by means of ganjing ‘clean’ is then neither unexpected nor in real contrast with other states which could have resulted from a washing action. This therefore reinforces the view that it is not the lexical content of the V2 which is really in focus in RVCs but rather the telic bound of completeness which it marks through its occurrence in the V2 position. The hypothesis I would like to develop in order to account for the changes observed in RVCs in classical Chinese is consequently that a critical imbalance arose in the relative importance of properties (a) and (b) in the resultative construction. From being a construction in which the lexical description of the result-state was both important and necessary (hence only literal V2 ’s predicating of the object could occur), I would like to suggest that V2 elements later came to be valued primarily for their aspectual bounding function, with property (b) being the identifying and licensing characteristic of any V2 . Once ‘literal’ V2 ’s (i.e. those V2 ’s genuinely able to predicate of the object in RVCs)

Restructuring and the development of functional categories 

could be licensed in virtue of an aspectual bounding role in RVCs, such a development would have significantly allowed for ‘phase’ (i.e. non-predicative) V2 ’s to begin to occur in RVCs as well, licensed purely by their ability to add a telic bound to the activity described by V1 . Such a hypothetical process of change, if correct, implies the following three stages in early RVCs: (27) Step 1: a. Only literal V2 ’s occur in early [V1 obj V2 ] resultatives. b. Literal V2 ’s provide both properties (a) predication of the object, and (b) addition of a telic bound. c. Literal V2 ’s are licensed primarily due to property (a). d. At this point V2 elements which do not predicate of the object (i.e., phase V2 ’s) do not occur in [V1 obj V2 ] resultatives, and literal V2 ’s are predicated of the object. (28) Step 2: Literal V2 ’s which still provide both properties (a) and (b) come to be optionally licensed simply in virtue of providing property (b) and have a predominantly aspectual role in RVCs, encoding telicity. (29) Step 3: Licensing simply in virtue of property (b) allows phase-type V2 ’s, which do not supply property (a) at all, to also occur in early [V1 obj V2 ] resultatives.

If the above account is able to offer a principled and plausible explanation for the changes observed in RVCs in the Wei/Jin dynasty periods, the question now is how such a change might have been translated into the syntactic structure? If V2 elements came to no longer predicate of an object inside the VP and instead provided a more general telic bound to the whole predicate indicating the completion of an action, it can be argued that in terms of scope and configurationality, the V2 came to be interpreted as applying its function to the entire VP rather than just a sub-part of it (i.e., just the object). I would therefore like to suggest that the first important change which occurred in RVCs and which introduced phase V2 ’s into the construction was that the V2 position became re-interpreted as a functional completive Aspect head, and that the V1 and its object NP came to be projected as an argument of the V2 -Aspect head in a leftward specifier position of AspP, as in (30):

 Zoe Wu

(30) Early V1 -NP-V2 AspP Asp’

VP V1

Object

Asp0 (=V2)

(30) can be argued to accurately represent the hypothesized new relation between Asp/V2 , V1 and the object, and the development that Asp/V2 now in fact selects for the combination of V1 and the object as its single argument. In support of such an analysis and the assumption that Asp/V2 verbs come to select for a VP comprising the V1 and its object, it is revealing to note that language textbooks tend to list phase V2 elements as characterizing the whole of particular actions rather than particular object-types, as for example in (31): (31) verb dao

cuo zhao

literal meaning arrive

be wrong *arrive (no longer in use as single verb)

meaning as V2 indicates the continuation of an action up to a certain point or time to carry out an action incorrectly completion or attainment of an action

This basic characterization of phase V2 ’s as being appropriate for certain classes of situations also frequently results in textbooks actually including lists of the V1 verbs which are commonly used with each V2 , and naturally corresponds to the assumption that it is indeed the V2 which selects for the VP and its V1 head, rather than having any direct relation to the object of that V1 , as represented in the structure (30). The list extract in (32) is from T’ung and Pollard (1997): (32) V2 frequently occurs with: V1 -jian (sensory perception) kan (look), ting (listen), peng (collide). . . -kai (detachment) da (hit), la (pull). . . -zhao (attainment) zhao (search), mai (buy), shui (sleep). . .

‘Literal’ (i.e. potentially predicative) V2 ’s are importantly also hypothesized to have allowed for their reanalysis as V2 -Asp heads occurring in (30), and indeed the appearance of phase V2 ’s in RVCs is argued to have been critically dependent on the ability of literal V2 ’s to be licensed in RVCs primarily in virtue

Restructuring and the development of functional categories 

of their aspectual contribution to the construction and thus ‘open the door’ to the occurrence of non-literal phase V2 ’s (as per (28)). Such an assumption however should not be taken to imply that descriptive V2 elements occurring in Asp0 in a structure such as (30) would have automatically ceased to contribute any lexical content in the construction. It is well observed that when lexical elements grammaticalize as functional elements, they may continue to contribute lexical content in a structure for potentially long periods of time (see e.g., Bybee, Perkins, & Pagliuca 1994). Consequently, it can be asserted and maintained without contradiction that literal V2 ’s came to be formally licensed as functional elements in the Aspect head, just like phase V2 ’s, but nevertheless continued to provide certain descriptive content in RVCs. The suggestion that such descriptive V2 ’s allowed for their use in a primarily functional role as markers of aspect might however be independently questioned on the grounds that functional elements are commonly taken to comprise small closed sets rather than open-classes of items such as stative verbs. Here two points can be made in defence of the hypothesis that literal V2 came to permit their use as primarily functional morphemes (i.e. aspect markers). First of all, while it is true that certain functional heads may allow for their instantiation by a very limited set of morphemes (e.g. the T0 or D0 heads in many languages), this is actually not always so, and there are also instances where functional heads may be realized by significantly large numbers of different overt morphemes. One good example of this is the classifier position/CL0 which is taken to select for NPs (and head a CLP) in many languages of the world. In the languages of Southeast Asia in particular such classifier systems may often have a very large membership. Consequently, the existence of a potentially large class of functional elements in RVCs made up by literal/descriptive V2 elements is not something which should be taken to be questionable due to any assumed size restrictions on classes of functional morphemes. Secondly, the actual membership of the literal/descriptive V2 set is significantly not fully open, and so shows the common restricted-membership property of many other functional categories. Hence although one might assume that any descriptive verb should be able to occur as a V2 in an RVC, this is actually far from being true, as examples such as (33) to (40) show both in English and in Chinese, a fact also noted in Lüdeling (1997) for German: (33) a. *He cried me sad. b. *He laughed me happy. c. He laughed himself insane. (34) He ran me tired/*exhausted.

 Zoe Wu

(35) Sally painted the door red/*beautiful. (36) We laughed the speaker off the stage/*embarrassed. (37) Robert ran clear of the door/away/*exhausted. wo. (38) ??ta xiao-le-le he laugh-happy-le me intended: ‘He laughed and as a result I became happy.’ (39) ?*ta (shuo gushi) shuo-ku-le wo. he say story say-cry-le me intended: ‘He told a story which made me cry.’ (40) *ta ku-beishang-le wo. he cry-sad-le me intended: ‘He cried so that I became sad.’

Essentially it would appear that the secondary predicates which occur most easily and regularly in the above resultative constructions in both English and Chinese are those which are to some extent canonical and very general realizations of the action of V1 . Thus in (35) the very general adjective ‘tired’ occurs legitimately where the more particular ‘exhausted’ cannot. In other cases the resultative forms are semi-idiomatic as in (36) and do not allow other seemingly natural secondary predicates. Many other examples also occur where the V2 encodes a natural outcome of the action of V1 which might even be predicted from the content of V1 – thus the action of ‘washing’ naturally leads to the object washed being ‘clean’, and the ‘riding of a horse’ might also naturally result in the horse being tired. To the extent that the set of V2 elements is actually not open and may often be semi-predictable from the content of the V1 , this reinforces the assumption that the primary focus of RVCs is indeed on indicating completion of the action of V1 , rather than describing the particular type or result state. Where a non-canonical result state predicate occurs this will tend to attract the focus away from the emphasis on completion and so there will be two competing focal points, resulting in the oddity felt in many RVCs. The idiomatic nature of many resultatives also suggests that the result state may be so commonly expected from application of the V1 that the result state and the V1 become a quasi-lexical unit. Interestingly in other types of resultative constructions, the balance between properties (25a) and (25b) may be the other way round, and secondary predicate type elements seem to be licensed primarily in terms of their description of the result state. Here one finds that the set of secondary elements is indeed open and there are significant contrasts with the RVCs considered so far.

Restructuring and the development of functional categories 

For example, the English ‘so-that’ resultative with its natural focus on the type of end-state allows in any type of result state predicate in striking and direct contrast to the examples in (33) to (40), as seen in (41) to (43). Similarly, and of considerable importance here, Chinese has a second type of resultative construction which makes use of a morpheme ‘de’ to embed a resultative clause in a way which resembles English ‘so that’ resultatives. Significantly the Chinese de-resultative allows for the easy and natural use of those descriptive V2 elements which cannot occur in RVCs. Examples (44) to (46) therefore contrast directly with the earlier unacceptable cases of RVCs in (38) to (40) and indicate that RVCs and de-resultatives should be treated as two quite different types of resultative construction: (41) a. He cried so much that I got sad. b. He laughed so much that I became happy. (42) He ran so much that he got exhausted. (43) We laughed at the speaker so much that he got embarrassed. (44) ta xiao-de wo hen le. he laugh-de I very happy ‘He laughed and as a result I became happy.’ (45) ta (shuo gushi) shuo-de wo ku-le. he say story say-de I cry-le ‘He told a story which made me cry.’ (46) ta ku-de wo hen beishang. he cry-de I very sad ‘He cried so that I became sad.’

Furthermore, because the focus of such de- and ‘so that’ resultative constructions is more clearly on the description of the result state, such resultatives do indeed require that the secondary predicate predicates of the ‘object’ NP, in contrast to the situation in bare English resultatives and Chinese RVCs. Consequently non-literal phase V2 ’s cannot occur in de-resultatives as (47) to (48) show, with similar effects occurring in English ‘so-that’ resultatives (49) and (50): (47) *wo kan-de shu wan-le. I look-de book finish-le intended: ‘I finished reading the books.’ (48) *wo zhao-de shu dao le. I search-de book arrive le intended: ‘I searched and found the books.’

 Zoe Wu

(49) I looked the number up./*I looked (at) the number so that it was up. (50) I threw him out./*I threw him so that he was out.

To briefly summarize now, Sections 2–4 above have attempted to understand and analyze the first major diachronic change which occurred in Chinese RVCs and the switch from a construction which required that all V2 elements predicate directly of the object to RVCs in which this restriction on the V2 element no longer obtained. It was suggested that this was due to a shift in the way that V2 elements were formally licensed in RVCs, and from an original form in which the V2 occurred licensed as a genuine secondary predicate, the V2 underwent re-interpretation as instantiating an aspectual head encoding a telic bound on the event depicted by the V1 and its object. The syntactic realization of this change was argued to be that V2 elements became basegenerated in a discrete Asp-head selecting VP as its argument projected in a leftward specifier position. Finally, certain differences between RVCs and deresultatives were highlighted, offering further support for the suggestion that the primary function of V2 elements in RVCs is aspectual and the addition of a predictable/canonical telic bound to the activity described by the V1 . Having developed an account of early RVCs this far, we are now in a position to consider and make sense of the second important change in Chinese RVCs which transformed the construction into its modern-day form: re-positioning of the V2 element and the creation of verbal clusters. Section 5 will offer an account of how and why this important re-structuring took place and suggest that the change should be similarly attributed to the functional-aspectual properties instantiated by V2 .

. Object/V2 re-positioning and directionality Around the time of the Song dynasty, RVCs underwent a reconfiguration process and V2 elements began to occur in their modern position preceding the object and right adjacent to the V1 . This change is schematized in (51): (51) V1 NPobject V2 → V1 V2 NPobject

The obvious question which arises here is what actually triggered this diachronic re-positioning and what an understanding of the historical repositioning of V2 elements might be able to tell us about the synchronic structure of V1 -V2 forms? Two recent syntactic accounts of modern Chinese RVCs, Zhou (1994) and Sybesma (1999), both suggest that movement of V2 to V1 actually

Restructuring and the development of functional categories 

occurs in modern-day RVCs, and hence that the historical source structure [V1 Object V2 ] is in fact still present in the base-generated structure of modern RVCs. V1 -V2 clusters both historically and synchronically would therefore result from V2 -movement over the object NP. However, while one might attempt to invoke verb-raising as a possible mode of explanation for the development of V1 -V2 clusters, neither Zhou (1994) nor Sybesma (1999) offers any understanding of what might actually cause the movement of V2 to V1 , and there seems to be no obvious motivation for such movement. Phonologically the V2 elements are not clitic-like and so cannot be claimed to raise for any encliticization needs. Syntactically, the scope of V2 elements as completive aspectual markers is also the entire VP and not just the V1 , as indicated in (30). It is therefore not clear why the V2 should need to attach itself to the V1 as this might seem to represent a reduction in its scope from the entire VP to just V0 . As it is therefore difficult to see what could possibly motivate any movement of the V2 elements to attach to the V1 in RVCs, I would like to suggest here that the change in (51) in fact did not take place due to any movement but occurred via a rather different mechanism of structural re-analysis and was triggered by factors relating to word order and directionality.2 Earlier it was suggested that V2 elements became re-analyzed as the heads of an Aspect projection, predicating of the VP in the specifier of AspP as in (30). Such a structure is in certain ways similar to that often assumed for unergative intransitive verbs which are taken to project their sole argument NP in a specifier position. Now, however, I would like to argue that such a suggestion needs to be re-assessed and that the label ‘specifier’ here is actually rather deceptive. Consider again the basic structure which has been proposed for AspP, (30) repeated below: AspP

(30) VP

Asp’ Asp0

What needs to be reflected on is the way in which the term ‘specifier’ is understood and the way in which a specifier is licensed in the projection of a head. Canonically it might seem that a specifier has the following general properties: (52) Specifier properties a. a specifier is an XP position

 Zoe Wu

b. the XP in the Spec position is either directly selected by the Y0 -head of YP if base-generated there, or agrees with the Y0 -head if moved to the Spec position. c. the specifier closes off the YP d. the specifier is built into the YP after the complement of Y0 is combined with Y0 .

Complements have the general properties in (53): (53) Complement properties a. a complement is an XP b. the XP is directly selected by the Y0 c. the complement is combined with its selecting head Y0 as the first branching sister of Y0 .

Abstracting away from the issue of directionality in the projection of specifiers and complements, it might seem that in one sense a specifier is only really different from a complement in virtue of being built into a projection critically after the complement of the head has been combined with the head. Both specifiers and complements are maximal projections commonly selected by the head of a projection; the specifier simply happens to be inserted into the structure at a point later than the complement. While such an ordering relation does indeed result in a clear hierarchical difference when both complements and specifiers are projected as in (54) (assuming a Spec-initial complementfinal pattern for illustration), if a head were not to project any complement, the difference between a specifier and a complement becomes less obvious. If indeed one follows assumptions commonly-made in recent work that a node is not projected unless it is genuinely branching (i.e., there should be no vacuous non-branching nodes (see Chomsky 1995a and references therein)), then configurationally the difference between complements and specifiers reduces further, as illustrated in (55): (54)

Restructuring and the development of functional categories 

(55)

Now, considering the structure for AspP in (30), Asp0 is a head which is suggested to have a specifier which hosts the single argument of Asp0 (the VP) and will therefore never have a complement branching off Asp’ to its right. If one assumes that the Asp’ node should therefore not be projected (due to the no vacuous branching nodes constraint), (30) should instead actually be represented as (56): AspP

(56) VP

Asp0

I would now like to suggest that in such a structure, where the head Asp0 importantly never projects any complement XP off an intermediate Asp’ level, that the leftward branching XP (i.e., the VP) can no longer be determined as a specifier but is instead itself interpreted as a leftward-branching complement. In other words, one might suppose that a specifier is a relation which may possibly depend on the potential existence of a complement XP, so that an element can only be functionally determined as a specifier if there may also occur a hierarchically lower complement XP. Here in the AspP projection, the VP is essentially so similar to a complement of Asp0 that although we have been referring to it as a specifier, in fact in the structure in (56) the VP is functionally interpreted as a complement. The AspP structure in (30/56) is therefore critically interpreted as being a head-final projection. Chinese being a dominantly head-initial language (see, among others, Simpson & Wu 2002), the head-final nature of AspP can therefore be taken to be a structure which significantly conflicts with the general direction of selection in the language.3 In the work of Lehmann (1973), Vennemann (1974), Hawkins (1979) and through into the work of Chomsky (1981), and other more recent work, the assumption is made that there is constant pressure towards uniformity in the directionality of selection in a language and that languages are indeed canonically either head-initial or head-final. At least among the former authors, the

 Zoe Wu

view is also commonly held that if a language somehow exhibits a structure which goes against the basic directionality of selection in a language, there will be a general pressure for the language to re-align such a structure in harmony with the canonical direction of selection. Vennemann (1974) talks of a Principle of Natural Serialization as a force which operates to restore consistency in languages which become ‘inconsistent’ (i.e., not uniformly head-initial or head-final) as the result of external influence or disruptive language-internal change. Lehmann (1973) also argues that if inconsistency creeps into a language then there will be pressure for this inconsistency to be corrected, and Hawkins (1979) proposes that a Principle of Cross-Category Harmony will similarly lead to languages re-aligning any non-uniform directionality. Commenting on the above proposals, Mallinson and Blake (1981) point out that correction of this type may be explained in terms of natural economy, and that it leads to a reduction in category/word-specific rules: A disharmonic language requires more category-particular ordering rules and thus the grammar of such a language is more complex. (p. 417)

Much research and investigation has consequently shown that languages do tend to follow either a consistent head-initial or head-final ordering and that there would seem to be internal pressure for all categories to conform to the basic directionality present in a language. Against such a background, the sentence-final, head-final Aspect projection (56) argued to have arisen in early Chinese can now be suggested to be a structure in potential conflict with the general directionality of head-initial functional/complement selection in Chinese, and therefore under pressure to be either eliminated or re-aligned in some way, should the necessary circumstances and means be available. I would now like to suggest that the appropriate conditions for a satisfactory reanalysis did indeed become available in the Song dynasty, and that a Minimalist approach to the licensing of morphology is critical to seeing how such a reanalysis could in fact be formally effected and successful. In regular Minimalist approaches to morpho-syntax (Chomsky 1993, 1995b), X0 -level functional morphemes corresponding to syntactic head positions are base-generated either as independent word-level elements or alternatively as affixes attached to other lexical elements. In the latter case, licensing of the functional affix is suggested to occur either overtly if there is pre-Spell-Out movement to the checking functional head, or covertly via LF raising. With Chinese RVCs, the V2 element has been argued to have been originally inserted into Asp0 as a free standing functional morpheme. This however results in a structure which has the problem that it is functionally determined as be-

Restructuring and the development of functional categories

ing head-final and so in disharmony with the head-initial parameter-setting in Chinese. Critically however, there is a way in which such an exceptional structure potentially could be reanalyzed as being head-initial. Supposing it were to be possible for V2 elements to somehow be re-interpreted as aspectual suffixes attached to the V1 in the lexicon, it could then be assumed that the V1 -V2 unit is inserted directly into V0 and that the V2 -Aspect suffix is subsequently licensed at LF by covert movement to the higher Aspect head. Significantly it then also becomes possible for the relevant Aspect projection to occur generated as a fully regular head-initial functional projection selecting its VP argument in the canonical rightward direction, as represented in the structure (57). In such a scenario and reanalysis, at LF the V1 -Asp unit would raise to the licensing Asp0 -head, as represented by the broken line:4 (57) wo kan-wan shu. I read-Asp book ‘I read up/finish reading the book.’ TP Spec

T’

woi T0

AspP Asp’ Asp0

VP NP ti

V’ V

NP

kan-wan

shu

Re-analysis of V2 -Asp elements as suffixes attached to V1 in the lexicon and licensed via LF raising to Asp0 would therefore in theory allow for a grammaticalized sentence-final AspP to be re-interpreted as a regular harmonic head-initial structure and thus eliminate the hypothesized conflict in directionality. What is critical for the success of such a further development of the RVC structure is however surface V1 -V2 adjacency, as without this it should not be



 Zoe Wu

possible for V2 elements to be plausibly reanalyzed as suffixes added to V1 in the lexicon. This might therefore seem to be problematic for such a line of analysis, as the regular base position of objects in early Chinese RVCs is between the V1 and V2 breaking up their adjacency. However, significantly in the Song dynasty, conditions existed which conspired with inherent properties of RVCs to actually result in natural adjacency of V1 and V2 . Li and Shi (1998) note that during this time, many V1 -NP-V2 occurrences were replaced by a range of forms, all of which displaced the object to a different position in the syntactic structure. This included object topicalization, preposing of the object via the ba-construction, and preposing the object in the verb-copying construction. In the ba-construction, objects occur before the VP and are marked by the functional morpheme ba. With ‘verb-copying’, the main verb of a clause is fronted with its object/other arguments and a copy of the verb is retained in its base VP-internal position. When employed in an RVC, all of these strategies give rise to sequences in which the V1 and the V2 naturally occur adjacent to each other, as schematized in (58): (58) a. Object Subject V1 V2 (object topicalization) b. Subject ba-Object V1 V2 (ba-preposing) (verb copying) c. Subject [V1 O] V1 V2

Such common preposing of the object in RVCs can be suggested to have a natural functional explanation. Earlier it was suggested that the focus in RVCs is centred on the completion of the event represented by V2 . The object of the verb is therefore normally not in focus and is old presupposed information and most naturally placed in topic or pre-verbal position via one of the strategies indicated in (58). Another obvious discourse possibility is that presupposed objects would occur pronominalized, and as the most common pronominal representation form for inanimate objects (also frequently for animate objects too) is via a zero/phonetically covert pro form, this would similarly have resulted in V1 and V2 elements being pronounced adjacent to each other [V1 pro V2 ]. The net result of all the common displacement/zero representation strategies noted above was ever more common positioning of the V2 directly following the V1 in RVCs. It can now be argued that such frequently occurring linear adjacency of V1 and V2 would have very naturally allowed for the V2 to be re-analyzed as a suffix attached to V1 rather than being base-generated in an independent head position. Consequently, highly suitable conditions for a re-analysis of V2 in a way which would automatically permit a head-initial realignment of AspP were indeed available in the language at the time when the


second important change in the surface structure of RVCs took place. I would therefore like to suggest that general pressure for uniformity in the direction of selection combined with this frequent occurrence of linearly-adjacent V1 -V2 led speakers in the late-Tang and Song periods to re-analyze V2 -Asp as a suffix on V1 in the way outlined, this ultimately resulting in structures which do conform to the canonical direction of selection in Chinese. We have therefore now arrived at a reasoned understanding of the second historical change in Chinese RVCs, and importantly how it relates in a direct way to the first change examined in Section 4. The analysis proposed can also be suggested to naturally underlie and apply to the continuation of V1 -V2 RVC forms in the modern construction, so that synchronic RVCs can be analyzed as consisting of V2 elements which are formally licensed as completive aspectual suffixes when combined with V1 roots. The suggested account has consequently achieved the joint goals set out earlier of offering explanations both for the historical re-positioning of V2 elements and the current syntactic status of V1 -V2 RVCs, and done this in a way which links the two together in a principled way. The analysis developed over the course of the paper has taken the three properties of RVCs identified as critical in Section 2, notably (a) the lack of necessary predication between V2 and the NP ‘object’, (b) the telic bounding property of V2 , and (c) the historical re-positioning of V2 , and suggested how these interact in important ways to result in the present day V1 -V2 resultative. Considering the development of RVCs since the Qin and Han dynasties, it was suggested that there were in fact two critical stages to the re-analysis process. First, the paper has suggested that increased focus on the completive function of V2 led to its re-analysis as instantiating a sentence-final Aspect head and allowed in non-predicative phase V2 elements. Secondly, I suggested that a further important step of the grammaticalization process was the reanalysis of V2 as an aspectual suffix attached to V1 in the lexicon and raised to Asp0 for licensing in the covert syntax, this being driven by pressure for the AspP to be interpreted as a head-initial projection.5 Finally, it can be noted that in such an analysis, V1 -V2 resultative pairs are essentially formed rather like complex-predicates, combining the properties of two discrete morphemes in the lexicon, prior to licensing in the syntax. Consequently it can be suggested that any descriptive properties which a V2Asp may have in addition to its critical aspectual function may be simply added to those of the main V1 and can be applied directly to an object of V1 in a direct way from within the V0 position.



 Zoe Wu

. Summary and general conclusions If we now stand back from the analysis presented and attempt to highlight the key issues at play, a number of general points can be made. The initial goal of the paper was to arrive at an analysis of the underlying structure of modernday RVCs and in particular the status of the V2 elements which occur in the construction, with the following questions relating to the V2 being central to the investigation: a. Why did initial restrictions on the type of verb that could occur in the V2 position change in the Wei/Jin dynasties? b. Why did the V2 subsequently undergo repositioning in the Song dynasty? c. Why are there certain different restrictions on the type of V2 which can occur in RVCs in modern Chinese (i.e. typically canonical, expected results if a V2 is ‘descriptive’)? At the heart of the paper then has always been issues concerning the V2 and its changing status in Chinese, and we have tried to arrive at a synchronic analysis of the V2 informed by the past developments and patterning associated with V2 elements. In many ways the core issue in this investigation has been the division which exists between lexical and functional categories, and a re-consideration of the balance which exists between the descriptive and grammatical roles an element may play. What we have arrived at is the claim that certain ‘lexical’ categories may actually be used and re-employed to encode a primarily functional role, and that a range of apparently descriptive, lexical verbs now occur reanalyzed in RVCs as aspectual morphemes licensed by a higher functional projection (AspP). Furthermore, it can be suggested that this re-employment of lexical items may have the potential to give rise to a new functional category and projection in a language, in Chinese arguably resulting in a completive aspect projection which previously did not exist in the language (in the sense that previously there do not seem to have been any clear overt instantiations and evidence for such a projection). In arguing for such a view, what we therefore hope to have done is to show that it is crucially how elements are used/allow themselves to be used which determines whether they are ‘lexical’ and occur in N/V/Adj-type lexical head positions, or whether they are primarily functional and occur in or are licensed by purely functional projections. Centrally important here also was the assumption that lexical elements may be valued for a grammatical role in a construction in addition to their lexical content, and that new functional categories may come into existence when the valuation of a lexical element’s grammatical role becomes higher than that of its descriptive


contribution. A critical re-valuation of the balance between descriptive and grammatical content associated with the use of V2 elements has been argued to have driven the development of completive aspect in RVCs, and the imbalance between such properties was shown to still be visible in a comparison of modern RVCs and de-resultatives in Chinese. In closing the paper, it can be added that whilst it would certainly be possible to arrive at the paper’s analysis of modern day RVCs looking only at synchronic patterns in Chinese, a consideration of the diachronic patterning of RVCs has really provided the important initial impetus for exploring such a hypothesis and also strengthens the general plausibility of the account which has subsequently been arrived at. The present study therefore confirms again how important a role may be played by diachronic studies in the formulation and support of hypotheses about synchronic syntax. Finally, along the way, we can note that two other general suggestions concerning patterns of diachronic development have been made. First of all, the paper has offered an account of how affixation might arise between two nonadjacent elements not as the result of any movement and cliticization of a weak element to a stronger host, but via the regular removal of material which intervenes between two elements, causing natural adjacency and an environment conducive to morphological reanalysis. Secondly, the paper presented an argument supporting the potentially important role that directionality may have in language change, and how morphological reanalysis may result from the pressure for a uniform direction of selection in a language.

Notes . Note that the homophonous morphemes le occurring in post-verbal and sentence-final position in (1), (2) and other examples in the paper are commonly taken to be two different elements. “Verb le” is assumed to be either a perfective aspect marker or a past tense inflection, whereas “sentence(-final) le” is regularly described as being a morpheme which adds “present relevance” to a sentence. Because there is not full agreement on the status of these elements, I simply leave them glossed as LE here, in both cases. An extensive discussion of verb le is provided in Wu (to appear) in which it is argued that verb le has grammaticalized from an original source as a completive aspect marker to become a marker of both perfective aspect and past tense. . A second conceivable phonological way to attempt to account for the re-positioning of V2 elements might be to try to relate it to the development of bi-syllabic words in Chinese. Certain decay in the syllable structure of early Chinese resulted in an increased tendency towards bi-syllabic and bi-morphemic words in the Zhou and Han dynasties. It might there-



 Zoe Wu

fore be suggested that V2 elements attached themselves to V1 ’s as part of this change towards bi-syllabic words. However, a comparison of the time periods when these changes occurred indicates that such an explanation cannot be maintained. It is well-reported that the critical change from [V1 Object V2 ] to [V1 -V2 Object] resultatives only came about in the Tang dynasty (618-907 AD) and that [V1 -V2 Object] only fully replaced [V1 Object V2 ] sequences in the Song dynasty (960-1279 AD), which is well over 700 years later than the Han period and the time of the change in syllable structure. This casts in serious doubt any possibility that early phonological changes leading to necessary bi-syllabism were the trigger for the change from [V1 Object V2 ] to [V1 -V2 Object]. Instead it seems that [V1 Object V2 ] forms in fact existed quite legitimately for many centuries after the general switch to bi-syllabic forms in the Han period, and therefore that the change to V1 -V2 resultatives must have some other non-phonological trigger and explanation. . See here also Dryer (1992), Kiparsky (1996), and Fuß and Trips (2002) for discussion of the observation that similar VO-Aux type orders are cross-linguistically very rare. . Note that such raising would indeed occur at LF, given the observation that at Spell-Out V1 -V2 pairs occur to the right of VP-adverbs, and hence remain inside the VP. . For a comprehensive discussion of how the perfective aspect-marker -le combines with V1 -V2 units making use of Smith’s (1997) two-tiered inner/outer aspect model, see Wu (Chapter 6 in press).

References Bybee, J., Revere, P., & Pagliuca, W. (1994). The Evolution of Grammar. Chicago: University of Chicago Press. Chomsky, N. (1981). Lectures on Government and Binding: The Pisa Lectures. Dordrecht: Foris. Chomsky, N. (1993). “A minimalist program for linguistic theory.” In K. Hale & S. J. Keyser (Eds.), The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger (pp. 1–52). Cambridge, MA: MIT Press. Chomsky, N. (1995a). “Bare Phrase Structure.” In G. Webelhuth (Ed.), Government and Binding Theory and the Minimalist Program (pp. 383–439). Cambridge, MA: Blackwell. Chomsky, N. (1995b). The Minimalist Program. Cambridge, MA: MIT Press. Dryer, M. (1992). “The Greenbergian word order correlations.” Language, 68(1), 81–138. Fuß, E. & Trips, C. (2002). “Variation and change in Old and Middle English. On the validity of the Double Base Hypothesis.” Journal of Comparative Germanic Linguistics, 4, 171– 224. Givón, T. (1976). “Topic, pronoun, and grammatical agreement.” In C. Li (Ed.), Subject and Topic (pp. 149–188). New York: Academic Press. Hawkins, J. A. (1979). “Implicational universals as predictors of word order change.” Language, 55, 618–648. Hoekstra, T. (1992). “Aspect and theta theory.” In I. Roca (Ed.), Thematic Structure, its Role in Grammar (pp. 145–174). Berlin: Foris/Mouton.

Restructuring and the development of functional categories 

Kiparsky, P. (1996). “The shift to head-initial VP in Germanic.” In H. Thráinsson, S. D. Epstein, & S. Peter (Eds.), Studies in Comparative Germanic Syntax II (pp. 140– 179). Dordrecht: Kluwer. Lehmann, W. P. (1973). Historical Linguistics: An Introduction. New York: Holt, Rinehart and Winston. Li, C. & S. Yuzhi. (1998). “Hanyu dong-bu jiegou de fazhan yu jufa jiegou de shanbian” [Development of the ‘V-complement’ construction and changes of syntactic structures]. Studies in Chinese Linguistics, 2, 83–100. Beijing: Beijing Language and Culture University Press. Li, C. & Thompson, S. (1981). Mandarin Chinese: A Functional Reference Grammar. Berkeley: University of California Press. Li, Y. (1990). “On V-V compounds in Chinese.” Natural Language and Linguistic Theory, 8, 177–207. Lüdeling, A. (1997). “Strange resultatives in German: new evidence for a semantic treatment.” In R. Blight & M. Moosally (Eds.), Texas Linguistic Forum 38. The Syntax and Semantics of Predication (pp. 223–235). Austin, Texas: Dept. of Linguistics, University of Texas at Austin. Malison, G. & B. Blake (1981). Language Typology. Cross-linguistic Studies in Syntax. Amsterdam: North-Holland. Simpson, A. & Wu, Z. (2002). “IP-Raising, tone sandhi and the creation of S-final particles: Evidence for cyclic Spell-Out.” Journal of East Asian Linguistics, 11(1), 67–99. Smith, C. (1997). The Parameter of Aspect. Dordrecht: Kluwer. Sybesma, R. (1999). The Mandarin VP. Dordrecht: Kluwer. T’ung, P. & Pollard, D. E. (1997). Colloquial Chinese. London: Routledge. Vennemann, T. (1974). “Analogy in generative grammar: The origin of word order.” In L. Heilmann (Ed.), Proceedings of the 11th International Congress of Linguists 2 (pp. 79–83). Societa editrice Il Mulino, Bologna: Bologna. Wu, G. (1999). “The origin of the Mandarin sentence-final particle le.” Ms., Leiden University. Wu, Z. Grammaticalization and Language Change in Chinese: A Formal View. In press with Routledge Curzon, London. Wu, Z. “A Minimalist approach to the re-grammaticalization of morphology: Chinese verbal le as aspect and tense.” To appear in Linguistic Variation Yearbook, Vol. 3, Amsterdam: John Benjamins. Zou, K. (1994). “Resultative V-V compounds in Chinese.” In H. Harley & C. Phillips (Eds.), The Morphology-Syntax Connection: MIT Working Papers in Linguistics 22 (pp. 271– 290). Cambridge: MIT Press.

Index

A Abney, S. 6 Abraham, W. 24, 65, 75, 91 absolute state 17 accusative, see Case adjacency 24, 69, 70, 72–76, 79–81, 89, 93, 94, 107–109, 111, 211, 212, 215 adjacency requirement (between C0 and the subject) 69, 70, 74, 76, 79, 109, 111 adjunct 69, 107–110, 114, 125, 134–136, 171 adjunction 92, 93, 108, 135, 186 adverb 14, 70, 106, 135–137, 143 event-related adverb 135, 136, 143, 147 licensing 137, 138 manner adverb 137 modal adverb 137 phrase (AdvP) 135 VP-adverb 14, 144 Afrikaans 107–111 Agree 69–73, 77, 80, 92 agreement (Agr) 6, 11, 13–16, 37, 38, 40, 46–48, 52–56, 61–95, 108–111 Agr-on-C 63, 65, 67, 71, 73–75, 80–83, 89, 93, 94 Agr-on-T 71, 73, 80, 82, 89, 93, 94 complementizer agreement 20, 59, 61–63, 67–75, 80–82, 86, 88, 89, 92–94 default 109, 111, 118, 120, 121, 125, 127 dissociated Agr-morpheme 59, 72–74, 76, 80, 82, 89, 93 features 127

marker 66, 77, 78, 94 morpheme 62, 69, 72–74, 76, 77, 79, 80, 82, 92–94, 125 morphology 20, 23, 33, 46–48, 54, 59, 63, 65–67, 69, 79, 81–84, 86, 89, 90, 93, 94, 102, 105, 106, 108, 110, 111, 117, 124–127 morphology, loss of 46, 48 morphology, grammaticalization of 13, 20, 59, 62–66, 75–91, 94–96 object agreement 6, 7 paradigm 20, 59, 62, 63, 83–86, 89, 106, 111 paradigm, defective 84, 86 phrase (AgrP) 16, 17, 20, 31, 35, 44, 48, 55, 56, 108–111, 116, 125–127, 154, 178, 181 rich agreement 105, 108–111, 114, 116, 117, 126 Alexiadou, A. 12–14, 20, 31, 34, 35, 56, 125, 181, 187 Allen, C. 8, 45–47, 119–122 Altmann, H. 60, 61, 65, 68, 75, 93, 95 American Structuralism 3 Anagnostopoulou, E. 35 argument order 8, 101, 106, 118, 119, 122, 123, 125, 127 variation, see variation argument structure 78, 80 Ariel, M. 90 Arteaga, D. 49, 53 Aspect 22, 174, 175, 194, 201, 203, 207, 209–216 Austronesian 13 auxiliary 108, 109, 140

 Index

B background 140, 146, 147, 150, 153, 154 Bare Phrase Structure 6 Bavarian 20, 59–70, 73, 74, 78–86, 88–95 Bayer, J. 59–63, 65, 68, 70, 79, 81, 90, 92 Behaghel, O. 139, 140, 152, 153 Bennis, H. 68, 92 Berwick, R. C. 8, 9 Besten, H. den 92 bilingualism 11, 115 Binding Theory 24 Blake, B. 210 blocking effects 12, 20, 59, 62, 82–84, 90, 91 Blocking Principle 83–86, 90, 91 Bloomfield, L. 3 Bošković, Z. 182, 183, 187 Bobaljik, J. 14, 15, 23, 76, 93, 94, 109, 112, 113, 115, 124–126 Bopp, F. 2 Borer, H. 6 Brugmann, K. 2 Bulgarian 44, 56, 104 Bybee, J. 13, 90, 203

C Campbell, L. 10, 155 Cantonese 21, 169, 173–177, 179, 180 Cardinaletti, A. 20, 31–33, 35–37, 41, 42, 52–56 Carstens, V. 34, 68–70, 94 Case 8, 50, 52, 92, 109, 132, 178, 187 accusative 6, 7, 120, 121, 128, 132, 150, 178 dative 50, 120, 121, 128 checking 56 feature 69, 70, 118–121, 127, 178 genitive 24, 35, 49–51, 56, 121, 153 licensing 69, 143 morphology 20, 39, 102, 118–124, 127 morphology, loss of 47, 118–124, 128

nominative 5–7, 50, 59, 60, 120, 121, 128 Catalan 18 Chinese 22, 162, 169, 170, 172, 173, 177, 178, 191–193, 195–198, 200, 203–206, 209–215 Chomsky, N. 3, 4, 6–8, 10, 11, 22, 49, 69, 70, 76–78, 85, 92–94, 103, 133, 172, 208–210 Cinque, G. 6, 7, 36, 38, 56, 135, 174 Clark, R. 3, 5, 6, 9, 10, 77, 85 Classical Greek 2 clitic 16, 59–66, 77–83, 89–91 clitic possessor, see possessor object, see object subject, see subject comparative method 2, 18 comparative construction (in Bavarian) 70, 71 complement 22, 74, 91, 131, 134, 135, 139, 141, 145, 147, 148, 151–153, 157, 165, 166, 170, 174, 177, 208–210 complementizer 107, 108 complementizer agreement, see agreement phrase (CP) 4, 5, 64, 72, 74, 81, 104, 108, 109, 143, 181, 183 complex-predicates 213 Comrie, B. 90, 155 ConcordP 16, 17 construct state 17–19, 24 core grammar, see grammar CP, see complementizer cue 4, 10, 23, 113–115, 126 Cysouw, M. 90

D Danish 15, 110–112, 116, 117 dative, see Case definiteness 7, 32–35, 40, 48, 49, 127 Delbrück, B. 146 demonstrative 37, 39, 41–44, 47, 48, 51, 55, 56, 153 Demske, U. 32, 33, 39, 40, 45, 49–52

Index 

determiner 18, 31, 33, 38, 40, 43, 47, 48, 51, 52, 55, 56, 140, 152, 153, 157 phrase (DP) 22, 24, 31–35, 37, 38, 41–44, 48, 49, 56, 78, 92, 107, 119, 120, 127, 141, 156, 157 Diesing, M. 154 diffusion 11, 23, 95 directionality 191, 192, 194, 206–211, 215 dissociated Agr-morpheme, see agreement Distributed Morphology 63, 67, 68, 73, 84 do-support 23, 94, 115 Double Base Hypothesis 139, 144, 145, 147 double object construction 119–121, 124, 126, 127 DP, see determiner Dresher, E. 9 Dutch 107, 108, 125–127, 134, 135, 138, 140–143, 147 E economy 9, 75, 77, 78, 85, 90, 210 Ellegård, A. 23, 115, 134 ellipsis 32, 33, 36, 37, 47, 53–55 Embick, D. 72, 76 Emonds, J. 106 expletive 126, 132 empty expletive (pro) 80, 109, 110, 126, 127, 182 English 1, 4–5, 8, 9, 14, 20, 21, 23, 24, 31–34, 36, 49, 52–56, 67, 76, 80, 86, 93, 103–105, 108–111, 118–123, 125, 127, 131–135, 137–140, 143, 146, 153, 164, 166, 167, 195, 199, 203–205 Middle English (ME) 33, 44–48, 110, 120, 122, 128, 134, 139, 144, 145, 147, 181 Old English (OE) 3, 5, 32, 33, 37–39, 41–46, 48, 51, 53, 56, 110, 119, 134, 139–143, 144–147, 150, 151, 156, 157, 181, 182, 187

EPP 21, 161, 175, 179–182, 185 checking 35, 185 feature 21, 22, 161, 162, 179, 180, 185, 186 Ernst, T. 137 event-related adverb, see adverb extraposition 145, 146, 157

F Falk, C. 8, 132 Fanselow, G. 12–14, 125, 181, 187 finiteness 6 focus 17, 21, 75, 140, 149, 152, 154, 155, 162, 167, 176–178, 212 broad 154 contrastive 117, 154 narrow 154 phrase (FocP) 16, 17, 74, 75, 93, 154, 155, 176 restructuring 155 fossilized movement 22, 161, 162, 178, 180, 185 Frascarelli M. 155 French 5, 8, 16–20, 24, 31, 34, 36, 37, 41, 49, 52–56, 84, 90, 105, 113, 115, 126, 182–184 Old French (OF) 5, 24, 33, 53, 54 Frisian 68, 107, 108, 125–127 Fujimoto, S. 182, 184 functional category/head 16, 17, 24, 74, 76, 77, 162, 174, 179, 180, 185, 210, 214 Fusion 76

G Gelderen, E. van 24, 93 gender 39, 50, 52, 91 genitive, see Case German 4, 15, 20, 21, 31, 34, 36, 37, 49, 52–55, 59, 64, 80, 91, 107, 108, 125–127, 131–134, 137–143, 147, 148, 150–152, 155–157, 186, 203

 Index

Middle High German (MHG) 51, 150 Old High German (OHG) 33, 41, 49-51, 56, 64, 87, 88, 91, 139, 140, 150–153, 155, 157 Germanic 2, 14, 20, 21, 23, 40, 68, 69, 71, 86, 89, 93, 94, 102, 106–109, 111, 114, 117, 118, 123, 126, 127, 139, 140, 146, 154 Givón, T. 13, 24, 78, 192 grammar core grammar 21, 131, 133, 146, 148, 150, 151, 156 object of linguistic study 11 periphery 131, 133, 148 grammar change 10, 11, 102, 131, 132, 149 grammar competition 12, 103, 115–118, 122–124, 128, 132, 133, 139, 143, 157 grammaticalization 7, 20, 23, 24, 31, 33, 47–49, 52, 54, 55, 59, 62–65, 67, 82–84, 86, 89–91, 94, 95, 153, 213 Greenberg, J. 13, 140 Grewendorf, G. 69, 74, 75, 80, 91, 104 H Haeberli, E. 8, 20, 23, 48, 74, 101, 107, 108, 110, 116–118, 125–127, 181, 182 Haegeman L. 15, 68, 69, 92 Haider, H. 135, 137 Hale, M. 7, 11, 23 Halle, M. 22, 63, 67, 68, 72, 73, 75, 76, 82, 83, 90, 92, 94 Harley, H. 90 Harris, A. 10, 155 Hawkins, J. 24, 140, 209, 210 head complement parameter 131, 134, 135, 139, 145, 147, 148, 151 head movement 14, 15, 68, 70, 175 Head Movement Constraint (HMC) 175 head-final 142, 143, 187, 191, 192, 209–211 head-initial 142, 143, 163, 174, 191, 192, 209–213

Hellendoorn 69, 70, 74, 92, 93 Hinterhölzl, R. 21, 131, 135, 136, 138, 143, 146, 149 Holmberg, A. 14, 106, 108, 117 Hopper, P. 10, 24, 33 Hróarsdóttir, T. 132, 139 I I-language 10 Icelandic 14, 126, 127, 132, 133, 139, 140 inflection 6–8, 14–16, 37, 40, 50, 61, 67, 68, 71, 80, 81, 84, 89, 121, 215 inflectional morphology 7, 14, 15, 20, 59, 86, 90, 91, 101, 105, 106, 118, 123, 124 rich inflection 15, 39, 51 weak inflection 51 information-structure 152 input 4, 9, 10, 20, 21, 23, 77, 78, 114, 181 intervention effect 70 inversion 5, 6, 20, 42, 59, 64, 66, 69, 75, 82, 86, 87, 89, 101, 106, 114, 120, 122, 126, 127, 144 IP 4, 5, 16, 43, 64, 81, 104, 108, 126, 136, 138, 139, 142, 143, 146, 149, 152 Italian 4, 5, 18, 20, 32, 35, 36, 43, 68, 84, 89, 90, 103, 104 J Japanese 104, 182–184 Jespersen, O. 2, 17, 44 Jones, W. 2 Julien, M. 76, 187 K Kayne, R. 136, 147 Keenan, E. 7, 24 Kemenade, A. van 1, 8, 23, 140, 142, 143 Kinyarwanda 13 Kiparsky, P. 8, 22, 23, 63, 68, 82, 216 Koopman, W. 6, 119, 142 Kroch, A. 1, 3, 7, 11, 12, 21, 23, 82, 86, 102, 103, 114, 115, 132, 141, 142

Index 

L language acquisition 1, 3, 4, 9, 11, 24, 77, 81, 84, 85, 90 logical problem of 8 language change 1–4, 7, 8, 10–16, 19, 21–24, 33, 47, 48, 78, 85, 86, 102, 112, 131–133, 148, 182, 215 abrupt 5, 8, 10–12, 21, 132 gradual 8, 10–12, 21, 131, 132 logical problem of 8 language contact 9, 102, 139, 140, 156 language faculty 11, 13, 103 Larson, R. 17, 136 Late insertion 67, 71, 84 Latin 2, 17, 24, 50, 140, 150–152 learner 1, 9, 10, 19, 21, 62, 76–80, 82, 89, 103, 113–116, 124, 156, 178 least effort 9, 78, 85 left periphery 81, 93 Lehmann, C. 24, 209, 210 Lexical Parameterization Hypothesis 6 lexicon 84, 85, 90, 148, 211–213 light predicate raising (LPR) 131, 143, 145–148 Lightfoot, D. 1, 3–5, 7–11, 16, 17, 23, 78, 86, 113, 132, 133 Longobardi, G. 7, 17–19, 24, 32, 34, 35, 38, 56 Lower Bavarian 63, 65, 66, 92, 93, 95

M Malagasy 13 Mallinson, G. 210 Mandarin 13, 22, 170 manner adverb, see adverb Manzini, R. 6, 9 Marantz, A. 67, 68, 72, 73, 76, 82, 92–94 markedness 9, 133, 156 Middle English (ME), see English Middle High German (MHG), see German Mitchell, B. 41, 76 modal adverb, see adverb modal particle 74, 75, 93

modal verb 162–169, 171–178, 180, 181, 185 Mood 76, 87, 88, 90, 166, 174, 175 phrase (MoodP) 186 morphological change 7, 112, 114, 116, 117, 128 Morphological Merger 73, 76, 77, 79, 93 Morphological Structure (MS) 67, 70–73, 77, 79, 92–94 morphology 2, 7, 14, 15, 21, 39, 68, 73, 91, 101, 110–127, 162, 180, 181, 192, 210 agreement morphology, see agreement case morphology, see Case movement, see syntactic movement N N-movement 15, 41, 42 N-to-D movement 24, 38 negation 17, 93, 94, 113, 164, 166, 175, 180 phrase (NegP) 16, 175 negative concord 16, 17 neogrammarian 2, 12 Niyogi, P. 8 nominative, see Case Norwegian 110, 111, 116–118, 126, 127 Notker 132, 150, 153 nP 20, 31, 34, 35, 55, 56 NP 31, 34, 35, 37, 43, 55, 56, 104, 193–199, 201, 205, 207, 213 NP-movement 15 number 14, 20, 34, 39, 46, 48–50, 52, 56, 59, 62, 63, 82, 83, 85–91, 127, 187 phrase (NumbP) 34 O object

17, 22, 119–122, 124, 126, 127, 135–138, 147, 149, 167, 171–173, 175–177, 180, 181, 186, 187, 192–202, 204–207, 212, 213, 216

 Index

object agreement, see agreement clitic 74, 75 Old English (OE), see English Old French (OF), see French Old High German (OHG), see German optionality 21, 131, 132, 149 Ouhalla, J. 7 output 9, 67, 72, 133, 173 P p-movement 176, 178, 186 parameter 3–6, 8–10, 12, 23, 101, 103–106, 110, 123, 131, 132, 134, 135, 139, 145, 147, 148, 151, 155, 156 default setting 4, 9, 23, 85 parameter setting 3, 103, 104, 106 parametric change 4, 5, 7–12, 21 passive, impersonal 80 Paul, H. 2, 7, 18 person 14, 20, 39, 44, 46, 49, 50, 59–66, 71, 82, 83, 85–92, 120, 121, 127 Pesetsky, D. 69, 135, 136, 138 Pesetsky’s paradox 135, 136, 138 PF 70–72, 75, 76, 119, 125, 133, 138 phase 69, 187, 193 phi-features 44, 69–71, 77, 92 phonological phrase 148, 149, 155 Pintzuk, S. 1, 8, 23, 37, 41, 42, 132, 139–142, 145, 146, 157, 187 Platzack, C. 8, 14, 106, 125 Poletto, C. 68, 90 Pollock, J.-Y. 6, 106 Polo, C. 120, 122 possessive adjectives 20, 31, 33, 35, 41, 47, 53, 54 possessive determiner 52 possessive pronoun, see pronoun possessor 32–35, 127 clitic 20, 31, 35–37, 42, 43, 45, 47–49, 52, 53, 55 strong 35, 41, 56 weak 35, 37, 43, 44 preposition 17–19 phrase (PP) 18, 54, 56, 70, 74, 138

Preservation of Argument Structure 78, 80 Primary Linguistic Data 84, 113 Principles and Parameters 3, 4, 6, 12, 101, 103 pro 62, 78–80, 82, 89, 92, 182, 186, 198, 212 pro-drop 5, 8, 20, 59, 61, 62, 65, 78–82, 86, 88, 89, 105, 106, 110 probe-goal mechanism 69–71, 92 pronoun 5, 13, 15, 16, 46, 47, 60, 61, 63, 64, 76–78, 87, 89, 156, 181, 182 possessive pronoun 31, 32, 39–41, 44, 45, 49–52, 55 possessive pronoun, tripartite distinction of 20, 33, 54 subject pronoun, see subject prosodic constraints 148, 149 Proto-Indo-European 2 Proto-Romance, see Romance

R reanalysis 3, 13, 17, 19–21, 59, 62–64, 75–81, 84, 87, 89, 93, 94, 132, 146, 150, 155, 161, 162, 179–181, 184, 191, 192, 202, 210, 211, 213 categorial reanalysis 19, 75, 78 morphological reanalysis 215 remnant movement 167, 186 residual verb second, see verb second Resultative Verb Construction (RVC) 191, 193, 199, 200, 203, 211–213 Ritter, E. 34, 49, 90 Rizzi, L. 1, 4, 7, 93, 103–105 Roberts, I. 1, 3, 5, 6, 8–10, 14, 23, 24, 33, 77, 78, 85, 94, 106, 143 Robinson, O. 150, 151 Rohrbacher, B. 14, 106 Romance 17–19, 24, 35, 56 Proto-Romance 17–19 Roussou, A. 3, 24, 33, 77, 78, 85, 94 Russian 38, 56

Index 

S Sanskrit 2 Santorini, B. 115 Sapir, E. 3, 7, 118 Saussure, F. de 3 Scandinavian 8, 107, 110, 115–117, 139 Schlegel, F. von 2 Schoorlemmer, M. 31, 32, 56 scrambling 135, 138, 147, 149, 186 Semitic 17, 24 Shlonsky, U. 68 Simpson, A. 16, 17, 21, 161–163, 165, 167, 169, 170, 172, 182, 183, 186, 187, 209 SOV 15, 79 Spanish 18 specifier (Spec) 17, 20, 22, 31, 34, 35, 42-44, 48, 52, 55, 56, 74, 75, 104, 109, 116, 125, 126, 154, 155, 167, 170, 171, 175, 179, 186, 201, 206–209 Spell-out 125, 149, 172, 183, 184, 186, 187, 216 Starke, M. 35 structural adjacency 73–76, 80, 93 Structuralism 3 stylistic fronting 133 stylistic rule 131, 133, 145–147, 155 subjacency 103, 108 condition 4, 5, 104 subject 13, 20, 68–74, 78–83, 89, 92–94, 107–111, 114, 116, 117, 125–127, 133, 134, 152, 164, 165, 181, 186, 192, 194, 197, 198, 212 clitic 20, 59-66, 79, 80, 83, 87, 89, 93, 94 pronoun 15, 16, 60, 78 Subset Principle (during language acquisition) 9, 23 Subset Principle (determining Vocabulary Insertion) 83 SVO 14, 15, 21, 163, 181, 182, 187 Swedish 8, 106, 110, 111, 116–118, 126, 127, 132 Sybesma, R. 194, 195, 206, 207

syntactic change 5–7, 12, 43, 47, 111, 112, 114, 116, 128, 151 syntactic movement, loss of 9, 162, 181, 182, 184–186 syntactic variation, see variation

T Taiwanese 21, 161, 162, 169–173, 175–177, 179–181 target grammar 8, 10, 11, 77, 80 Taylor, A. 3, 142 telicity 22, 191, 193, 199–201, 206, 213 Tense (T) 6, 7, 23, 88, 92, 94, 109, 111, 125, 187, 215 phrase (TP) 72, 74, 75, 108–110, 125, 126, 186 Thai 21, 161–164, 168, 169, 172–177, 179 Theta-role 78 Theta-theory 75, 78, 80 Thráinsson, H. 109, 112, 124, 126 topic 13, 64, 75, 78, 212 phrase (TopP) 44, 52, 74, 75, 93 topic left dislocation 13 Torrego, E. 69 TP, see Tense Traugott, E. 10, 24, 33 trigger 77, 111, 161, 162, 178, 179, 185, 186 trigger experience 4, 8–10

U Universal Grammar (UG) 1, 3, 4, 10–12, 14, 24, 77, 86, 91, 101, 103, 104 underspecification 82, 91 Universal Base Hypothesis 131, 136, 147

V Vance, B. 7, 8 variation 11, 12, 23, 116, 117, 119–125, 149, 151 argument order 20, 118–125, 127

 Index

syntactic variation (cross-linguistic) 4, 7, 24, 101, 103–110, 112, 123, 125, 127 (basic) word order 102, 133, 134, 139, 140, 147, 149, 151, 154, 157 Vennemann, T. 209, 210 verb movement 1, 8, 76, 106, 110, 125 V-to-C movement 62, 64, 93 V-to-I movement 14–16, 23, 113, 115, 126 verb raising 141 verb second (V2) 5, 6, 10, 15, 22, 23, 64, 66, 79, 80, 82, 86, 87, 91–94, 106, 107, 114, 126, 141, 143, 146, 147, 181, 182, 187, 192–207, 210–216 residual 108 verbal mood 87, 88, 90 Vikner, S. 14, 15, 106–108, 111, 126 Vocabulary Insertion 67, 72, 76, 77 VP 17, 22, 23, 134–139, 142, 163–168, 172–178, 180, 201, 202, 206, 207, 209, 211, 212, 216 VP-adverb, see adverb VP-intraposition 131, 138, 145, 147 VP-remnant 176, 180, 186 vP 15, 35, 69, 70, 72, 76, 80, 187 VSO 15 W Warner, A. 1, 125 Watanabe, A. 182, 184 Weiß, H. 60, 61, 65, 66, 68, 79, 91

West Flemish 68, 69, 92, 107–110, 125, 126, 141 Wexler, K. 6, 9, 10 wh-movement 104, 181–184 wh-questions 104, 183 Wood, J. 37–44, 55 Word building constraint 77, 79 word order 20, 21, 24, 114, 117, 122, 131, 140, 183, 194 change 21, 22, 131, 132, 134, 150–152, 155, 187, 191, 193, 194, 206, 207 unmarked 131, 146–148, 155 variation, see variation within DP 24, 33, 37, 38, 41, 42, 44 Wu, Z. 7, 16, 17, 22, 170, 172, 186, 191, 197, 198, 209, 215, 216 X X-bar theory 6 XP-subject order 20, 106–111, 113–118, 123, 124, 126, 127 Y Yiddish 107, 108, 115, 125–127, 154, 155, 157 Z Zubizarreta, M. L. 176 Zwart, J.-W. 61, 68, 92, 143

In the series Linguistik Aktuell/Linguistics Today the following titles have been published thus far or are scheduled for publication: 15 ROHRBACHER, Bernhard Wolfgang: Morphology-Driven Syntax. A theory of V to I raising and prodrop. 1999. viii, 296 pp. 16 LIU, Feng-Hsi: Scope and Specificity. 1997. viii, 187 pp. 17 BEERMAN, Dorothee, David LEBLANC and Henk van RIEMSDIJK (eds.): Rightward Movement. 1997. vi, 410 pp. 18 ALEXIADOU, Artemis: Adverb Placement. A case study in antisymmetric syntax. .. 1997. x, 256 pp. 19 JOSEFSSON, Gunlög: Minimal Words in a Minimal Syntax. Word formation in Swedish. 1998. ix, 199 pp. 20 LAENZLINGER, Christopher: Comparative Studies in Word Order Variation. Adverbs, pronouns, and clause structure in Romance and Germanic. 1998. x, 371 pp. 21 KLEIN, Henny: Adverbs of Degree in Dutch and Related Languages. 1998. x, 232 pp. 22 ALEXIADOU, Artemis and Chris WILDER (eds.): Possessors, Predicates and Movement in the Determiner Phrase. 1998. vi, 388 pp. 23 GIANNAKIDOU, Anastasia: Polarity Sensitivity as (Non)Veridical Dependency. 1998. xvi, 282 pp. 24 REBUSCHI, Georges and Laurice TULLER (eds.): The Grammar of Focus. 1999. vi, 366 pp. 25 FELSER, Claudia: Verbal Complement Clauses. A minimalist study of direct perception constructions. 1999. xiv, 278 pp. 26 ACKEMA, Peter: Issues in Morphosyntax. 1999. viii, 310 pp. 27 RŮŽIČKA, Rudolf: Control in Grammar and Pragmatics. A cross-linguistic study. 1999. x, 206 pp. 28 HERMANS, Ben and Marc van OOSTENDORP (eds.): The Derivational Residue in Phonological Optimality Theory. 2000. viii, 322 pp. 29 MIYAMOTO, Tadao: The Light Verb Construction in Japanese. The role of the verbal noun. 2000. xiv, 232 pp. 30 BEUKEMA, Frits and Marcel den DIKKEN (eds.): Clitic Phenomena in European Languages. 2000. x, 324 pp. 31 SVENONIUS, Peter (ed.): The Derivation of VO and OV. 2000. vi, 372 pp. 32 ALEXIADOU, Artemis, Paul LAW, Andre MEINUNGER and Chris WILDER (eds.): The Syntax of Relative Clauses. 2000. vi, 397 pp. 33 PUSKÁS, Genoveva: Word Order in Hungarian. The syntax of Ā-positions. 2000. xvi, 398 pp. 34 REULAND, Eric J. (ed.): Arguments and Case. Explaining Burzio’s Generalization. 2000. xii, 255 pp. 35 HRÓARSDÓTTIR, Thorbjörg: Word Order Change in Icelandic. From OV to VO. 2001. xiv, 385 pp. 36 GERLACH, Birgit and Janet GRIJZENHOUT (eds.): Clitics in Phonology, Morphology and Syntax. 2001. xii, 441 pp. 37 LUTZ, Uli, Gereon MÜLLER and Arnim von STECHOW (eds.): Wh-Scope Marking. 2000. vi, 483 pp. 38 MEINUNGER, Andre: Syntactic Aspects of Topic and Comment. 2000. xii, 247 pp. 39 GELDEREN, Elly van: A History of English Reflexive Pronouns. Person, Self, and Interpretability. 2000. xiv, 279 pp. 40 HOEKSEMA, Jack, Hotze RULLMANN, Víctor SÁNCHEZ-VALENCIA and Ton van der WOUDEN (eds.): Perspectives on Negation and Polarity Items. 2001. xii, 368 pp. 41 ZELLER, Jochen: Particle Verbs and Local Domains. 2001. xii, 325 pp. 42 ALEXIADOU, Artemis: Functional Structure in Nominals. Nominalization and ergativity. 2001. x, 233 pp. 43 FEATHERSTON, Sam: Empty Categories in Sentence Processing. 2001. xvi, 279 pp. 44 TAYLAN, Eser Erguvanlı (ed.): The Verb in Turkish. 2002. xviii, 267 pp. 45 ABRAHAM, Werner and C. Jan-Wouter ZWART (eds.): Issues in Formal German(ic) Typology. 2002. xviii, 336 pp.

46 PANAGIOTIDIS, Phoevos: Pronouns, Clitics and Empty Nouns. ‘Pronominality’ and licensing in syntax. 2002. x, 214 pp. 47 BARBIERS, Sjef, Frits BEUKEMA and Wim van der WURFF (eds.): Modality and its Interaction with the Verbal System. 2002. x, 290 pp. 48 ALEXIADOU, Artemis, Elena ANAGNOSTOPOULOU, Sjef BARBIERS and Hans-Martin GÄRTNER (eds.): Dimensions of Movement. From features to remnants. 2002. vi, 345 pp. 49 ALEXIADOU, Artemis (ed.): Theoretical Approaches to Universals. 2002. viii, 319 pp. 50 STEINBACH, Markus: Middle Voice. A comparative study in the syntax-semantics interface of German. 2002. xii, 340 pp. 51 GERLACH, Birgit: Clitics between Syntax and Lexicon. 2002. xii, 282 pp. 52 SIMON, Horst J. and Heike WIESE (eds.): Pronouns – Grammar and Representation. 2002. xii, 294 pp. 53 ZWART, C. Jan-Wouter and Werner ABRAHAM (eds.): Studies in Comparative Germanic Syntax. Proceedings from the 15th Workshop on Comparative Germanic Syntax (Groningen, May 26–27, 2000). 2002. xiv, 407 pp. 54 BAPTISTA, Marlyse: The Syntax of Cape Verdean Creole. The Sotavento varieties. 2003. xxii, 294 pp. (incl. CD-rom). 55 COENE, Martine and Yves D’HULST (eds.): From NP to DP. Volume 1: The syntax and semantics of noun phrases. 2003. vi, 362 pp. 56 COENE, Martine and Yves D’HULST (eds.): From NP to DP. Volume 2: The expression of possession in noun phrases. 2003. x, 295 pp. 57 DI SCIULLO, Anna Maria (ed.): Asymmetry in Grammar. Volume 1: Syntax and semantics. 2003. vi, 405 pp. 58 DI SCIULLO, Anna Maria (ed.): Asymmetry in Grammar. Volume 2: Morphology, phonology, acquisition. 2003. vi, 309 pp. 59 DEHÉ, Nicole: Particle Verbs in English. Syntax, information structure and intonation. 2002. xii, 305 pp. 60 TRIPS, Carola: From OV to VO in Early Middle English. 2002. xiv, 359 pp. 61 SCHWABE, Kerstin and Susanne WINKLER (eds.): The Interfaces. Deriving and interpreting omitted structures. 2003. vi, 403 pp. 62 CARNIE, Andrew, Heidi HARLEY and MaryAnn WILLIE (eds.): Formal Approaches to Function in Grammar. In honor of Eloise Jelinek. 2003. xii, 378 pp. 63 BOECKX, Cedric: Islands and Chains. Resumption as stranding. 2003. xii, 224 pp. 64 BOECKX, Cedric and Kleanthes K. GROHMANN (eds.): Multiple Wh-Fronting. 2003. x, 292 pp. 65 MANNINEN, Satu Helena: Small Phrase Layers. A study of Finnish Manner Adverbials. 2003. xii, 275 pp. 66 GROHMANN, Kleanthes K.: Prolific Domains. On the Anti-Locality of movement dependencies. 2003. xvi, 372 pp. 67 MIŠESKA TOMIĆ, Olga (ed.): Balkan Syntax and Semantics. 2004. xvi, 499 pp. 68 BREUL, Carsten: Focus Structure in Generative Grammar. An integrated syntactic, semantic and intonational approach. 2004. x, 432 pp. 69 KISS, Katalin É. and Henk van RIEMSDIJK (eds.): Verb Clusters. A study of Hungarian, German and Dutch. 2004. vi, 514 pp. 70 AUSTIN, Jennifer R., Stefan ENGELBERG and Gisa RAUH (eds.): Adverbials. The interplay between meaning, context, and syntactic structure. 2004. x, 346 pp. 71 GELDEREN, Elly van: Grammaticalization as Economy. 2004. xvi, 320 pp. 72 FUSS, Eric and Carola TRIPS (eds.): Diachronic Clues to Synchronic Grammar. 2004. viii, 226 pp. 73 CARNIE, Andrew, Heidi HARLEY and Sheila Ann DOOLEY-COLLBERG (eds.): Verb First. On the syntax of verb initial languages. Forthcoming 74 HEGGIE, Lorie and Fernando ORDONEZ (eds.): Theoretical Perspectives on Clitic and Affix Combinations. Expected Winter 04-05

Diachronic Clues to Synchronic Grammar (Linguistik Aktuell Linguistics Today)

Diachronic Clues to Synchronic Grammar (Linguistik Aktuell Linguistics Today)

Experimental Pragmatics Semantics (Linguistik Aktuell Linguistics Today)

Multiple Wh-Fronting (Linguistik Aktuell Linguistics Today)

Parentheticals (Linguistik Aktuell Linguistics Today, 106)

Argument Structure (Linguistik Aktuell Linguistics Today)

Phase Theory (Linguistik Aktuell Linguistics Today)

Causatives in Minimalism (Linguistik Aktuell Linguistics Today)

Cyclical Change (Linguistik Aktuell Linguistics Today)

The Grammar of Focus (Linguistik Aktuell Linguistics Today)

The Grammar of Focus (Linguistik Aktuell Linguistics Today)

Continuity and Change in Grammar (Linguistik Aktuell Linguistics Today)

Current Issues in Generative Hebrew Linguistics (Linguistik Aktuell Linguistics Today)

Theoretical Approaches to Universals (Linguistik Aktuell Linguistics Today)

Quantifier Scope in German (Linguistik Aktuell Linguistics Today)

Advances in Comparative Germanic Syntax (Linguistik Aktuell Linguistics Today)

Balkan Syntax and Semantics (Linguistik Aktuell Linguistics Today)

The Derivation of Anaphoric Relations (Linguistik Aktuell Linguistics Today)

The Emergence of Order in Syntax (Linguistik Aktuell Linguistics Today)

Grammaticalization As Economy (Linguistik Aktuell Linguistics Today, LA 71)

The Copy Theory of Movement (Linguistik Aktuell Linguistics Today)

The Linguistics Enterprise: From knowledge of language to knowledge in linguistics (Linguistik Aktuell Linguistics Today)

The Grammar of Repetition: Nupe grammar at the syntax-phonology interface (Linguistik Aktuell Linguistics Today)

Adverbials and the Phase Model (Linguistik Aktuell Linguistics Today, 177)

The Structure of Stative Verbs (Linguistik Aktuell Linguistics Today)

Perspectives on Negation and Polarity Items (Linguistik Aktuell Linguistics Today)

Movement Theory of Control (Linguistik Aktuell Linguistics Today)

Copular Clauses: Specification, Predication And Equation (Linguistik Aktuell Linguistics Today)

Agreement Systems (Linguistik Aktuell Linguistics Today, Volume 92)

Scrambling and the Survive Principle (Linguistik Aktuell Linguistics Today)

Syntatic Aspects of Topic and Comment (Linguistik Aktuell Linguistics Today)

Diachronic Clues to Synchronic Grammar (Linguistik Aktuell Linguistics Today)