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

Grammaticalization as Economy Linguistik Aktuell/Linguistics Today Linguistik Aktuell/Linguistics Today (LA) provid...

Author: Elly Van Gelderen

56 downloads 1672 Views 2MB 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

Grammaticalization as Economy

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 71 Grammaticalization as Economy by Elly van Gelderen


Elly van Gelderen Arizona State University

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 Gelderen, Elly van Grammaticalization as economy / Elly van Gelderen. p. cm. (Linguistik Aktuell/Linguistics Today, issn 0166–0829 ; v. 71) Includes bibliographical references and index. 1. Grammar, Comparative and general--Grammaticalization. 2. Economy (Linguistics). 3. Linguistic change. P299.G73 G45 2004 415.22-dc21 isbn 90 272 2795 0 (Eur.) / 1 58811 552 6 (US) (Hb; alk. paper)

2004046405

© 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

Acknowledgements

xiii

Notes for the reader

xv

List of tables

xvii

Part I Chapter 1 Introduction 1. Layers and clauses3 2. Grammaticalization across clauses5 3. Grammaticalization in a Generative Framework8 4. Economy10 5. Outline and methodological points13 Chapter 2 Economy 1. Spec to Head18 1.1 Pronouns19 1.2 Negation26 1.3 D-words27 2. Late Merge28 3. Conclusion33

3

17

Part II Chapter 3 The structure of CP and the Layer Parameter 1. The structure of CPs38 1.1 The Clause38

37

vi

Table of contents

2.

3. 4.

5.

6. 7.

1.2 The CP40 1.3 The split CP41 1.4 The ‘cartographic’ approach and some problems44 Cross-linguistic differences in the split CP46 2.1 Infinitivals46 2.2 Differences in finite clauses: Topics47 2.3 Double complementizers48 2.4 Concluding observations49 No split in the embedded OE clause51 Syntax and semantics55 4.1 Factives, non-factives, emotives, and non-emotives56 4.2 Assertive: A new term for non-factives and emotive-factives60 4.3 A structure62 Verb classes in the history of English64 5.1 Do matrix verbs matter?65 5.2 The fact that and it69 The frequent absence of (embedded) CP in Indo-European and Germanic71 Conclusion75

Chapter 4 Spec to Head: The rise of the (embedded) CP 1. Relatives in ModE: From Spec to Head77 2. From Spec to Head in OE relatives81 2.1 Old to Middle English81 2.2 That from specifier to head82 2.3 The introduction of a new specifier87 3. Complementizer that: From Spec to Head89 4. The complementizer whether: Between Spec and Head92 5. Other relatives96 5.1 Spec to Head96 5.2 No movement at all99 6. Conclusion99 Chapter 5 Late Merge: The rise of the split CP 1. Finite complementizers in ME101 1.1 For101

77

101

Table of contents

2. 3. 4. 5.

1.2 Till105 1.3 That105 Nonfinite complementizers in ME107 The left periphery111 Topic incorporation114 Conclusion117

Chapter 6 More Late Merge: Heads to higher Heads and Specs to higher Specs 1. From lexical to grammatical head120 1.1 N to P to C120 1.2 V to C123 1.3 From A to C123 1.4 P to C124 2. From grammatical to grammatical head126 2.1 From C to C126 2.2 D to C127 2.3 I to C128 3. From Spec to Spec129 3.1 Relatives129 3.2 Adverbs: From VP to CP130 4. Conclusion131

119

Part III Chapter 7 The IP, VP-shell, and their layers 1. The structure of IP and the VP-shell136 1.1 IP136 1.2 The VP-shell137 1.3 The cartographic approach142 2. Cross-linguistic differences regarding the IP142 3. IPs as reduced CPs; VPs as reduced IP144 3.1 From CP to IP to VP144 3.2 Characteristics of (CP and) IP-deletion145 3.2.1 DP-Raising and Scrambling: CPs versus IPs145 3.2.2 Binding: CPs and IPs146 3.2.3 Control and to: IPs versus VPs148

135

vii

viii Table of contents

3.2.4 Clitic Climbing: IPs and VPs151 4. Patterns of grammaticalization152 5. Conclusion154 Chapter 8 Changes in modals and have: Competition for ASP-hood 1. ModE Analysis156 1.1 Deontic and epistemic modals: Preliminary remarks156 1.2 Modals as ASP157 1.3 Complements160 1.3.1 Perfect and progressive auxiliaries160 1.3.2 Statives162 2. Double modals164 3. From V to ASP: Competition with ge166 4. Modals and have: Taking over the ASP position170 4.1 OE170 4.2 ME171 5. Late Merge and concluding remarks178 Chapter 9 Perception verbs and ASPect 1. PV complements and aspect179 2. PVs in ModE and Dutch: ASP and v182 2.1 Two kinds of see182 2.2 Additional evidence187 2.3 Other languages with a PV split190 3. OE and ME193 4. Conclusion199 Chapter 10 Aspect: The Tense Aspect Parameter and inner to outer aspect 1. The simple present and the progressive: The Tense Aspect Parameter201 1.1 ModE202 1.2 The progressive in other Germanic languages203 1.3 From aspect to tense207 2. Changes in ASP208 3. Other aspectual markers: do?211

155

179

201

Table of contents

4. Inner aspect: The E(xtent) Phrase212 4.1 English213 4.2 Dutch verbal prefixes215 4.3 OE prefixes218 4.4 E markers220 5. Giorgi & Pianesi: The demise of the infinitival ending and aspect221 6. Conclusion and further research224 Chapter 11 Late Merge: Heads to higher Heads 1. From lexical to grammatical229 1.1 N (to A/P) to I229 1.2 V to I230 1.3 A to I: Adverbs of time as T, of mood as M, and of manner as ASP231 1.3.1 Temporal adverbs231 1.3.2 Germanic MPs232 1.3.3 OE233 1.4 P to higher positions236 2. From grammatical to grammatical: From (P to) ASP to M to C237 2.1 The lower to238 2.2 Higher to241 2.3 Adverbs vs. Negation: Specifier vs. head245 2.4 Other accounts246 2.5 Conclusion247 3. Conclusion248

229

Part IV Chapter 12 The Layer Parameter and Pronominal Argument Languages 1. The Layer Parameter251 2. Characteristics of PAL/Polysynthetic Languages252 3. OE as (partial) PAL255 4. The switch to Middle and Modern English258 5. Conclusions259

251

ix

x

Table of contents

Chapter 13 Conclusion 261 1. Grammaticalization as Economy261 1.1 Head Preference, Late Merge, innovation, and prescriptivism261 1.2 Individual changes262 1.3 Clauses and layers263 2. Apparent counterexamples to uni-directionality264 2.1 High heads to low heads and heads to specifiers?265 2.2 Subjunctives and infinitives267 3. Principles and Parameters271 Notes

275

References

287

Index

307

Acknowledgements

I am very grateful for comments and discussion on content and form with Werner Abraham, Mariana Bahtchevanova, James Berry, Harry Bracken, Eleni Buzarovska, Claudia Felser, Eloise Jelinek, Bettelou Los, Anette Rosenbach, Lynn Sims, Chen Chen Sun, Johanna Wood, and several anonymous reviewers of the entire text or parts of it. Peter Soewarno’s help with Indonesian, Anju Kuriakose’s with Hindi, Chiara Lage’s with Italian, and Monica von Eggers’s with Swedish are much appreciated. Kees Vaes and the typesetter deserve many thanks. Parts of this work have been presented at the ICHL (International Conference on Historical Linguistics) 2001, TiLT (Tools in Linguistic Theory) 2002, NRG (New Reflections on Grammaticalization) 2002, LASSO (Linguistic Association of the Southwest) 2000 and 2002, GLAC (Germanic Linguistics Annual Conference) 2003, SHEL 2000, DGfS (Deutsche Gesellschaft für Sprachwissenschaft) 2000, and WECOL (West Coast Conference on Linguistics) 2003. I would like to thank the audiences at these conferences for helpful comments.

Apache Junction, AZ June 2004

Notes for the reader

I provide two kinds of glosses. The one is a word-for-word gloss, using abbreviated symbols and the other, enclosed in single quotation marks, provides a freer translation. I do not always provide both since the meaning is often clear from the word-by-word gloss; and sometimes a word-by-word gloss is redundant. In cases of ungrammatical sentences, I often leave the freer translation out. The glosses only list morphological features such as accusative or dative where this is relevant for the argumentation or meaning. Please see the list of abbreviations below for explanations of these. English will be divided into Old English, or OE, from 450–1150, Middle English, or ME, from 1150–1500, Early Modern English, or EModE, from 1500–1700, and Modern English, or ModE, from 1700 up to the present. C14 will mean fourteenth century; eC15 early fifteenth century; lC16 late sixteenth century; etc.

Abbreviations a acc anno ANS art asp aux BCE BNC

c cf. cl conj

Adjective or Adverb accusative Case year Algemene Nederlandse Spraakkunst (see bibliography) article Aspect auxiliary Before Common Era British National Corpus, quoted with BNC Manual abbreviation (www.thetis.uk) complementizer ‘confer’, or ‘see’ classifier conjunction

cp CSE d dat Diss DOE dp e

ECM e.g. EMOD

Complementizer Phrase Corpus of Spoken English, see Chapter 1, Section 6 Determiner dative Case Dissertation Dictionary of Old English (=electronic) Determiner Phrase Extent (see Chapter 7), i.e. internal aspect, or Event as in Reichenbach (1948) Exceptional Case Marking ‘for example’ Early Modern English (1500– 1700)

xiv Grammaticalization as Economy

emph ep erg Fin foc gen Gmc G&P HC H&K H&T I i.e. IE impf inf int IP K&K loc ME ModE MP MTA n negp n.d. nf np OE OED OHG OTA p PA PAL past

emphatic Extent Phrase (see Chapter 7) ergative Case Finite, part of the CP (see Chapter 3) focus genitive Case Germanic Giorgi & Pianesi Helsinki Corpus (see bibliography for periods) Heine & Kuteva Hooper & Thompson Inflection ‘that is’ Indo European Imperfective aspect Infinitive Intimate (Korean) Inflection Phrase Kiparsky & Kiparsky Location Middle English (1150–1500) Modern English (1700-) Mood Particle or Mood Phrase Mood Tense Aspect Noun Negation Phrase no date Non Finite Noun Phrase Old English (–1150) Oxford English Dictionary (see bibliography) Old High German Oxford Text Archive Preposition when independent; plural affix in gloss Pronominal Argument Pronominal Argument Language past tense

p(er)f pf poss progr PV PVC QP r S sh SOV Spec subj subord s.v. SVO t T TMA TP V V2 VoiceP vP VP vs. 1 2 3 7 * % ?

» or > < # |

perfective aspect perfective features possessive progressive aspect Perception Verb Perception Verb Complement Quantifier Phrase Reference Singular in glosses, or Speech in Reichenbach (1948) Subject Honorific (Korean) Subject Object Verb Specifier Subjunctive Subordinator sub verbo ‘see entry for’ Subject Object Verb trace Tense Tense Mood Aspect Tense Phrase Verb Verb-second Voice Phrase, used for passives and transitives The highest verb, often a light verb, in the VP shell Verb Phrase versus, ‘as against’ first person second person third person and in OE ungrammatical or proto-form unattested marginal becomes derives from number indication line division

List of tables

Table 2.1. Table 2.2. Table 3.1. Table 3.2. Table 4.1. Table 5.1. Table 7.1. Table 7.2. Table 8.1. Table 10.1. Table 11.1. Table 11.2. Table 13.1. Table 13.2. Table 13.3.

First and third person singular pronouns versus nouns in CSE-FAC21 Cliticization to pronouns in CSE-FAC25 Split CPs51 English complements64 That versus who in the CSE79 Ps in the Prologue of the Cotton Cursor Mundi109 EC complements in the BNC149 PC complements in the BNC150 Properties of English modals164 Changes involving ASP and E224 Split vs. non-split infinitives in spoken corpora246 Grammaticalization as Late Merge248 First person changes in Case in HC272 First person singular coordinated pronouns in HC and spoken BNC273 Coordination and person in the CSE-FAC273

Part I

Chapter 1

Introduction

1.

Layers and clauses

Independent clauses that are semantically related to each other tend to become structurally connected over time. A coordinator such as and can be used to connect two independent clauses or it can become a subordinator, used to embed one clause into the other. Further merging involves the embedded clause starting to have a dependent tense and to lack an independent subject. A final step is for one of the main verbs to become an auxiliary. Contrary to this (simplified) and uni-directional view of clausal change, it has been argued that there is very little historical evidence for this kind of grammaticalization. For instance, Harris and Campbell (1995: Chapter 10) provide some instances but urge caution, and Newmeyer (1998: 275) discusses counterexamples to the unidirectionality of this kind of grammaticalization. In this book, I will show that historical evidence can be found for these unidirectional changes and that they are expected within a grammatical theory constrained by Economy Principles. I argue that two Economy Principles drive grammaticalization. Change can also go the other way, generally for external reasons, which reverse or impede the effects of the Economy Principles. Clausal structure can be divided into an outer layer (the Complementizer Phrase, or CP), an inner layer (the Inflectional Phrase, or IP), and a layer that contains the verb and its arguments (the traditional Verb Phrase, or VP). Accordingly, there are different mechanisms of cross-clausal grammaticalization. The book is structured in terms of these three layers and the two Economy Principles. Previous work has shown that grammaticalization causes reanalysis, from preposition to complementizer and from lexical verb to auxiliary (Roberts 1993; van Gelderen 1993; 1998; Haspelmath 1997; Abraham 2004; Roberts & Roussou 2003). Dividing up the clause into CP, IP, and VP, I will show (a) that these labels are cover terms for categories to be expanded into three layers, (b) what the different paths of grammaticalization for each of these layers are, and (c) that languages choose different ‘layer strategies’. Regarding the expansion of

4


CP, I follow Rizzi (1997) in distinguishing a Force and a Finite Phrase in addition to Topic and Focus phrases. As for the IP, this will be divided into Tense, Mood, and Aspect. The VP can be seen in terms of a VP shell (as in Larson 1988), with a VP, vP, and at least one functional category (to be made clear below). The paths of CP grammaticalization are different from those of the IP and VP. The CP changes often involve phrases becoming heads as well as heads becoming higher heads (‘Late Merge’), whereas the ones for IP and VP typically utilize head to head changes. The principles underlying these changes will be discussed below, first briefly in Section 4 and then in more detail in Chapter 2. The philosophy behind this book is that languages differ in the setting of a macroparameter as to which layer(s) is(are) ‘favored’ and which layer will be expanded, e.g. CP into Force and Finite; IP into Tense, Mood, Aspect; VP as a shell with inner aspect. In Universal Grammar, therefore, there is a set of functional categories arranged in an order that doesn’t vary, but, unlike in Cinque (1999), not all languages make use of them all: (1) The Layer Parameter CP/IP Italian, Modern English, Modern Greek VP Mandarin, Chantyal,1 Nunggubuyu,2 Navajo

In Chapter 12, it is shown that (1) may follow from a larger parameter, namely one suggested by the work of K. Hale, Jelinek, and Baker. It is obvious that the definition of parameters, as in (1), is not the usual one, but is as in Baker (1995). In Chapter 10, I use a ‘traditional’ parameter to account for major changes in the tense-aspect system. My use of ‘principles’ also needs some explanation. Principles are “language-invariant statements” (Chomsky 1995: 25) and include locality conditions on movement (movement can’t be too far), Full Interpretation (each phrase must have an interpretation), and an Economy of Derivation Principle (avoid “superfluous steps”, Chomsky 1995: 27). The principles I assume in Section 4 and Chapter 2 below belong under Economy of Derivation. The theoretical framework will be (relatively conservative) Minimalism. I have tried to make it readable for linguists outside this framework. For instance, I have been very cavalier with notions such as feature checking since it isn’t crucial for the analysis. Critical for my analysis are theories of phrase structure and economy and these are outlined in Chapter 2. The notion of ‘phase’ (prominent in Chomsky 2001a) will be used in certain arguments, but just to describe that the VP shell and the CP have something in common. There are

Introduction

ultimately only two stages/phases in the derivation that are relevant for interpretation, namely CP and VP (to be precise the VP-shell), as in (1), cf. Diesing’s (1992) restrictive and nuclear distinction. Descriptively, however, it is useful to use IP during the stage(s) when the CP reduces to a VP. The outline of this introductory chapter is as follows. In Section 2, I provide further background on the terminology of ‘clause union’ and some examples of the different stages. Section 3 explores the tension between the generative framework the book uses and grammaticalization, and Section 4 formulates a model of how (some) instances of grammaticalization can be seen as Economy. The final section gives an outline for the entire book and describes the texts and corpora used. It also defends that use as being helpful with the exploration of our internal grammar (also known as competence or I-language).

2. Grammaticalization across clauses “Grammaticalization across clauses” is the title of a chapter in Hopper & Traugott (1993). Other terms for cross-clausal grammaticalization are “interlacing”, “clause-union”, “Satzverknüpfung”, “serialization”, “restructuring” (Rizzi 1978), “coherent infinitive” (Bech 1955), “auxiliation” (Benveniste 1968), “unisententiation” (Vennemann 1984: 635), and “verb-chaining”, to name but a few. The constructions that these terms apply to are varied. I will argue that there are basically three kinds of changes, one involving the outer or discourse layer, one involving the grammatical layer, and one involving the lexical verbs and arguments. In generative terms, these layers are called CP, IP, and VP respectively. Many different changes can be described as cross-clausal grammaticalization. Schematically, some of its stages can be represented as in (2), going from least to most grammaticalized. However, many more and different gradations are possible.3 Typically, in the most combined clauses, there is a shared subject, a dependence of one tense or mood on the other, and the absence of a complementizer: (2) a. b. c. d. e. f. g.

[clause] [clause] [clause] and/then [clause] [clause [that/wh clause]] [clause [SUBJ clause]] [clause [(C) (DP) non-finite clause]] [clause AUX V] [clause Affix-V]

5

6


Representative examples of each stage are given in (3). The syntactically unconnected clauses in (3a) from Yagua, an Amazonian language, show three syntactically independent clauses that are related in meaning. A fairly loose coordinate structure is given in (3b) from Old English (hence OE), as representative of stage (2b). In (3c), two versions of the same Middle English (hence ME) text are provided, showing the change from (2b) to (2c). In (3d), from Icelandic, the interpretation of the reflexive shows that the embedded clause is transparent. This indicates a clause-union. (3e), shows two examples from ME, representing the change from a subjunctive to an infinitival, as from (2d) to (2e). In (3f), a change from Early Modern English to Modern English (hence ModE) is provided, showing the change from (2e) to (2f). In (3g), from Sanskrit, a stage is shown where the desiderative is an affix, representative of stage (2g). Note that most languages contain all these constructions, but that through grammaticalization, one may change to the other. (2) is descriptive and, in this book, I will attempt to explain a number of the changes from (a) to (g) in structural terms: (3) a.

Sa-jaachíy. Sa-jaachíy. Sa-cááNNsiiy-tée-níí 3S-throw:spear 3S-throw:spear 3S-terminate-emph-3S munun~u jásiy savage there ‘He threw a spear twice and killed the enemy there.’ (from Payne 1985: 224, with only a morpheme representation given) b. 7 þy ilcan geare mon ofslog Æþelbald Miercna cyning on and the same year man killed Æþelbald Mercian king at Seccandune 7 his lic liþ on Hreapadune 7 Beornræd feng Seckington and his body lies in Repton and Beornræd ascended to rice 7… to throne and ‘And the same year Æþelbald, the Mercian king, was killed at Seckington, and his body lies in Repton; and Beornræd took the throne; and …’ (Anglo-Saxon Chronicle A, anno 755, Thorpe 1861: 86) c. þenne he þe treoweðe alre best on. þenne bi-swikes tu heom then he you trusts all best on then betray you him (Layamon, Caligula 1705) » Wan hii þe troueþ alre best. þan þou heom bi-swikest when he you trusts all best then you him betray ‘When he trusts you the best, (then) you betray him.’ (Idem Otho)

Introduction

d. Jon segir að Maria elski sig Jon says that Maria loves-subj him(=Jon) ‘Jon says Maria loves him.’ e. Comaunde him that he come order him that he come » I command him to come. (Visser 830) f. I saw her corall lips to moue (Shakespeare, Shrew I i 175) » I saw them move. g. devah somam pipa-sa-ti god-nom soma-acc drink-want-3S ‘The god wants to drink soma.’ (from C. Lehmann 1988: 203)

The changes represented in stages (2a) to (2c) involve the outer layer, the CP, namely the connection between the clauses, whereas the ones in (2d) to (2g) involve the inner layers, the IP and VP, namely the increasing interdependence of the two verbs (e.g. modals, causatives, see Givón 1991: 82 for a list of references). My aim is not to argue that the direction from (2a) to (2g) is diachronically correct, but to explain some of these changes through two Economy Principles. For instance, some of the ones that go from (2a) to (2c) are discussed in Part II, and some of the ones from (2e) to (2g) in Part III. There are a few apparent counterexamples to the change from (c) to (d) in English and Greek.4 I discuss this in the concluding chapter. Changes on occasion skip one of the stages in (2) and it could also be argued that there are in fact two strategies, as in (4), namely coordination and subordination, since there are numerous examples of (a) changing to (c) and of coordinated sentences changing to VP-coordination, the latter being closer to subordination than sentence-coordination. In addition, as will be shown in Chapter 6, there is some evidence for the two parallel branches of (4) rather than the single representation of (2):5 (4)

clause + subordinate strategy / [clause [that/wh clause]] [clause [SUBJ clause]] [clause [(C) (DP) non-finite clause]] [clause AUX V] [clause Affix-V]

clause coordinate strategy \ [clause] and/then [clause] [clause and VP] [clause V+V]

7

8


Coordinators become subordinators and vice versa but the other changes are unidirectional. This is better accounted for in (4). Also, there are languages that show little or no coordination, e.g. Chitimacha (Swadesh 1946: 331)6 and ones that show less subordination, e.g. OE (see Chapter 3) and Bambara (Givón 1979: 202). Givón (1995: 39) shows that conjoined structures are very frequent in (English) oral-informal discourse (namely 86% coordinate, where writtenformal has 36%). In this book, I will focus on the subordinate branch of (4), i.e. the left side. In most cases, there is no difference between arguments and adjuncts in terms of cross-clausal grammaticalization. For instance, the for-complementizer in English first develops with adjuncts (He laughed for he felt happy) but is quickly extended to argument clauses (She would like for us to do that) and the complementizer that is used for relative as well as for complement clauses. The Akkadian complementizer kiima ‘because’ develops into a complementizer for embedded complements (see Deutscher 2000: Chapter 4). Absolutive clauses (in Latin, Greek, and English, see Haiman & Thompson 1984) have the same syntactic realization, e.g. participles, as argument clauses do. There’s been a lot of debate about the terms paratactic, hypotactic, coordination, and subordination. As Stockwell & Minkova (1991: 402) say, “[i]t would make sense to use the terms ‘parataxis’ and ‘hypotaxis’ to label logic-semantic relations, and the terms ‘coordination’ and ‘subordination’ to label formally marked syntactic relations”. Rynell (1952) reviews a lot of the literature, and comes to a similar conclusion. Hopper and Traugott (1993: 170) use the features [±dependent] and [±embedded]. Paratactic constructions are neither; hypotactic ones are dependent but not embedded; and subordinate ones are dependent and embedded. In the remainder of the book, I use coordination for (2b) and subordination for (2c–e), but occasionally use paratactic and hypotactic as well. In short, one of the ways to describe clause-union is through stages, as in (2) or (4). One can categorize these stages by degree of clausal grammaticalization. I will now briefly discuss the tension between grammaticalization and a structural approach as in transformational grammar.

3. Grammaticalization in a Generative Framework Grammaticalization and Generative Grammar have had an uneasy relationship. Proponents of generative grammar see syntax as autonomous (see e.g.

Introduction

Chomsky 1957) whereas advocates of grammaticalization see meaning and function as the determining factors behind syntactic structure and of course behind change. Proponents of functionalism argue that the meaning of a verb such as see predicts that it will become an evidential marker. The meaning and function of a verb such as report or say also predict that the syntactic representation will be one with a full CP, as in (5a), not with an IP or VP, as in (5b): (5) a. I reported that she left early. b. *I reported to leave early.

Evidence for an autonomous syntax approach is provided by phenomena such as reflexives. The Binding Theory states that an anaphor, such as himself in (6), must be c-commanded by its antecedent: (6) [[George W’s] father] voted for himself in 2000.

Using the traditional definition of c-command (Reinhart 1983: 18), only George W’s father c-commands himself, not George W, and that is exactly how speakers interpret the sentence. So, this rule operates structurally regardless of pragmatic circumstances such as there might be in (6). In (6), it has to be the father that was voted for, even though George W was running for office in that year, and it would make more pragmatic sense if it would express that the son was voted for. The emphasis on function and meaning has prompted the one side to say there are no structural representations (e.g. Hopper 1987, cited in Newmeyer 1998) and the other side that there is no grammaticalization (Newmeyer 1998: 226). Lightfoot (1999: 83) argues that languages change gradually but that grammars change abruptly, meaning there is little place for grammaticalization in a theory of change. Many have attempted a middle position, e.g. Roberts (1993) regarding main verbs becoming auxiliaries, and Rochette (1988) regarding the relationship between syntax and semantics in sentential complementation. The use of theta-roles is a good example of the interplay between representation and meaning in a generative framework, since the meaning of a verb will determine whether an agent or experiencer occurs. Newmeyer (1998: e.g. 30–1) gives references for the different generative positions. The ‘middle’ position involves a continued assumption of the autonomy of syntax but with an interaction between structure and meaning. Semantic aspects of a verb may come to be seen as syntactic features. For instance, saw includes in its meaning ‘visually witness an event’. This semantic aspect becomes a syntactic feature, and as a result, a specific syntactic category is triggered where the feature is checked, e.g. through movement. As to whether

9

10


or not syntactic change is gradual or ‘catastrophic’, as Lightfoot (1999) calls it, I want to argue that the principles and parameters of Universal Grammar work differently. The two Economy Principles I use are obviously principles, and it appears that they do not lead to the immediate reanalysis of certain words into different categories. Roberts & Roussou (1999; 2003) undertake something similar to what I have attempted in this book and in other works (e.g. van Gelderen 1998; 2004), namely to account for grammaticalization in structural terms. They (2003:198–199) characterize the changes in three ways where individual changes are concerned: (a) loss of movement, as with English modals, (b) reanalysis due to loss of morphological endings, as in the creation of modal particles, and (c) a combination of the two. In the chapters below, I assume that the changes in (a) and (b) are due to Late Merge and those in (c) to the Head Preference Principle, two principles I’ll describe more in Section 4 and Chapter 2.

4. Economy Within the generative tradition (e.g. Chomsky 1986), syntactic structures are built up using general rules, such as that each phrase consists of a head (X in (7)), and a complement (ZP in (7)) and specifier (YP in (7)): XP

(7) YP

X′ X

ZP

In early work, this schema is quite strict, e.g. specifiers and complements are always full phrases. With the introduction of (minimalist) bare phrase structure in the early 1990s, this changes. A verb and a pronoun object can merge, as in (8), with one of the two heads projecting, in this case V: VP

(8) V see

D it

In Chomsky (1995: 246), the ‘bareness’ becomes even more extreme and the words project, as in (9):

Introduction

(9)

the the

book

Phrase structures are built using merge and move. ‘Merge’ combines two items, e.g. the and book, of which one projects into a higher level and transmits its categorial features. The VP domain is usually seen as the thematic-layer, i.e. where theta-roles are determined. One can think of theta-dispersion as a motivating factor behind merge, or having the structure determine the thetaroles (as in Hale & Keyser 1993; 2002). After functional categories such as I and C are merged to VP, ‘move’ (e.g. Chomsky 1995: 250) raises heads and phrases so that features such as agreement and Case7 can be checked in the IP and CP layers. Both Head–Head checking and Spec–Head checking occur. In this book, using general Minimalist principles, I argue that checking between two heads, also referred to as incorporation, is more economical than between a specifier and a head. This is formulated in van Gelderen (2001a; 2004) as (10), and is most obvious for pronouns, as will be shown in the next chapter: (10) Head Preference or Spec to Head Principle: Be a head, rather than a phrase.

Moving just features would even be more economical, as Chomsky (1995:262ff.) suggests, but this is only possible for covert (LF) movement since features on their own are unpronounceable at PF. Stated as in (10), the principle holds for merge (projection) as well as move (checking).8 Once a lexical item is a head, it can disappear because it typically merges with a head for checking reasons (see e.g. changes discussed concerning negatives in Chapter 2). In a number of historical changes, words go from Spec to head (to being lost), as I show in Chapter 4, and principle (10) accounts for that neatly. There is a movement in current generative work to do away with headmovement, and only to have remnant phrase movement (e.g. Mahajan 2000). This implies that what looks like a head is really a full phrase with everything but the head moved out. The empirical evidence is extremely difficult to obtain. Besides, it does make the grammar more complex. The phenomena described in this book could be characterized in this framework as well: an empty XP with just an X is cheaper to move. I will continue to use (10), however. Within recent Minimalism, there is a second economy principle, namely (11), (see e.g. Chomsky 1995: 348). Merge, as in (8), “comes ‘free’ in that it is required in some form for any recursive system” (Chomsky 2001b: 3) and is “inescapable” (Chomsky 1995: 316; 378):

11

12


(11) Merge over move.

Principle (11) says it is less economical to merge early and then move than to wait as long as possible before merging. This reduces to (12): (12) Late Merge Principle: Merge as late as possible.9

Only non-theta-marked elements can wait to merge outside the VP (Chomsky 1995: 314–5). I will therefore argue that if, for instance, a preposition has fewer semantic features and is less relevant to the argument structure (e.g. to, for, and of in ModE), it will tend to merge higher (in IP or CP) rather than merge early (in VP) and then move. Like (10), Late Merge is argued to be a motivating force of linguistic change. Both (10) and (12) are principles of Economy that work in grammars of speakers, as well as in leading children to build their grammars in a particular way (early child language is characterized by heads and by lexical material over grammatical).10 However, language does not only change in the direction guided by these Economy Principles. Jespersen (1921: Chapter 14, §6), in discussing the ‘Ease Principle’, which mainly deals with articulatory ease of pronunciation, puts it this way: “the correct inference can only be that the tendency towards ease may be at work in some cases, though not in all, because there are other forces which may at times neutralize it or prove stronger than it”. It may be that the new utterance is hard to understand, for instance, and is not economical for the hearer. Lightfoot (1979: 384) sees borrowing and expressivity as external factors. He distinguishes (p. 405) between “changes necessitated by various principles of grammar” and those “provoked by extragrammatical factors”, hence, between necessity and chance. In this book, I provide examples of cyclical changes of both chance (e.g. which prescriptive rule is enforced) and of necessity (e.g. Spec to Head). All of the changes show the interesting interaction, talked about in e.g. Jespersen (1921: Chapter 14, §6) as a ‘tug-of-war’, between economy and innovation. Economy eliminates specifiers; innovations reintroduce them (e.g. reinforcing not and wh-relatives); and prescriptive rules either stop their introduction (e.g. the ban on multiple negation) or try to stop their change to head (e.g. rendering whether phrasal by adding or not). This ‘tug-of-war’ is described mainly in Chapters 4 and 11. Under this view, grammaticalization is uni-directional, caused by structural factors.

Introduction

5. Outline and methodological points In this book, I will examine changes involving clauses that are related to each other. Part I (Chapters 1 and 2) provides an introduction to the main points of the book as well as to the Economy Principles. In Part II (Chapters 3–6), the CP is the focus, and in Part III (Chapters 7–11), the IP and VP are. Part IV (Chapter 12) argues for the Layer Parameter, and Chapter 13 contains a conclusion and a discussion of some apparent counterexamples. I’ll describe the chapters in a little more detail here. In Chapter 2, I outline the ‘Head Preference’ and ‘Late Merge’ Principles more and provide examples. These two principles account for many cases of clausal grammaticalization in structural terms. In Chapter 3, I look at the structure and characteristics of a clause, a CP and a split CP. I list criteria, most of which are known from the literature, but also argue that factive and non-factive complements — the terminology will be adapted — differ in whether the complement is a split/ complex CP or a simple one. This too differs across languages. I show that earlier English makes more use of independent clauses. The changes to ModE involve the introduction of complementizers; the change towards a more frequent use of infinitives; and the incorporation of topics into the clausal structure. The introduction of the basic CP often involves the Spec to Head change, whereas that of the split CP involves Late Merge. Thus, in Chapter 4, I use the ‘Head Preference’ Economy Principle to examine several types of changes involving the CP, among them, relatives, that-clauses, and whetherclauses. In Chapters 5 and 6, I use the ‘Late Merge’ Economy Principle to account for grammaticalization involving the CP and the split CP. I show that prepositions become complementizers, and that the latter end up triggering a split CP by moving to a higher C (Force as in Rizzi 1997). Chapter 7 explains the structure of the IP and vP/VP the way Chapter 3 does for the CP. In Chapter 8, I use a Late Merge account to explain changes in deontic modals (from V to v to ASP(ect)); Chapter 9 does the same for perception verbs (from V to v to ASP). Since Chapters 8 and 9 examine certain aspectual changes, in particular the reanalysis from inner to outer aspect, Chapter 10 examines the general changes in aspect. Not only does this chapter argue that an outer aspect is introduced and that a number of items compete for this position (e.g. the prefix ge-), it also looks at the fundamental differences among languages in whether the unmarked aspect is moment or interval (as in Cowper 1999). This difference is expressed by means of the Tense-Aspect Parameter. In Chapter 11, I return to Late Merge and provide a systematic list

13

14


of all of the lexical and functional categories that grammaticalize into auxiliaries (or I). Chapter 12 formulates macroparameters relevant to cross-clausal grammaticalization, in particular the Pronominal Argument Parameter and the Layer Parameter. Chapter 13 forms a conclusion to the book, with a list of the major changes, a discussion of some apparent counterexamples, and a return to the terminology of principles and parameters. I show that fast change is suggestive of a parameter resetting, but that slow change (e.g. the noun pronoun split) is indicative not of a change in a principle but of a change in pronominal status. Apart from data available in the linguistic literature (e.g. Heine & Kuteva 2002, here H&K, Visser 1963–1973, quoted as ‘Visser’ with page numbers only, and Mitchell 1985, quoted by volume and page numbers), I have made use of several corpora, namely the 100-million word British National Corpus, or BNC (thetis.bl.uk), the Helsinki Corpus of older English texts (HC, see Kytö & Rissanen 1988 for a description of this corpus), the Cobuild Corpus (titania.cobuild.collins.co.uk/form.html), and the 2-million word Corpus of Professional Spoken American English (www.athel.com). The latter is abbreviated here as CSE and contains transcriptions of committee meetings (COMM), faculty meetings (FACMT), and White House briefings (WH). The choice of which to use depends on the size of the query. For instance, if looking at all instances of saw, it is more convenient to select the smaller corpus. I have also made use of individual electronic-texts, made available by the Oxford Text Archive and the Dictionary of Old English project (DOE). If a sentence is provided without reference to Visser, the Oxford English Dictionary (hence OED), or other reference works, they are taken directly from these electronic sources. Mostly, I have provided bibliographical references for the primary sources (e.g. King Alfred’s works), but for some I have just used electronic resources (e.g. Jane Austen’s novels). If not listed in the bibliography (e.g. Asham’s The Scolemaster or Colville’s Boethius), they can be found in the DOE reference list or the HC one. I didn’t want to clutter up the bibliography with works I use only once or twice as primary sources. Generative linguists often view work with corpora and statistics not as indicative of the I-language (or competence) but rather of the E-language (also known as performance). Wasow (2002) reviews Chomsky’s views on the use of quantitative data. Some of the quotes he chooses show that what Chomsky has in mind is word choice or concordances. For instance, the choice of Nevada over New York seems irrelevant to underlying linguistic structures in:

Introduction

It seems that probabilistic considerations have nothing to do with grammar, e.g. surely it is not a matter of concern for the grammar of English that ‘New York’ is more probable than ‘Nevada’ in the context ‘I come from —’. (Chomsky 1962: 128)

In this book, I try to get at subtle grammatical changes and I find that data obtained from corpora and other electronic texts show systematic differences that are indicative of features and parameters of the internalized grammar. In contrast to synchronic generative linguists, historical generative linguists have embraced work with e.g. the Helsinki Corpus, the Dictionary of Old English, the Penn-Parsed Middle English, and the Brooklyn parsed Old English corpus (e.g. Pintzuk 1999, Wood 2003). Notable exceptions exist of historical linguists not making use of these corpora, e.g. Lightfoot (1999). As mentioned, Lightfoot argues that change in language is gradual, but change in grammar is abrupt. The latter is due to parameter resetting. His interest in abrupt changes may be the reason for his non-use of the corpora.

15

Chapter 2

Economy

In the previous chapter, I have briefly mentioned two Economy Principles that will be used in the remainder of this book as mechanisms of grammaticalization. One is the ‘Spec to Head Principle’, and the other is the ‘Late Merge Principle’. Spec to Head has not been discussed much in language change. It provides an explanation of the origins of complementizers and relative markers. Late Merge is mainly concerned with why heads go to higher positions in the tree (e.g. V > I), when they grammaticalize. This kind of grammaticalization has frequently been examined (Haspelmath 1989; van Gelderen 1993; IJbema 2002; Roberts & Roussou 2003), but not from the Late Merge point of view. As for other Economy Principles, Chomsky (1995: 265) argues that movement carries along as much as is absolutely necessary. In English, the most wellknown case of this is preposition stranding. There is tremendous prescriptive pressure ‘not to end a sentence in a preposition’,1 and therefore, pied-piping, as in (1), is found in formal writing, rather than the expected stranding, as in (2): (1) About who(m) were you talking? (2) Who were you talking about?

Variation between (1) and (2) is noticed (but not condemned) by early grammarians such as Bullokar in his 1586 grammar. Two centuries later, only (1) is allowed (e.g. Coar 1796).2 Preposition stranding is a good example of the tension between internal and external forces. Chomsky does not go into the question of head versus phrasal movement but the conclusion seems obvious. In cases where it is possible to strand, stranding is obligatory. Thus, it is cheaper to move a pronoun as a head than as a phrase, and checking will proceed through Head to Head rather than through Spec to Head. The outline is as follows. In Section 1, three cases of Spec to Head are discussed, namely, pronouns, negative markers, and demonstrative/possessive elements. In Section 2, Late Merge is outlined with examples of prepositions becoming complementizers.

18


1.

Spec to Head

The Principle formulated in the previous chapter is repeated here as (3): (3) Spec to Head or Head Preference Principle Be a Head, rather than a Phrase.

Stated as in (3), the principle holds for merge (projection) as well as move (checking). I will not in this book be concerned with objects, and I have therefore formulated (3) more narrowly, namely, in terms of a Spec rather than a phrase, but in principle this should hold for the merging of objects as well. I sometimes refer to the Principle in (3) as ‘Spec to Head’ and occasionally as the ‘Head Preference Principle’ and ‘Head-over-Spec’. Principle (3) predicts that structures will be built as in (4a) rather than as in (4b): VP

(4) a. D

V′

You

V

D

see

her VP

b. DP

V′ NP

D

see You

DP

V D

NP her

In fact, in Chomsky (1995: 246), it is suggested that labels, such as N, V, NP and VP, are superfluous and that lexical items, such as see, need not be labelled as heads, as in (4), but that instead we have (5): see

(5) You

see see

her

Economy

Both (4a) and (5) are in accordance with (3). One of the possible problems is the Head Movement Constraint: if her in (5) needed to move, it would have to cross another head position (not allowed according to Rizzi’s 1990 Relativized Minimality). This problem is a broad one and I will not attempt a solution here. Clitics in a non-bare phrase structure face the same problem, so whatever works for them will work for (5) as well. To provide evidence that a principle such as (3) is indeed operative, I’ll examine pronouns in ModE, negation, and D-words across time. 1.1 Pronouns I will argue that, unlike pronouns, nouns do not have the option to be constructed as the pronoun in (4a) above. Nouns, if they are to be argumental, have to have a D(eterminer). Pure NPs occur as predicates, vocatives, and adverbials, but not as subjects or objects, see e.g. Higginbotham (1985), Rothstein (1983), and Longobardi (1994). This is the reason why most researchers assume a D even in languages that do not have an overt one. Comparing pronouns and nouns gives (6ab): (6) a.

DP

b.

D

D

she

N coyote

When pronouns are coordinated or modified, they lose the ability to be heads. A possible structure for a coordinate pronoun is as in (7). This is the stage in the derivation when the pronoun has just been combined with the coordinator and, and with the pronoun or the coordinator can project, i.e. the result can be a Coordinator Phrase (as in Munn 1992; van Gelderen 1997a) or a DP, but never just a head: (7)

DP/CoP Co

D

and

I

The phrase marker in (7) will combine with a D or DP but this won’t make any difference for the final result which will be a phrase. This structural difference

19

20


between nouns and pronouns has implications for Case checking. (Uncoordinated) pronounds can check via head to head, whereas nouns will check via Spec to head. In contrast to (6ab), Cardinaletti & Starke (1995: 36) follow an older literature and analyze pronouns as being of three kinds: clitics are ‘deficient heads’, weak pronouns are ‘deficient XPs’, and strong pronouns are ‘nondeficient XPs’ (XPs being full phrases). In their discussion of, for instance, French, they argue that “the strong variant can be used only if the deficient variant is not accessible” (p. 33 bold type omitted), e.g. if an adverb separates it from a verb or when coordinated. The weak pronoun “remains an XP on the surface …, while … resisting coordination or modification” (p. 36). Being an XP while resisting modification seems incompatible and I will therefore reformulate Cardinaletti & Starke’s three-fold distinction as a two-fold one: elements are either X or XP, but whenever possible X. The first piece of evidence that pronouns typically function as heads, i.e. incorporate, involves modification. For instance, they are not preceded by adjectives or modified by (restrictive) relative clauses. The reason is that speakers prefer to construct them as heads. So, nice they is not possible,3 and relative clauses occur very infrequently.4 From the Corpus of Spoken Professional American English (hence CSE), I have selected the 161,000-word Faculty Meetings’ section, but the other parts are very similar. These transcripts show a split between nouns and pronouns. Thus, nouns such as faculty, departments and school(s) are coordinated over 10% of the cases, e.g. faculty occurs 353 times and is coordinated with and,5 as in (8), 62 times, i.e. 17.6%: (8) to try to tap into what students and faculty have an interest in doing. (CSE-FAC97)

Faculty occurs much more often than 353 but I have disregarded the modifying uses of faculty, as in (9): (9) but for the grieving faculty member who feels that he or she was dismissed … (CSE-FAC95)

For the noun school(s) in the same part of the CSE, the percentage coordinated is 16.4, namely 55 instances of school(s) with nine coordinated. In the same part, students are coordinated 51 times out of 367, which is 13.9%. Departments occurs 52 times of which 12 are coordinated, i.e. 23%. This use is very different from that of pronouns, as Table 2.1 shows. Thus, first person singular pronouns are coordinated less than 1% of the time.

Economy

Table 2.1.First and third person singular pronouns versus nouns in CSE-FAC, X-square is 418.061 p < .001 for first persons against nouns and 33.340 p < 0.001 for third persons against nouns uncoordinated coordinated I and me s/he and him/her faculty, students, school(s), and departments

3024 227 693

20 (=.66%) 4 (=1.73%) 134 (=16.2%)

total 3044 231 827

Personal pronouns can be ‘pushed into’ being full phrases through coordination, as in (10).6 Then, the Case is often a default Case such as the accusative in (10): (10) Huck and me ain’t cry-babies. (Mark Twain, Tom Sawyer, Chapter 16)

This is true in most colloquial usage, as (10) represents (through an author’s filter), and (13) below, and provides more evidence that checking in (10) is different from that with non-coordinated pronouns. Other, less direct, evidence that pronouns are heads is found in varieties of English, e.g. North East Yorkshire (Cowling 1915, cited in Chapman 1995) where agreement on the verb disappears if the pronoun is adjacent, indicating that the pronoun has ‘become the agreement’, and that means it is a head in T(ense) or I(nflection). In these dialects, there would be a choice of either a verbal ending or an adjacent pronoun. Chapman (1995) divides English dialects of the British Isles into three groups: those in southwest England and East Anglia where no inflection is left; those in southeast England where the -s ending is used throughout the paradigm; and those of the ‘North Country’, where “the use of -s as a marker of all persons singular and plural in the present is determined by the type of subject with which the verb agrees” (Chapman 1995: 36). If a subject pronoun is immediately adjacent to the verb, as in (11), there is no -s; if it isn’t, the -s occurs, as in (12): (11) I tell him not to. (12) I often tells him. (both from Chapman 1995)

This suggests that the pronoun is the agreement.7 The variety also shows that pronouns need not incorporate if, for instance, they are coordinated. Therefore, the agreement in (13) is as it would be if the pronoun was non-adjacent:

21

22


(13) him and me drinks nought but water. (also from Chapman 1995)

So, pronouns are heads in (11), but phrases in (12) and (13). A related tendency in this variety is that only the pronouns that are adjacent to the verb get nominative Case, as in (11), indicating that the nominative is for heads that check via adjunction to the verbal head. This tendency is not absolute, as (12) shows. Wolfram et al. (1999: 70) demonstrate how these same tendencies are reflected in American dialects. For instance, agreement tends to be lacking with pronominal subjects. Evidence from code switching that pronouns are heads involves subjects and finite verbs. In many languages (cf. Jake 1994; Nortier 1990), a switch between a full DP subject and a verb, as in (14) between English and Dutch, results in an acceptable switch, but in the case of a pronoun and verb, as in (15) again between English and Dutch, it does not: (14) Those awful neighbors schijnen altijd herrie te moeten maken8 those awful neighbors seem always noise to have to make ‘Those awful neighbors seem to always have to make noise.’ (15) *They schijnen altijd herrie te moeten maken they seem always noise to have to make ‘They seem to always have to make noise.’

This can be explained, as in e.g. MacSwan (1999), in terms of a prohibition against language mixing at the word level in the case of a pronoun and a verb. In the languages where pronouns and verbs can code switch, the pronouns are emphatic (see e.g. Jake 1994), i.e. phrasal. Examples of lexical (or emphatic) pronouns that can switch between French and Moroccan Arabic are French moi ‘me’ in (16) and Moroccan Arabic nta ‘you’ in (17): (16) moi dxlt I went-in-1S ‘I went in.’ (17) nta tu vas travailler you you go work ‘You go to work.’ (from Bentahila and Davies 1983: 313)

In (17), the non-lexical French pronoun tu is included as well, reinforcing the idea that the pronoun nta is really lexical and is like a full DP.

Economy

The same occurs in Dutch/Moroccan Arabic Code Switching, as Nortier (1990) has shown using naturalistic data. Code Switching occurs between the emphatic subject pronouns and a Dutch verb, as in (18) and (19): (18) humaya vergelijken de mentaliteit met de Islam they compare the mentality with the Islam ‘They compare the mentality with Islam.’ (19) ana ik vind ’t zo’n knuffeldiertje I I find it such a cuddly-little-animal ‘As for me, I think it is so cuddly.’ (from Nortier 1991: 165–6)

In (19), as in (17) above, the ‘real’ pronoun is repeated in Dutch. The reverse switch never occurs, i.e. a Dutch pronoun with a Moroccan Arabic verb, since Dutch does not have emphatic pronouns (in the framework of van Gelderen 2000, pronouns would be checking uninterpretable features whereas DPs and emphatic pronouns have interpretable features). Another piece of evidence that pronouns incorporate (i.e. move as heads) is that when pronouns are separated from the verb by an adverb as in (20) or interjection, as in (21), they are often repeated. This is reminiscent of subject doubling in e.g. Northern Italian and colloquial French: (20) they apparently — they’re involved in this. (CSE-WH96B) (21) We find they, you know — they both work comfortably, to be honest with you. (CSE-COMR6B97)

In coordination of verbs, as in (22), the pronoun is often doubled as well, unless the two verbs form a semantic unit, as in (23a), or a special subordinate, as in (23b): (22) a. but I do. And I bet Judith does. (CSE-COMR797). b. Yes I do, and I am absolutely convinced … (CSE-FACMT97) (23) a.

And I hope and expect that we could continue to improve… (CSE-WH94) b. when I try and work with elementary teachers and try to map stories … (CSE-COMR6A97)

Without formulating a theory of ellipsis, it is hard to arrive at relative numbers for the two constructions. I will leave that for further research. It is my impression that first person doubles more often than third person, however. If pronouns can choose to be heads or full phrases, depending on the syntactic environment, one would expect languages with different phonological

23

24


forms for these. Even though English has no special phrasal forms, there is a form which is only a head, namely ye, the reduced form of you where the vowel is a schwa. In English, ye is a clitic head9 and (24) with the vowel reduced to schwa is grammatical but the coordinated (25a) is not. Instead, the full form is required, as in (25b): (24) Ye can’t go there. (25) a. *Mary and ye can’t go there. b. Mary and you can’t go there.

In Swedish, Josefsson (2000: 738) argues that certain pronouns occur in the specifier of DP, as in Han den gamle vaktmästeren ‘he the old janitor’, even though regular Swedish pronouns are heads. This means there are two pronominal forms in Swedish, heads and specifiers. Languages such as Arabic and French have special phrasal (emphatic) forms and special head (clitic) forms, see (16) and (17) above. The French phrasal first and second person forms are moi ‘me’ and toi ‘you’; the heads are je ‘I’ and tu ‘you’, as will become clear below. In French and Arabic, to name but a few, the argument can be made that the head form is an agreement marking. Lambrecht (1981) makes a strong case for Non Standard French as a pronominal argument language (to borrow Jelinek’s term, or PAL), since (26a) is ungrammatical and needs to have the subject clitic repeated, as in (26b). Even in Standard French, (26a) is marginal and (26b) is preferred: (26) a. *Je lis et écris ‘I read and write.’ b. Je lis et j’écris I read and I-write ‘I read and write.’

So, (26) in French shows that pronouns are further on their way to becoming agreement than the pronouns in (22) and (23) are in English. A last argument is that if pronouns check as heads, they might be more often cliticized with an auxiliary. This is indeed the case as Table 2.2 shows for cliticization of am, will, would, has etc. It occurs quite frequently with first person pronouns, as in (27), but it never occurs with faculty, student, department, and school: (27) I’m concerned that this perception came across. (CSE-FAC95)

Economy

Table 2.2.Cliticization to pronouns in CSE-FAC (significant between first and second and first and third at p < .001)

I you he

uncliticized

cliticized

total

2037 1176 128

685 (=25.%) 162 (=12.1%) 19 (=12.9%)

2722 1338 147

Note that, in Table 2.2, the first person is more often a head. I won’t go into that further here, but this observation corresponds to similar differences in coordination, mentioned in Chapter 13. If pronouns check features in a Head–Head configuration, they can start to be seen as agreement. Standard English is not that far (yet), but further along perhaps with first than with third person. Pierce’s (1994) data comparing French and English child language shows that French subject pronouns, in accordance with Lambrecht’s (1981) observations, are much further along to being agreement markers than English ones. French children only use pronouns with finite verbs (which in the singular have the same null-form) but DPs occur with non-finite verbs. Coordinated pronouns are not used and non-emphatic pronouns are also repeated in (28), showing they are needed for proper checking: (28) Moi je sautes et je descends me I jump and I go-down ‘As for me, I jump and go down.’ (Pierce 1994: 329)

English children, in contrast, use pronouns with infinitives as well, coordinate pronouns, and need not repeat them in sentences such as (28).10 The view that agreement represents an argument, as Jelinek (1984) and Willie (1991) have argued for e.g. Navajo, is similar to the view that the agreement on the verb in Spanish and Italian licenses the (subject) argument (e.g. Safir 1985). Recent work by Taraldsen (1992) and Ordonez & Trevino (1999) argues that the (subject) DP in languages such as Spanish is an adjunct, not an argument. However, as Jelinek (2001) makes clear, all nominals are adjuncts in pronominal argument languages, and that is certainly not the case in (English and) French. In the latter, only definite nominals can be adjuncts. So, even though English pronouns may be developing towards agreement markers, they are not at that point yet. They do, however, act like heads rather than full phrases.

25

26


1.2 Negation Jespersen (1917) describes what has come to be known as Jespersen’s Cycle where historically the negative element attached to the verb disappears and the other negative element takes over. Latin non dico goes to jeo ne di to je ne dis pas and je dis pas. These changes are described as cases of specifiers becoming heads. Using (3), Jespersen’s Cycle can be accounted for straightforwardly. See van Gelderen (2004) for more details. Assume negation is checked in a NegP, as in (29). The change then involves grammaticalization whereby a head ne becomes a clitic n- (prefixed to a verb). This change is followed by the introduction of a new XP, such as no thing, in specifier position: (29)

NegP Neg′ Neg no thing

...

[+ Neg] ne

Subsequently, the XP in the specifier becomes a head not and the above changes could start over again. What stops the cycle in English is the prescriptive rule (perhaps the most prevalent of all prescriptive rules) against multiple negation, since for the specifier to be filled again, there would have to be a second negative. For some of the admonitions against the use of multiple negation, see Kirszner & Mandell (1992: A37); Quirk & Greenbaum (1973: 186); Swan (1980: 182). These rules have existed at least since the 18th century, and are very strong. Nevertheless, speakers continue to produce violations. For instance, Anderwald (2002: 104 ff.) examines the BNC and finds considerable numbers of multiple negation in spoken British English (namely 14.3% of negatives). Other languages have a choice between placing negation in the head or the specifier. For instance, a typologically very different language, Korean, has short and long negation. The literature doesn’t talk about this variation in terms of Spec–Head variation (see e.g. Park 1992 who considers the preverbal negative an adverbial), but it is easily put in that framework. Both kinds involve the negative element an(i), but short negation involves an(i) in preverbal Spec position, as in (30a), and long negation involves an(i) in postverbal head position, as in (30b):

Economy

(30) a.

b.

NegP

NegP

Neg′ an(i)

VP

Neg′ Neg

VP

V′ DP

Neg V′

V

DP

an(i) V-ci

In the case of (30a), both object and subject move out of the VP (if both are there), and the verb moves to T and C (not shown here) to check other features. This results in (31a). In (30b), the V is blocked by the head an(i) from moving to T and C, and an auxiliary/light verb -h- is used. The result is (31b): (31) a.

apenim-un an ka-sy-e father-nom not go-sh-int ‘Father is not going.’ b. apenim-un ka-ci an-h-usy-e father-nom go-inf not-do-sh-int ‘Father is not going.’ (Sohn 1999: 390, glosses adapted)

Under (3), (30b) and (31b) should be more economical. The additional insertion of the light V may make it less economical, however. Korean speakers (Min Young Kim p.c.) find (31b) formal and awkward in speech; Sohn (1999: 390) says “the former being slightly less formal than the latter”.11 1.3 D-words As argued in Wood (2003), but see also Lyons (1999), there are a number of instances where elements in the DP change from Spec to Head. Wood shows that, as the demonstrative becomes an article in OE, a shift from specifier to head occurs. An additional shift of the possessive from ME specifier to ModE head also occurs. Two orders predominate in the OE DP: a demonstrative followed by a possessive followed by a noun, as in (32), and possessive followed by a demonstrative and an adjective and a noun, as in (33): (32) sio hira lufu that their love ‘their love’ (Pastoral Care 147.10, from Wood 2003: 115)

27

28


(33) Min se leofesta freond my that dearest friend ‘my dearest friend’ (Apollo 14.9.14, from Wood 2003: 113)

(32) is less common and Wood argues, basing herself on Mitchell and Traugott, that the construction is appositive. In (33), the possessive is in the specifier, and the demonstrative has become the head. Later on, in ModE, both the (now) article and possessive pronoun are in the head position and hence (33) is no longer grammatical, even though (32) is, with appositional structure. The numeral one undergoes a similar change, from Spec to Head, according to Wood (2003), and so do heads like such since they have their origin in the combination of a Spec so and a head lik. Concluding Section 1, there is quite some evidence that the Spec to Head Principle is relevant to merge and checking. In Chapter 4, I will show how it is relevant to syntactic change, in particular in cross-clausal grammaticalization.

2. Late Merge Within recent Minimalism, there is a second economy principle, namely (34), see Chomsky (1995: 316; 378) and Chapter 1 above: (34) Late Merge Merge as late as possible.

The intuition behind this principle is that fully lexical words such as Ns and Vs are merged first since they are relevant to theta-marking (Chomsky 1995: 314–5). Grammatical words such as auxiliaries and prepositions are ‘needed’ later in the derivation and therefore either moved there or merged late. Child language as characterized by the Root Infinitive Stage (see Rizzi 1994) typically uses only lexical categories since functional categories are not yet part of the child’s (working) lexicon. So, the child uses early merge without violating Late Merge. Using (34), I will argue that if, for instance, a verb is less relevant to the argument structure, it will tend to merge higher (e.g. IP or CP) rather than merge early (e.g. in VP) and then move. This can happen when a V merges with another V. Since all the head to head changes are to higher positions, this principle can be considered the driving force behind this form of grammaticalization. In work on grammaticalization and reanalysis, there has been a lot of emphasis on the change from lexical to auxiliary verbs (e.g. have and will), from

Economy

prepositions to complementizers (e.g. for and like), and from verbs to complementizers. As a lexical head is reanalyzed as a grammatical (or functional) head, it ‘climbs’ up the tree, as it were (see also IJbema 2002). Modals (Traugott 1972, Lightfoot 1979), perfect marking, progressive (Heine & Reh 1984), infinitive markers and complementizers (van Gelderen 1993; 1998) fit into this, and I will come back to these in detail in Chapters 6 and 11. In Chapter 6, which deals specifically with Late Merge as it applies to clausal grammaticalization, I also show that there are cases of a Spec of a lexical category changing to the Spec of a higher functional category. This too is a result of Late Merge. Chinese aspect markers such as le have grammaticalized from verbs. In Chapters 8 and 9, I will argue the same pattern can be seen in English modals and perception verbs. The perfective marker le in Chinese has been discussed a lot (e.g. Li & Thompson 1981; Bisang 1992; Sybesma 1999). It derives from two different verbs: liao meaning ‘to complete’ among other meanings, and lai ‘to come’ (Sun 1996: 85; 178; Shi 2002). These are typically elements in the light verb position that come to be generated higher. Both Wu (2000) and Lord (1993: Chapter 8) provide instances of serial verbs becoming auxiliaries. Chomsky’s (1995; 2001ab) preference of ‘merge over move’, reformulatable as ‘Merge late so that you don’t have to merge as well as move’, as in (34) above, provides a ready explanation. At the time of the change of main verbs to auxiliaries, verbs move to I and/or C to check features of tense and agreement, and having the auxiliary merge rather than merge and move in e.g. (35) is more economical: (35) ne mahte he wið leasse gref habben arud us? not could he with less grief have saved us ‘Couldn’t he have saved us with less pain?’ (Ancrene Wisse, 106r)

Roberts & Roussou (2003) use an insight similar to Late Merge to explain grammaticalization affecting C, T, and D. They couch it in a theory of features and markedness. Following Borer (1984) and Chomsky (1995), they assume all variation between languages is due to the lexicon. Functional categories can be specified as to whether or not they have a PF representation (F* for Roberts & Roussou 2003: 29) which can be achieved either by Merge or by Move. As an example, in English wh-questions, Q* is specified as MOVE, since the wh-word has to move, but as MERGE in languages with a question particle.12 The reason for historical changes is through a ‘simplicity metric’ (Roberts & Roussou 2003: 201): a moved structure will not be as simple (not their words) and the child learning the language will reanalyze it. I will use a less formal approach,

29

30


i.e. without using features, and will provide a few examples here of how Late Merge accounts for change. As is well-known, for, the preposition of location and purpose, as in OE (36) to (38), comes to be used as a complementizer by early ME, as in (39) and (40): (36) þæt he for eaxlum gestod that he before shoulders stepped ‘that he stood in front of …’ (Beowulf 358) (37) for werefyhtum … ond for arstafum usic sohtest for fighting and for support (you) us sought ‘You wanted us to help fight.’ (Beowulf 457–8) (38) forþan ic hine sweorde swebban nelle therefore I him sword-dat kill not-want ‘Therefore I don’t want to kill him by sword.’ (Beowulf 679) (39) Locrin 7 Camber to þon scipen comen. for to habben al þa æhte Locrin and Camber to the ships came for to have all the goods ‘Locrin and Camber came to the ships to take all the goods.’ (Layamon, Caligula 1113–4) (40) moche he lofde echn(e) cniht. þat lofde for to segg(e) riht much he loved every knight that loved for to say truth ‘Much he loved every knight who loved to say the truth.’ (Layamon, Otho, 5523)

So, between OE (36) and ME (40), for has been reanalyzed from P to C, i.e. is located higher in the tree. This means that, out of the original meaning of for, the benefit and future features become grammaticalized. At first sight, these constructions don’t seem to be cases where the P actually moves from the lower PP to C, as with the verb from VP to I in (35) above, and where Late Merge would be simpler than merge and move. I will argue they are nevertheless instances of Late Merge. My explanation follows Kayne’s (2001; 2002) account of certain prepositions. He argues that prepositions such as ‘of ’ are merged outside the VP. Chomsky (1995: 232) states the same insight as follows: “overt movement of β targeting α […] is possible only when α is nonsubstantive [read functional]”. Adapting that to (36) to (40), I argue that once a preposition is no longer fully lexical but is used to express purposive, as in (37), and dative Case, the PP of which it is the head moves outside the VP (to the Spec of CP). Once it moves outside VP, it can be reanalyzed as a complementizer indicating purpose, e.g. in (40). This accounts for the change from P to C. Prepositions are notoriously ‘in-between’ categories, i.e. can be either lexical or functional.

Economy

In many early English sentences, for as head of the PP is indeed at the beginning of the sentence, e.g. as in (41) and in (42):13 (41) þa wile þe ich hæuede mi kinelond. lueden me mine leoden. then while that I had my lands loved me my people for mine londe 7 for mine feo. mine eorles fulle to mine cneo for my land and for my property my earls fell to my knees ‘While I had my lands my people loved me. All my barons knelt at my knees for my land and property.’ (Layamon, Caligula 1733–4) (42) þu Šef þeseluen for me to lese me fra pine ‘you gave yourself to me to release me from pain.’ (Wohunge 88–9)

According to van Dam (1957: 6), this fronting occurs regularly in OE. In (42), for is ambiguous between P and C, and hence the language learner ends up reanalyzing the P as C, and the DP as a topicalized element. In this connection, it is remarkable that the first instances of that-deletion listed in the OED (entry for that II 10) are as in (43) and (44), from the 14th century, i.e. where a for-phrase has been fronted and can serve as C: (43) I dred me sare, for benison He sal me giue his malison I dread me sore for blessing he will me give his curse ‘I am afraid, instead of a blessing he will give me cursing.’ (Cursor Mundi, Cotton, 3665) (44) Joab … slowh Abner, for drede he scholde be … ‘Joab killed Abner, out of fear that he should be …’ (Gower, Confessio I. 263)

I think these are not cases of that-deletion at all but cases of a PP moving to the CP, resulting in for being reanalyzed in a higher position. This means that it doesn’t have to merge and move, but can just merge late. So these cases are indeed similar to (35) above, with the change going from merge and move to Late Merge. Other evidence is that, in some sentences, for clearly ‘belongs’ in the lower VP, e.g. in (45): (45) Seoð seide Cordoille. for cuð hit is me nouþe true said Cordoille for known it is me now ‘Cordoille spoke true: because it is now known (to) me.’ (Layamon, Caligula 1727)

In Chapter 5, I come back to for in particular where it introduces a finite clause.

31

32


Van Craenenbroeck (2002) discusses a similar situation in South Brabant Dutch where the preposition van ‘of’ becomes a complementizer indicating aspectual/ evidential dissociation. There are three stages, all still present in the dialect: one where van is a clear P that is merged as P and doesn’t move, a second where it is merged as P but then moves, and a third stage where van is merged high in the tree. To show ‘Late Merge’ (not mentioned in van Craenenbroeck), the crucial stage is the second one. Basically following Kayne (1999; 2001), the structure for a sentence such as (46) starts out as in (47) (with a RootP rather than a VP-shell): (46) Freddy probeert van den auto te repareren Freddy tries of the car to repair ‘Freddy tries to repair the car [but won’t succeed].’ (p. 49) PP

(47) Spec

PRO den auto te repareren

P′ P van

RootP DP

Root′

Freddy

Root probeer

The subsequent stage in the derivation is one where van moves to C (right above PP) and the RootP to the specifier of CP. After that stage, a VP is added to the CP and the subject Freddy and the verb probeer move to their ultimate positions. Van Craenenbroeck’s final tree is the somewhat unorthodox (48):

Economy

vP

(48) Freddym

V′ v

probeerl

CP C′

Spec [tm tl]i

C vank

PP Spec

PRO den auto te repareren

P′ tk

ti

Hoekstra (1995) has a similar analysis of van being outside the VP and having a predicate move to its specifier. So, like English for, Dutch van ends up in a higher position. There are many similar cases of P becoming C that will be discussed in Chapter 6. Some of the formulations, in other frameworks, come quite close to Late Merge. For instance, Heine & Claudi (1986: 105) describe a stage in the change from P to C, quite like (41) and (42), as “[f]or some time, the emerging embedded clause is confined to that position in the sentence which is occupied by the adpositional phrase from which it emanated. Later on, it tends to gain more syntactic independence…”. Psycholinguistic evidence that the PP and CP are very similar is provided in e.g. Josefsson & Håkansson (2000). Emonds (2001) uses a different version of Late Merge to account for Romance restructuring verbs. He argues that, due to the semi-lexical nature of verbs such as Italian cominciare ‘begin’, and French laisser ‘let’, the latter can merge late (but don’t need to) and are then allowed to be ternary branched structures. In answer to the question why these semi-lexical verbs do not always merge late, he appeals to pragmatic effects (2001: 58).

3. Conclusion In this chapter, I have provided evidence for two Economy Principles, namely Spec to Head and Late Merge in both synchronic grammars and language change. Before looking at historical changes involving the CP in the light of these principles, I will first discuss the structure of the CP in Chapter 3. In

33

34


Chapter 4, I will use the first Economy Principle, Head-over-Spec, and in Chapters 5 and 6, the second, Late Merge. Part III is organized the same way. After first looking at the structure of IP and VP in Chapter 7, I apply the Late Merge principle to account for modals in Chapter 8, perception verb complements in Chapter 9, and infinitival to in Chapter 11.

Part II

Chapter 3

The structure of CP and the Layer Parameter

Clause-union entails the breaking down of the independent CP-layer. This starts by making the C dependent on the tense or mood of the higher clause, and results in embedding. Before examining clause-union, however, it is necessary to look at some of the characteristics of clauses across languages. A typical clause will include a subject, an (independent) tense, and a complementizer (C), and is represented by a Complementizer Phrase (CP). In addition to a CP, since Rizzi (1997), split CPs increasingly account for topicalization and other phenomena typical of the edge or periphery of the clause. I discuss cross-linguistic differences regarding topics and double complementizers indicating that not all languages split the CP equally. I then show how OE differs from ModE in having fewer actual positions. In ModE, there are also differences among clauses as to how transparent a CP is. This often depends on the matrix verb; e.g. the traditional factive verb complement is not transparent. This should come out in the structure as well. I relate matrix verb and type of CP selected. In ModE, all verbs have split finite CPs but, in the case of certain verbs (e.g. regret), one of the Cs is filled making that complement less transparent. OE displays very little difference among verb complements, fitting with its lack of a split CP in embedded clauses. The outline is as follows. I start with a discussion of the structure of the CP and split CP in Section 1. In Sections 2, 3, and 6, I examine cross-linguistic differences. Languages display different levels of complexity where the CP is concerned. This justifies the Layer Parameter of Chapter 1. Sections 4 and 5 connect the structural differences from Sections 1 to 3 to differences in verbal complementation. Certain complements are quite opaque and their reduced forms lack a layer of the CP, but are still clauses. Other complements are transparent and their reduced forms are IPs. The conclusion will be that clauses consist of three basic layers, but speakers, on the basis of the linguistic data surrounding them, choose (‘set their parameters’) as to how expanded a particular layer is, and what the main layer(s) is/are. In Chapters 4 and 5, I show how the CP and split CP are relevant to the Economy Principles.

38


1.

The structure of CPs

In 1.1, I briefly point out some of the characteristics of a traditional clause. In 1.2, I (again briefly) review the reasons for assuming a basic C(omplementizer) P(hrase) and, in 1.3, for expanded CPs based on the structure suggested by Rizzi. In 1.4, I examine Cinque’s expanded CP structure, and justify the use of Rizzi for the remainder of this book. 1.1 The Clause A lot has been written on what constitutes a clause cross-linguistically, see e.g. Shopen (1985), and Haiman & Thompson (1988). In English, main clauses consist of a marking of tense and mood and a subject, whereas subordinate clauses may lack one or more of these. Yet, both can be CPs, as in (1ab): (1) a. [He must leave]. b. I wanted very much [for him to leave].

In the case of independent clauses, as in (1a), the C will not always be overt; the C of a dependent clause, as in (1b), depends on that of the higher clause and if there is no C,1 the tense or reference of the subject makes the sentence independent. Since main clauses indicate mood, assumed to be in CP, I assume every main clause is a CP, even in English, where there isn’t that much evidence for it. In languages with second-position effects, there is more evidence. Tense is (usually) a by-product of finiteness. The latter is complex, and may differ slightly in different languages. In this section, I show that the CP unlike the IP is an independent unit even though it may lack tense, as in (1b). This is different from e.g. Noonan’s (1985: 51) definition of a complement as only being sentence-like if the Case of the subject and the tense are the same as in a regular main clause. I’ll use two criteria to show that the CP is a syntactically independent unit: movement and the impossibility of DP-extraction. These are in accordance with the theory of phases proposed in Chomsky (2001a), namely that the CP and vP are phases but that IP is not. In Section 2.1 below, I’ll show that CPs as in (1b) are the smallest CPs and cannot be expanded. As to the first criterion, CPs can be moved, as in (2), but IPs, as in (3), cannot. (Note that that in (3) is a C not D): (2) [That Mary hates soccer] I don’t believe/regret t.


(3) *[Mary hates soccer] I don’t believe/regret that t. (sentences from Corver & van Riemsdijk 2001: 2, but ‘regret’ added)

If one accepts Chomsky (2001a) that only CP and vP (and DP) are phases, this fits. Not only is CP-movement allowed in sentences such as (2), it is required in a number of constructions. This requirement is formulated as the Case Resistance Principle in Stowell (1981): the head of a phrase that assigns Case cannot remain in a Case marked position. Since C is the head of the CP, whether or not the CP moves depends on the character of the C. It can be either +N (i.e. nominal) or -N (i.e. non-nominal). Wh-elements are +N and need not move, as in (4); that is -N and needs to move, as the ungrammaticality of (5) shows:2 (4) We were talking about [who we should help]. (from Stowell 1981: 392) (5) *We were talking about that we should help.

A second diagnostic is extraction. Passivization is impossible when a CP is present, as in (6a), since the DP crosses at least two boundaries (under an early version of restrictions on movement). The non-moved version in (6b) is well-formed: (6) a. *Musharrafi was regretted/believed [that she met ti]. b. It was regretted/believed [that she met Musharraf].

This is the same with infinitival CPs with an overt C. Wh-questions, on the other hand, occur since wh-pronouns ‘escape’ through the specifier of CP, as in (7), where that is in the head position. If the Spec is filled, as in (8), such movement is impossible: (7) Whoi did you forget/believe [that she met ti]? (8) *Whoi did you forget/believe [how she met ti]?

The judgments here are the same for factive and non-factive verbs, but there are major differences between these two kinds of verbs to which I turn in Section 3.3 If a clause minimally involves a complementizer to count as a clause, the lack of a complementizer that results in ECM constructions, such as (9a), predicts that the connection between believes and the IP is not as water-tight as that between a verb and its CP complement. This lack of the complementizer (and tense) means that the Case of the subject of the embedded non-finite IP can be ‘assigned’ by the higher verb, as in (9a); the IP doesn’t ‘count’ as a barrier to Case. That’s why the construction is called ECM, Exceptional Case Marking. The IP doesn’t count as a barrier for passives either, as (9b) shows:

39

40


(9) a. He believes IP[himself to be so nice]. b. Hei was believed IP[ti to be very nice].

Having discussed what defines a clause, I will now turn to the structural representation of the clause. 1.2 The CP The C and the precursor of the CP, namely the S¢, are present in representations early on, e.g. Chomsky (1955 [1979]: 278; 1957). Den Besten (1983) is one of the first to advocate that the finite verb in so-called Verb-second (hence V2) languages moves to C and is therefore in complementary distribution with an overt complementizer. This shows the parallel between Germanic V2 and auxiliary-fronting in English interrogatives, reducing the differences between the different languages. Chomsky (1986) and others suggest that the complementizer C is the head of a Complementizer Phrase CP, bringing functional categories in line with general rules of phrase formation, and accounting for the many second-position phenomena. The tree with a single CP, as in (10), is very familiar by now since it clarifies certain word orders in languages across the world: (10)

CP C′ C

IP I′ I

VP

Apart from accounting for V2, the empirical advantages of (10) are many, e.g. accounting for the order of [wh – that], where the wh-element can be a phrase, as in (11), which occurs in varieties of English: (11) it’s down to the community in which that the people live. (BNC-KNF 569)

As is well-known, topics do not trigger V-movement in English, as (12ab) shows, and they don’t in OE if the topic is a pronoun (Kroch & Taylor 1997: 310 ff.). They do in the other Germanic languages, e.g. (13) from Swedish:


(12) a. Stewed pears we always avoid. b. *Stewed pears avoid we always. (13) Kompott äter vi aldrig stewed-fruit eat we never ‘We never eat stewed fruit.’

In English, topics do not trigger auxiliary movement. I will not elaborate on the difference between main verb and auxiliary here. However, what triggers movement receives a nice explanation under a more expanded CP, as I’ll go into now. 1.3 The split CP During the last decade, E. Hoekstra (1993), Rizzi (1997), and Cinque (1999), to name but a few, have suggested a more expanded functional projection to accommodate the complementizer and other material appearing on the left edge of the sentence. In e.g. Lasnik & Saito (1992) and McCloskey (1991), topics and sentence adverbs are seen as adjoined to the main clause. Adjunction is not very restrained and therefore it is good to eliminate it from the grammar. One way to do this is to think of the ‘adjoined’ topicalized and focused elements as occupying separate functional categories, as part of an expanded CP, as in (14): (14) … Force … (Topic) … (Focus) … Fin IP (from Rizzi 1997: 288)

Regarding (14), Rizzi (1997: 283) argues that “the complementizer system [is] the interface between a propositional content (expressed by the IP) and the superordinate structure (a higher clause or, possibly, the articulation of discourse, if we consider a root clause)”. Therefore, the CP contains elements that look outside which he calls Force (following Chomsky 1995), and those that look inside, which he refers to as Finite. In addition, as mentioned, the CP in (14) (optionally) accommodates Topic and Focus. I am assuming a topic is old information, and a focus is new, but not much hinges on this since there are very few sentences with both in the CP in English. See Rizzi (1997) for more. Rizzi (2001: 289) assumes (15), where ‘Int’ is interrogative. However, some of the Topic positions can be eliminated by allowing a head to move to the different head positions: (15) Force (Topic*) Int (Topic*) Focus (Topic*) Fin IP

41

42


I will generally stick to (14). Returning to the difference between (12) and (13) above, the split CP provides at least a descriptive account: in English, the head of topic does not need to be filled, but in Swedish, it does.4 Recent work using a split CP to account for Verb movement is, for instance, den Dikken (2002). It also provides a description for the fact that even though topics generally precede wh-elements in main-clauses, in embedded sentences the reverse holds (e.g. Kiss 1995: 12), as (16) shows: (16) ?I don’t know what to Mary we should give. (from Den Dikken 2003: 83)

This means that the wh-element moves from a lower (Focus) to a higher (Force) position in embedded sentences such as (16). This move to the ForceP will be shown to result in wh-clauses being able to follow prepositions. As will be shown below, in language change, the complementizer that also changes from being in the lower part of CP to the higher part. Languages in which both topic and focus can appear sentence-initially display the sequence of (14), e.g. (17), from Zulgo, a Chadic language, and (18) from Modern Greek: (17) mekele ka, ]gat na azla si]gwe ya mekele top he foc he-took money foc ‘As for Mekele, it is he who took the money.’ (from Haller & Watters 1984: 30, but tones left out) (18) Ti ðulja mu se kanenan ðen tin embistevome the work mine to no one not it entrust ‘I don’t entrust my work to anyone.’ (from Condoravdi & Kiparsky 2001: 4)

In Bulgarian too, the topic and focus are separate. In (19), the topic is filma and the focus is Marija. The focus is always followed by li. This means li is in the head and the focused DP is in the Spec of FocusP: (19) Filma Marija li gleda film Marija foc watch ‘As for the film, is it Marija who is watching it?’

In this book, I am concerned with cross-clausal grammaticalization and hence I will not consider main clauses, except in general terms, e.g. in Section 6. Notice, however, the remarkable word order in English and the difference between main and embedded clauses, as in (20) and (21), see also (16):


(20) a. To Leonard what should we say on his birthday? b. *What to Leonard should we say on his birthday? (21) a. ?I wonder what to Leonard we should say. b. I wonder to Leonard what should we say.

The asymmetry between (20b) and (21a), with the wh-element in Force in subordinate but not main clauses, is unexpected. Rizzi maintains that, in Italian, the finite complementizer che is in Force but that the infinitival di occupies Fin. The evidence for this comes from topicalization and left dislocation. The finite complementizer precedes the Topic, as in (22), whereas the infinitival one follows, as in (23), both from Rizzi (1997: 288): (22) Credo che il tuo libro, loro lo apprezzerebbero molto believe that the your book they it appreciate much ‘I believe that they would appreciate your book a lot.’ (23) Credo, il tuo libro, di apprezzarlo molto believe the your book for appreciate-it much ‘I believe I like your book very much.’ (Chiara Lage p.c.)

A tree for a sentence such as (22) would look like (24): VP

(24) V

ForceP Force′ Force

TopicP Topic′

credo

che

Topic

IP

il tuo libro

loro lo . . .

Poletto (2000: 129) says that speakers accept che on the other side of the topic as well, which may mean optional movement of che from Fin to Force. She (2000: 174) argues that either verb or complementizer move this way depending on the setting features of the matrix verb. This is reminiscent of movement of what in (16) and shows some clauses are split.

43

44


Not all languages have such an expanded CP and I’ll briefly look at that now, and more in Section 2 and Chapter 5. The English finite complementizers that and for behave like che, as (25) (taken from McCloskey 1991) and (26) show. This is less true for factive verbs, as in (27), a fact which will become important later on: (25) She maintains that Irish stew she sort of likes t. (26) … for Irish stew I sort of like. (27) ?I regret that Irish stew she sort of likes t.

This means that and finite for are in Force not in Fin.5 Historically, this has probably always been the case for finite for, which is first attested as a finite complementizer around 1200, as in (28) and (29). Since topics and adverbs follow for, the latter is in Force. That is now in Force but in Fin probably until about 1500, as (30) shows, with when preceding that (more in Chapter 4 about this): (28) For hardely I hym heete ‘For indeed I threaten him.’ (York Plays, 11, 286) (29) For frenshippe we haue foune ‘Because friendship we have (indeed) found.’ (York Plays, 10, 12) (30) She loved Arcite so | That [when that he was absent any throwe] Anon her thoghte her herte brast a-two ‘She loved Arcite so much that when he was absent she thought her heart would break.’ (Chaucer, Anelida & Arcite 377.92–4)

Further evidence for for in ForceP and that in FinP between 1200 and 1500 is the frequent occurrence of for that, as in (31): (31) and forþi we clepeð him fader for þat he us feide here and therefore we call him father because that he us fed here ‘and we call him father because he fed us here.’ (Trinity Homilies HC-ME1)

This historical change of that from Fin to Force will be discussed in Chapter 5 as an instance of Late Merge. 1.4 The ‘cartographic’ approach and some problems Cinque (1999), in what has been coined the ‘cartographic’ approach (see e.g. Rizzi 2002; Cinque 2002), argues that CP and IP do not suffice because adverbs


need to be accommodated. The CP needs to accommodate speech act adverbs (frankly, honestly), evaluatives ((un)fortunately), evidential adverbs (allegedly, evidently), and modal affixes in certain languages. The full range of the three CP-adverbs is given in (32). I have added an epistemic IP-adverb as well: (32) Mood speech act Mood evaluative Mood evidential Mod epistemic frankly fortunately allegedly probably (from Cinque 1999: 107)

Cinque (1999) does not use Rizzi’s categories, unfortunately, but testing the compatibility of these adverbs with topics and focus, one finds (33) and (34). Even though (33) seems slightly odd, it is acceptable to native speakers with honestly in ForceP and those books in the topic. (34) nevertheless presents a problem in Cinque’s approach, since frankly is higher in the tree than surprisingly, but the two cannot occur together as in (34): (33) ?Honestly, those books, he should have read before class. (34) *Frankly, surprisingly, he read those books.

The examples Cinque gives with multiple adverbs are as in (35), but one of the adverbs is inside the IP: (35) Honestly, I am unfortunately unable to help you. (p. 33)

This suggests that the English CP is restricted. In the 100 million word British National Corpus (hence BNC), frankly and fortunately never co-occur; neither do fortunately and allegedly; or allegedly and frankly (even though the adverbs occur frequently by themselves).6 Thus, matrix CPs such as (33) and (34) show that the CP rarely contains a ForceP with the sentence adverb and a TopicP with a topic. Cinque says very little about subordinating conjunctions, which is surprising since these originally motivated the use of C and CP (Chomsky 1986). The BNC gives three instances of that preceding frankly, as in (36):7 (36) She has told Paul that frankly she’s lapping up the attention. (BNC KBF 8830)

This should not be possible under Cinque’s approach since the specifier of the Speech Act adverb, frankly, should be highest, and in (36), that is. Sentences such as (36) are perfectly compatible with Rizzi’s account where that is in Force and the adverb in the Topic position. The ungrammaticality of (38) and (39) below is again in accordance with Rizzi but not with Cinque. In the remainder of this book, I will further argue that Rizzi’s split CP is correct.

45

46


2. Cross-linguistic differences in the split CP In this section, I will show that there is significant variation in whether and how languages expand the CP. Non-finite clauses typically do not (contra Rizzi), since they are reduced clauses (Section 2.1), and neither do relative clauses. There is quite a lot of other variation, e.g. in whether topics are incorporated (Section 2.2), and whether there are multiple complementizers (Section 2.3). In concluding (Table 3.1), I make a distinction between evidence for the split CP from actual complementizers and from topic/focus material present in the CP. 2.1 Infinitivals As mentioned above, Rizzi (1997; 2001: 288) argues that the infinitival di occupies Fin, since it follows the topic, as in (37), from Rizzi: (37) Credo, il tuo libro, di apprezzarlo molto believe the your book for appreciate-it much ‘I believe I like your book very much.’

My focus is not Italian but other non-split possibilities exist, e.g. having di in I. Rizzi (2001: 287) allows for the possibility of a single head in CP: “[w]e may think of Force and Finiteness as two distinct heads closing off the complementizer system upward and downward respectively (and perhaps coalescing into a single head in the simple cases)”. In English, the infinitival complementizer for in (38) is similar to Italian di in (37) in that a topic cannot follow it. The reason is that for needs to be adjacent to the subject for Case reasons. However, it is unlike di in that a topic cannot precede it either, as the ungrammaticality of (39) shows. This is unexpected if for is in Fin: (38) *I expect for [her homework] her to do. (39) *I expected [her homework] for her to do.

If we test the position with other non-finite complementizers, the result is the same, i.e. topics cannot precede them, as in (40) and (41), or follow them, as in (42), where the trace indicates the original position of the topic: (40) *He swam the channel i[to England] in order to get ti. (41) *I wonder i[to Alaska] whether to go ti. (42) *I wonder whether i[to Alaska] to go ti.


Thus, non-finite complementizers in English do not fit readily in the expanded CP. It seems as if only the non-finite complementizer can be present, not a topic or focus.8 As we’ll see in the next section, only finite clauses have multiple complementizers as well. This is indicative of their reduced clausal status. 2.2 Differences in finite clauses: Topics Not all languages have such an expanded CP as in (24) above and I’ll briefly look at that now where topics are concerned. Abraham (1997: 39) argues in a similar vein that there is no TopP or FocP between CP and IP in German, whether the C is finite, as in (43), or non-finite, as in (44). I have given glosses for what they might mean in English: (43) a. *Ich hoffe daß dein Buch sie würden es sehr mögen I hope that your book they will it much like ‘I hope that they will really like your book.’ b. *Ich hoffe dein Buch daß es ihnen gefällt I hope your book that it them pleases ‘I hope that they are pleased with your book.’ (44) *Ich hoffe dein Buch es schätzen zu können ‘I hope to be able to appreciate your book.’ (from Abraham 1997: 39)

In Dutch, the same is true: (45a) and (45b) are comparable to Italian (22) and (23) above respectively, but are ungrammatical, and so is (46) with an adverbial clause: (45) a. *Ik geloof dat jouw boek ze waarderen I believe that your book they appreciate ‘I believe they appreciate your book (very much).’ b. *Ik geloof jouw boek dat ze waarderen I believe your book that they appreciate ‘I believe they appreciate your book (very much).’ (46) *Ik ging naar de winkel dat boek om te kopen I went to the store that book for to buy ‘I went to the store to buy that book.’

These sentences show that the CP in Dutch and German is not split. I will come back to this in Section 4.

47

48


2.3 Double complementizers Even though the lack of embedded topics in Dutch, as in (45), provides evidence for an unsplit CP, there is still evidence for a split CP in that multiple complementizers occur with certain verbs, as with the negative in (47). Extraction is possible, as in (48): (47) Ik weet niet wie of dat gedaan heeft I know not who if that done has ‘I don’t know who has done it.’ (48) Wie weet ik niet of dat ’t gedaan heeft? who know I not if that it done has ‘About whom don’t I know if s/he has done it?’9

Double complementizers are only possible with verbs such as vergeten ‘forget’, betwijfelen ‘doubt’, and afvragen ‘wonder’. Of, like for in English, seems quite restricted and indicates a doubt or unreal situation, as in (49): (49) Ze gedraagt zich of dat ze hier de baas is she behaves herself as-if that she here the boss is ‘She behaves as if she is the boss.’ (from ANS 664 but with dat added)

So even though Dutch does not allow embedded topics, as (45) shows, it has a small class of verbs that allows double Cs, indicative of a split CP. Barbiers (2002) argues that in Dutch, there is a ForceP in factives which allows wh-embedding, as in (50), but not in non-factives, as in (51): (50) Jan zal meedelen welke boeken dat Marie leest Jan will announce which books that Marie reads ‘Jan will tell us which books Marie is reading.’ (51) *Jan zal denken welke boeken dat Marie leest Jan will think which books that Marie reads (Barbiers 2002: 50–51)

He argues a ForceP complement is present when the complementizer of ‘whether’ is, namely in factives, as in (52), and that it is absent in non-factives, as in (53): (52) Jan zal meedelen of (dat) Marie deze boeken leest Jan will announce if (that Marie these books reads ‘Jan will tell us if Marie is (really) reading these books.’


(53) *Jan zal denken of (dat) Marie deze boeken leest Jan will think if (that Marie these books reads (Barbiers 2002: 50–51)

I will come back on the use of factives but I think (51) and (53) are not relevant since they do not subcategorize for wh-complements at all, let alone split ones. Rephrasing (53) as (54) with a verb that does allow a wh-complement results in a grammatical sentence, as pointed out to me by a reviewer: (54) Ik vroeg me af welke boeken (dat) Marie heeft gelezen I asked myself of which books (that Marie has read ‘I wondered which books Mary read.’

ANS (p.645; 1151ff.) talks about the difference between dat and of as certain versus uncertain. This meaning is also present in (49) and (52) above, and in (55): (55) Het scheelde geen haar of hij was verdronken it missed no hair or he was drowned ‘He almost drowned.’ (ANS 648)

In (55), of would be in Force and was in Fin. Adding a Topic by changing (55) into (56) results in another grammatical construction. This shows that the finite verb was has moved to Fin and of is in Force: (56) Het scheelde geen haar of gisteren was hij verdronken it missed no hair or yesterday was he drowned ‘He almost drowned.’

So, Dutch allows an expanded CP with those matrix verbs that have a double C. This class cannot be characterized just as factive, however, as Barbiers argues, but as having irrealis mood (see (107) below for a possible structure). 2.4 Concluding observations Because of the incompatibility of topics with non-finite for, as in (38) and (39) above, I assume ModE non-finite CPs have only an unsplit CP, as in (57), with the head filled:

49

50


(57)

CP C′ C

IP I′

for

I

VP

For finite constructions, Rizzi’s structure provides insight since finite for is in Force in (26), preceding the topic. OE, as I will indicate in the next section, shows no evidence of a split CP, but certain verbs in Dutch (shown above) have two complementizers, one in Force and one in Fin. Thus, there is quite a lot of variation. Apart from not using a split CP, I assume that an infinitival C is typically connected with [+future] that provides a purposive meaning, as in van Gelderen (1993: Chapter 5). Thus, in (58), the doing would be future with respect to the liking, which is a volition verb: (58) I would like for you to do your homework.

Verbs that are epistemic, such as believe, never have infinitival Cs, as (59) shows, and lack the purposive meaning. The tense of their complements cannot be future but needs to be ‘anchored’ to the tense of the matrix verb: (59) *I believe for him to be nice.

If the non-finite C for in English has future, it should be in Rizzi’s FinP, but since there can only be one C element in the embedded clause, as (30) and (38) above show, the name is less important. I will suggest that the presence of these special features connected to infinitival for in Fin is a reason why the CP cannot be split. A non-finite is too dependent on the main clause to have a strong layer in between. There is some evidence that in Modern Greek there are three Cs, as argued in Roussou (2000). Since oti can either precede or follow the topic/focus in sentences with tha in the lowest C, there may be three Cs in (60), or tha may be in T: (60) nomizo (ta mila) oti (ta mila) ðen tha ta fai o petros think-1S (the apples) that (the apples) not fut eat-3S the Peter ‘I think that Peter won’t eat the apples.’ (Roussou 2000: 76)


If there are in fact three Cs, Greek might receive a better explanation under Cinque’s theory, as I now show. For English, however, I argue that two Cs are better. In Table 3.1, I summarize the possibilities so far for Italian, ModE, Modern Greek, and Dutch. I have added OE and ME to the table, the evidence for which I provide in Section 3 below.

Table 3.1.Split CPs (parentheses indicate limited occurrence)

multiple Cs topics in CP

Italian/English

Dutch

Greek/ME

OE

no yes

(yes) no

yes yes

no no

3. No split in the embedded OE clause Having shown that there is no evidence for a split CP in some languages and some constructions, I’ll argue that OE has no split embedded clause but, in some varieties, shows a split main clause. I will then show what happens to the structure in ME: namely double complementizers are introduced and the topic is placed in the embedded CP. This chapter is concerned with the structure of the CP, not with the mechanisms that bring this about. Chapter 5 will discuss the changes in the history of English in terms of the Late Merge Principle. In OE, wh-questions always trigger V2. This means they are integrated in the clause-structure, i.e. through the CP. There is some verb-movement with topics as well. In addition, clitics follow topics in southern OE varieties, resulting in Verb-third, (see also Kroch & Taylor 1997; Lightfoot 1999: 155–7). Roberts (1996) argues that clitics are in the Spec of FinP and the verb in Fin; in wh-questions, the V is in Focus. I adapt these ideas in (61) by having the V in Fin with topics and the clitic as a head in Top,10 but just in the southern OE main clause:

51

52


ForceP

(61)

Force′ Force

TopP Top′ Top

⎧ wh ⎫ ⎨ ⎬ ⎩ þa ⎭

V

FinP

topic

Fin′ Fin clitic

...

V

Note that, as in Dutch and German, there is no evidence for such an expanded main clause CP in northern dialects, as shown by the examples in Kroch & Taylor (1997: 321). In the northern glosses, the verb is always in second position even with topics. In OE, most subordinate clauses are Verb-final, hence show no evidence of V-to-C movement, but once in a while there is one that looks as if it has a topic with a verb moved to C, as in (62), from Pintzuk (1991): (62) þæt his aldres wæs ende gegongen that of-his life was end come ‘The end of his life had come.’ (Beowulf 822, from Pintzuk 1991: 187)

Pintzuk claims OE has two strategies: movement of the V to C as well as to I. She takes (62) as evidence for V to I movement with the topic in the Spec of IP. (62) could also be analyzed as VP raising of ende gegongen or in a number of other ways. The evidence that (62) would have a split CP is not overwhelming, however, as I’ll show below regarding the lack of double Cs and then in Section 5 where complementation in general is concerned. Van Kemenade (1997: 333) mentions an example of an embedded topic, namely (63): (63) Gregorius se trahtnere cwæð þæt forði wolde drihten getrahtnian… Gregorius the translater said that therefore wanted God to-translate ‘Gregorius, the translator, said that therefore God wanted to translate …’ (van Kemenade 1997: 333)


There are in OE sequences of what look like double Cs, as in (64): (64) dide he swa oþ þæt hie comon to þære stowe did he so till that they came to that place ‘He did until they came to that place.’ (Homily S19, Palm Sunday, Schaefer 1972)

However, a word or phrase never occurs in between oþ and þæt. The two words are also often written together, indicating a change from Spec to Head. The OED says, oþ is a “conjunctive adverb”, i.e. still in the specifier, as in (65): CP

(65) oþ

C′ C

...

þæt

As to the change from Spec to Head, in sequences of oþ/oð and þæt, oþ is merged with the C very frequently. Checking all forms of oþþe (i.e. all possible þ/ð combinations) gives 7008 in the DOE corpus. Not combined with the C, oþ and oð occur 3410 times. This tendency to combine morphologically is expected if specifiers become heads. The other C is less frequently merged, since there are only 547 instancs of oþþæt (again with all combinations of þ/ð). This may be due to the phonological weight of þæt as opposed to þe. Unexpected is the sequence oþ þæt þe and this occurs only once (with oþ þætþe also occuring once). So, oþ is not a second C. The same analysis can be given to þeah þæt/þe sequences. So that would be another instance. Diessel (1999: 125) mentions that this pattern often occurs in other languages (as well as a combination of a demonstrative and a particle). I will return to the structure of complementizers in Chapters 4 and 5. Now, I’ll turn to the introduction of the split CP. The first instance of for as a C that the OED provides is from 1123, from the Peterborough Chronicle. After this, for that, as in (31) above and (66), and till that, if that, while that also start to occur: (66) I trowe I loved hym best, for that he | Was of his love daungerous to me ‘I believe I loved him the best because he was with his love standoffish to me.’ (Chaucer, Wife of Bath Prologue 112.513–4)

The structure of (66) is different from that of OE oþ followed by a complementizer since the latter is often combined with the complementizer. Unlike oþ, for

53

54


doesn’t occur in this C capacity till the 12th century. For in (66) is in Force and that remains in Fin, but the CP is split, at least in adverbial clauses. It is harder to date the introduction of embedded topics exactly. Russom (1976) describes topicalizations in texts of the Gawain-poet. These texts are from the middle of the fifteenth century. Russom shows the construction is productive and has a variety of topics, e.g. infinitives, as in (67), PPs as in (68), adverbials, and of course DPs, as in (69): (67) To lysten þat watŠ ful lufly dere to listen that was very lovely pleasant ‘which was most delightfully pleasant to hear.’ (Pearl 880, from Russom 1976: 22) (68) For wyn in his hede þat wende for wine into his head that went ‘because of the wine that went to his head.’ (Gawain 900, Russom 1976: 22) (69) Relande in by a rop, a rode þat hym þoŠt reeling in through an intestine a road that him seemed ‘reeling in through an intestine, which seemed a road to him.’ (Patience 270, Russom 1976: 22)

In these examples, the topic precedes that, indicating the latter is in Fin. A little later, e.g. in Chaucer (lC14), as in (70) and (71), in Caxton (lC15), as in (72), and in the York Plays (lC15), that precedes the topic: (70) Ther may swich cause ben … | That hardily thou wolt thiselven saye there may such reason be that certainly you want yourself say ‘There may be such a reason that certainly you yourself want to say …’ (Chaucer, Troilus & Criseyde 577.1305–6) (71) And thus I lyved ful many a day | That trewely I hadde no ned | Ferther than … ‘And thus I lived fully for many days so that really I had no need other than …’ (Chaucer, Book of the Duchess 345.1252–3) (72) Blanchardyn answerd, that [for no drede nor fere that he had of hym] he shuld kepe … ‘Blanchardyn answered that not for dread or fear of him should he …’ (Blanchardyn 84/3, from Kellner 1890: xciv)

Since that precedes the topic, that is in or moves to Force by 1400. Many of these do not have that, as in (73), but are clear topics:


(73) But of his face I can not seyn the hewe | For sikerly his face shon so bryghte ‘But of his face I couldn’t see the color, because truly his face shone so brightly.’ (Chaucer, Legend of Good Women 594.162–3)

There is another set of examples, that are less clearly topicalizations, but that I will discuss nevertheless. Kroch & Taylor (1997: 315) argue that, in the fifteenth century northern Rule of St Benet, there are sentences such as (74), with the topic in front of that (Russom’s examples in (67) to (69) are different from (74) in that multiple topics can occur in (68)): (74) I sal yu lere þe dute of god, his wille þat Še may do I shall you teach the duty of God his will that you may do ‘I’ll teach you the duty to God, so that you may do it.’ (Benet 2.5, from Kroch & Taylor 1997: 315)

Kroch & Taylor’s examples are all of adjunct clauses, and reminiscent of similar (spoken) Dutch adjunct clauses, as in (75), and those that Mitchell (I: 789–90) cites from OE, as in (76). These could have the topic in Spec CP: (75) Gek als ie was deed ie zoiets natuurlijk crazy as he was did he such of naturally ‘The crazy man he was, he did such a thing of course.’ (76) Uncuð þeh ic wære ðonun come [ic] … unknown though I were from-there come [I ‘Though unknown, I come from there.’ (Alfred, Soliloquies 26.21, from Mitchell I: 789)

Having given structural evidence for the introduction of the split CP in ME, I’ll now turn to verbal complementation. First (in Section 4), I’ll provide an account of the ModE situation, and then (in Section 5), I’ll link this to the conclusions about OE that I’ve reached in the present section, namely that a split CP does not occur in OE.

4. Syntax and semantics Matrix verbs (partly) determine the structure of the verb in the complement. Thus, cross-linguistically, causatives typically do not have CP complements and verbs of saying do not have VP-complements. This is of course not completely true since languages choose to be more CP-, IP-, or VP-oriented, as I’ll discuss in Chapter 12. The connection is questioned in e.g. C. Lehmann (1988: 183)

55

56


who argues that “the semantic nature of the relation between the two clauses … does not appear to be constitutive of cross-linguistically valid types of clauselinkage”. In this section, I examine a possible relationship between the meaning of a verb and the presence of a split CP. In 4.1, I’ll first look at the traditional effects of factive and non-factive matrix verbs on the structure of their complements. Factives have several defining characteristics. For instance, it is harder to move out of their complements through passivization, raising, and wh-movement (hence, they are also called non-bridge verbs). Kiparsky & Kiparsky (1970), hence K&K, divide factives into emotives and non-emotives, mainly because of the ability of emotives to have a for-to and subjunctive complement. In 4.2, using arguments put forward by Hooper & Thompson (1973), hence H&T, and others and using the BNC, I show the most typical ‘factive’ characteristics appear with emotive factives (e.g. regret), but not with the non-emotives (also called semifactives, e.g. forget). H&T suggest ‘assertion’ is the relevant distinction between verb groups. The terminology in this area is daunting, and I will not add to it. None of these approaches uses the split CP to account for the differences: a split CP, in particular one that can have the fact that, creates more barriers in the complements of emotive factive verbs. I suggest such a structure in 4.3. 4.1 Factives, non-factives, emotives, and non-emotives The distinction between factives and non-factives is well-known since the work of K&K (1970). I will provide some background on this, as well as on K&K’s distinction for factives as being either emotives or non-emotives that has been accepted in a number of later works. K&K (1970: 159), talking about complements to factive verbs, call their independence a “syntactic insulation”. They (1970: 162) show that the complements to factive verbs are more insulating than those of non-factives. The difference between them is both semantic and syntactic. With factive verbs, as in (77), the truth of the complement clause is presupposed by the speaker, unlike with non-factives, as in (78): (77) I forgot that he left a key. (78) I believe that he left a key.

A partial list of factives, non-factives, and ambiguous verbs is given in (79) to (81):11


(79) Factive: regret, be aware of, grasp, comprehend, ignore, forget, make clear, resent, know,12 realize, deplore, see (80) Non-factive: suppose, assert, allege, assume, claim, believe, conclude, say, think (81) Both: anticipate, acknowledge, suspect, report, remember, admit (cf. K&K 1970: 145; 163)

Syntactic differences involve Negative Raising (allowed with non-factives), it-pronominalization (with factives), ECM-constructions (with non-factives), paraphrasing through the fact that (factives), passive raising (with factives), embedded topics (marginal with factives), and wh-extraction of the subject and adjunct (with non-factives). K&K divide the factives and non-factives into emotives and non-emotives, as in (82): (82) Emotive: factives: regret, resent, deplore, be important non-factives: intend, prefer, be unlikely Non-emotive: factives (later called semi-factives): forget, be clear, be aware of, know, see, realize, discover non-factives: say, suppose, conclude, seem, be likely (K&K 169–170, but some added)

The main reason for this division is that emotive complements can be for — to clauses or have subjunctives, even though, as I’ll show later, not all of them do. H&T (1973: 474ff.) add other reasons. Thus, we have criteria to distinguish factives from non-factives and emotives from non-emotives. Table 3.2 later on in this section provides a summary, with an additional class added. In 4.2, I’ll argue that the verbs that disallow Neg-raising, and can have it or the fact that preceding the complement have a split CP with a Force specified with a mood feature. This mood feature will be [+realis] for factive emotives and [−realis] for non-factive emotives. I will now discuss the data differentiating these classes. Non-factive verbs allow Negative Raising from an embedded clause such as (83a) to a higher clause, as in (83b). This does not occur with factive verbs, as in (84), where (84a) and (84b) do not mean the same. Both kinds of clauses may have an overt complementizer but only factives, as in (83), block percolation:

57

58


(83) a. It’s likely that he won’t lift a finger. b. It’s not likely that he’ll lift a finger. (84) a. It bothers me that he won’t lift a finger. b. It doesn’t bother me that he will lift a finger. (adapted from K&K p. 162)

Horn (1989: 323–4) argues that even though factives never have Negative Raising, the opposite is not true, i.e. not all non-factives allow it (e.g. verbs of saying don’t). Horn concludes that a pragmatic or semantic reason behind Negative Raising may be more likely than a syntactic reason, but there are many problems since verbs such as want allow it, wish less so, and desire not at all (p. 321). There are other complications: some verbs allow it with infinitival but not finite complements (e.g. wish). It-pronominalization and paraphrasing through the fact that are possible with factives, as in (85a) and (86a), but not with non-factives, as in (85b) and (86b): (85) a.

They regretted it that Hittite became extinct. (no examples of regret it in the BNC) b. *They believed it that Hedwig had brought that message.

(86) a.

Many Americans regret the fact that they now discuss sport with the passion that 200 years ago they brought to everyday debates… (BNC ABD 752) b. *Many Americans believe the fact that they now discuss …

H&T (1973: 481) show that the subclass identified by K&K as non-emotive cannot have the fact that introduce their complements, as (87) and (88) show: (87) *I see the fact that the Bruins lost. (88) *I know the fact that you’re not speaking to me. (from H&T 481)

A search in the BNC confirms this for see but not for other non-emotives such as realize and discover where two instances of each occur, as in (89) and (90): (89) … was quick to realize the fact that Cubism had not been the product of … (BNC GUJ 863) (90) I thought that Sarah had discovered the fact that we were both, …, being unfaithful to her. (BNC HD6 746)

This is also observed by e.g. Reis (1977: 179) and I will leave this unresolved, and assume these verbs pattern with see and other non-emotives that don’t allow paraphrasing with ‘the fact that’.


ECM and passive raising constructions occur with non-factives, as in (91) and (92), and embedded topics are mostly ungrammatical with factives, but grammatical with non-factives, as in (93).13 Stowell (1981) and Authier (1992) link the possibility to have embedded topicalization with affective inversion, as in (94): (91) a. I believe him to be nice. b. *I regret him to be nice. (92) a. He was believed to be pro-French. (BNC CRK 998) b. *He was regretted to be pro-French. (93) a. John believes that this book Mary read often. b. ?John regrets that this book Mary read often. (94) a. I believe that never in my life have I done that. b. ?I regret that never in my life have I done that.

Again, not all factives work alike. The non-emotive factive know allows ECM, as in (95), which occurs very frequently, but not the non-emotive factive forget: (95) nor did I ever know him to be disloyal. (BNC CD9 1620) (96) *I forgot him to have done that.

Emotives can have subjunctive and for-to clauses, as (97) and (98) show: (97) Asserting, what was palpably untrue, that ‘There are probably few people in India who do not sincerely regret that you should have made it impossible for any government to leave you at liberty’, he … (BNC C90 1004) (98) I regret for you to be in this fix. (K&K 169)

Rosenbaum (1967: 53) shows that for-to complements can also be preceded by it, as in (99): (99) Everyone would prefer it for you to come early. (from Rosenbaum 1967: 53)

However, the BNC has no complements as in (99) with prefer, regret, or resent. Native speakers find the sentence marginal. Those who do say it would have a split CP, with it in Force and for in Fin. Having shown that factives versus non-factives is too broad a division, I’ll discuss the assertive versus non-assertive difference, and will conclude that this distinction is convenient for certain complementation patterns. Not all verbs fit perfectly, as also observed by H&T, Horn and others.

59

60


4.2 Assertive: A new term for non-factives and emotive-factives It has long been noted (Vikner 1995: 70 for an overview; Haider 1986; Thráinsson 1979) that non-factives allow V2 order in their complements in German, but that factives don’t. The same is true in Dutch, as (100ab) shows: (100) a.

Ik geloof gisteren is ie gevallen I believe yesterday is he fallen ‘I believe he fell yesterday.’ b. *Ik betreur gisteren is ie gevallen I regret yesterday is he fallen

This is similar to English topics, in (93) above, repeated here as (101). They are allowed with non-factives, but disallowed with factives: (101) a. John believes that this book Mary has read many times. b. ?John regrets that this book Mary has read many times.

In English (101a), the complementizer that appears in Force and this book in the topic. This is true in Danish as well, as in (102) from Vikner (1995: 67): (102) Vi ved at denne bog har Bo ikke læst we know that that book has Bo not read ‘We know that Bo hasn’t read that book.’

This means Danish has a split CP with at in Force and har in Fin. Since English doesn’t allow V2, and the topic would have to be in a separate position in the expanded CP, it means that the CP is split in cases such as (101a), with that in Force and this book in the Topic position. In German and Dutch, many argue that V2 is movement of the topic to the Spec of CP and of the finite verb to Fin. Therefore, (100) could have that structure too, i.e. of a single CP. This would mean complements to non-factive verbs in German and Dutch would be a single CP, but complements to factive ones would be split and allow for double Cs.14 Meinunger (2002) shows that the verbs that allow V2 complements in (91) above are those of saying, thinking, and the non-emotive factives (verbs of ‘discovery’). The ones that don’t are emotive factives, negative verbs, such as nicht glauben ‘not believe’, imperatives, questions, and causative verbs, such as verursachen, ‘cause’. He follows H&T in referring to these as assertive versus non-assertive.15 The volitional verbs, he argues, show the characteristics of the non-assertive verbs. He relates this to H&T’s observations on root transformations in English. These are allowed with almost the same matrix verbs except that verbs of wanting are left out. I agree with this distinction, with the addition


of features in Force (typically something of ForceP). Meinunger’s verb classes end up as follows with semi-factive an alternative name for non-emotive factives:16 (103) Assertive: saying thinking semi-factive

Non-assertive: emotive factive (resent, regret) negative verbs causative volition

Assertive verbs allow V2 in their complements, lack subjunctives, and display root transformations; non-assertive verbs show the opposite characteristics in that they do not allow V2, have subjunctives, and lack root transformations. In 4.3 below, I will argue that non-assertives have a mood specification in Force, namely [+realis] for factives and [−realis] for volition verbs (I am leaving out causatives and negative verbs). Assertives have an unspecified Force which leaves it free for V2 etc. How does this new division fare with K&K’s prototypical features, such as the fact that and it after the verb? A (semi-)factive such as see should have the fact that under K&K’s account. As mentioned above, this is not so clear-cut. In the BNC, there are 115,200 instances of see of which 4854 of see that but none where a form of see is followed by the fact that. Realize(d), another (semi-)factive, has two instances (see (89) above), and the other semi-factives have some too. As I also mention above, this is problematic for the division, any division in fact. The semantics of the verbs interfere: verbs of sensation, such as see, are incompatible with facts. I would like to suggest that, as in the case of which verbs select for to, there is an element of chance, as to whether and how the ForceP is filled. The basic insight of (103) is that assertives allow a reduction of the CP, e.g. raising and ECM, but no indication of mood in the CP. Nonassertives do the opposite, showing the CP-layer is more robust. Meinunger doesn’t rely on a split CP, but I’ll argue that the (finite) CP is always split with predicates such as regret, and that there is a the fact that blocking the ForceP. This ForceP would also block movement to the topic position in English, perhaps because the entire proposition is presupposed. English complements to assertives would also be split since topicalization is allowed after that, but as I’ll argue, their Force has no mood features. Most Dutch and German complements are single CPs since topics cannot appear after dat/daB, as in (104) and (105) (see 2.3 above for evidence for a split with some verbs):

61

62


(104) *Ik geloof dat deze boeken zij niet waarderen I believe that these books they not appreciate ‘I believe they don’t appreciate these books.’ (105) *Ik geloof dat deze boeken waarderen zij niet I believe that these books appreciate they not

If German and Dutch assertive complements also had split CPs, (104) and (105) wouldn’t be ungrammatical. 4.3 A structure A structure for non-assertive factives is given in (106), with the ForceP ‘blocked’ by a covert or overt element. This structure is a hybrid between a nominal and a clausal structure, and is proposed for it with extraposed clauses in e.g. Stroik (1991) and van Gelderen (1997a): ForceP

(106) Spec

(the fact) (it)

Force′ Force

...

that

It is not clear that there exists a syntactic account for Negative Raising, but one could perhaps argue that a ForceP, as in (106), would somehow stop this process. (106) accounts in a straightforward way for (85) and (86): the extra nominal material would be placed in ForceP. Since infinitival complements are reduced clauses, non-assertive factives with a structure as in (106) will never reduce to an IP completely. This explains why ECM-verbs and passives are not possible (cf. e.g. Massam 1985), but for-to infinitives and subjunctives are. Thus, him in (91a) but not in (91b) can be Case marked; he in (92a) can move to the subject position of the main clause since there is no double CP boundary in between, but not in (92b). Volition verbs, also non-assertives, have a structure as in (106) but where the CP is specified for irrealis and therefore never reducible since an irrealis/future marker for is present. The revised structure for non-assertives is therefore (107), with factives having [+realis] features and non-factives [−realis]:


VP

(107)

ForceP

V regret/intend

Spec

Force′ Force

...

[±realis]

In addition, not discussed above, (106) and (107) might also connect the fact that both wh-clauses, as in (108), and the fact that/it that, as in (109), occur as prepositional objects: (108) I was aware of what he said. (109) They were aware of the fact/it that India and Pakistan were talking to each other.

In descriptive terms, prepositional objects would select a full CP with the Spec filled either by a wh-element, it, or the fact. As is also well-known from H&T (1973) and Authier (1992), complements to the fact cannot have an (in Authier’s terms) iterated CP, as in (110). This fits with the structure in (106) and (107): (110) *The fact that never has he had to borrow money makes him very proud. (from H&T 1973: 466)

An idea similar to (106) is proposed in Varlokosta (1994) who argues that certain clauses in Greek “are not real complements of the matrix verb but stand in a paratactic relation to an empty nominal object of the subcategorizing verb (italics kept)” (1994: 4). Greek also displays a difference in complementizers for the different clauses: pou for opaque complements (e.g. after ‘regret’), oti for those less opaque (e.g. after ‘think’), and na for those even less so (e.g. perception verbs). Having a more complex CP, as in (106), accommodates these. Assertives have a structure as in (111) with just that in the head of the ForceP and no modality features. This lack enables the extended CP to be deleted in the non-finite clause:

63

64


VP

(111)

ForceP

V Spec believe/discover

Force′ Force

TopicP

(that)

...

To recap the English situation, non-assertive factives, such as regret, occur with the fact that and it that. This shows their structure is as in (106) or (107). A few assertive ones, such as realize, do too but this remains unresolved. The other verbs do not and have a more transparent complement, as in (111). ECM is typical for assertives (except verbs of saying which never reduce, e.g. say, claim, assert). Within the ECM verbs, there are two kinds, assertive verbs lack a C in the infinitive but non-assertive volition verbs have one, in keeping with the more explicit character of their CP complement. Please see Table 3.2 for a summary.

5. Verb classes in the history of English In OE, there is not much evidence for a split CP: there are no double complementizers17 or embedded topics. This shows the CP is not split. In this section

Table 3.2.English complements Assertive

Neg Raising the fact it that Emb Top ECM Raising Subjunctive for-to

Non-Assertive

semifactive (discover)

think (believe)

say

factive (regret)

volition (want, intend)

no some some yes yes yes no no

yes no no yes yes yes no no

no no no yes no yes no/yes no

no yes yes no no no yes yes

yes no no no yes no yes yes


(in 5.1), I will show that the differences between verb classes, so obvious in Table 3.2, do not exist in OE, thus lending extra support to my claim for an unsplit CP. For instance, subjunctives occur as complements to both assertive and non-assertive verbs; ECM and for to complements do not occur; and neither does Negative Raising. Then (in 5.2), I will provide evidence for when the split CP is introduced, again compatible with claims in Section 3 above. Prepositions followed by that-clauses appear a little later in the 15th century. The fact that, as in (34) above, and it that do not appear regularly until the 19th century. Negative Raising starts slowly in late ME (Fischer 1998). 5.1 Do matrix verbs matter? I’ll first show that assertive and non-assertive complements are very similar to each other in OE since they both take an indicative and subjunctive complement. This is unexpected given the modern situation since assertives typically lack subjunctives in ModE. It isn’t until late ME when non-assertives start to allow for-to complements, when verbs of belief acquire ECM complements, and not until the 19th century that the fact that appears in the complement clause. Negative Raising isn’t attested in OE either (see Traugott 1992: 270; Fischer 1998: 62) but starts in late ME. So, it seems that at the same time a split CP develops, as argued in Section 3, a reduction of the complement (ECM) also starts. I’ll go into each of these points in more detail now. In English, the core set of non-assertives, the emotives, consists of loans from French, e.g. resent, regret, deplore. OE counterparts are sorgian ‘be sad’ and sweorcan ‘be troubled’. When they occur with CP complements, these are either indicative or subjunctive, as in (112), an analytical subjunctive in fact: (112) hi þa sorgodon þæt hi sceoldon heora gewunan forlætan they then grieved that they should their habits leave ‘they grieved they had to leave their customs.’ (Gregory’s Dialogues, Hecht, 104.18–20)

Volitional and causative verbs are also non-assertive and show single CPs, as in (113) and (114). This CP complement can be either indicative, as in (113), or subjunctive, as in (114): (113) Ic gedoo ðæt ðu forgietsð I make that you forget ‘I’ll make you forget.’ (Alfred, Pastoral Care 207.11)

65

66


(114) Ic wille … þæt þu forgyte þæt ic þe nu secge I want that you forget that I you now say ‘I want you to forget what I am telling you now.’ (Byrhtferth’s Manual 154.14, Visser 841)

Assertive complements such as to the semi-factive forgietan occur with a clausal complement in the indicative, as in (115), and subjunctive, as in (116). The latter is surprising given the situation in ModE summarized in Table 3.2: (115) hig forgæton ðæt hig hlafas namon they forgot that they breads took ‘They had forgotten to take bread.’ (OE Gospel, Matthew 16.5, from Visser 832) (116) ðæt hie forgieten hwider hie scylen that they forget whither they should-subj [go] ‘That they forget where to go.’ (Alfred, Pastoral Care, Hatton 387.14)

Another semi-factive is ongietan ‘to understand’. In the Dictionary of OE texts, it occurs most often in the Pastoral Care, and is often complemented by a that clause. There is one instance where it might have a structure as in (106) above with willan comparable to fact in ModE: (117) he ongiete ðone ufancundan willan ðæt he hit don scyle he understands the divine will that he it do should ‘He understands that it is divine will that he should do it.’ (Pastoral Care, Hatton 51.8)

So, the historical situation may show that the semi-factive assertives are closer to the non-assertives. This is also the situation in ModE. Perception verbs are assertive. In OE, see occurs mainly as a mental perception verb, as in (118), or with a ‘see to it’ meaning, as in (119): (118) Gesihst þu nu þ þa rihtwisan sint laþe 7 forþrycte seest thou now that the virtuous are hated and oppressed ‘Do you see that the virtuous are hated and oppressed.’ (Alfred, Boethius, 9.24–5) (119) Gesih ðu þæt (nænigum menn ðu coeðe see you that (to) man you talk ‘See that you say nothing to any man.’ (Lindisfarne Gospel, Mark 1, 44, from Visser 838)


These complements are subjunctive, which is unexpected given the modern situation. There is a possible 13th century direct perception example, i.e. (120), but it is interesting that the first form of ‘see’, sehe, is in the subjunctive and hence the ‘seeing’ is less direct. The second form, seoð, is a present plural which is more direct: (120) Hwa þat sehe þenne hu þe engles beoð isweamed, who that see-subj then how the angels are grieved þat seoð hare suster swa sorhfulliche afallet that see-pres their sister so sorrowfully overthrown ‘Whosoever that should then see how the angles are grieved, who see their sister so sorrowfully overthrown.’ (Hali Meidhad, Titus, Furnivall, 23, 233–5)

A last class of assertives I’ll look at is the class of saying (see Warner 1982: 189 that verbs of thinking work the same). Verbs of saying occur with a subjunctive (as well as indicative) complement in OE, as in (121), unexpected under the division proposed above: (121) þæt hi cwædon þæt he God wære that they said that he God were ‘That they said he were God.’ (Homilies I 190.32, from Mitchell I: 8)

Farkas (1992): 70) notes the same for Romance languages, and says there is a change of meaning, e.g. between (122) and (123) from Romanian. The indicative “reports an assertion” whereas the subjunctive “reports a directive”: (122) Ion a spus ca˘ Maria a plecat Ion has said that Maria has left ‘Ion said that Maria left.’ (123) Ion a spus ca Maria sa˘ plece imediat Ion has said that Maria that-subj leave immediately ‘Ion said that Maria should leave immediately.’ (from Farkas 1992: 70)

Giorgi & Pianesi (1997: 217) argue that languages differ as to how the indicative and subjunctive differentiate the degrees of realistic and non-realistic. Germanic uses indicative for totally realistic contexts whereas French uses it even for non-realistic contexts. Quer (1998: 93) also shows that there is a lot of variation in which verbs take a subjunctive complement. So far I’ve shown that in OE assertive and non-assertive complements do not have a different form: both allow subjunctives. In the remainder of this

67

68


section, I will show that when for-to and ECM complements start to occur in Early ModE, a difference comes about. As is well-known from K&K (see Section 4.1 above), non-assertive factives such as be important and regret allow for-to complements. This use “is hardly ever met with before” the late 19th century, according to Visser (2244–45). There are, however, instances he mentions, as in (124), and (125) from the OED. The earliest I have found is (126): (124) and wishing for those hands to take off his melancholy bargain. (1681 Dryden, from Visser 2248) (125) as I expect for my reward to be honoured with Miss Sophia’s hand as a partner. (1766 Goldsmith, s.v. interest) (126) moche he lofde echn(e) cniht. þat lofde for to segg(e) riht much he loved every knight who loved for to say truth ‘Much he loved every knight who loved to say the truth.’ (Layamon, Otho, 5523)

The history of important is worth noting in that its use as a predicative adjective with a complement is very late. It too is a loanword from French but quite a late one, namely late 16th century according to the OED (the verb import is from the early 16th century), and is initially only used attributively. Even 18th and 19th century texts, such as (George) Berkeley, (David) Hume, (Emily) Bronte, and (Jane) Austen only contain attributive use, so the subjunctive use, as in (127), must have arisen late. Some early American texts (e.g. Samuel Adams’s 18th century writing) just have subjunctives, as in (127), not yet infinitival complements: (127) Some of our military gentlemen have, I fear, disgraced us; it is then important that every anecdote that concerns a man of real merit among them, and such I know there are, be improved, as far as decency will admit of it, to their advantage and to the honor of a colony, which, for its zeal in the great cause, well as its sufferings, deserves so much of America. (Samuel Adams, letter to E. Gerry 1775, Cushing ed., University of Virginia e-texts)

George Washington (in the 37 volumes of his work written between 1745 and 1799; University of Virginia E-text) uses important for with an infinitive seven times, as in (128): (128) ’Tis almost as important for us to know what does not happen as what does happen. (1780, Volume 19)


In short, the for-to complement arises quite late. There has been a long debate (Callaway 1913, Zeitlin 1908, Lightfoot 1979, Miller 2002) as to whether OE has verbs with ECM complements. The evidence shows, I think, that it does not. It is not particularly important for the question of whether or not OE has a split CP, and I will therefore not go into it here, but see van Gelderen (1993; 1997a; 2000). As to Negative Raising, Fischer (1998) provides evidence that the earliest instances are from the ME3 and ME4 periods of the HC, i.e. 1350–1500. Her earliest example with think is as in (129) from the 15th century: (129) I cannot thinke that he hath informed us all truely. (Paston Letters, from Fischer 1998: 71)

This would seem to imply all complements are split CP, going against the other evidence. Fischer, however, suggests other reasons for the CP not being available, and I’ll follow that account. Hence, it can be argued that a differentiation in complement types arises quite late. 5.2 The fact that and it The introduction of complements starting with the fact that is also quite late. In general, the phrase occurs quite rarely (see Jespersen 1926: 31–2). According to the OED, the phenomenon starts in the 19th century, and the first instances are factive, as in (130) to (132), but slow to get started (e.g. none in Darwin’s Origin of Species or in Jane Austen’s novels): (130) I would not agree to the fact that ennui prevailed. (1803, s.v. fact) (131) … ought to be made aware of the fact that among the reigning Sovereigns, [they] have not … (1851, s.v. fact) (132) We cannot ignore the fact that aeroplaning is beginning to progress as a pastime. (Observer 1927, s.v. aeroplane)

There are, however, some early instances, as in (133), with evyll rather than fact: (133) We have done evyll that we have not taken surete ‘We have done wrong that we aren’t in safety.’ (1489 Caxton, s.v. that)

The texts in which these occur interestingly also have a that-clause after a preposition, as in (134), similar to (109) above. This might provide some evidence for the beginning of a structure as in (106) above in the 15th century:18

69

70


(134) I shalle not leue the goo, withoute that thow hold to me ‘I shall not let you go without your holding me.’ (1484 Caxton, s.v. that) (135) After that I understande by your wordes ‘after I understand through your words.’ (1475 Caxton, s.v. after)

When do sentences with an it, as in (85) above, start appearing? Heralding objects occur in OE, as Visser calls them, but are problematic in terms of their analysis. Visser’s examples with it (p. 460) occur with all kinds of verbs, as with the semi-factive understand in (136), and others such as hear and believe, as in (137). This is unexpected if they are really like the structure in (106). In addition, the heralding objects typically precede the verb, as in (136) and (137), and hence do not contribute to an expanded CP: (136) and þæt georne understandan, þæt ðær symble heofonlicra engla and that eagerly understand, that there feast of-heavenly angels neawest bið nearest is ‘and eagerly to understand that there is a feast of heavenly angels near.’ (Wulfstan Polity 252.15, from Visser 460) (137) þæt ne gelyfdon þte liffruma in monnes hiw… ahafen wurde that not believed that source-of-life in man’s form… elevated was ‘who didn’t believe that the source of life had been raised up in the form of a man.’ (Crist 656, from Visser 460)

If OE is a more paratactic language than its modern counterpart, one expects an independent object with a separate clause. Sentences such as (136) are therefore not like (85) above, but contain two separate clauses. The introduction of the modern variant of (136), one that provides evidence for (106) is quite late. It doesn’t occur in the Early ModE section (up to 1710) of the Helsinki Corpus; in Shakespeare’s First Folio (1623), there are none, but there is an instance of a prepositional object as in (138): (138) I will sweare by it that you loue mee. (Much Ado IV, i, 277)

Having shown in Section 3 that OE has no split CP but that it starts gradually in the 13th century by having double complementizers, then embedded topics in the late 14th, I show in the present section facts concerning verbal complementation that lead to the same conclusion: for to starts very gradually from the 13th century, and ECM is ME as well. The introduction of it/the fact becomes available after the split CP is there. In the next section, I’ll go even further. Not only is there no evidence for a split CP, there is often no CP either.


6. The frequent absence of (embedded) CP in Indo-European and Germanic Curme (1931 etc), Jespersen (1921), Mathesius (1909 but published as 1975), and later Vachek (1961) notice that ModE makes frequent use of non-finite clauses. Much of this research can be criticized for expressing value judgments such as ‘elegant’ and ‘clumsy’. The question, however, is an interesting one and can be rephrased as whether or not there is a macroparameter. This will be addressed in Chapter 12. In order to make a transition to the next chapter, where it is argued that a CP is introduced, I need to say something briefly about the relative lack of CP in OE (even though it is sometimes present where it isn’t in ModE, see e.g. (113)). This section shows some constructions without a noticeable CP, mainly embedded CPs but some main clauses as well. I will be brief since this section just serves to set the stage for the introduction of complementizers. Even though a real CP is questionable, most scholars agree, however, that “some kind of hypotaxis occurred even in Indo-European” (see references in Rynell 1952: 24; Hettrich 1988: 29–30 for a review of in-between phases). Luraghi (2000: 47ff.) sees a real increase in complex sentences between Old Hittite (before 1550 BCE) and Middle Hittite (after 1550 BCE). The examples provided in this section show (i) a lack of CP, and (ii) the many topic positions. In the next chapters, there will be more evidence of the rise of the CP. As is well-known, languages such as Sanskrit are highly paratactic (e.g. Gonda 1971: 137ff). Verbs of saying most often lack a complementizer, and use an adverb-like element iti, as in (139), while the ‘embedded’ clause has the form of a main clause (see also Whitney 1889: 405): (139) nerpo (a)ham iti vadati king (I-am thus he-says ‘He says that he is the king.’ (from Goldman & Sutherland 1987: 67)

In Sanskrit, a pronominal yad functions as complementizer, after verbs such as ‘say’, ‘think’, and ‘believe’ (see Speyer 1886: 359). There also seems to be a consensus that IE has no relative pronoun, see e.g. Greenberg (1990: 125). As Kiparsky (1995) argues, the frequent correlative construction, as in (140), shows that the two clauses are independent and one is not embedded in the other (see also W. Lehmann 1980: 120): (140) yam yajnam adhvaram visvatah that-acc sacrifice-acc (and) service-acc everywhere-from

71

72


paribhur asi sa id devesu gachati surrounding you-are that-nom focus at-gods it-goes ‘The sacrifice and service that you surround, it reaches the gods.’ (Kiparsky 1995: 156, but glosses modified and diacritics left out)

The gloss for (140) could easily be ‘You surround the sacrifice and service. It reaches the gods’, i.e. without a syntactic link between the clauses. Kiparsky argues that Indo-European has “two left peripheral operator positions” for a topic and a focus element (p. 141) but lacks a C.19 A restructuring takes place in Germanic and a C is introduced. Regarding main clauses, if some of the lexical items on the left are really topics, their case should be absolute, i.e. not determined by the verb of the main clause. Meillet (1937:359) says “il n’y aurant pas en indo-européen de ‘rection’ d’un mot par un autre, comme il y en a en latin par example; l’autonomie du mot est le principe qui commande la structure de la phrase indo-européene” [there is in IE no government of one word over the other as there is, for example, in Latin; a word’s autonomy is the guiding principle regarding the IE structure of the sentence]. W. Lehmann (1978), but not (1993), uses this property to argue that IE is a topic-prominent language in the sense of Li & Thompson. However, Luraghi (2000:92) says complex sentences are not frequent in Hittite. Relatively little seems written on wh-questions in Indo-European (e.g. nothing in W. Lehmann and Whitney on the word order). In Sanskrit, however, the k-element is often at the beginning, sometimes following an adverb. Verbmovement is never triggered and, if a question word is a focus, the word order fits that pattern. I’ll now turn to Germanic. Lenerz (1985) argues that Old High German lacks a C and that the non-occurrence of Verb-movement in certain constructions can be explained in this way: (141) ek Hlewagastir holtingar horna tawido I H. of-wood horn made ‘I, Hlewagastir, made a horn of wood.’ (Gallehus’s Horn)

Early Germanic (runic) inscriptions are often Verb-final, as in (141). The same is true for OE and Old High German, as in (142), and (143) from Middle Dutch: (142) er sliumo sar tho zin sprah he quickly at once them spoke ‘He spoke to them quickly at once.’ (Otfrid, IV 16, 36, from Lenerz 1985: 107)


(143) Van watre hi algader seep of water he completely dripped ‘He was dripping wet.’ (Ferguut 731, from van Gestel et al. 1992: 103)

Faarlund (1990: 21ff.), looking at early Germanic runic inscriptions, argues that the SOV order is basic but that the V moves in the main clause. Eythórsson (1996: 109) argues similarly that “there are some cases that require the assumption of movement of the verb” in early Germanic, so this means C may have occasionally been used. Eythórsson (2003) shows that of 13 instances of early runic, six involve some kind of verb-movement. Relative clauses in Old High German do not have a relative particle but use subordinate word order (see Erdmann 1874), and this too points towards a lack of a CP. However, some scholars say that “[r]elative clauses were fully developed in the Gmc languages at the time written monuments began to appear” (Kivimaa 1966: 27). Some constructions show the extended topic possibilities in these older versions (compared to present-day counterparts). It is not clear how tightly (if at all) these topics are incorporated in the CP. (144) is from Old High German and (145) from Middle Dutch. Note, however, that the verb moves, even though V2 does not occur, another indication that the CP may not be used: (144) Des nahtes an minemo bette uorderota ih minen uuine at night at my bed demanded I my lover ‘At night in my bed I longed for my lover.’ (Williram 48, 1, from Lenerz 122) (145) Ende als du mi aenroeps in dijnre temptacien ic sal di and if you me call in your temptation I will you moederlic hulpen motherly help ‘If you call me in your temptation, I will help you.’ (Van Gestel et al. 97, from Stoett 221)

In OE, there are paratactic constructions, but there is usually V-movement in the main clause, as shown in Section 3. In northern texts, this V-movement provides evidence for a single CP in the main clause, but in southern texts, there may be a split. However, even as late as ME, there are texts where sentences are very unconnected to each other. Hence, embedding is not required. For example, 13th century Brut, as in (146), and 14th Century AŠenbite of Inwyt, as in (147), show many such examples:

73

74


(146) An preost was on leoden. laŠamon wes ihoten. he wes leouenaðes a priest was among people Layamon was called he was Liefnoth’s sone. liðe him beo drihten. he wonede at ernleŠe. at æðelen are chirechen. son kind him be God he lived at Areley at lovely a church ‘There was a priest living here, called Layamon. He was the son of Liefnoth, may God be him kind. He lived at Areley, at a lovely church.’ (Layamon, Caligula 1–3, from the facsimile) (147) þe þridde boŠ of auarice: is roberye. þet heþ uele smale roten. þe uerste is ine kueade exequitors of bekuydes. þe oþer is ine kueade lordes by he knyŠt oþer oþer þet be-ulaŠeþ þe poure men: þet he ssolden loki. ‘The third bough of avarice is robbery. That has many small roots. The first is in evil executors of bequests. The other is in bad guardianship by the knight or others that plunder poor people: that he should guard against.’ (Dan Michel’s Ayenbite of Inwyt, p. 38)

These texts are of course not without subordination, as the relative in (147) shows, but the sentences certainly sound very ‘short’ to our ears. I’ll come back to relatives in the next chapter, but some scholars see OE relatives as adjoined and not embedded (O’Neil 1977 as quoted in Hopper & Traugott 1993: 190). In Chapter 1, I mention the change from (148) to (149). These are from different versions of the same text, with (148) having more archaic features in general. In (148), the two sentences are not formally connected, but in (149) they are: (148) þenne he þe treoweðe alre best on. þenne bi-swikes tu heom Then he you trusts all best on. Then betray you him. (Layamon, Caligula 1705) (149) Wan hii þe troueþ alre best. þan þou heom bi-swikest when he you trusts of-all best then you him betray ‘When he trusts you the best, (then) you betray him.’ (Layamon, Otho 1705)

According to Österman (2001: 244), “[t]he change of the relative and conjunctive adverbs from th- to wh- … took about eight hundred years”. The origins of wh-form are in OE but th-form persists until Early ModE. Using the HC, Österman (2001: 253) provides the numbers of each relative, and between ME2 and ME3, wh-forms start to prevail over th-ones, i.e. around 1380. As to the position of both relatives and adverbs, it is most likely the specifier position as I show in the next chapter. Some evidence that they are in the specifier position is that there that/as sequences occur also, as in (150), with as in the head:


(150) Þer ase þeos þincges beoð þer is riht religiun there as these things are there is right religion ‘where these things exit, the religion is right.’ (Ancrene Riwle 12, from the OED entry of thereas)

So, in some of the (older) languages discussed, a complementizer or relative pronoun can be lacking. In addition, the material on the left of the sentence is not always integrated in the sentence, i.e. is not an argument of the verb of the main clause, so not connected by movement. Additionally, V2, i.e. movement of the verb to C, does not (always) occur. This points towards a CP not being (obligatorily) present.

7. Conclusion Clauses consist of three layers: the CP, IP, and VP. In this chapter, I have primarily examined the structure of the CP. It turns out that speakers set their parameters for split or non-split depending on how expanded a particular layer is, in this case CP, but we’ll see the same for IP and VP. Using structural arguments, I show (in Section 2) that in ModE the CP is restricted to a ForceP and TopicP in the case of finite complements and to a non-split CP in the case of a non-finite clause. OE (in Section 3) and other languages are more restricted and select a non-split CP as complement. The evidence is the lack of multiple Cs and embedded topics. There are major differences between verbs in the kinds of complements they select. In English, some verbs allow raising out of their complements, and others have subjunctives. These are well-known from the extensive literature (e.g. K&K, H&T). I argue (in Section 4) that all finite complements have a split CP in English but that in the case of the intransparent complements (to nonassertive verbs), the structure is as in (107) above. This structure accounts for the presence of the fact/it that, impossibility of embedded topics, as well as for the properties of reduced clauses. Verbs that have a structure as in (107) reduce in a different manner from those that have (111) and typically those verbs are nonassertives. In doing so, I have indicated some cross-linguistic variation both in where complementizers are placed (Force or Fin) as well as the ‘expandability’ of the CP. I then (Section 5) show that in OE, verbs do not differ in these ways and argue that that’s another indication that embedded clauses are single CPs. Finally (Section 6), I demonstrate that languages such as OE and Old High German often do not utilize a CP. The next chapters show how languages change in complementation patterns and how they acquire CPs and split CPs.

75

Chapter 4

Spec to Head The rise of the (embedded) CP

Before it is possible for the CP to reduce as a result of grammaticalization, the CP has to be there. As seen in the previous chapter, Indo-European languages such as Sanskrit, Hittite, Greek, and Latin can be quite paratactic, i.e. clauses are fairly independent of each other. This correlates with their having little peripheral structure. With the introduction of material on the left, subordination becomes possible, and the CP is introduced. In this chapter, I show how the CP is used: by incorporation of a pronoun as the Spec of CP and subsequent changes of this element from Spec to Head. Section 1 shows how the Head Preference Principle works with relatives in ModE. As expected given the Economy Principle of Chapters 1 and 2, in spoken English, the head that is preferred over the wh-pronoun in specifier position. In Sections 2 and 3, I’ll discuss the well-known introduction of a complementizer derived from a demonstrative pronoun (stage 2c of Chapter 1). I will argue that the demonstrative starts to occupy the specifier of CP and then, in accordance with the Economy Principle, becomes a head. In addition to the complementizer and relative pronoun in English, I will discuss changes in whether in 4. In 5, I’ll examine some complementizers in other languages that change from specifier to head. The Head-over-Spec Principle is the instigator of internal change. As mentioned in Chapter 1, once a specifier has become a head, new specifiers can be introduced. With relative clauses, the wh-pronoun is the innovation, due probably to contact with French (see Section 2.3). Apart from language contact, there are other external factors. Some of these actually stop the change from specifier to head (see Sections 1 and 4).

1.

Relatives in ModE: From Spec to Head

Many people have shown that demonstratives become complementizers or relative markers (Hopper & Traugott 1993; Heine & Kuteva 2002:106–7; 113; 115–6,

78


hence H&K). In this section and the next, I look at these as instances where a specifier becomes a head, in accordance with (3) of Chapter 2, and is then (sometimes) reinforced again by a new element in the specifier position for languageexternal reasons. I first sketch the ModE situation and show that, even though heads are preferred, there are (prescriptive) forces at work that favor specifiers. In ModE, relatives are formed using that or a wh-pronoun in the case of restrictives, as in (1), and a wh-pronoun in the case of, relatively infrequent, nonrestrictive relative clauses, as in (2). A wh-element and that are not allowed to occur together in standard English, as (3) shows, but occur in certain varieties of English, e.g. in Belfast-English, as in (4), as in other varieties of Germanic: (1) The woman that/who lives next door is about to travel to Mull. (2) Clinton, who was president for two terms, is thinking of running again. (3) *The woman who that I met yesterday is great. (4) I wonder which dish that they picked. (Henry 1995: 107)

Like interrogatives, relatives are involved in checking the features of the C (see Rizzi 1990 for an early version of the Wh-Criterion). Using a non-expanded CP model (there being no evidence that relatives are split in ModE), the wh-element would be in the specifier of CP and that in the head,1 as in (5), and they cannot occur together (in ModE): (5)

CP wh

C′ that

IP

Assuming that in (3) is the head of CP accounts for the absence (again in standard English) of (6) and (7), where of that and that’s are XPs: (6) *The woman of that I saw a picture … (7) *The woman that’s picture I saw …

There could in principle also be evidence for that being in C and the wh-word in Spec from extraction. If that were the head, extraction should be possible, but if that were in Spec, it should block the extraction. This can’t be tested, however, since relative clauses are islands (part of the complex NP constraint), at least in English. In ModE, there is a strong tendency in relative clauses to use heads, such as that (and as), rather than specifiers, such as wh-pronouns, or to have no marker

Spec to Head

at all (Dekeyser 1996: 299). This is shown in Table 4.1 for the 2-million CSE. This corpus was searched for the string [article + noun + relative-pronoun]. The CSE is a relatively formal spoken corpus, and the numbers are even more extreme in other studies of spoken varieties. Montgomery & Bailey (1991: 155) analyze relative clauses in academic writing as well as speech. Out of 200 relative clauses in speech, 138 have that, 36 have no pronoun, and the remaining 26 (or 13%) have a wh-form. The writing sample, as expected, is different. There are 22 clauses with that, 6 with no pronoun, and the remaining 172 (or 86%) are wh-forms. Even for human antecedents, the number of wh-pronouns is very low, namely 18% in the spoken sample, as opposed to 67% in the written sample (there were very few instances). The same tendencies to avoid the wh-form have been observed by others, e.g. Romaine (1982) and Poussa (2002). Miller (1993) shows that in Scottish English, a preposition and relative pronoun are often replaced by that, as in (8), or even just a wh-form, as in (9): (8) I haven’t been to a party yet that I haven’t got home the same night. (Miller 1993: 112) (9) a cake where you don’t gain weight.

In American English, constructions such as (10) and (11) occur, compatible with having where function as a head:2 (10) That house where nobody lives in. (real estate agent, overheard by EvG) (11) Stay where you’re at and I’ll come where you’re to. (overheard by Lynn Sims p.c.)

(8) to (11) are different from the resumptive pronoun cases, as in (12) and (13), also discussed by Miller. I’ll examine the latter in Section 5.2 below and in Chapter 6. These too avoid using a full phrase in the specifier: (12) The road which I don’t know where it leads.

Table 4.1.That versus who in the CSE the N

a(n) N

that wh-form

5637 (=82%) 1199 (=18%)

1758 (=81%) 414 (=19%)

total

6836 (=100%)

2172 (=100%)

79

80


(13) The girl that her eighteenth birthday was on that day. (Miller 1993: 111)

There are several reasons why the change to only using heads has not been completed. Many prescriptive rules concern relatives and these favor wh-relatives over that. For instance, the rule that who is to be used for humans. Fowler (1926[1950]: 716) says “at present there is much more reluctance to apply that to a person than to a thing. Politeness plays a great part”, and a more recent guide says: “who refers to people or to animals that have names. Which and that usually refer to objects, events, or animals and sometimes to groups of people” (Kirszner & Mandell 1992: 381). Anecdotal reports from English composition teachers say they often correct sentences such as (14), and (15) is from an e-mail,3 but would be ‘wrong’ prescriptively speaking: (14) People that eat the above foods, intending to increase their protein consumption, can unintentionally eat too much fat. (15) The people that you should contact are …

If the wh-preference is indeed a prescriptive rule, the difference between written and spoken, mentioned before, is completely expected. There is a second prescriptive rule that favors wh-relatives, i.e. specifiers, namely, the rule against preposition stranding. Sentences that end with prepositions, such as (16), are judged to be incorrect, and (17) is preferred: (16) I met the woman who I had seen a picture of. (17) I met the woman of whom I had seen a picture.

The figures given in Montgomery & Bailey (1991: 156) for the spoken and written samples mentioned above are again interesting in showing that the written sample more closely mirrors the prescriptive norm. In speech, 86% of prepositions are stranded, whereas in writing, 7% are. Since (17) is only possible with wh-relatives, this second prescriptive rule (indirectly) also favors wh-relatives. Dekeyser (e.g. 1996) has argued that this preference for wh-pronouns in formal registers exists even in OE. He compares the translated Bede and the non-translated Chronicle and says that the wh-pronoun “became the hallmark of more sophisticated registers” (p. 293). Note also that very few wh-elements are Case-marked. For instance, in the CSE, there is one instance of a noun followed by whom, as in (18), but hundreds with who: (18) and that the president, whom I think you’ve all heard on this subject, is — he has … (CSE-WH97B)

Spec to Head

So, perhaps who is becoming a head in (16), not marked for Case. This occurrence of who without Case marking may be the reason it is a head ‘competitor’ to that. That would explain why the Case is being levelled towards ‘nominative’ who whereas the other pronouns lose Case distinctions by all becoming ‘accusative’ forms, such as him, us, and me, as in (19), where erm expresses a hesitation: (19) me and my mother have erm arranged it all. (BNC-KC8 920)

There is psycholinguistic evidence that this is indeed the case. Tait et al. (1995) report on aphasic patients who do better comprehending wh-questions with who than with which DP. They conclude “An alternate explanation … [is that t]he difference between who and which DP questions then reduces to the difference between head and phrasal movement” (1995: 79).

2. From Spec to Head in OE relatives Turning now to the historical data, I show that there are innovative cyclical changes where specifiers become heads, in accordance with (3) of Chapter 2, and where new specifiers are introduced. I will first describe the situation in OE, and then argue that the relative complementizer that derives from the specifier that. After the change from specifier to head is complete, a new specifier is introduced. 2.1 Old to Middle English For OE, it is widely accepted in the literature that the regular relative complementizer is þe (e.g. Quirk & Wrenn 1955: 72; Allen 1977; Mitchell II: 243). Another relative clause marker is the demonstrative pronoun (se, þat etc.) with þe, as in (20), or without þe (in the glosses below, þe is translated as ‘that’): (20) Æghwæþres sceal scearp scyldwiga gescad witan worda ond worca every shall sharp shield-fighter difference know words and works se þe wel þenceþ the that well thinks/judges-3S ‘Every sharp shield fighter, who judges well, must know the difference between words and works.’ (Beowulf 287–9)

There is evidence that the demonstrative originally belongs to the main clause since the case in OE is often that of the main clause, unlike in ModE. Hock

81

82


(1991: 342) mentions that “similar patterns are found in Old Norse and traces also in Gothic”. Allen (1977: 84–5) says that “there are a few examples where the relative pronoun ‘attracts’ into the case of the head noun phrase”, as in (21), where þone is accusative even though it functions as subject in the relative clause: (21) Ic wat wytodlice ðæt ge secað ðone haeland ðone ðe on rode I know truely that you seek the-acc savior that-acc that on cross ahangen waes hung was ‘I know that you seek the savior who was crucified.’ (Matthew 28.5, from Allen 1977: 87)

The use of þe dies out in the thirteenth century when it is replaced by þat. Forms of hwa ‘who’ start to be used as relative pronouns later in ME, at the point when þat has taken over from þe already. Since forms such as hwa(m) are already used as interrogative pronouns, i.e. are in Spec CP, it is a logical choice to fill up the Spec for expressive purposes (see also Harris & Campbell 1995: 298; H&K 2002: 251). In structural terms, the following change occurs: (22) a.

fi

CP

b.

CP

C′ C se/þam/ þat...

þe

C′ ...

C

...

who/whom that

The specifier that becomes a head after 1050 and, after that change, a new wh-specifier is introduced, starting in the 12th century, for reasons unrelated to Economy, but possibly due to external reasons. For instance, Mustanoja (1960: 192; 194) suggests Latin influence. In effect, these new reinforcements counter the effects of Economy. In the next subsection, I provide evidence for the Spec to Head change and in the section after that for the introduction of the wh-element. 2.2 That from specifier to head In the previous chapter (Section 2.2), I argue oþ þæt is a sequence of an adverb in the Spec position and a complementizer head. I will argue the same for the

Spec to Head

demonstrative followed by the complementizer sequence, but will provide more evidence and data. The evidence that that is in the Spec is three-fold. (a) That occurs in addition to the C þe, as in (23), without there being evidence that that is a second head. (b) That allows Pied-piping, as in (24), showing that it is part of a phrase that moves to Spec CP. (c) V2 occurs, as in (25), showing that that cannot be in C: (23) eall þæt þe styraþ and leofaþ all that that stirs and lives ‘everything that stirs and lives.’ (Alfric, Genesis 9,3, from Grossmann 1906: 26–7) (24) þæt is seo lufe embe þæt he wite … that is the love about that he knows ‘that is the love he knows.’ (Alfred, Soliloquiorum 341: 32 ASC 344 A13, from Kock 1897: 35–6) (25) on an igland þæt is ut on þære sæ þæt is Meres ig haten on an island that is out on the sea that is Mere’s island called ‘on an island out in the sea that is called Mere’s island.’ (Anglo Saxon Chronicle A, anno 895, see Bean 1983: 91)

When did the change take place? Sentences such as (23) occur up to HC OE3 (period up to 1050); (24) is rare, so it is hard to use as an argument; and (25) becomes the norm since V-2 disappears and is therefore also difficult to use. However, (26) shows that by HC ME1 (1250s) a wh-word and that co-occur, indicating that that is in C: (26) to whom þat miŠth þee, Kyng, slen in þis fiŠth, He shulde haue half… to whom that might thee king slay in this fight he should have half ‘King, whoever slays thee in this fight should have half.’ (Kyng Alisaunder I, 219, from HC ME2)

I will now go into these three sets of arguments in more detail. One of the arguments that þat occupies the specifier position is the presence of combinations such as þatte, þat þe, and þat þat, as in (23) above, where the first þat is in Spec and þe or þat is in the head position (I take the writing of the scribes as indicative of word status). In (27), Spec to Head may already have taken place: (27) and wundor godes þætte on þam cnihtum gecyþed wæs and miracle of-god that-that to the youths made-known was ‘and God’s miracle that was made known to the youths.’ (Daniel 470–1, from Grossmann 1906: 26–7)

83

84


The reason that is not a head in ForceP in (23) is that there is no evidence in OE for a split CP (see previous and next chapter) but also that once the split CP is introduced, that is in Fin, not in Force. The change from Fin to Force occurs after the CP is split. These sentences occur very frequently. As (23) shows, that is used not only with neuter antecedents but (at least from the 9th century on, see Kock 1897: 30; Wülfing 1894: 407–8; Dekeyser 1986: 99) also with a masculine, feminine, or plural antecedent. I take that to indicate that that is losing its gender features. This makes it easier to reanalyze to head later on. The second argument involves preposition stranding. Allen (1977) argues that the demonstrative se always takes along the preposition, but that þe and þat never do (see her pp. 83, 76 and 102–5 respectively). This shows that already in OE, þat is in C (her terminology is pre-CP and pre-Spec/head, but I’ll put it in current terminology), as well as þe. This view is similar to Traugott (1972: 153) who says that the modern relative that is not directly derived from the demonstrative se, seo, that. I will argue that þat is sometimes in C and sometimes in the Spec of CP, as a demonstrative, and that the ME ‘take-over’ is by the demonstrative þat. Mustanoja (1960: 188) suggests the same. Showing that þat can be a specifier is sufficient to argue against Allen and Traugott, but is not sufficient to argue that the relative complementizer that develops from it. I’ll give arguments for both. Thus, Allen’s main piece of evidence that þat is the C head is that prepositions are never taken along in relatives with þat and þe, indicating that both move to C not to Spec. There are, however, sentences such as (28), also from the Chronicle and mentioned in Grossman (1906: 39) and (24) above, repeated as (29), from Kock (1897: 35–6), that show þat is in Spec CP, since umbe þæt is a full phrase. This use continues up to the 17th century (the use of the abbreviation þ “is the cause of much difficulty”, but is seen by Mitchell II: 3–4 as abbreviating þæt (see also Campbell 1959: 12): (28) þa þa hi þyder comon. 7 umbe oþer þing gesprecon hæfdon. then then they there came and about other things spoken had umbe þ hi sprecan woldon about that they speak wanted ‘Then they came there and spoke about other things than those they wanted to speak about.’ (Anglo Saxon Chronicle A anno 1070, Thorpe 1861: 344)

Spec to Head

(29) þæt is seo lufe embe þæt he wite … that is the love about that he knows ‘that is the love he knows about.’ (Alfred, Soliloquiorum 341: 32 ASC 344 A13)

The third argument that relative that is in Spec involves word order. Bean (1983: 92) and Dekeyser (1986) show that the word order with þe is mainly SOV, expected if þe is in C (because the Verb cannot move to C), whereas with the demonstrative pronoun, it is SVO. Bean (1983: 110–1, n 3) says that the latter may reflect the paratactic nature of the relative clause introduced by a pronoun. Wülfing (1894:395) also uses the word order to decide what the status of the demonstrative is: independent pronoun or relative. Bean and Wülfing do not mention þat in particular. Due to the rarity of examples with þat and considering that V-movement is often optional, it is difficult to obtain evidence from word order for the different status of þat and þe, but notice that the verb is final in (27), and second in (25) above, repeated as (30): (30) on an igland þæt is ut on þære sæ þæt is Meres ig haten ‘on an island that is out in that sea (which) is called Meres Island.’

Showing that þat is sometimes in Spec CP, as in (24) to (30), is not sufficient to argue that the relative complementizer in C in fact derives from the demonstrative in Spec CP. There could be other reasons, e.g. French influence, as Einenkel (1916: 119) claims, or analogical extension, since OE that is also a complementizer for embedded sentences functioning as arguments or adverbials. An argument for the claim that relative that derives from the demonstrative (with neuter singular features) is based on van Gelderen (1997a: 76–9). That is the OE relative with singular neuter antecedent (Grossmann 1906: 38). Hence, the agreement is singular in OE, as expected. In ME, that becomes the generalized relative, no longer tied to a singular DP but still with third person singular features: “[t]he Relative (perhaps it does not signify by inflection any agreement in number or person with its antecedent) frequently (1) takes a singular verb, though the antecedent be plural, and (2) the verb is often third person, though the antecedent be in the second or first” (Abbott 1872: 167, italics deleted), as in (31) to (36), mostly taken from van Gelderen (1997a), with the relevant parts in bold: (31) and suggeð feole þinges … þat næuere nes i-wurðen and say many things that never not-was happened ‘and say many things that never happened.’ (Layamon, Caligula 11472–3, from Mätzner 1864: 142)

85

86


(32) and it am I | That loveth so hoote Emelye the brighte and it am I that love-3S so intensely Emelye the bright ‘and it is me that loves Emelye the bright so intensely.’ (Chaucer, Knight’s Tale 48.1736–7) (33) Ye yeve good counsel, sikirly | That prechith me alday that … you give good council certainly that preach-3S me all day that ‘In telling me that … you certainly give me good council.’ (Chaucer, Romaunt of the Rose 741.5173–4, from Wilson 1906: 47–8) (34) we ladys and jentil women in this contrey that is wedows we ladies and noble women in this country that is widows ‘We, ladies and women of nobility, in this country that are widows.’ (Paston Letters III, 338, from Carstensen 1959: 83) (35) With sighes of loue, that costs the fresh bloud deare ‘With sighs of love that cost the fresh youth dearly.’ (Shakespeare, Midsummer Night’s Dream III, 2, 97) (36) Thou that so stoutly hath resisted me you-S that so bravely has-3S resisted me ‘You that have so bravely resisted me.’ (Shakespeare, 3 Henry VI, II, 5, 79, from Visser, p. 91)

This agreement pattern can be explained if that keeps the features of its demonstrative ancestor, namely third person singular. The ‘lack of concord’ is due to the features of the relative complementizer. Even in modern spoken English, ‘mistakes’ such as these occur, where that rather than the antecedent determines the agreement: (37) There are other things you talked about that is not on the tape. (Christopher Darden, 2 March 1995, O. J. Simpson trial)

Sentences (38) to (40), on the other hand, are more modern in that that does not have independent person and number features as a result of its grammaticalization to a head. These are again taken from van Gelderen (1997a): (38) I am the second sonne of old Sir Rowland, That bring these tidings ‘I am the second son of old Sir Rowland that brings this news.’ (Shakespeare, As You Like It, V, 4, 159) (39) what myster men ye been | That been so hardy for to fighten heere ‘what kind of men you are that are so bold as to fight here.’ (Chaucer, Knight’s Tale 48.1710–1)

Spec to Head

(40) Two woful wrecches been we, two caytyves | That been encombred of oure owene lyves ‘Two woeful wretches are we, two wretches, that are burdened by our own lives.’ (Idem, 1717–8, cf. Stoelke 1916: 48–50)

Thus, the relative when it functions as subject often triggers the ‘wrong agreement’, as shown in (31) to (36) above, i.e singular third person neuter. This is not always so, as (38) to (40) show. Wilson (1906: 45) comments that “Chaucer’s skill in handling relative clauses referring to personal pronouns of the first and second persons is shown in that he violates the principle of concord but rarely”. Cases where Chaucer does (antecedent and verb are in bold) are (32) and (33) above. Shakespeare does too, as is shown in (35) and (36) above. The explanation I give for the variation is that the relatives start out with features (singular third person neuter) because they are originally demonstratives in the Spec of CP. They gradually lose these features due to a process of grammaticalization, at which point, there is a reanalysis of that from Spec CP to C. The reason for the grammaticalization lies in Economy: it is ‘cheaper’ to be a head. 2.3 The introduction of a new specifier If Economy is the reason behind the Spec to Head change, the question arises why a new element is introduced in Spec CP again, namely the wh-pronoun, as in (41): (41) the est orisonte, which that is clepid comounly the ascendent the eastern horizon which that is called commonly the ascendant ‘The eastern horizon, which is commonly called the ascendant.’ (Chaucer Astrolabe 669.17–8)

Dekeyser (1986: 100–101) lists the first wh-relatives. They appear in the 12th century and are always PPs, as in (42) and (43), quoted in Dekeyser, and in (44) and (45). Allen (1977: 197–9) has similar examples from the Peterborough Chronicle and Homilies, also from the 12th century: (42) þ hi næfdon na on hwam hi fengon swa rædlice that they not-had no on what they caught so readily ‘that they would have nothing which to seize upon.’ (Anglo Saxon Chronicle, E anno 1085, Thorpe 1861: 352)

87

88


(43) þurh ungewædera for hwan eorþ westmas wurdon swyþe amyrde through bad-weather for why earth fruits became very damaged ‘through bad weather through which their crops became very damaged.’ (Idem anno 1110, Thorpe: 369) (44) forðæm hie nyton mid hwam hie hit ðe forgielden because they not-know with what they it thee recompense ‘because they don’t know how to recompense you.’ (Alfred, Pastoral Care Hatton 323.23–4) (45) ðonne mon mæg ongietan of hwam hit æresð com & for hwæm then man may understand of what it first came and for what ‘Then we can understand whence it arose and why.’ (Idem 241.16–7)

In OE, the preposition is never stranded (cf. e.g. Allen 1977) and cases that look like stranding, such as (46), involve a head complementizer, i.e. not a moved object of the preposition: (46) Seo gesyð ðe we god mid geseon the sight that we God with see ‘The sight with which we see God.’ (Alfred, Sol. 67.6 from Allen 1977)

The reason behind introducing wh-relatives in constructions such as (42) to (45), rather than as subject or object relatives, is not clear. What is clear is that they are phrases in the specifier position (e.g. hwam in (45) is a full phrase and does not move on its own). These new forms fill the gap that is left by that becoming a head. As to what the origin of the new specifier is, as mentioned, Mustanoja cites Latin influence, Romaine (1982) shows that the introduction of the wh-pronouns was stylistically influenced, and Rydén (1983) shows both Latin and French influence. The latter shows that the first instances of who occur in epistolary idioms that are very similar to those in French letters of the same period. For instance, in many of the collections of letters from the fifteenth century, the same English and French formulaic constructions occur, such as in (47) from Bekynton and (48) from the Paston Letters: (47) a laide de Dieu notre Seigneur, Qui vous douit bonne vie et longue. with the-help of God our lord who us gives good life and long ‘With the help of God, our Lord, who gives us a good and long life.’ (Bekynton, from Rydén, p. 131) (48) be the grace of God, who haue yow in kepyng ‘by the grace of God, who keeps you.’ (Paston Letters 410, Davis p. 655)

Spec to Head

Bergs (2002), based on Rydén, argues that who is initially restricted to a deity antecedent (as in (48)), constituting a lexical innovation. The consensus seems to be that wh-relatives are introduced for reasons of language contact and possibly innovative tendencies.4 In conclusion to Section 2, the relative head that develops out of a specifier in accordance with the Economy Principle, providing a structural explanation for this kind of grammaticalization. Other specifiers, such as who/m and which, are introduced, and we see Economy interacting with innovations, making cyclic change possible.

3. Complementizer that: From Spec to Head As is well-known from Lockwood (1968) and Traugott (1972), the origin of the Germanic complementizer that is the demonstrative pronoun (see also H&K 2002: 107). A (made up) sentence such as (49a) could be reanalyzed as (49b): (49) a.

þa cwæð he þæt. Beowulf ferde then said he that Beowulf went. ‘Then said he that: Beowulf went.’ b. þa cwæð he [þæt Beowulf ferde].

I will not go into this further (see Mitchell II: 14; Visser 775 ff.), but will show evidence that the demonstrative that becomes the complementizer is in the specifier position and later becomes a head, in accordance with Principle (3) of Chapter 2. One of the arguments that þat is in Spec, at least in some of the examples, is the presence of combinations such as þatte, þat þe, and þat þat, as in (50) and (51), where one that is in Spec and þe/þæt in the head position. This argument is similar to that made in the previous section for relatives. Grossmann (1906:26–7) has numerous examples, among them (50), where þæt is represented as þ (in accordance with common orthographic practice, see Campbell 1959: 12; Mitchell II: 3–4): (50) þa leton hy sume. þ þ mycel unræd wære þ hy then supposed they some that that much folly were that they togedere comon together came ‘Then some thought that it would be great folly that they should engage in battle.’ (Anglo-Saxon Chronicle D anno 1052 Thorpe 1861: 314)

89

90


(51) forðam wearð ylda bearnum undyrne cuð gyddum therefore became to-elders to-children not-hidden known through-tales geomore þæt þe Grendel wan hwile wið Hroþgar sadly that that Grendel fought while against Hrothgar ‘Therefore, all mankind found out in sad tidings that Grendel fought against Hrothgar.’ (Beowulf 149–151)

This use is frequent in OE, e.g. there are 70 instances in Orosius, but it stops by OE3 of the HC. The last text to use it frequently is the Durham Ritual, from the early 10th century. As mentioned, many of the examples where that is followed by þe occur in relatives, as in (23) above, repeated here as (52): (52) eall þæt þe styraþ and leofaþ ‘everything that stirs and lives.’ (Alfric, Genesis 9,3)

The non-relative use, as in (50) and (51) is also frequent, seven times in Beowulf. Apart from (51), these are (53) to (58): (53) monig oft gecwæð þæt te suð ne norð … oþer … selra nære many often said that that south nor north other better not-was ‘It was often said that no better one could be found north or south.’ (Beowulf 858) (54) þæt gesyne wearþ … þ te wrecend þa gyt lifde that obvious became that that revenger then still lived ‘Then it became clear that a revenger was still alive.’ (Beowulf 1255–6) (55) wen ic talige gif þæt gegangeð þæt | ðe gar nymeð expectation I maintain if it happens that that spear takes [the lord] ‘I firmly believe if it comes to pass that the lord dies …’ (Beowulf 1845–6) (56) ond þu þin feorh hafast þþe | sea-geatas selran næbben … and you your life keep that-that Geats better not-have [than you] ‘and if you are still alive that the Geats won’t have anyone better than you.’ (Beowulf 1849–50)5 (57) Ne bið swylc cwenlic þeaw … þeah ðe hio ænlicu sy þ te not is such queenly manner though that she excellent be that that freoðuwebbe feores onsæce peace-weaver of-life deprive ‘That is not the behavior of a queen, even though she is beautiful, she as a peacemaker should not deprive people of their lives.’ (Beowulf 1940–2)

Spec to Head

(58) ne wene ac wæs wide cuð þte ongenðio ealdre besnyðede not expect but were widely known that Ongentheow of-life deprived ‘(I) don’t expect (this), but were it widely known that Ongentheow killed…’ (Beowulf 2923–4)

Allen (1977: 129) says that she “know[s] of no evidence … to suggest that ðætte was anything more than a variation of ðæt during the literary period”. However, Zupitza’s facsimile edition of Beowulf shows þæt and þe are quite separate most of the time, as I have indicated in (51) to (58). Klaeber’s edition renders them as þætte except in two instances, namely (55) and (56) above.6 This is probably the case in other editions too, e.g. in the Concordance to the Anglo-Saxon Poetical Records, there are 79 instances of þætte, and 10 of þæt þe. It is unfortunate editors don’t heed the manuscript more. If scribes continue to write them as separate, this means that in the scribes’ minds the two are separate, since the language is not standardized. If þe is the generalized complementizer, these constructions show that that is in the specifier position, expected if it is a demonstrative originally. This use dies out by later OE, perhaps around 1000. The first (but the only one mentioned in the OED) occurrence where þæt is already clearly in C is (59), from the late 9th century: (59) gif ge nu gesawan hwelce mus þ wære hlaford ofer oðre mys if you now saw which mouse that were lord over other mice ‘If you saw which mouse was lord over other mice.’ (Alfred, Boethius, 35.30–31, from the OED, but reference as in Sedgefield)

Allen (1977: 121–2) shows that OE, like many of the other Germanic languages, violates the that-trace filter, as in (60): (60) Ac hwæt saegst ðu ðonne ðæt sie forcuðre ðonne sio ungesceadwisnes? but what say you then that is more-wicked than is foolishness ‘But what do you say is more wicked than foolishness.’ (Alfred, Boethius XXXVI.8, from Allen 1977: 122)

The examples she gives do not have þætte, but checking þætte I find none that have genuine extractions from a complement clause. This provides additional evidence that that is in the specifier position, blocking extraction even though the orthography might suggest it is in the head. There are some topicalizations, as in (61), and some extractions from relative clauses, as in (62), but these aren’t counterexamples:

91

92


(61) Hu micel þæt is to secganne þætte Leoniða … gebismrade how great it is to say that Leonidas put-to-shame ‘How great it is to say that Leonidas put to shame …’ (Orosius, Bately 47.9) (62) Hwa is nu ðære ðe gesceadwis sie & to ðæm gleaw sie ðæt he who is now there that wise is and to whom wise is that he swelces hwæt tocnawan cunne ðætte nyte ðætte on gimma gecynde such what distinguish can that not-know that in of-gem family carbunculus bið dio[r]ra ðonne iacinctus? carbuncular is more-costly than iacinctus ‘For who is there, who is wise and experienced enough to distinguish such things, who does not know that in the class of gems the carbuncle is more precious than the jacinth.’ (Pastoral Care 411.25–8, translation as in Sweet)

If there were such extractions, this would mean þætte would be in C, i.e. not blocking wh-movement. I have argued that that’s not the case, so the absence of such data is expected. In conclusion to Section 3, there is evidence up to the year 1050 that the complementizer that is in the specifier position, but that it changes to a head after that, in accordance with the Spec to Head Principle. Once it is a head, it also starts deleting around 1250 (see the OED entry for that II 10), and, as we’ll see later, sequences such as for that and whether that (with that in Fin) become possible for a short period (from 1200 to 1500). Then as a result of further grammaticalization (and in accordance with Late Merge), that changes from Fin to Force (see Chapter 3 for evidence that is in Force).

4. The complementizer whether: Between Spec and Head7 Whether has a number of functions throughout the history of English. The OED has it as originating from an IE form of ‘who’ with a comparative suffix. In OE, it is an adverb, as in (63), as well as a pronoun, and a conjunction as in (64): (63) hwæþre me gyfeþe wearð þæt ic aglæcan orde geræhte however me granted became that I wretch spear-dat hit ‘However, I managed to hit the wretch with my spear.’ (Beowulf 555–6) (64) þær se snotera bad hwæþer him alwalda æfre wille … there the wise waited whether him almighty ever would

Spec to Head

wyrpe gefremman change accomplish ‘There the wise one waited whether the almighty would ever grant him change.’ (Beowulf 1313–5)

Structurally, whether is both a head and a phrase in OE and, unlike previous instances, it does not consistently become a head but either disappears or is ‘forced’ (by prescriptive rules) to remain a specifier. The discussion again illustrates how the general principle of Head over Spec interacts in a variety of ways with other (sometimes non-linguistic) rules in the language. The criteria used to determine the status of whether are (a) co-occurrence with C heads such as þe and þæt, and (b) V2. Both show that whether can be in Spec, but since V2 is not always triggered, it is seen in the head position as well. As to the first criterion, sentences such as (65) and (66) are very frequent all through OE: (65) wisson hwæðre þæt þæt gewit ne mæg mod onwendan knew whether that that mind not can mind change ‘They knew, however, that they couldn’t change someone’s mind.’ (Meters of Boethius 26.100–1) (66) hwæðer þe for his aŠene oððe for his maŠæ whether that for his own [sins] or for his kin ‘either for his own sins or for those of his kin.’ (Belfour Homilies p. 58 l. 23)

Whether introduces a direct question, as in (67) to (70). In this yes-no question function, whether causes the verb to be in second position, even as late as the 18th century. This shows whether is in the specifier position of CP and the verb in C: (67) Hwæðer wæs Iohannes fulluht þe of heofonum þe of man whether was John’s baptism that of heavens or of man ‘Was the baptism of John done by heaven or by man.’ (AS Gospel, Matthew 21.25) (68) Hwæðer wille ge ðæt ic cume to eow, ðe mid gierde ðe mid whether will you that I come to you or with rod or with monnðwære gæste? gentle spirit ‘Do you want that I come to you, with a rod or with gentleness of spirit?’ (Alfred, Pastoral Care 117.7–8)

93

94


(69) Whether hadst thou rather be a Faulconbridge, … ‘Had you rather be a Faulconbridge?’ (Shakespeare, John I, i, 134) (70) Whether doth doubting consist in embracing the affirmative or negative side of a question? (Berkeley, Hylas I, 173, 10 from 1713)

After the 18th century, whether disappears and verb movement to C suffices. Assuming question features are checked in CP (also referred to as the Wh-Criterion, see Rizzi 1990: 65ff), this is initially done via Spec–Head agreement, but later by Head-checking of the verb. Whether has a number of other functions, most prominently as pronoun and as complementizer. (71) and (72), taken from the OED, show pronominal use. Like the direct question whethers, the pronouns are always in the specifier position in OE and ME, since they are phrasal and trigger V2: (71) Hwæðer þara tweŠra dyde þæs fæder willan? which the-gen two did the-gen father will ‘Which of the two did the father’s will?’ (Anglo Saxon Gospel, Matthew xxi. 31) (72) hweðeres fere wult tu beon? ‘Whose companion will you be?’ (Ancrene Riwle 284)

This pronominal use of whether disappears gradually and is replaced by which. The use of whether as complementizer, as in (73) and (74), has survived up to now. Whether initially occupies either the specifier, as in (73), shown by the verb being in C, or the head C, as in (74): (73) nast ðu hwæðer beoð þæs rican mannes ban. not-know you whether be of-that rich man’s bone Hwæðer þæs þearfan or of-that poor ‘Don’t you know if they are of the rich man’s bone or of the poor man’s.’ (Alfric Homilies I Thorpe 256, Clemoes, p. 324) (74) ða cwædon … hwæðer ænig man him mete brohte then said [the disciples] whether any person him food brought ‘Then said [the disciples] has anyone brought him food.’ (AS Gospel John 4, 33)

Later examples are (75) to (78), with whether possibly in the head in (75) and in the specifier in (76) to (78). The reason I say ‘possibly’ is because there might be a split CP at this time with whether as a head in Force in (76) to (78):

Spec to Head

(75) I know not whether Heauen will haue it so ‘I don’t know if heaven will have it this way.’ (Shakespeare 1 Henry 4, III, ii, 6) (76) to witen whether that alle thise thinges maken … to know whether that all these things make ‘to know if all these things make …’ (Chaucer, Boethius 433.160, HC ME3) (77) whether that I lyve other dye whether that I live or die ‘whether I live or die.’ (Malory, Morte d’Arthur, Vinaver edition p. 203, from HC ME4) (78) Then iudge … if I haue done amisse: Or whether that such Cowards ought to weare This Ornament of Knighthood, yea or no? ‘Then judge if I have done something wrong, or whether such cowards should really wear this ornament of knighthood.’ (Shakespeare, 1 Henry 6, IV, i, 29)

Under the ‘Spec to Head’ Principle, one would expect whether to become a C head in present day English, but this hasn’t happened. In the case of heads, e.g. that in (79), wh-movement moves via the embedded Spec CP and the sentence is grammatical. This is not so with whether; it isn’t considered a head since wh-movement across it is blocked, as (80) shows, hence, it must be in specifier position: (79) Who did you think that he met? (80) *Who did you wonder whether he met?

Snyder (2000) reports on a psycholinguistic experiment in which speakers who hear (80) start accepting the construction as grammatical, meaning that they can analyze whether as a head with the right trigger. Sobin (1987: 58) reports on asking native speakers for acceptablity/rejection judgments on sentences such as (80), and the rejection rate is around 75%. The reason, however, that whether has not become a head, I will argue, is prescriptive pressure. For instance, Kirszner & Mandell (1992), in their writing guide, say that whether or not is used ‘when expressing alternatives’ (1992: A26) and would be forced to be a phrase and be a Spec.8 In the CSE, whether is immediately followed by or not in 18% of the instances (to be precise, there are 257 instances of whether or not and 1141 of just whether). There is also direct evidence to the speaker that whether is in the specifier position since sequences such as (81) and (82) are encountered, even in formal speech, as in (81), and writing in (82):

95

96


(81) I just wondered whether that as a next step we might look to see why this seems to be the case. (CSE-FACMT97) (82) The local authority will know whether if they let the council house to the tenant. (BNC-FC3-80)

Regarding whether, historical changes show interesting variety. In the case of yes-no questions, the specifier whether disappears and the verbal head suffices to check the features, in accordance with the ‘Spec to Head’ Principle; in the case of embedded yes-no questions, whether has remained a specifier. I argue that the latter is caused by prescriptive pressures. If whether needs to be followed by or not, it has to be a specifier.

5. Other relatives Relatives in other languages go through the same stages that English has gone through (even though Diessel 1999: 120 mentions the relative rarity of pronominal relatives, there are numerous examples in H&K: 113ff.). The ‘relative’ in Arabic can be seen as a demonstrative not yet in Spec CP, having moved from a position in the relative clause. In Chinese, the relative de is said to have evolved from a determiner di (Wu 2000, Simpson & Wu 2002; n.d., Shi 2002: 9), but I won’t go into this here. Gothic is at the same stage as OE with an optional demonstrative in the specifier position and a complementizer head; and Dutch (and German) are at the stage between OE and ME with a demonstrative used as specifier (with an optional complementizer) and only a wh-pronoun in PPs. In Norwegian dialects, the wh-element in Spec has become the Head, a development that can be expected in English too, with the loss of cases (see Section 1 above), and in Indonesian, the wh-element is avoided (except as a PP in formal speech) and a general complementizer is used. In Russian, the relative, (interrogative), and complementizer are of the same form too, namely ˇcto. I’ll look at these in some more detail in 5.1, and then in 5.2 turn briefly to languages without any movement. 5.1 Spec to Head Standard Arabic represents a stage where the definite article or demonstrative pronoun has not changed to Spec CP (yet), and is in the Spec of DP, in the case of relatives, as in (83):

Spec to Head

(83) ra’ay-tu r-rajul-ayni lladd-ayni qatalaa ‘asad-an I-saw the-men-dual-acc that-dual-acc killed lion ‘I saw the men that killed a lion.’ (from Haddad & Kenstowicz 1980: 142)

In (83), the relative has the same case as the noun in the main clause, in this case accusative, even though its function in the relative clause is subject, indicating it is not in the Spec CP. This stage is reminiscent of early OE (see (21) above). Gothic uses a demonstrative pronoun, e.g. sa in (84), and/or a complementizer ei (“beziehungspartikel” in Eckardt 1875: 15). In (85), the demonstrative and complementizer are combined possibly showing Spec to head: (84) Iþ unleds sums vas namin haitans Lazarus; sah atvaurpans now poor who was name-dat called Lazarus; that lying vas du daura is, banjo fulls. jah gairnida saþ itan was at door his, sores full. and yearning enough eat ‘There was a poor person called Lazarus, who was lying at his door with sores and hungry.’ (Luke 16.20, from Eckardt 1875: 13, translation mine) (85) Aþþan all uskiusaiþ þatei goþ sijai gahabaiþ but all prove that-that good is hold ‘Prove all things; keep what is good.’ (I Thessalonians 5,21)

Ei (sometimes þei) can also combine with personal pronouns, and originates in a demonstrative particle itself (Eckardt 1875: 15; Windisch 1869: 205; 209; 248). Hence, the IE demonstrative (cf. Sanskrit ja) becomes a head in Gothic. In OE and ME, as I have shown in 2.1, the cycle goes on and the demonstrative (specifier) becomes the head again. In Dutch and German, for reasons that are unclear, there hasn’t been a change from specifier to head. In Dutch, for instance, there is a demonstrative die/dat ‘that’ agreeing with the antecedent in gender, that serves as a relative in (86). Die is most likely in Spec CP because of the ungrammaticality of (87) and of the improved grammaticality of (88): (86) Hij die daar heen gezonden is doet zijn werk goed he who there to sent is does his work well ‘He who was sent there does his work well.’ (87) *Hij wie die ik gezien heb … he who that I seen have ‘He who was seen by me.’

97

98


(88) ?De man die dat ik hier zag … the man that that I here saw ‘The man that I saw here.’

There are in Dutch, as in ME, wh-forms with prepositions (and indirect objects), as in (89) (wie for humans and waar for other antecedents): (89) De man (aan) wie ik het verhaal vertelde …. the man (to whom I the story told ‘The man I told the story to …’

This pronoun is clearly phrasal because of the preposition and because it is possible to insert a complementizer dat after wie (in non-standard varieties, as well as e.g. Middle Dutch). The change in Dutch is a change (see van der Wal 2002) from demonstrative to wh-pronoun, but no change towards a head. There are Norwegian dialects (e.g. Taraldsen 1985) in which the wh-element when it is monosyllabic does not bring about V2, as in (90), but does when it is a phrase, as in (91). This is explained if the wh-element has changed from Spec to Head blocking Verb-movement, in accordance with the Economy Principle: (90) Ka dokker sa what you say ‘What are you saying?’ (91) Ka for nokka sa dokker what for something say you ‘What kind of thing are you saying?’ (both from Taraldsen 1985: 21)

Many other languages display this preference of Head over Spec in varying forms. Indonesian, for instance, has both strategies, similar to ME and Dutch. When the relative is a subject or object, yang, as in (92) is used, and when it is a prepositional object siapa and mana are used, as in (93). The latter is, according to Kwee (1965: 105) “very rarely used in every-day speech”, and Suwarno (p.c.) says that “the people I hear using these have been influenced by English”. Instead, the head yang, as in (94), replaces the specifier: (92) Itulah rumah yang kita cari that house that we seek ‘the house which we seek.’ (93) Orang kepada siapa saya berikan uang itu adalah tukang-kebun saya. man to whom I gave money the is gardener my ‘The man to whom I gave the money is my gardener’.

Spec to Head

(94) Orang yang saya berikan uang itu adalah tukang-kebun saya. man that I gave money the is gardener my ‘The man that I gave the money is my gardener.’ (sentences from Kwee 1965: 105, my glosses)

So, as in ModE, the head strategy of (94) is preferred, showing not an actual change of a lexical item from Spec to Head, but a disuse of the Spec option. See H&K (2002: 254) for other examples of relatives becoming complementizers. 5.2 No movement at all There are a number of languages (see Keenan & Comrie 1977) where relatives have a complementizer and a resumptive pronoun, as in (95) from Farsi, and (12) above: (95) in haman mardist ke asbi az u xaridam this same man-is that the-horse from him I-bought ‘This is the same man that I bought a horse from.’ (Lambton 1953: 76)

This is of course another way of obeying the head-preference option. I’ll come back to this in Chapter 6, Section 3.1.

6. Conclusion The main part of this chapter is devoted to the introduction of the CP. The Economy Principle of ‘Head over Spec’ explains why a demonstrative pronoun changes to a complementizer. This change represents a change from Spec to Head, and can be observed in a number of other developments: relatives, complement clauses, and interrogatives. It can also be observed in other languages. It is a language internal change. There is, however, a counter movement to innovate otherwise possibly ‘bland’ head constructions in relatives. The source of these is typically external, in the case of the wh-element probably French or Latin. In addition, there are a number of prescriptive rules that stop the changes prompted by the Economy Principles.

99

Chapter 5

Late Merge The rise of the split CP1

In Chapter 3, I showed that embedded clauses in OE are single CPs, not split ones. Multiple complementizers and embedded topics are considered evidence for the introduction of a split CP in the late 14th century. In the present chapter, I will elaborate on both phenomena, on complementizers in Sections 1 and 2, and on topics in Sections 3 and 4. I argue that a split CP is triggered through Late Merge. For, till, and that become merged higher and change from Fin to Force, traditionally seen as their grammaticalization. The incorporation of (multiple) topics then follows. So the split CP is a result of the Late Merge Economy Principle. Non-finite clauses, not surprisingly, do not end up with an elaborated complementizer system. This makes sense since non-finites are more dependent in other ways as well. A split CP makes a clause more independent. In Section 1, I first examine if for could have undergone the same development from Spec to Head as some of the complementizers discussed in Chapter 4. I conclude it is always a Head, and then show how its development from Fin to Force contributes to the rise of the split CP. I also examine till and that and show that the same change happens. In Section 2, I examine non-finite complementizers and show how they do not result in split CPs. In Section 3, I examine the left periphery in older languages and show how topics are incorporated in OE and ME after for starts to merge late. In Section 4, I briefly look at left peripheral positions becoming integrated into the argument structure.

1.

Finite complementizers in ME

1.1 For In Chapter 2, Section 2, I mention for as an example of a change from a Head to higher Head in accordance with the Late Merge Principle. However, the earliest use of for as a sentence connector is as part of a phrase in the Spec position. In

102 Grammaticalization as Economy

this section, I therefore first examine whether for could have gone from a Head in the Spec of a CP to a C Head. The conclusion is that it never occupies the Spec position on its own as an adverb (only as head of a PP) and that the origin of the complementizer for is a head (preposition). According to the OED, for starts functioning as a C around 1200. Van Dam (1957: 40–1) finds earlier instances, but doubts if the ME use is a continuation of the OE one. In OE, for is part of compounds such as for ðon ðe, for ði, for ðæm ðe, as in e.g. (1) and (2). With ðe present, there is usually no V-second, indicating ðe is in C: (1) ac for þæm þe hie us near sint, we … ne magon … but for that that they us close are we not may ‘but for those that are near to us, we can’t …’ (Orosius, Bately 122.18–9) (2) soð ic cuedðo iuh to forðon onfengon mearde hiora verily I say you to for-that receive [they] reward their ‘Verily, I say to you that therefore they’ll receive their reward.’ (Lindisfarne Gospel, Matthew 6.16)

In (1), for is part of the PP in Spec CP, as in (3), and as in other instances where the Spec is filled (see relatives in Chapter 4), a head ðe (and later that) is optional: (3)

CP PP

C′

P

DP

C

for

ðam

(ðe)

...

If for were in Spec on its own before becoming a head (or a separate head in a split CP), one would expect for ðe to occur and this occurs only once in the entire DOE Corpus, namely (4). This, however, could be analyzed as an adverb: (4) Forþe se Scyppend gesceop ærest þone middaneard swylce he þæm therefore the creator created first the earth so he for-the men hus getimbrode people house built ‘The creator first created the earth so the people would have a home.’ (Alfric’s Alcuini interrogationes 7.49)

There is a similar singular occurrence of for that in (5), lacking in three other versions of the same sentence (see e.g. (2) above).2 In (4) and (5), the more likely

Late Merge 103

analysis, however, is one where for is a regular preposition, not a conjunction: (5) soþ ic eow sæcge for þ hiæ onfengun heora lean verily I you say for that they receive their reward (Rushworth Gospel, Matthew 6.16)

Van Dam (1957: 53) finds one additional instance of (4) in Wærferth’s translation of the Dialogues of Gregory (MS O, 330.7, but another manuscript has forþon ðe). So these are highly infrequent considering that there are thousands of for þon ðe variations in the DOE Corpus (for instance, 2694 for-demonstrative compounds followed by ðe/ðe, 1652 combinations of for with an orthographically separate demonstrative followed by ðe/þe, as in (1)). The virtual absence of (4) in OE shows not only that for is not a Spec in the CP but also that an expanded CP is absent. Additional evidence for the unsplit nature of the CP is that topics do not follow or precede for. Combinations with for as head of a phrase, as in (1), (2) and (6), typically show a Spec-like behavior since they trigger V2 when not followed by ðe, as in (6): (6) for hwon secest ðu sceade sceomiende? for why seek you shade ashamed ‘Why do you seek shade in shame?’ (Genesis 3, Krapp 1931 ed)

The structure is therefore as in (3), with for part of the Spec, but with secest in C. As mentioned in Chapter 3, for is introduced ‘on its own’ in the 12th century. One of the OED’s first examples is (7), but the first instance is from 1137 (from the Peterborough Chronicle). It is at that time a general purpose complementizer and can replace whether, as in (8), also from the OED: (7) for heo synd godes gesceafte ‘because they are God’s creatures.’ (Alcuin’s V&V 115) (8) witen for it dede ðis witterlike know whether it did this certainly ‘know if it did this for sure.’ (c1250 Gen&Ex 2651)

“The use of for as a conj[unction] has not been found earlier than the 12th c” (OED, s.v. for). For marks causation in OE but as a preposition, e.g. for þon ‘therefore’. Jucker (1991) provides a very detailed description of the situation around 1230. He shows that at the time for is becoming a complementizer on its own, e.g. in the Ancrene Wisse (pp. 207–8), the verb never moves to C (not Jucker’s terms but mine), indicating that for occupies the head position.


Once for occurs as a C, it almost immediately grammaticalizes to Force and triggers a split CP, as also shown in Chapter 3, Section 3. The evidence for that is the occurrence of for followed by that, as in (9) and (10), which occurs frequently in the HC ME section (namely 111 times), even though only as causal adjunct not as complement: (9) we clepeð him fader for þat he us feide here ‘we call him father because he feeds us here.’ (from HC ME1, OE Homilies, Morris p. 25) (10) for þat he hadde islehŠe moche of hire cunne for that he had slain much of their people ‘because he had slain many of their people.’ (Layamon, Otho 5453)

According to the data in the HC, for starts to precede that in ME1, i.e. from 1150 on. This is confirmed by the first example in the OED (entry of for that) which is from 1200. The use becomes very rare by the time of Shakespeare but the OED lists 18th century examples of a perhaps fossilized form. For in Force regularly precedes if in Fin, as in (11): (11) For Šif we here Šerneþ wonie mid Greckes for if we here agree live with Greeks ‘because if we agree to live here among the Greeks.’ (Layamon, Otho, 483)

The opposite orders that for and if for never occur introducing a clause (in the HC at least). Forte (that) ‘until’ gets used as a conjunction with finite clauses in 1200 for the first time (again from the OED entry), and here too very early it moves to Force, e.g. in Layamon’s Otho version (but not in the earlier Caligula version): (12) forte þat he come to Maximian to Rome until that he came to Maximian to Rome ‘Until he came to Rome to Maximian.’ (Layamon, Otho 5746) (13) forte þat ich mihte bet borewes a-winne until that I might better boroughs win ‘Until I might the better conquer boroughs.’ (Layamon, Otho 7700)

In Layamon, there may be additional evidence that for is in Force since negative verbs move to C when for is there, as in (14) and (15), from both versions: (14) for nes he neuer þi fader because not-was he never your father ‘because he was never your father.’ (Layamon, Caligula 1146)

Late Merge 105

(15) for nis þar no kinelond because not-is there no kingdom ‘because there is no kingdom.’ (Layamon, Otho 6660)

This may show that for is in Force and an extended CP is triggered when negation needs to have scope. A question with this is why regular verbs don’t move in the same way and why only the negative triggers the split. In short, the main evidence that OE for is in the head position of a PP (which itself is in Spec) is that (a) it occurs as a head in sentences such as (1) and (6), and (b) it does not occur before another C or before a topic, as also shown in Chapter 3. In ME, for becomes a head and actually helps to expand the CP. The evidence is that it precedes þat and þe which occupy C at that time. I’ll turn next to another complementizer that changes in the same way, namely till. 1.2 Till The situation with the complementizer till is similar to the one with for. The OED lists (16) as the first example, probably composed in 1154: (16) til hi iafen up here castles ‘till they gave up their castles.’ (Anglo Saxon Chronicle E, anno 1137)

From then on, the sequence til(l) that starts to occur, as in (17) and (18). In the OED, the earliest examples are from Ormulum, i.e. 1200. This is confirmed by the HC: (17) Anan till þatt itt cumenn wass again till that it come was ‘Again till it had come.’ (Ormulum, Introduction l. 3) (18) til ðat he ðe heuene feð till that he the heavens reaches ‘until he reaches heaven.’ (Bestiary, Natura Aquile)

Till for and that till do not occur. This shows that till, like for but unlike that, is in Force. I’ll now turn to that. 1.3 That The occurrence of (10) above shows that that is not in Force, but in Fin. This is the situation at least till about 1500. However, once the split CP is available and


Force is not occupied by for, that moves there. The well-known who that sequences, as in (19) from the 15th century, do not provide evidence either way since who could be in the Spec of ForceP with that in the head, or they could be Spec and Head of a lower Phrase. (20), also from the 15th century, is clearer since the topic as for such mony that … follows that, indicating that is in Force: (19) or who that dothe it I wyll paye ‘or who that does it I will pay.’ (Paston Letters #346, anno 1471) (20) And I told him that, as for such mony that shuld …, I wold … ‘and I told him that, as for such money that should …, I would.’ (Paston Letters #75, anno 1465)

So, that moves to Force in ME, as (20) shows, where the relevant that precedes the topic. In the North, as (21) shows, this is true as well since that precedes the (northern) complementizer at: (21) þat att Še toke me, take it you þere ‘That where you took me, may it take you there.’ (York Plays 32, 298)

A century later, as shown in (22), the modern order, with that in Force, prevails: (22) I am sorrie that with better speed and iudgement I had not quoted him. (Shakespeare, Hamlet, II, i, 111)

The number of constructions where that follows a relative pronoun, as in (19) above, becomes very limited in that period, e.g. four instances in Shakespeare.3 I will not go into the reasons why ModE enforces the doubly filled comp filter (but see Rizzi 1990; H. Koopman 1996; 2000). The ModE situation is one where that and finite for are both in Force, as shown in (23) and (24), repeated from Chapter 3: (23) She maintains that Irish stew she sort of likes t. (24) … for Irish stew I sort of like.

A possible scenario for the changes in the finite CP system is that because there is a shift of the tense features from C to I in ME (and hence the loss of V-second), tense features are no longer in Fin but in I, and that is reanalyzed as a marker of a subordinate clause, i.e. a Force, and a split CP is triggered. This brings that in line with finite for, and shows that change is typically towards a higher position. As mentioned in Chapter 3 as well, the non-finite complementizer for becomes syntactically restricted in that it cannot occur together with a focus or topic, as (25) shows:

Late Merge 107

(25) a. *I expect for [her homework] her to do. b. *I expected [her homework] for her to do.

I’ll examine this more in the next section. In conclusion, there is evidence for two finite complementizers, one in Force and one in Fin, with one clearly changing from Fin to Force, i.e. to a higher position.

2. Nonfinite complementizers in ME In this section, I show that where the finite complementizer expands the CP in ME, the infinitival ones do not. In ME, there is an increase in infinitival markers: for, to, forto, at, and till are all used. I will show that in terms of the CP structure, for and till may first be in the ForceP (indicating purpose) and are then reanalyzed as Cs in unsplit CPs. Some examples of various complementizers and infinitive markers from an early 14th century text, Cursor Mundi (Cotton version), are given in (26) to (30)4 showing the varieties of uses: (26) þis ilk bok is es translate … | For the commun at understand this same book [it] is translated for the people to understand ‘This book has been translated so the (common) people can understand it.’ (Cursor Mundi, Cotton, 232; 236) (27) Til all oure bale ai for to bete | Oure lauerd has made þat maiden till all our sorrow always for to heal our lord has made that maiden ‘Our lord has created that woman in order to heal our sorrow forever.’ (Cursor Mundi, Cotton, 105–6) (28) For wen þow traistes[t] wenis at be … ‘because when you surely think to be.’ (Cursor Mundi, Cotton, 59) (29) To lith a quil his word til her ‘to listen a while to hear his words.’ (Cursor Mundi, Cotton 5330)5 (30) Man yhernes rimes for to here ‘People yearn to hear rhymes.’ (Cursor Mundi, Cotton 1)

These constructions provide evidence for a split infinitival CP, as in (31), a simplified structure for (27), (32), a structure for (30), and (33) for (26). They continue to occur well into ME. In (31), it would be unlikely that the topic is in Spec IP, and till in an unsplit C, since there can be at least two topics, as in (27).


Thus, till is in Force, for in Fin (unlike in finite clauses), and the CP is split: (31)

ForP For′ TopP

For

Top′

till all oure Top

FinP

bale

Fin′ Fin

IP

for

I′ I

VP

to

V bete6

(32)

TopP Top′ rimes

Top

FinP Fin′ Fin

IP I′

for I

VP

to

V here

Late Merge 109

(33)

FinP Fin′ Fin

IP I′

for the commun

I

VP

at

V understand

The most common, both as a preposition and infinitive marker, is to. In the Prologue to Cursor Mundi, it occurs as a preposition 13 times and as an infinitive marker (not preceded by for) 28 times, as in (34): (34) And openlik bigan to preche ‘and openly began to preach.’ (Cursor Mundi, Cotton, 175)

The situation can be summarized in a table (Table 5.1). Thus, in ME, there is a real increase in complementizers that derive from prepositions. In the Cursor Mundi, this is certainly the case and here, there may be evidence for a split infinitival CP. In this text, till could be in Force and for in Fin. At and to are infinitival markers, most likely in I. The consolidation of the non-finite CP as unsplit occurs before 1600. In Chaucer’s English, i.e. around 1400, the patterns [for to V O], [for S to V O], and [O for to V] occur, suggesting that the non-finite CP is still expanded, since an object precedes for. Some 200 years later, in Shakespeare, there are 32

Table 5.1.Ps in the Prologue of the Cotton Cursor Mundi

to at til/l for for to

P

Inf

Fin/Force

Total

13 3 5 8 –

28 3 – – 7

– – 1 12 –

41 6 6 20 7

110


occurrences of for to left, and a few [for to V O], but no objects precede for. The regular construction is [for S to V O], as in (35): (35) Leauing thy trunke for Crowes to feed vpon. (2 Henry 6, IV, x, 90)

The rarity of preposed objects is evidence for a change towards a non-split (non-finite) CP. In other texts, e.g. Sidney’s 1580s Defence of Poetrie, Tourneur’s 1611 Atheists Tragedie, preposed objects no longer occur either. Having established that non-finite CPs may have been split briefly in late ME, I now turn to a possible reason for the demise of the split CP, namely the subcategorization of infinitival CPs with for by (deontic) verbs. As is shown in van Gelderen (1998), an infinitival clause with for is used as a complement to a verb for the first time around 1250. It is very rare though, since in a long text such as Layamon’s Brut (Otho version), there is only one clear instance, namely (36), as mentioned before: (36) moche he lofde echn(e) cniht. þat lofde for to segg(e) riht much he loved every knight who loved for to say truth ‘He loved every knight dearly who loved to say the truth.’ (Layamon, Otho, 5523)

In Chaucer, 150 years later, this construction becomes more frequent, e.g. (37) to (40): (37) Som man desireth for to han richesse some man desires for to have riches ‘Some men desire to be rich.’ (Chaucer, Knight’s Tale 42.1255) (38) Nedeth namoore for hym to go ne ryde needs nomore for him to go neg ride ‘He didn’t need to go out anymore.’ (Chaucer, Merchant’s Tale 158.1615) (39) That oother manere is whan men or wommen preyen for folk to avauncen hem, oonly for wikked flesshly affeccioun that they han unto the persone7 ‘The other way is when man or woman pray for people to help them only out of wicked devotion that they have for the person.’ (Chaucer, Parson’s Tale 315.785–6) (40) Assembled been, his answere for to heere assembled were his answer for to hear ‘were assembled to hear his answer.’ (Chaucer, Wife of Bath’s Tale 119.1029)

Late Merge

In (38) and (39), the position of the subject indicates that for and to surround the subject. The construction can be seen as evidence that the verb selects for as a complementizer with verbs of volition, such as desire in (37), that indicate future. As mentioned above, the CP is still ‘expandable’ with an object preceding the complementizer, even though most instances of split CP are with adjunct CPs, as in (40), not with complements.8 If subcategorization, i.e. the selection of a complement, is always of a sister constituent, with a split CP, the selection will be of the ForceP, not of the FinP. The future features in Fin can therefore not be directly selected, and this is what I claim happens. So, when verbs such as expect, hope, desire start having for clauses with future/irrealis features as complements, as in (36), an expanded CP becomes incompatible. This is indeed the modern English situation. In terms of conclusions reached about complements in Chapter 3, this means that nonassertive verbs typically have an expanded CP as their finite complement, but a single CP as their non-finite complement. Assertive complements typically have simpler CPs and their infinitival complements involve an IP not a CP. Concluding Sections 1 and 2, I have shown how multiple complementizers arise through Late Merge. I now turn to topics, which also indicate that a CP is split.

3. The left periphery In this section, I first show that certain languages do not incorporate topics into the sentence structure. After that, I come back to the ME situation where topics become incorporated due to the changes described in the previous two sections. Kiparsky (1995: 153ff.), based on work by M. Hale,9 argues that the basic IE sentence structure has two positions on the left, a topic and a focus position, but crucially no C or CP. So, in (41), from Sanskrit, ratham ‘chariot’ is in topic and ko ‘who’ in focus position: (41) ratham ko nir avart ayat chariot-acc who down rolled ‘who rolled out the chariot.’ (from Kiparsky 1995: 154)

Some of the arguments for the absence of a CP have been given in Chapter 3, Section 6, and they revolve around the non-embedded nature of subordinate clauses. Many times, clauses are not connected to an argument position in the main clause, as in for instance (42):

111

112


(42) ista virtus est … qui malum fert fortiter that courage is who evil endures bravely ‘It is courage for someone to bravely endure misfortune.’ (from Kiparsky 1995: 157)

As in OE, Classical Latin obviously has subordinate clauses. However, it isn’t until Late Latin, e.g. of the Vulgate Bible, that there is V2, as in (43), with a topic (see references in Haiman 1974: 69). So, as in OE, the use of the CP may have been more restricted: (43) In principio creavit Deus caelum et terram in beginning created god heaven and earth ‘In the beginning God created heaven and earth.’ (Haiman 1974: 69)

In OE, topics occur in left peripheral positions regularly in main/finite clauses. Since V2 occurs frequently in the main clause, sentences with an initial topic, as in (44) and (45), occur regularly as well: (44) Rihtlice wæs se swicola herodes fram þam tungelwitegum bepæht rightly was the false Herod from the astrologers deceived ‘Rightly, the false Herod was deceived by astrologers.’ (Alfric, Homilies I, Clemoes p. 240; Thorpe 118.34) (45) Geleafful wæs se hreofla þa he cwæð ‘Faithful was the leper when he spoke.’ (Alfric, Homilies I, Clemoes p. 242; Thorpe 122.10)

OE is not as absolute as e.g. German and Dutch are, and topics often do not trigger V2. This may be a sign that they are not incorporated into the sentence structure, unlike the wh-element. For instance, W. Koopman’s (1998: 137–8) figures for Alfred’s Orosius and Pastoral Care and Wulfstan’s Homilies show that only half of the topicalized PPs and objects result in a V2 order (when the subject is a full DP).10 In Alfric, however, the figures are much higher and topics generally ‘cause’ inversion, as shown in (44) and (45) above. The position of pronominal subjects confirms that topics are not in the same position as wh-elements: the topic is typically followed by the clitic whereas the wh-element is followed by the Verb and then the pronominal subject (see e.g. Eythórsson 1996). Multiple topics occur as well, as in (46) to (49), and occasionally cause V2, as in (46) and (47), but sometimes don’t, as in (48) and (49). Note that I use ‘multiple topics’ even though some could be topic and focus, as in (48), and fit in a Rizzian structure:

Late Merge

(46) & ðonne wyrð þurh Godes mihte sona deofol swyðe geyrged and then gets through God’s might soon devil very terrified ‘Through God’s power, the devil soon gets terrified.’ (Wulfstan 3, 176, 28, from Haeberli 2000: 118) (47) þa wæs in þa tid Uitalius papa þæs apostolican seðles aldorbiscop then was in that time Vitalius pope the apostolic see’s senior-bishop ‘At that time, Vitalius was chief bishop of the apostolic see.’ (Bede 1, 252, 1, again from Haeberli) (48) Him sona of heofena mihte com unasecgendlic myrhð … him at-once from heavens’ might came unspeakable joy ‘There came to him at once unspeakable joy through the power of heaven.’ (Byrhtferth’s Manual, 150.22, from Koopman 1997: 308) (49) Eft on ðære ylcan nihte æfter ðæs cempan martyrdome ferde again on the same night after of-the soldier martyrdom went Decius to … Decius to ‘Again on the same night after the soldier’s martyrdom Decius went to …’ (Alfric, Homilies I, 428.9, from Koopman 1997: 308)

Throughout OE, then, there is an increase in the use of CP. Thus, even though a C is not always present or utilized, left-peripheral positions, such as in (46) and (47), exist that are reanalyzed as part of the CP and split CP (see Chapter 3, Section 3). So, when languages gain a CP via a reanalysis of the topic and focus positions, the CP may be split. In OE, there is some evidence of a split CP in the main clause because the verb moves into a position in between the ‘topics’, as in (46) and (47) above. In ME, the embedded CP becomes split, and stays that way in the ModE finite clause, but not in the non-finite one. Kroch & Taylor (1997) have argued there is a dialect split in ME with northern texts behaving like strict V2 languages of the Dutch/German style and southern texts continuing the OE optional V2. Haeberli (2000: 129) links to that the possibility in southern ME to have an adverb split the finite verb from its subject, as in (50), but not in northern ME: (50) Se þicke is þrinne þe þosternesse þt … so thick is therein the darkness that (Sawles Warde 171.8, from Haeberli 2000: 123)

In (50), the verb would be in the expanded CP together with Se þicke and þrinne, but this would not be possible in the north where a split CP is not available. Northern texts are similar to Dutch, where (51) is ungrammatical:

113

114


(51) *Ik zei dat gelukkig Kim de wedstrijd zou winnen I said that fortunately Kim that race would win ‘I said that Kim would win the race, fortunately.’ (van Gelderen 1993: 47)

As mentioned earlier, sentences such as (52) in which topics follow finite for, indicate that for is in Force: (52) For frenshippe we haue foune ‘because we have found friendship.’ (York Plays, 10,12)

The presence of sentences such as (53), repeated from Chapter 2, contributes to the reanalysis of the object of the preposition as a topic: (53) þu Šef þeseluen for me to lese me fra pine ‘you gave yourself to me to release me from my pain.’ (Wohunge 88–9)

I will summarize the data on the CP in the history of English. Initially, the use of the left periphery is limited (as evidenced e.g. by optional V2 and paratactic constructions), but the CP is used and even split after 1300. It remains split for finite clauses (see 2.1 of Chapter 3) but not for non-finite ones. Most of the changes driving these developments are due to the Economy Principles formulated in Chapter 2. The demonstrative enables two sentences to become connected in the first place and then through the change from Spec to Head (see Chapter 4) to make the connection tighter. After becoming a head, the complementizer is then merged higher, contributing to the introduction of a split CP. Lastly, the topicalized phrase is incorporated, as a TopP.

4. Topic incorporation Text- and handbooks make frequent mention of ‘anacoluthic subjects’ (e.g. Visser 60ff.). These are constructions in which a topic occurs with a subject that ‘repeats’ it. Examples are given in (54) to (56): (54) Se seðe ealne ðone wisdom ðara uferrenna gæsta oferstigð & ær he who all the wisdom of-the higher spirits surpasses & before worolde ricsode on hefenum, hit is awriten on ðæm godspelle, Iudeas world reigned in heavens it is written in the Gospel Jews comon & woldon hine don niedenga to cyninge came and wanted him make by-force to king ‘He who surpasses all the wisdom of the higher spirits, and reigned in

Late Merge

heaven before the world was, it is written in the gospel that the Jews came and wanted to make him king by force.’ (Alfred, Pastoral Care, Hatton, 33.12–14 from Visser 61 but text as in Sweet) (55) Wærleas mon ond wonhydig, ætrenmod ond ungetreow, þæs ne false man and reckless venomous and untrustworthy of-him not gymeð god likes God ‘God does not like false, venomous, and untrustworthy men.’ (Maxims I, 161–3, from Visser 61) (56) he or they that hafe infoormed the Lordes wele of me, I am behold to hem ‘To those who have told the lords good things about me I am obliged.’ (Paston Letters #513, from Visser 61)

If topics, especially topic subjects, as in (54), are unintegrated in the sentential structure in OE, something happens to make them more integrated in ME. Prodrop occurs in OE (see van Gelderen 2000 and Chapter 12 below) and hence, even if an XP situated to the left of the sentence is not repeated by a pronoun, the XP can still be a topic. However, most instances of pro-drop in OE and ME are not in the context of a left-peripheral XP, but in terms of an already known topic, mentioned in another clause, e.g. as in (57), and the reason for leaving the subject out is stylistic. In (57), I have provided a long quote from d’Ardenne (1977) to give a feel for the paratactic nature of the text: (57) Costentin walde efter ant warpen him þe onne. ah se wide him Constantin wanted after and drive him out but so wide him weox weorre on euche halue ant no meliche inalont became war on every side and especially in a country ylirie hatte þ [tear] he etstutte Illyria called that [there he stopped þa maxen ce iherde þis þ he wes of him siker ant of his then Maxence heard this that he was of him secure and in his cunne car les. war king of þ lont þe lei into rome as manner/coming careless became king of the land belong to Rome as duden meast alle þe oðre of þe weorlde. did most all the other of the world Bigon anan ase wed wulf to we orrin hali chirche ant dreaien began anon as mad wolf to persecute holy church and draw cristenemen þe lut þ ter weren alle to heaðendom heaðene as he wes

115

116


Christians the few that there were all to heathenism heathen as he was summe þurh mu chele Šeouen ant misliche meden summe þurh fearlac some by large gifts and diverse rewards some through fear ‘Constantin wanted to follow him and drive him out, but wars became so wide-spread, especially in Illyria, that he stopped there. When Maxence heard that he was secure and free of worries that Constantin would come, he made himself king of the country that was subject to Rome as was most of the rest of the world. He began at once, as a mad wolf, to persecute the holy church and to draw the few Christian men that there were to heathenism, heathen as he was, some through large gifts and rewards and some through threats.’ (Katherine, Bodley, pp. 17–18)

If most topics, as in (54) above, were the subjects of the sentences they are connected with, one would expect the loss of pro-drop to necessitate topic incorporation. This is not the case as (55) and (56) show, and therefore topics continue to exist. There are a number of ways non-subject topics can be incorporated, e.g. through V2, as in (46) above. In Late ME, another device appears, namely as for. When as for/to is introduced, the main clause still often has no subject, as in (58): (58) And as to the tytell that I have to the Lordship of Gressam, schal with in short tyme be knoweyn And as for the title that I have to the Lordship of Gressam, this shall become known within a short time. (Paston Letters #95, from Visser 10)

Visser says that this use “was not uncommon in the 15th and 16th centuries” (p. 10). The first instances where the subject is connected to the as for-DP appear in HC ME4 (1420–1500), e.g. Chaucer has many, as in (59), and so do the Paston Letters, as in (60). These do incorporate the topic, probably in Spec TopP: (59) But as for me, I seye that yvele it sit ‘But as for me, I see that it ill befits one.’ (Chaucer, Clerke’s Tale 143.460) (60) As for þe seyd John Heylisdon, he was a por man born ‘As for John Heylisdon, mentioned before, he was born poor.’ (Paston Letters, p. 135)

The earlier texts (e.g. HC ME3) show a very paratactic use of as for, even as late as Chaucer, as in (61), since the topic is not connected to an argument in the main clause:

Late Merge

(61) Now as for to speken of the firste coveitise, that is concupiscence, … I seye forasmuche as man is nat obeisant to God, that is his Lord, therefore is the flessh to him disobeisant … ‘Now speaking of the first strong desire, namely concupiscence, …, I say that if a man does not obey God, his lord, he flesh does not obey him.’ (Parson’s Tale, p. 297)

In the 15th Century Paston Letters, the ‘as for X, X…’ occurs very frequently and most of the 426 ‘as for’ have a referent in the main clause. The frequency seems high compared to say the ModE CSE with 27 instances (over 2 million words whereas the Paston Letters is 250,000 words). Thus, texts such as Chaucer and the Paston Letters show a real shift toward incorporation of material on the left periphery that plays a role in the actual sentence. This may be evidence for the introduction of a TopP.

5. Conclusion This chapter is devoted to seeing how the introduction of the finite complementizers for, till, and that leads to a split CP. I argue that once these prepositions start functioning as complementizers, due to Late Merge, they then merge even later, from Fin to Force. The incorporation of topics follows. This chapter differs from Chapter 3 in that there the emphasis is on the synchronic description of OE and ME whereas here it is on the outcome of Late Merge, i.e. introducing a split CP. The next chapter will show that basically every category can grammaticalize to a C. This is also due to Late Merge.

117

Chapter 6

More Late Merge Heads to higher Heads and Specs to higher Specs

As argued in previous chapters, it is more economical for a lexical item to merge late than for it to merge early and then have to move. This Late Merge Principle accounts for the fact that, in grammaticalization, heads move to positions that are structurally higher. Lexical items that have more grammatical meaning do not need to be merged in the early stages and can wait. In Chapter 5, I have shown that prepositions such as for and till become complementizers, but also that within the complementizer category, one grammatical category changes to another, higher in the structure. Other such cases are demonstratives that become complementizers (often by changing from demonstrative in Spec position to Head D). In this chapter, I provide examples of each of the logically possible origins for the C: the four lexical categories, N, V, A, and P, and the three functional/grammatical ones, C, D, and I. (It isn’t relevant for my purposes whether prepositions are lexical or functional, as most recently examined in Baker 2003b). The changes discussed here are from head to head, and therefore not relevant to the other Economy Principle that prefers heads over specifiers. The literature on lexical heads becoming functional ones is enormous, unlike the one on Spec to Head, so I have selected a few instances of verbs and nouns that become complementizers, and prepositions that go to complementizers, and will discuss these only as they are relevant to clausal grammaticalization and the Late Merge Principle. Many of my examples are based on H&K (2002), Heine & Traugott (1991), and Traugott & Hopper (1993). I will first discuss the lexical categories N, V, A, P (Section 1) as they become C since that is still the focus at the moment. In a later chapter, I discuss the same categories as they become I (M, T, and ASP to be more precise). I also examine the grammatical categories (Section 2) as they become C. In Section 3, I look at some Spec to Spec changes, also explained by Late Merge.


1.

From lexical to grammatical head

1.1 N to P to C Nouns form the basis of many prepositions, e.g. English via < Latin via ‘road’, Italian senza ‘without’ < Latin absentia, French chez ‘with’ < Latin casa (cf. Vincent 1999), English instead < ‘place’ (see Schwenter & Traugott 1995), baka in Sranan < ‘back’ (see Plag 1998), tp ‘on’ < ‘head’ in Egyptian (Gardiner 1957 [1988]: 130) and Ewe, as in (1): (1) e-fe ta » e-ta he-of head ‘his head’ he-on ‘on him’ (from Heine & Reh 1984: 257)

H&K provide numerous other examples, e.g. ‘ear’ becoming a locative preposition in Tzotzil, a language spoken in Mexico, and in Finnish, ‘earth’ changing to the A and P ‘down’, ‘shoulder’ becoming ‘up’, and ‘heart’ becoming ‘in’ (H&K 2002: 271; 171). Not all of these Ps develop into Cs though. Based on unpublished work by Giuseppe Longobardi, Lightfoot (1999: 197 ff.) shows that casa undergoes movement from N to D in Latin and early French (chies at this point) and that, once it is in D, it can be incorporated as a P. This clearly is a case of Late Merge then. Hopper & Traugott (1993: 107) mention the use of beside as a P. Using the noun ‘side’ as a P is quite wide-spread across languages (cf. also H&K 2002: 271–2). According to the OED, beside(s) is two words in OE and only after 1200 does it become one word, as in (2) and (3). So, the change is not from N>P, but from PP to P, so this qualifies as a Phrase to Head change. Layamon is quoted as the first to use bisiden both in the function of preposition and of adverb: (2) ða com on angel of heuene to hem, and stod bisides hem ‘Then came an angel from heaven to them and stood besides them.’ (c1200 Trinity College Homilies 31, from the OED) (3) heom biforen wende. Brennes bisides. mid his folke him before went Brennes besides with his people ‘[Belin] went before them, Brennes besides them, with his people.’ (Layamon, Caligula 2583–4)

Probably depending on when the combination is fully reanalyzed as P, an additional of is necessary, e.g. ahead (of), aside (of) are both later than beside(s). Most of the nouns in English that become prepositional are part of a PP and it

More Late Merge

is the entire PP that reanalyzes as P. As I mentioned, this is different from the change of N to P directly. Similar African examples abound, as documented e.g. in Heine & Reh (1984). For instance, the relational noun ‘top’ with a preposition ‘of ’ is reanalyzed as a preposition in (4) from Swahili: (4) juu ya mlima >>> juu ya mlima top of hill on hill (from Heine & Reh 1984: 101)

Complex prepositions such as beside, on top of, and ahead of arise from a reanalysis of a [P+N+P] as P and those in (4) arise from a reanalysis of [N+P] as P. These are neither cases of Spec to Head nor of one Head becoming another. I will therefore only look at cases that originate more directly from heads. These are in fact difficult to come by for the N to P/C change. Many of the prepositions that originate in nouns in their turn become complementizers. There are also nouns that change to complementizers without first being prepositions. Japanese koto ‘that’ is said to derive from ‘thing’, Korean kes from ‘thing’, and the Hebrew complementizer ’asher/she is derived from ’athar ‘place’ (cf. Givón 1991: 259). Other instances occur in the African languages Ani, Baka, Kikuyu, and Lingala where ‘cause’ becomes a causal complementizer; in Nama and Ik, ‘matter’ becomes a regular C; and in Nama and Susu, ‘matter’ becomes a purposive C. The latter is also true for the non-related language Thai (H&K 210–212). According to the OED, both and and till derive from nouns meaning ‘end’. However, within OE, there is no evidence that they are still used as nouns with that meaning. Regarding and, there are initially two lexical items, the preposition and/ond and the conjunction end, the latter originally a locative form of the noun. By OE, there are very few ends left (in the DOE texts, there are 18 conjunctions end as against 16,369 instances of and), and none are ambiguous between the noun ende and the conjunction. I’ll quote the full etymology for and from the OED, since it shows some of the complexities and provides a nice grammaticalization account: OE. shows two forms: (1) and, ond (:-OTeut. *anda) OFris. anda, and, OS. ant, OHG. ant, Goth. anda-, and, ON. and-, prep. ‘against, fronting’; (2) end (:OTeut. *andi) OFris. ande, and, an, end, en, OHG. anti, enti, inti, unti, endi, indi, unde, MHG. unde, und, unt, mod.G. und, Du. en, conj.; cf. L. ante before, Gr. ντí against, Skr. (Vedic) antí over against, locative of antá ‘end, boundary, vicinity’, hence ‘on the frontier of, abutting on, fronting, facing.’ From the idea of opposition, juxtaposition, or antithesis, the word was used in the Teut.

121


langs. to express the mutual relation of notions and propositions. The general Teut. form of the conj. is *andi, of the prep. *and(a); in OE., with the early loss of end, and (ond) remained for both, but soon became obs. as prep., exc. in a few derivatives: … The levelling of OE. ond, end, under the single form and was no doubt helped by the fact that the conj. is nearly always unemphatic, so that the vowel is obscured and tends to sink to a mere voice glide ((6)nd). From the same cause the final d has from early times been often dropped, as now universally in the dialects, and commonly in familiar speech: bread and butter = bread ‘n butter.

If the origin of the conjunction is to be found in the locative of the noun ‘end’ (which e.g. in Sanskrit, since it is a masculine -ah noun, would be ante), this would be a good example of N to P. Conjunctions are notoriously ‘unstable’ and the IE ones, such as que in Latin and ca in Sanskrit, do not survive into Germanic. If and is a preposition in OE derived from a noun ‘end’, its prepositional meaning ‘before, in the presence of ’ with a dative case, as in (5), or ‘by the side of, in addition to’, as in (6), makes sense: (5) Hæfdon gleam and dream and heora ordfruman ‘had joy and mirth before their creator.’ (anno 1000 Cædmon 13) (6) Emb eahta niht and feowerum ‘after eight nights and four.’ (anno 1000 Menol. Grein, 211)

The use of and as conjunction is very frequent, and the change from N to C is expected under Late Merge. The OED’s first examples of coordinated sentences are (7) and (8): (7) Her Cerdic forþferde, 7 Cynric his sunu ricsode in-this-year Cerdic went-forth and Cynric his son ruled ‘In this year, Cerdic died and Cynric his son ruled.’ (Anglo Saxon Chronicle A, anno 534) (8) Aris ænd nim þæt cild, and his moder ‘Rise up and take the child and his mother.’ (c1000 Ags. Gospels Matthew ii. 20)

So, even though and changes from N to P to C, I have not found direct evidence of the N to P stage in OE. I will come back to and in Section 2.1. Till is very similar to and. It (probably) derives from a noun meaning ‘end’ (cf. German Ziel; Dutch doel) and becomes a P as well as a C. In OE, the use of til(l) as ‘end’ no longer occurs, and as a preposition it assigns dative, whereas in e.g. Old Norse (Gordon 1927), it assigns genitive, perhaps still reminiscent of its

More Late Merge

nominal use. It starts to be used as a conjunction in ME, and as described in the previous chapter, this occurs at the time when the split CP is triggered, as in (9): (9) Till þatt teŠŠ wærenn alde till that they were old ‘Until they were old.’ (Ormulum 126, from the OED)

So, neither till nor and provides direct evidence of a sentence where a nominal use is extendable to a prepositional use, but there is textual evidence of a change from preposition to complementizer. All of these are compatible with Late Merge. 1.2 V to C Verbs typically grammaticalize into auxiliaries, but there are some verbs that become complementizers. I am adding this section to show that all logical possibilities occur but am not adding any new data. Hence, I’ll keep this section very short. The Egyptian complementizer dd derives from a verb meaning ‘say’, and many Niger-Congo languages (e.g. Ga, Ewe, Gokana, and Igbo) derive it from ‘say’ (see Lord 1993). Lord (1993: Chapter 7) discusses sΕ in Twi as still having a great number of meanings: as verb ‘to resemble’, as comparative particle, as object marker, and as complementizer. Saramaccan, a creole spoken in Suriname that has preserved more African features (see Holm 1989: 440), has táa ‘say’ and mbéi ‘make’ as complementizers, according to Veenstra (1996: Chapter 6). Tibeto-Burman languages (see e.g. Matisoff 1991) do so as well. Saxena (1988:375) argues that the use in Tibeto-Burman is “due to Indic influence”. H&K (2002:261; 269) provide other examples from Baka, Lezgian, and Ewe. In English, new quotatives are arising, as in (10): (10) And he goes “…”, and she is all “…”

These are full sentences though and not just verbs. 1.3 From A to C In Chapters 3 and 4, the case of adverbs such as oþ and þeah is mentioned, changing from adverbs to complementizers as they combine with þe/þæt. In ModE, there are similar cases of adverbs that connect two sentences, namely yet, however, nevertheless, therefore, though, and so, as in (11): (11) “you are anxious for a compliment, so I will tell you that you have improved her.” (Jane Austen, Emma, I, Chapter 8)

123


Some grammarians see these as complementizers; others see them as adverbs. With the punctuation as in (11), the complementizer scenario is more obvious since so connects the two sentences. However, so sometimes appears at the beginning of a sentence, in which case it could be an adverb expressing the reason why something was done. This ambiguity has always existed, as (12) and (13), from OE show; (14) is an early pro-form: (12) sue he uuundra gihuaes ‘As he wonders did.’ (Caedmon’s Hymn, Northumbrian version) (13) Swa he selfa bæd ‘As he himself asked.’ (Beowulf 29) (14) ðær hie meahton swa where they might so ‘where they might (do) so.’ (Beowulf 797)

Fourquet (as mentioned in Rynell 1952: 21) distinguishes them by means of the word order: C if verb-final, A if V2 (see also Mitchell II: 291ff. for the difficulty in distinguishing them). If adverbs such as these have in fact become coordinators or subordinators, this change is not unexpected under the Late Merge principle. In Arabic, the adverb inna ‘indeed’ is related to the complementizers ’anna/ ’inna (cf. Caspari 1876: 231). A following DP is assigned accusative (the complementizer un is an exception since it triggers (subjunctive) verb-movement to its immediate right). See also Gardiner (1957 [1988]: 403). 1.4 P to C Emonds (1985) argues that P and C are the same category, and it is true that many (temporal) Ps can be C in English: after, before, since. I will assume they are different and that only some Ps become Cs. Note that, as quoted in e.g. Pinkster (1972: 148; see also Weber 1934: 94), some grammarians think there are no original Ps; they are all adverbs. This would make the change ultimately one of A to P to C. For has been discussed at length, and till has as well. The use of P seems to antedate that of C. This is true with the more recent like and also with the familiar infinitival marker to (see H&K 2002: 37 for examples in other languages of ‘to’ becoming a C). I’ll discuss like here. Especially since the 1980s, like has expanded its uses tremendously (e.g. Underhill 1988). It is sometimes claimed that it is the most frequent word in the speech of certain groups of speakers. Prescriptive grammarians are not too

More Late Merge

pleased with this development, but tend to focus on the use of like as a complementizer. The prescriptive rule is that like is a preposition and not a complementizer. This means that it can introduce a DP, as in (15), but not a clause, as in (16) to (22). Instead of like, as is used to introduce a sentence.1 Swan (1980: 73) is relatively low-key and says that “[i]n informal American English, like is very often used as a conjunction instead of as”: (15) Certainly, he is not like Mr. Knightley. (Jane Austen, Emma, I, Chapter 4) (16) Shop like you mean it. (advertisement) (17) Eat it as it is, right out of the pouch. Enjoy it like today’s astronauts do. (on a package of space food) (18) “People have never been down and out like they are today”, said Angela Alioto, a candidate for Mayor… (The New York Times, 26 August 1991) (19) “We just felt like this year we needed to get away for a while.” (The New York Times, 10 August 1991) (20) I felt like I could tell you anything. Now I don’t feel like I can anymore. (quoted in Tannen’s That’s not what I meant) (21) Winston tastes good like a cigarette should. What do you want: Good grammar or good taste? (an ad in the 1960s) (22) She forgot all about the library like she told her old man now. (Beach Boys’ song)

Except as complementizer and preposition, like is often used to mark direct speech, as in (23), focus, as in (24), or to soften a request or demand, as in (25). These uses are not accepted in formal speech either, even though some are old, as (26) and (27) show, quoted in the OED: (23) … So the other girl goes like: ‘Getting an autograph is like, be brave and ask for it’. So I got it. I just went up to him and he like. ‘O.K … (24) I couldn’t get to class because, well, like I had this accident on the freeway. (25) Tiffany, you, like, still owe me that $10. (26) Šon man is lyke out of his mynd. (Dunbar Poems, xix, 19) (27) all looking on, and like atonisht staring. (Spenser, Fairie Queen, iv, x, 56)

So, due to the Late Merge Principle, like in (16) to (22) is on its way to becoming a complementizer despite enormous prescriptive pressure.

125


2. From grammatical to grammatical head 2.1 From C to C In the previous chapter, I have shown how Late Merge makes a low C (Fin) change into a high C (Force). For most languages, we lack the information whether a C is low or high in the tree structure. I will therefore not consider those changes any further here. I will instead consider changes involving coordinating and subordinating conjunctions. In Chapter 1, I mention that the paths of clausal grammaticalization may go in two directions, a subordinating and a coordinating one. In this section, I will look at cases where a coordinator becomes a subordinator and back again. This makes sense if the two systems are parallel to each other and not one more basic than the other. Harris & Campbell (1995: 290) and H&K (2002: 43–4) give numerous examples from a variety of languages of the change from coordinator to subordinator. The change is well-known and I will therefore not provide further examples here. I will examine the opposite change, from subordinator to coordinator. If both coordinators and subordinators are heads of functional projections (the CP and Coordinator Phrase respectively) that are hierarchically similar, there is no Economy of Head over Spec or of higher Head over lower Head involved. If that indeterminacy exists, one expects a change in either direction and this is indeed the case as I’ll now show, first with an example from K. Hale (1983b) and then with examples from OE. K. Hale (1983b: 302) argues that the O’odham (same subject) conjunction c, as in (28), once was a subordinator: (28) Pi:wlo ’o ’am ’i’it hegai sa’i Pete aux there scoop that trash c gd hu to’aw heg ’eda walin and there away put that in barrel ‘Peter is scooping up that trash and putting it over there in that barrel.’ (from Hale p. 303; there is another proximate conjunction when the first clause is perfective)

O’odham is spoken in Arizona and in related languages such as Hopi, also spoken in Arizona, the subordinating form is q. On the basis of these and many other forms, Hale reconstructs Proto Uto-Aztecan *-ku ~ -ko as the origin for the coordinator in (28).

More Late Merge 127

In English, it is unclear in e.g. (29) if VPs are coordinated or if clauses are. When the subject is expressed, as in (30), it has to be clausal coordination: (29) John went hiking and saw lots of dogs. (30) John went hiking and Mary saw lots of dogs.

Ellipsis of the subject, as in (29), is a form of clause union. This is not so visible in English, but in many languages, the coordinator is different if the same subject is involved, or if a different subject is. For instance, in O’odham, the same subject coordinator is c, as in (28), but the switch-subject coordinator is ku, and in the Niger-Congo language Nawdm (Watters 2000: 218), there are different coordinators depending on same subject/different subject as well. Apart from O’odham, there are other examples of a subordinator becoming a coordinator. For instance, in OE, and does not always trigger V2, as in (31), possibly indicating that the coordinator is in the C position, normally reserved for complementizers: (31) 7 Hæstenes wif 7 his suna twegen mon brohte to ðæm cyninge 7 and H’s wife and his two sons man brought to the king and he hi him eft ageaf he them him back gave (Anglo Saxon Chronicle A anno 894, Thorpe 1861: 168, from Stockwell & Minkova 1990: 503)

Does this mean and is a subordinator? Mitchell (I: 694–710) seems to suggest so but Kiparsky (1995: 147ff.) argues that it isn’t. Stockwell & Minkova (1990) review the word order data of Bacquet (1962), Denison (1986), and Bean (1983), add their own, and say that, in Early OE, “conjunct main clauses lean toward verb-final order about one-third of the time, conjunct VPs almost half the time” (p. 508). I won’t go into this word order debate but, if and is at some point a subordinator, it is not so unexpected that it would become a coordinator again, given what I claim about parallel structures in Chapter 1. 2.2 D to C The English demonstrative that undergoes a change from specifier to head. I have been unable to find a case of a D-head becoming a C, only demonstratives in Spec CP become Cs. H&K (2002: 115) mention the case of the African languages !Xun and Sango and the English-based creole from Surinam, Saramaccan, where a D becomes a subordinator. However, it is not clear from the


data that this was a direct change and not via the specifier. For instance, in Saramaccan, the demonstrative form is dísi ‘this’ whereas the subordinator is the reduced form di ‘when’. Simpson & Wu (2002; and earlier work by these authors) describe the Chinese change of de from a demonstrative to a relative clause marker. They mention Lakota and Diegueño as other languages with the same phenomenon. So, the change of D to C may occur, but it is hard to know from the literature whether or not there is an intermediate stage in which the demonstrative is a phrase in the specifier position. 2.3 I to C The ME changes involving to from P to I have been discussed many times. Initially to is a preposition of location, and subsequently starts to mark dative Case (see van Gelderen 1993; 1998) and infinitival verbs (which are originally deverbal nouns, see e.g. Lightfoot 1979; Disterheft 1980). In OE, the infinitival to (like the corresponding words in German and Dutch) is never split from its verb, but, as is well-known, the split infinitive is introduced in the 14th century in English. In the days before the split IP into TP and AGRP, to is assigned to I, in complementary distribution with modal auxiliaries but not with aspectual ones. In Chapter 11 below, I will argue that to goes from P to ASP (by Shakespearean English) and then to M. In this section, I’d like to show briefly that it moves to C in ModE, but will return to this in Chapter 11. I’ll use I rather than the expanded I here since the emphasis is on the C. The evidence that to is in a position higher than I is: (a) to not occurs in addition to not to, as in (32ab), (b) to not to sequences, as in (33), occur frequently and are judged as reasonably grammatical by native speakers, and (c) complementary distribution between a clear C and to not, as in (34): (32) a.

It would be unrealistic to not expect to pay higher royalties. (BNC-CSS 245) b. It would be unrealistic not to show them to be human. (BNC-CBF 14312)

(33) — as a request to not to — (CSE-WH97A) (34) if you as a parent choose for your child not to participate, you may opt out. (CSE-COMR6B97)

The problem for the ‘merge over move’ principle is to explain why to would not always be merged in the higher position. Rather than see it as a counterexample,

More Late Merge 129

I’ll provide two possible reasons for why the lower position is preferred. One is that the rule against splitting infinitives is still very strong in standard English (see Fowler 1926 [1950]: 558; Quirk & Greenbaum 1973: 312). In Chapter 11 (Section 2.3), I provide some corpora data which show that informal corpora split infinitives more often, and American corpora more than British, indicating it is indeed a prescriptive rule. The prescriptive rule favors the lower position. The second reason is that typically not is a head and that in order for to to move across, not is forced to be a specifier. Even though this goes against the Spec to Head Principle, it is still possible in ModE for not to be a Spec since nothing else is in the specifier position (due to the double negation prohibition). However, that makes the move infrequent enough not to trigger Late Merge. So, two Economy Principles interact, and one ‘bleeds’ the other. In ModE, to occurs in either I or C. Under Late Merge, it is predicted in C but other principles stop it. In Chapter 11, I’ll show the interaction with other principles more clearly by comparing the behavior of non-negative adverbs with that of not. The former result in to moving to C more often.

3. From Spec to Spec 3.1 Relatives In Chapter 4 above, I examine relatives and argue that phrases in the specifier position change into heads. The sequence of events for relatives is: (a) that is in Spec and þe in the head of CP, (b) that becomes the head, (c) the specifier of CP is filled again by a wh-element that moves there. There is, however, a short period from 1400 to 1600 where wh-pronouns are base generated (i.e. merged) in Spec of CP, as in (35) and (36),2 with a resumptive pronoun in argument position: (35) þis is he | Which þat myn uncle swerith he moot be deed this is he who that my uncle swears he must be dead ‘This is he, who my uncle swears must be dead.’ (Chaucer, Troilus 498.653–4, from Visser p. 58) (36) I tell my sorrowes bootles to the stones. Who though they cannot answere my distresse, Yet in some sort they are better than the Tribunes ‘I tell my sorrows without success to the stones, who though they can’t answer my distress, are still in some way better than the Tribunes.’ (Shakespeare, Titus, III, i, 37, also from Visser, p. 60 but Quoted as in the First Folio)


Constructions (35) and (36) show that the wh-element is base generated (merged) in the Spec of CP, and this is expected given Late Merge. If a wh-element merges inside the VP and then moves to Spec CP, this is less economical. The reason that this stage is short and that constructions such as (35) disappear is that they interact with the change of specifiers towards heads. There is an alternative analysis to resumptive pronouns, namely Boeckx (2003), who argues that resumptive pronouns and their relative clauses are merged together and that the clause moves up by itself. Boeckx (2003: 46ff.) crucially shows that this is possible only when the complementizer does not agree with the resumptive pronoun, as with a non-inflecting complementizer. In sentences (35) and (36), the complementizer is a wh-element, which does agree with the pronoun, certainly at that time. Hence, the account that Boeckx offers is very elegant for languages with uninflected complementizers, but is not applicable to the stage of English represented in (35) and (36). Hence, I assume that the wh-form in (35) and (36) is base generated in the Spec of CP, in accordance with Late Merge. 3.2 Adverbs: From VP to CP Another Spec to Spec change is that from a VP-adverb such as actually to an IP or CP-adverb (cf. Traugott). Unlike Late Merge in (35), which is triggered by the relative needing to move anyway, this is not the case with VP-adverbs. The reason is that adverbs are added late to the derivation. Swan (1980: 296–7) mentions that there are two uses of an adverb such as hopefully: one is full of hope, as in (37), and the other use, as in (38), “shows the speaker’s attitude”, and means it is hoped. According to Swan, “[s]ome people consider the second use ‘incorrect’: (37) She sat there hopefully for someone to stop by. (38) Hopefully, hunger will cease to be a problem.

It is not clear why hopefully should have attracted all this attention. There are several other adverbs like it, e.g. those in (39) to (42): (39) Naturally, I’d like you to stay with us for a few days. (40) Amazingly, he arrived on time. (41) Fortunately, the bus wasn’t late. (42) Funnily enough, I’d been thinking about that.

More Late Merge

In (39) to (42), the adverbs all express the speaker’s attitude and this is a legitimate use of an adverb; they do not all have to modify the VP. The change from VP-adverb to CP-adverb is compatible with Late Merge. It may be triggered because adverbs are added late to the construction.

4. Conclusion In this chapter, I provide systematic examples of N, V, A, and P changing to C, and D, C, and I changing to C. The same will be done in Chapter 11 for the changes to I, even though there is less variety there. All these shifts follow the Late Merge condition. When discussing the change from Spec to Head in Chapter 4, it was shown that the change is sometimes hindered by an external factor. The change of to from P to C is also hindered by the ban against split infinitives, as well as the interaction with the Head Preference Principle. Two changes of specifiers becoming specifiers (of higher functional projections) follow the same condition to undergo merge as late as possible.

131

Part III

Chapter 7

The IP, VP-shell, and their layers

In this chapter, I will provide some background on the structure of the IP and VP. There are ways to distinguish CP from IP and IP from VP. The CP is relatively independent, which is expected if it is a phase as in Chomsky (2001a). As seen in Chapter 3, the character of the CP depends on whether it is split or not. The IP represents the grammatical layer of agreement, tense, mood, and aspect. Its characteristics depend on whether all of these features are present. If they are, the IP will be more resistant to e.g. clitic climbing and control. The transparency of the ‘clause’ to clitics, control and distance to to, and reflexives will shed light on whether the ‘clause’ is an IP or VP and on grammaticalization. The VP contains the lexical arguments that are merged first, as well as what is sometimes called ‘Aktionsart’ or inner aspect, and which I’ll refer to as E(xtent) Phrase. In the first section, I review some of the categories that have been seen as making up the ‘middle’ of the sentence, i.e. the area between the (expanded) CP and the VP. Cinque (1999), already discussed in connection with the CP, has an elaborate number of functional categories splitting the IP as well. Just as with CPs, Cinque’s system for IPs is too elaborate. Tense, Mood, and Aspect suffice for ModE, as will be argued in the second section (see also Chapters 8 to 10). I will also review the reasons for assuming a VP-shell. This shell, consisting of vP and VP, is universally present but some languages may add a ‘low’ or ‘inner’ Aspect Phrase that is part of the VP layer (see (8) below). In the third section, some criteria are provided on how to distinguish CPs from IPs from VPs. Two of these are mentioned in Chapter 3 (raising and ECM). The possibility to have raising and clitic climbing generally distinguishes IPs from CPs. Binding domains are shown not to distinguish CP and IP. Control and distance to to are important to differentiate VPs from IPs, but clitic climbing is not. In the fourth section, IP and VP-shell grammaticalization will be discussed together since they are often hard to separate. When an IP complement is reduced, this means that the matrix verb becomes an auxiliary and the construction changes from two clauses (bi-clausal) to one clause (or monoclausal). Chapters 8 and 9 provide examples of grammaticalization where two


clauses are reduced to one. The main difference between IP/VP and CP grammaticalization is (a) that the element that grammaticalizes is either V or P (or A) rather than almost all categories in the case of CP, as shown in Chapter 6, and (b) that only Late Merge is relevant with IPs, whereas with CPs both Economy Principles are.

1.

The structure of IP and the VP-shell

1.1 IP In Chomsky (1986), as mentioned in connection with the CP, functional categories are given structural representations similar to lexical categories. The auxiliary comes to be seen as I(nflection), projecting to an Inflection Phrase (IP), as in (1): (1)

IP I′ I

VP

Subsequently, Pollock (1989) and Chomsky (1989) expand the IP to accommodate both tense and agreement by using a separate TP and AGRP. Since there is no empirical evidence for this split into T and AGR in English, I won’t use it. In later work, Chomsky casts doubt on the existence of AGR. Other categories are added, accomodating tense, mood, and aspect. For instance, ASP(ect) has become a frequently used functional category in Tenny (1987), Speas (1990), Marácz (1990), van Gelderen (1993), Travis (2000), to mention but a few. The structure I assume for the expanded English IP is as in (2):


(2)

TP T′ T

MP M′ M

ASPP ASP′ ASP

vP/VP

Some authors argue that not all of these categories are present in every language, and I will argue the same. Chomsky (1995; 2001ab) contends that crosslinguistic variation is due to the lexicon. This means that the presence of certain items in the lexicon determines the structural representation, e.g. if auxiliaries are in the lexicon, the language will have an IP. The reasons why I enters that lexicon are worth exploring though. 1.2 The VP-shell In addition to expanding the IP, the VP has come to be regarded as a VP-shell with a light verb v and a (heavy) V. So a relatively traditional late 1990s tree looks like (3), ignoring the CP for the moment and expanding IP only minimally. Note that the subject starts out in the specifier of the vP:

137

138


(3)

IP I′ I

ASPP ASP′ ASP

vP v′ v

VP V′

Shei

may

have

ti

made

balls

V

PP

roll

down the stairs

In (3), make and roll are treated as belonging to one clause. The original reason (e.g. Larson 1988) for having a vP as well as a VP, together called the VP-shell, is to accommodate clauses with three arguments, such as double object verbs. This structure predicts that verbs have three arguments or less and that is what one finds cross-linguistically. Many verbs can also be thought of as consisting of a light verb and a heavy one (Radford 1997), e.g. roll in She rolled it there can be thought of as make roll, as in (4), and this is predicted by the VP-shell: vP

(4) She

v′ v (made)

VP it

V′ V

PP

roll

there

The IP, VP-shell, and their layers 139

If roll moves from V to v, a separate (pleonastic) form of make is not necessary. A v-verb (or light verb) like make is grammaticalized from a ‘real’ verb meaning ‘produce’. As I show in later chapters, verbs in v can grammaticalize further by starting to occupy higher functional positions. Typically the v expresses the process and its specifier is occupied by a DP with an Agent/Causer theta-role. VP expresses the result and its specifier is often a DP with a Theme theta-role. The difference between transitives, unaccusatives, and unergatives can be expressed through this structure as well. Following the spirit of Hale & Keyser (1993; 2002), unergatives, such as I laughed/smiled/coughed are formed by having the object N incorporate to ‘become’ the V; causatives such as I shelved the books; I thinned the paint have similar derivations, as in (5), where the A thin moves to V and to v, if the light v made is not present: vP

(5)

v′ v

VP V′ V

I

(made)

paint

AP thin

Hale & Keyser (2002: Chapter 7) tentatively connect aspect to this structure. Typically, derivations that start out with non-verbal categories will be stative. This accounts nicely for the ambiguity of experiencer verbs such as respect and fear. When they are projected as Ns, as in (6), they are stative, whereas if an additional verb is projected, as in (7), a non-stative reading occurs (see Hale & Keyser 2002: 208–213): (6) They respect the truth. (stative) (7) I gave my respect to Mary. (non-stative; from Hale & Keyser 2002: 210)

Cross-linguistically, there may be differences as to whether both v and V must be overt. Lin (2001) makes the case that in Modern Chinese, the light verb must be present, as I indicate in (1) of Chapter 1, unlike in English and Old Chinese. Many linguists assume more structure than just VP and vP. Other categories and phrases such as VoiceP, and Tr(ansitive)P occur in Jelinek (1997; 1998).


There is some evidence, e.g. from phrasal verbs in English, that there is an (inner) aspectual layer, as in (8), between the v and V, to account for (9) and (10), where the object moves to the specifier when it is definite. I will refer to this inner aspect as EP, Extent Phrase, for reasons I’ll explain below: (8)

vP v′ v

EP Ep′ E

Vp V′

turn

itj

offi tj

V

A

(be)

ti

(9) I turned it/the light off. (10) I turned off the light.

Inner aspect would account for the relationship that exists in many languages between movement, a certain Case, a certain aspect, and definiteness. I have called it E(xtent) Phrase, since it expresses the extent to which the action has been done and how much the object has been affected. Meinunger (1995:92ff.) lists some German sentences relevant in this respect which I list in Dutch. The perfect participle connected to hebben ‘have’ in Dutch (and German) is ambiguous between perfective and imperfective, for instance, in (11), where the object de bijbel is in the VP, either the Bible was read completely or parts of it were. In (12), however, where the object moves out of the VP (considering the adverb to indicate the left-boundary of the VP), the Bible has been read completely a number of times. The reason for the difference may be that a DP that moves to a higher position as in (12) moves to one (i.e. Spec EP) where terminative aspect is checked and must be complete: (11) omdat ik vaak de bijbel gelezen heb because I often the Bible read have ‘because I’ve read the Bible often.’


(12) omdat ik de bijbel vaak gelezen heb because I the Bible often read have

There is also an association between specificity, movement (cf. Diesing 1992), perfectivity, and a special kind of Case (see Abraham 1996ab; Kiparsky 1996; Leiss 2000). For instance, (13) sounds odd with the perfectivizing affix up and the unspecific object: (13) ?I ate up apples.

Once an EP has been selected (through up), the DP must be definite and must move out of the VP. The relationship between Case and movement is shown in (11) where the VP-internal DP only gets an (inherent) partitive Case as opposed to the VP-external DP in (12). In discourse terms, the specific DP in (12) moves out of the VP because it is the topic rather than the focus. Following Cinque (1993), who argues that the focus position is the most deeply embedded, the specific DP moves out of the VP because it is the topic. Moving through inner ASP might also account for the aspectual differences between fear/have fear discussed by Hale & Keyser (2002), mentioned above, and pairs such as blind/make blind; clean/make clean where the first of the pair is a permanent state. I see the inner aspect as connected to the Aktionsart of the verb and in Chapters 8 and 9, I will argue that there occurs a reanalysis in English from inner to outer aspect: aspectual affixes are replaced by auxiliaries and -ing. In Chapter 10, I will also suggest that ModE may be developing a new inner aspect, due to the increased frequency of resultative particles, as in (14), and particle verbs, as in (9) above: (14) They received it in.

Travis (2000) posits a structure similar to (8) for Tagalog and Malagasy and so does Sybesma (1999: 76) for Chinese verbal le. Since in in (14) is more of a functional element, one might expect Late Merge to make it merge later and, in fact, the recent development of verbs as in (15) seems to be suggestive of that: (15) Obviously we wish we wouldn’t have to outlay the dollars. (East Valley Tribune, 21 April 2003)

Other verbs are downplay, input, upgrade. These may point to having the aspect higher in (8), i.e. in ASP rather than A. I’ll come back to ASP in connection to modals, perception verbs, and particles. I will argue that originally the particle is in the inner aspect and then is reanalyzed as outer aspect. In Chapter 11, Section 4, I’ll compare Dutch and OE, and show how Late Merge comes into play.

141


1.3 The cartographic approach In (16), I list some of Cinque’s (1999) categories that are relevant for the IP, with the adverbs listed underneath: (16) Modepist Tpast Tfut Moodirr Modnecc Modposs Modvol probably once then perhaps

Modability/perm possibly intentionally [not provided in C]

Cinque (1999: 79; 81; 106)

However, many of these adverbs never co-occur, as I will show in more detail in the next chapter. For instance, in the BNC, there are no instances of perhaps and possibly next to each other; in the Cobuild Corpus, there is one of the unexpected order possibly perhaps, but with a pause in between. Rather than having all of the categories of (16), I argue that it is better to see them as three clusters (TP, MP, and ASPP) that are sometimes expanded, just like the CP. Section 1 has outlined how the IP and VP are split up: IP typically can be seen as Tense, Mood, and Aspect and VP as vP, VP, and (at least) one other functional category which I have called E(xtent).

2. Cross-linguistic differences regarding the IP I’ll now discuss tense, mood, and aspect across languages and then relate these to the structure in (2) above. The three are connected, and sometimes occur in complementary distribution. Thus a language with a lot of grammaticalized aspectual markers may have fewer tense markers. However, accounting for the changes between the different stages is far from easy since the changes occur in non-predictable ways. Other major changes, e.g. the loss of morphological Case marking and the loss of adjectival inflection, occur in many languages, result in similar changes, and have a clear account in a number of frameworks. For tense, mood, and aspect, the picture is less clear. As an example of the complexity and lack of agreement in the literature, take older English. The shift from Old to Middle to Modern English involves major shifts in the tense, aspect, and modal system, as well as a decrease in morphological tense and mood marking and an increase in auxiliaries. In OE, the present and past tense are marked on the verb but the presence of relative tenses such as the present perfect is contested (see Brinton 1988). Mood is expressed through subjunctive and imperative inflection, but not (as much) through modal auxiliaries. Aspect is expressed through prefixes on the verb, but


some people argue there is an early decay of that system. Aspect is also possibly expressed by a forerunner of the -ing form, but that too is debated and the form is often seen as an adjective (e.g. Mossé 1938). There are a limited number of auxiliaries, mainly forms of have, be, become, but usually not more than one at a time, and it is contested if they are auxiliaries or main verbs. If the system in OE is not clear, this is also the case for ModE, see for instance the discussions on the present perfect in Jespersen (1931), Comrie (1976), Julien (2001); the discussions of modals in Thráinsson & Vikner (1995) and Wurmbrand (2001); and the many possibilities that a structure such as Cinque (1999) provides. One of the problems with tense, aspect, and mood is that the theoretical framework is still being developed: the representation of inner aspect or Aktionsart, the structural representation of tense and mood markers. Aspect is perhaps the most basic, and used first by children (e.g. the progressive in English); deontic modals are acquired before epistemic ones. Bybee (1985) provides an extensive typological overview, based on a sample of 50 languages. In her sample, tense typically shows a ±past tense distinction (present in 36%), a ±future distinction (in 44%), and a ±anterior (no percentages are given). Frequent aspectual contrasts are perfective/imperfective and habitual/continuous. Mood is very complex (Bybee 1985: 191 ff.) and includes imperative, subjunctive, as well as interrogative, epistemic modality, and evidential (among others). The order of the morphemes in her sample (Bybee 1985: 34–5) is MTA and hence is different from the TMA found in creoles and so often claimed to be more universal (see Bickerton 1984, and Bakker et al. 1995). Bickerton (1984; 1981) identifies the typical tense in creoles as ±anterior, the mood as ±irrealis, and the aspect as ±non-punctual. Returning to the structure of the IP, in this book, I’ll show that there are three positions, T, M, and ASP, in that order. As is well-known (see e.g. Akmajian et al. 1979), the I(nflection) position in ModE expresses tense: if the sentence is finite, ±past tense is present; if non-finite, there is no tense. Mood is communicated mainly through modal auxiliaries and through movement to C (e.g. Had I gone there, I would …; Did he go?). See also Frajzyngier (1995) for the view that complementizers are part of the modality system. Mood, tense, and finiteness are often connected: (a) if a modal is present, tense is not expressed separately, (b) the subjunctive mood makes use of the tense system, and (c) the (admittedly controversial) irrealis mood marker to is non-finite. Therefore, the evidence is that mood is expressed through modals and to in M but moves to T, but that C can be involved as well (as argued in Chapter 11).


Hence, there is no one-to-one relationship between the nodes in the tree and the actual categories. This is not unexpected, see work on Dutch and Swedish that argues tense is in C (Platzack 1986; 1998: 92, van Gelderen 1993; 1997a). This shows that tense and mood are expressed (or checked) but that features can be in different positions in different languages. In Turkish (Slobin & Aksu 1982), there are two past tense markers for direct and indirect evidence, again linking tense and mood. As for (grammatical) aspect, in some languages, tense and aspect are difficult to separate. In short, ModE often expresses M and T together, but M can also be seen as a C element (in Rizzi’s system, C is probably Force). Movement to check the relevant features explains, for instance, why an element merged as M needs to move to T (e.g. epistemic modals move from M to T, see Chapter 8) and why an element merged as T moves to C (to, see Chapter 11). Aspect is never situated as ‘high’ up in the tree and this explains the MTA or TMA sequence, found in so many languages.

3. IPs as reduced CPs; VPs as reduced IP For simplicity’s sake, I will ignore the v/V distinction in this section. I will first briefly look at the grammaticalization pattern from CP to IP to VP, as outlined in Chapter 1, and then try to distinguish CPs from IPs from VPs in terms of scrambling, control and distance to to, clitic raising, and binding. The latter end up determining whether a clause is an IP or VP. There is cross-linguistic variation; for instance, in certain languages IP is a barrier to binding but in others it is not. 3.1 From CP to IP to VP If a complementizer is absent between two independent verbs, the connection between the verbs becomes closer, especially if no subject (in Spec IP) and no I-element intervene between them. In (17), representing the change from (2c) to (2d) of Chapter 1, the complementizer is deleted in (17b) and the subject, the Spec IP, is ‘shared’ between the higher and lower clause in (17c): (17) a. You want CP[that IP[he goes]]. b. You want IP[him to go]. c. You want IP[PRO to go].


In (17b) and (17c), to occupies I(nflection), so the ‘middle’ is still present. Sentence (18) shows the change from (2d) to (2e), with the embedded IP deleted: (18) a. You want IP[to go]. b. You [wanna go].

The presence of a subject and an independent tense are important for defining a clause (see also Evers 1990). In the case of the loss of IP, as in (18b), the tense will be a dependent one, and the higher verb becomes an auxiliary. Whether or not verbs can lose the IP depends on the semantics of the verbs as well. Keeping the CP and IP separate is sometimes difficult since the subject and tense are relevant to both. In the next section, I’ll provide some diagnostics for IP-deletion, raising and ECM, already mentioned in Chapter 3, and clitic climbing, scrambling, and binding. It turns out that the classes of verbs that allow CP-deletion (e.g. believe does) in Chapter 3 do not also automatically allow IP-deletion (e.g. believe does not). 3.2 Characteristics of (CP and) IP-deletion In Section 1 of Chapter 3, I discussed some differences between the presence of a simple CP and that of a split CP. Here, I will do the same for the difference between IP and CP (in 3.2.1 and 3.2.2), and also for that between VP and IP (in 3.2.3 and 3.2.4). The focus is on what defines the clause. 3.2.1 DP-Raising and Scrambling: CPs versus IPs CPs can be moved in English but IPs can’t (cf. Section 1 of Chapter 3). IPs in turn are more transparent and allow DP-raising, as in (19), whereas CPs do not, as in (20): (19) Hei was believed IP[ti to be nice]. (20) *Hei was wanted CP[IP[ti to go]].

In Chapter 3, I have discussed complements to believe in terms of certain CPs reducing to IPs and then allowing raising and ECM. They are the strongest indications for having an IP over a CP. In German, scrambling is possible from certain clauses, as in (21), but not from others, in (22). One could attribute that difference to a difference in clause structure, with the CP absent in (21) but not in (22):

145


(21) weil die Mariai der Hans [ti zu besuchen] versprach because the-acc Mary the-nom Hans to visit promised ‘because Hans promised to visit Maria.’ (22) *weil die Mariai der Hans den Peter [ti zu besuchen] bat because the-acc Mary the-nom Hans the-acc Peter to visit asked ‘because Hans asked Peter to visit Maria.’ (from Hinterhölzl 1999: 1)

Wurmbrand (2001:317ff) provides a detailed overview of restructuring verbs in a number of languages. The main criteria she uses to distinguish restructuring from non-restructuring are scrambling and movement, similar to (21) and (22) above. Cross-linguistically (Wurmbrand 2001:343–5), modals, aspectuals, and motion verbs are the most likely to be restructuring. In some languages, ‘seem’, ‘appear’, ‘try’, ‘dare’, and ‘manage’ are frequent restructuring verbs. These verbs have semantic features that are grammaticalizable and therefore Late Merge can apply. 3.2.2 Binding: CPs and IPs One would expect binding domains to determine clause-hood. This turns out not to be the case. Lasnik & Saito’s (1992: 110) observation that (23) is grammatical points towards not even a finite CP being a binding domain: (23) John thinks CP[that himselfi Mary likes ti].

Typically, the presence of a complementizer and finiteness, as in (24), render a clause opaque for binding. Compare (24) with the complementizer-less grammatical infinitival complement in (25). This is not the case in (23), however: (24) *John wants that himself wins. (25) John wants himself to win.

Sentence (26) is a well-known example showing that an intervening subject, without a complementizer and tense, makes the clause opaque as well, so as also obvious from (23), a CP isn’t relevant: (26) *John wants Mary to vote for himself.

Nevertheless, binding is sensitive to the presence of just a complementizer, so that even a non-finite clause, as in (27), blocks himself from finding an antecedent in the higher clause, unlike in (28): (27) *He wanted for himself to win that contract. (28) He wanted himself to win that contract.


Once there is a complementizer, however, there is also a subject in (Modern) English. So, the subject seems more basic for binding and the binding domain is therefore weaker than an actual clause. The binding domain can be non-finite without a complementizer, as in (26), as long as it has a subject. Added to this problem for English is that, in other languages, full CPs do not constitute binding domains. In Japanese, it is not, as (29) shows, and the same is true in Chinese (30) and Korean (31), to name but a few: (29) John-waj [Bill-gai zibun-oi/j nikunde iru to] omotte iru John-top [Bill-nom self-acc hates is that thinks is ‘John thinks that Bill hates him/himself.’ (30) Zhangsanj renwei [Lisii hai le zijii/j] Zhangsan think [Lisi hurt asp self ‘Zhangsan thought that Lisi hurt himself.’ (from Ho 1999) (31) Minca-nunj [Yongho-kai caki-luli/j miweha-n-ta-ko] sayngkakha-n-ta Minca-top [Yongho-nom self-acc hate-pr-decl-c think-pr-decl ‘Minca thinks that Yongho hates her/himself.’ (from Choi 1999)

One of the ways to account for this is to define the clause in terms of having agreement which Japanese, Chinese, and Korean do not have. The reason for this is that they are not IP-oriented. So, what makes a clause opaque for binding is different in different languages. In English a subject can do so, even if the clause is a complementizerless infinitival. In some, if the clause is subjunctive, the reflexive can ‘look outside’. The Subset Principle (Manzini & Wexler) says that the minimal domain can be selected, and simplified, it looks like (a) to (d), with (a) the most restricted: (32) Minimal domain contains: a. a subject, or b. a tense, or c. an independent tense (i.e. not a subjunctive), or d. a root tense.

Japanese would set the parameter for zibun as (d), Catalan and Icelandic as (c), since subjunctives are transparent, English as (a) for himself.1 This means the binding domain cannot really be used to define a clause. Since the late 1980s, there have been attempts to explain cross-linguistic differences through the status of the anaphor and whether or not LF-movement applies. However, (32) is nicely descriptive and parallels (2) of Chapter 1.


Thus, differences between CP and IP seem to boil down to being able to move the CP but not the IP and whether or not the clause is transparent. Transparency can also indicate that the ‘clause’ is a VP, however. Control and distance to to distinguish between IP and VP, as I’ll now show. 3.2.3 Control and to: IPs versus VPs Landau (2000) divides infinitives into exhaustive and partial control, EC and PC respectively. PRO in the complement of verbs of exhaustive control must be exactly the same as the subject of the higher verb. In terms of cross-clausal grammaticalization, this means that EC verbs are more grammaticalized and that their complement is less clausal. For instance, intransitive gather is a verb that requires a plural subject, and therefore an EC verb such as manage cannot be its matrix verb, if the subject of manage is singular, as in (33). This is not the case with PC verbs since their PRO is free. It can refer back to the subject of prefer or to someone else and hence, (34) is grammatical: (33) *The chair managed PRO to gather at 6. (34) The chair preferred PRO to gather at 6. (both from Landau, p. 5)

Landau (2000: 6) divides infinitival complements into 7 classes, depending on the main verb, as in (35), but with EC and PC as the main division: (35) EC: implicative (manage), aspectual (begin), modal (need, able) PC: factive (hate, regret), propositional (claim, believe), desiderative (want, prefer), interrogative (wonder)

He notes, based on work by Stowell, that the PC verbs have an independent tense, whereas the EC ones do not. I will consider the EC-verbs with their complements as mono-clausal, but PC-ones as bi-clausal. A structure for EC-verbs would be as in (36), with manage or begin in ASP moving to M to pick up to and then move to T:


(36)

TP T¢ T

MP M¢ M

ASPP ASP¢

to ASP

vP

Ï manage ¸ Ì begin ˝ go Ó ˛

...

Looking in the BNC, one finds independent evidence for this structure in that the EC-verbs are mostly immediately followed by to. This is indicated in the column under to in Table 7.1. The few where an adverb or (pro)noun intervenes, indicated in the column under [X to], are (still) bi-clausal. As can be seen from the data, need is less often immediately followed by to and is therefore a PC-verb. It is the only one where a DP (subject of the lower V) can intervene between it and to, another indication it is a PC-verb. With PC-verbs like hate, want, and prefer, a DP can be found in between, and the data are listed in Table 7.2. PC complements immediately precede to less frequently than EC-ones do (the difference between begin and hate is statistically significant at p < .001, and the same for manage and hate, but, as expected, not between begin and manage).4 Certainly in cases with a noun in between the PC verb and to, the

Table 7.1.EC complements in the BNC

begin able manage need

to

X to2

Total

2628 28268 948 21967

14 (=.53%) 237 (=.83%) 16 (=1.66%) 1159 (=5%)

2642 28505 964 23126


Table 7.2.PC complements in the BNC

hate want prefer

to

X to

Total

313 286863 1360

23 (=6.85%) 4611 (=16%) 308 (=18.5%)

336 33297 1668

structure would be bi-clausal, as in (37), with a simplified inflectional layer: (37)

VP V

TP

want

T′ him

T

vP

to

...

PC verbs are also analyzable as (36), however, and are probably changing in that direction, as predicted by Late Merge. I will now give an example of such a change. Krug (2000: Chapter 4) examines the change from want to to wanna. Since this verb expresses volition it is prone to modalization, and Krug argues that this is precisely what is going on. In terms of the structure that means want changes from V, as in (37), to ASP in (36). How can we measure the changes from bi-clausal to mono-clausal structures? Krug shows that, first, there is an increase in frequency of the use of this verb, especially since the 19th century. Second, the use of want in the spoken BNC is three times as frequent as in the written BNC, and this fits with a general trend towards informality (Krug 2000: 137). Historically, the verb want occurred mainly preceding a DP and the noun want mainly preceded of and a DP. “Between 1710 and the 1990s a complete reversal of the situation has taken place” (Krug 2000: 137) and want is most often followed by a to infinitive, according to Krug. The use of the noun complement becomes infrequent. Another indication of the grammaticalization of wanna is that want and to are rarely separated by an adverb. Krug gives the data for 2 adverbs (actually and really, p. 139), but does not provide data for want DP to, which according to Table 7.2 occurs quite frequently. This is


confirmed by the CSE. Out of 4506 instances of want in the corpus, 1236 (=33%) are not immediately followed by to, but have a pronoun or full DP in between (never actually or really though). This is a marked difference from Table 7.2 and is due to the latter including only one intervening word and the data from the CSE having a full DP or one word between the verb and to. So separability from to, as with PC verbs, indicates a looser connection between the matrix verb and the embedded clause. Language change is towards inseparabilty and a loss of the intervening IP. 3.2.4 Clitic Climbing: IPs and VPs An unclear diagnostic for the presence or absence of IP is clitic-climbing. In (22) above, I assume that the presence of a CP stops the clitic from climbing. As is well-known, clitics can ‘climb’ in Romance languages with certain verbs, as in Italian (38), but not with others, as in (39): (38) Piero tii verrà a parlare ti di parapsicologia Piero you will-come to talk about parapsychology ‘Piero will come to speak to you about parapsychology.’ (39) *Piero tii deciderà di parlare ti di parapsicologia Piero you will-decide to talk about parapsychology ‘Piero will decide to talk to you about parapsychology.’ (from Rizzi 1978: 113)

This might mean that one of the two verbs has become an auxiliary, so that the IP boundary has disappeared. Myhill (1988:230) certainly links up the status of a verb with the possibility of clitic climbing: “estar, ir and andar [‘be, go, walk] … convey meanings of progressive and future when CC [Clitic Climbing] is used”. However, there are languages where clitic climbing can cross certain complementizers resulting in a marginally acceptable sentence, e.g. SerboCroatian, as in (40): (40) ?Milan ga želi da vidi Milan him want-3S comp see-3S ‘Milan wants to see him.’ (from Progovac 1993)

So, even though the Spanish data seem to suggest that clitic climbing is easiest when the main verb is becoming an auxiliary, this is not absolute. In some languages, crossing a CP is allowed (Serbo-Croatian); in some, the CP is a barrier to clitic climbing (German); and in others, the IP is (Romance).

151

152


In conclusion to Section 3, there are a number of constructions that are only possible if an IP or VP is involved. Clitic Climbing generally shows that the boundary across which the clitics ‘climb’ is an IP, not a CP. Obligatory Control is typical of an IP rather than a CP boundary. Closeness to to differentiates between IP and VP complements. Binding data are not conclusive. I have also shown in 3.2.3 that there is a continuum as to which verbs allow IP-deletion more readily, given as left-most in (41): (41) modal – aspectual – seem/appear – irrealis – propositional – factive

This will be relevant in, for instance, the next chapter since the left-most verbs will more easily change to auxiliaries, i.e. to higher positions in the tree. Factives such as regret (as also seen in Chapter 3) are the most intransparent and hence less prone to grammaticalization.

4. Patterns of grammaticalization In Chapter 2, I outlined two economy principles that account for clausal grammaticalization. Where IP and VP are concerned, it is primarily the second of these two that is relevant, namely Late Merge. The head to head changes affecting I and V are different from similar changes affecting the C. For instance, as described in Chapter 11, the sources for I are mainly V or P. In Chapter 2, I also mentioned Chinese as showing a change from V or v to ASP. Having outlined the tree structures for VP-shells in Section 1 above, I will now talk about this change in more detail, relying crucially on the work of Wu (2000). Liao ‘finish’ is originally a verb that comes after the object, as in (42), and ‘re-position[s] itself right-adjacent to the verb as an aspect-marker and reduce[s] into -le” (Wu 2000: 17), as in (43). The latter change happens during the Song Dynasty, between 960–1279 (Wu 2000: 378): (42) V Object liao (43) V-le Object

Liao is a resultative in (42) and becomes “reanalyzed as an aspectual suffix” (Wu: 380) on the first verb. This means an outer ASP is ‘created’ to accommodate the new ‘auxiliary’. Sybesma’s (1999: 75ff.) analysis of contemporary resultative le, as in (44), involves two Small Clauses (XP and YP), with le as the head of the higher one, as in (45):


(44) Zhang San ca-gan-le boli Zhang San wipe-dry-le glass ‘Zhang San wiped the glass dry.’ (Sybesma 1999: 76) (45)

VP V ca

XP X le

YP DP

Y

boli

gan

(Sybesma 1999: 76)

This is of course reminiscent of a VP-shell analysis, with le as inner aspect, as in (46): vP

(46) v ca

EP E le

VP DP boli

V′ V

A gan

Le could, however, also be in an outer ASP with gan moving to V to ca and then to le. Once le is reanalyzed as aspect marker, other resultatives can start to fill the VP-shell. Ba is another instance of a verb grammaticalizing to an (I think) aspectual marker. Ba marks a DP as definite/specific and needs to occur together with the aspectuals le or zhe (Zou 1995) and can only occur with achievement and accomplishment verbs. Ba is originally a verb meaning ‘to hold’, but becomes ambiguous, as in (47) from the 8th Century: (47) zui ba zhu-gen-zi xi kan drunk ba dogwood careful look ‘While drunk, I carefully took the dogwood and carefully looked at it’, or ‘While drunk, I carefully looked at the dogwood’.

153

154


(example taken from Lord 1993: 115, but based on a number of works by Li & Thompson)

Sybesma (1999: 163ff.) analyzes ba as the head of a CAUSEP, following an idea of Hashimoto (1971). I think it can also be analyzed as an aspect marker. In Modern Chinese, ba only marks a definite object that precedes the verb; it no longer is an independent verb. The same development occurs with jiang, which in historical texts (Sun 1996: 60) is used interchangeably with ba. Jiang has also become an object marker, but with a different geographic spread than ba. In the two chapters that follow I present case studies where verbs move from v to ASP. In Chapter 11, I then look at possible changes in this area more systematically, i.e. what moves from V to v, v to ASP, ASP to M, M to T, and T to C. I will then do the same from A to P to ASP to M to T to C. Spec to Head could be shown to be relevant to cross-clausal grammaticalization, if e.g. subordinate clauses had more clitic subjects than main clauses. I have not noticed that in any data.

5. Conclusion In this chapter, I have examined the layers in the IP, namely, T, M, and A. The ones in the VP include vP, VP, and E(xtent)P. I have shown that it is sometimes difficult to distinguish VP from IP from CP because verbs are in the process of grammaticalization and have an ambiguous structure. Raising out of a complement indicates the presence of an IP or VP and being able to move a clausal complement means the complement is a CP. If a verb is often immediately adjacent to to, it is probably an auxiliary, not divided from its main verb by a full IP. Clitic Climbing and Binding are less decisive. The next chapters are case studies of grammaticalization. Chapter 8 will show modals going from V to v to ASP, and Chapter 9 will argue that saw in English has become an evidential in ASP. Chapter 10 looks at aspectual markers moving to higher positions.

Chapter 8

Changes in modals and have Competition for ASP-hood

Perhaps the most radical of cross-clausal grammaticalizations is the one where one of the verbs becomes an auxiliary with the result that the two earlier clauses become one. As is very well-known, this happens to modals (as well as many other verbs) between OE and ModE. In this chapter, I examine the parts of this change that are relevant to Late Merge, i.e. prompted by it. In particular, I show that one of the stages that the modal goes through is from V to ASP(ect). Since much has been written on the change from V to Auxiliary, I will only briefly discuss the first stage in which the modal is a full verb, but will focus on the change to ASP. Most accounts (e.g. Roberts & Roussou 2003), while acknowledging that there may be more functional categories, argue that the change is from V to T. I will be more precise. This change will show many parallels to Chinese le and ba grammaticalization, as briefly mentioned in Chapter 2 and in more detail in Chapter 7. Since I argue that (deontic) modals merge as ASP, this raises the question as to what modals have in common with aspect. Comrie’s (1976) book on aspect doesn’t mention the connection and neither does Palmer’s (1986) book on mood and modality. However, in many languages there is such a connection. Spanish has a construction where an (originally perfect) auxiliary indicates obligation, and so does a typologically very different language, Yagua, where Payne (1995: 68) argues “-Maa indicates ‘perfect’ when following the verb, but … an obligative sense when following any preverbal element”. Payne (p. 242), in fact, makes the connection from Yagua to English I have to go. Ought to derives from the verb agan ‘to have’ in a similar manner, and expressions such as Be done at 1 are relicts of an earlier perfective. I’d like to argue that (deontic) mood and aspect are mutually exclusive: either of the two is expressed in (outer) ASP. In Section 1, I will provide an analysis of modals in ModE, with the deontic modal base generated in ASP and the epistemic one in M, both moving to T. In Section 2, I’ll account for double modal varieties within the framework sketched


in Section 1. Sections 1 and 2 are based on work in van Gelderen (2003). The remainder of the chapter will focus on deontic modals, and the change to ASP rather than on epistemic modals and the further change to M. Many linguists have worked on the historical origins of modals (Traugott 1972; Lightfoot 1979) and this work will not be reviewed in any detail. Instead, in Section 3, I’ll look at the role of the perfectivizing ge- prefix, and demonstrate that modals often combine with a verb which has that prefix. This is unexpected given that ModE deontic modals do not occur together with aspectual have. Section 4 considers the role of have as an ASP element, and an analysis of the changes in terms of Late Merge is provided in Section 5.

1.

ModE Analysis

Having indicated in the previous chapter that the traditional IP can be divided into a M(ood), ASP(ect), and T(ense) Phrase, I now argue in favor of a structure for modal verbs in ModE making crucial use of these categories. Deontic and epistemic predicates differ in that the latter but not the former can have auxiliaries and stative complements. This is accounted for if epistemics select an ASPP, but if deontics are in ASP in complementary distribution with other aspectuals. Double modal varieties of English typically have at most two modals with the epistemic preceding the deontic. This fits if one modal is in ASP and the other higher in the structure, as I show in Section 2. In standard English, modals are restricted to finite contexts and double modals no longer occur. 1.1 Deontic and epistemic modals: Preliminary remarks The division of modals is an old and debated one. Palmer (1986) summarizes the different terminologies and divisions. Without going into that debate too much, I argue in this chapter that structurally there are two kinds of modals in English and, following e.g. Lyons (1977), I will refer to one as deontic, and the other as epistemic. The deontic modal typically describes volition, ability,1 or permission, whereas the epistemic modal is concerned with belief and attitude of the speaker. Shall and can only function as deontics; may, will, would, should, could and must function as epistemics and deontics; might is only an epistemic.2 The class of modals (except for will) is characterized as preterite-present, i.e. the members derive from verb forms which as their present tense have originally preterite forms (OED: “present tense has the inflection of a strong preterite”,

Changes in modals and have

entry for may). Preterites indicate finished action and are hence aspectual. That modals indicate aspect at one stage in their development does not mean they necessarily do so later on. There might, however, be a connection. If deontics are lower in the tree, i.e. connected to ASP, and epistemics are higher, i.e. in a category between T and ASP, this suggests that elements ‘climb’ higher up in the tree as they grammaticalize, i.e. as they change from lexical to grammatical elements (see also IJbema 2002). An explanation for this grammaticalization from verbal head to auxiliary head can be found in Chomsky’s (1995; 2001a) ‘merge-over-move’ principle. If a verb must move to T, then it is ‘cheaper’ to merge it there right away. 1.2 Modals as ASP In this section, I present tree structures for deontic and epistemic modals; in the sections that follow, I will provide arguments for these structures. The structure I suggest for the deontic modal is as in (1): base generated as the head of the ASPP and moving to T via M. Quite a number of linguists have argued that deontic modals are lower in the tree than epistemic ones (e.g. Picallo 1984, 1990 for Catalan; Avram 2000 for Romanian; Cinque 2001: 14 for Turkish),3 without making the connection to ASP. Making this connection, as in (1), accounts for the complementary distribution between modals and aspect, discussed in Section 1.3 below, and makes the deontic into an aspect marker, which is not surprising given its history:

157

158


(1)

TP T¢ T

MP M¢ M

ASPP ASP¢ ASP

vP v¢ v

VP V¢

can

she

V read

DP a book

In (1) and (2), I assume (as in Chomsky 1995) that there is no separate AGR(eement) Phrase, but nothing really hinges on this. The structure for an epistemic modal, I suggest, is as in (2):4

Changes in modals and have 159

(2)

TP T¢ T

MP M¢ M

AspP Asp¢ Asp

vP v¢ v

VP V¢

might

be

she

V reading

DP a book

There is a lively debate, e.g. Thráinsson & Vikner (1995), Wurmbrand (2001: 188ff.),5 Reis (1999), and Axel (2001),6 as to whether or not epistemics are raising verbs. I will argue they originate in M, and as can be seen from adverbs and negation (see below), they then move to T. As mentioned in the previous chapter, Cinque (1999) provides an incredibly rich functional structure that is included in every sentence of every language. The evidence for these functional categories comes from adverbials in the specifiers of the positions. In (3), repeated from the previous chapter, I list some of the categories that are relevant to modals, with the adverbs listed underneath: (3) Modepist Tpast Tfut Moodirr Modnecc Modposs Modvol probably once then perhaps

Modability/perm possibly intentionally [not provided in C]

Cinque (1999: 79; 81; 106)

I have already indicated some problems with this approach. For modals, (3) would predict a preference for an epistemic modal being followed by an irrealis or possibility adverb, whereas a deontic modal might be preceded by such an


adverb. It could be that later movement eradicates the difference, but Cinque doesn’t talk about this. In the trees I advocate, there should be no differences since both epistemic and deontic modals move to T. This turns out to be correct, as I’ll now show. The numbers in the BNC with perhaps and deontic can are: 139 can perhaps and 7 perhaps can(not). This is unexpected under Cinque, but expected if the deontic moves to T. With epistemics, the numbers are: 73 might perhaps and 9 perhaps might, indicating that deontics and epistemics end up in the same position. With possibly, the numbers are: 76 might possibly, 3 possibly might, 147 can possibly, and 134 possibly can. The reason the latter combination is so high is because all are fixed elliptical expressions after comparatives in as DP possibly can, and in conditionals if DP possibly can. It may be that these deontics don’t move from ASP to T since these are irrealis. I’ll not go into this further. Both deontic can and epistemic might are typically followed by the adverb showing they end up in the same position. To account for these data, Cinque would have to argue that all modals move to or via T (which he seems to do, e.g. p. 87), but that takes the sting out of the proposal. A second argument that all modals end up in T, even though they do not seem to be checking tense and agreement there, is the cliticization with the negative n’t. If the NegP is below TP (see Ouhalla 1990), the modal moves to the head of Neg on its way to T, just like the auxiliaries have, be, and do that move to check their features. In the sections that follow, I will assume that modals end up in T, because of the evidence just provided, but will examine other aspects of (1) and (2), starting with the complements the two kinds of modals select. This shows that the base positions of the modals differ even if the surface ones do not. 1.3 Complements In this section, I show that deontic modals behave differently from epistemic ones with respect to the complements they take. This difference is accounted for if the former are in ASP and the latter in M. Modals in M are not in complementary distribution with elements in ASP but modals in ASP obviously are. In 1.3.1, I’ll discuss modals preceding auxiliaries, and in 1.3.2, modals and stative verbs. 1.3.1 Perfect and progressive auxiliaries The most compelling argument that ASP is involved in the structure for modals is that deontic modals cannot occur with perfective and progressive auxiliaries, as (4) and (5) show. If the latter auxiliaries are related to ASP (move through it), in keeping with their aspectual character, this suggests a structure as in (1)


above with deontic modals in ASP, in complementary distribution with the aspectual auxiliaries:7 (4) *I can have read that book. (deontic and perfective) (5) *I can be swimming. (deontic and progressive)

With epistemic modals, combinations such as in (4) and (5) are grammatical because the modals would be in M and the auxiliaries in ASP: (6) He must have read that letter. (7) He must be looking for that letter.

There are some interesting complications in that the three past tense deontic modals could, would, and should8 do appear with have, unexpected with deontics. Their interpretation is different, however. In English, the ‘normal’ present perfect, as in (8), is incompatible with past time adverbials. This is not the case when a deontic modal precedes have, as (9) shows: (8) *I have made him ill yesterday. (9) You should have made him ill yesterday.

Have in (8) needs to move from ASP to T where it checks its present tense features, resulting in a ‘present relevance’ interpretation. This is incompatible with the time adverbial. English have following a past tense deontic (and epistemic) modal,9 as in (9), displays behavior quite unlike its use without a modal, namely the meaning is a real past without present relevance. Thus, an adverb as in (9) is completely grammatical, unlike in (8) when have is not the complement of a modal. This interpretational difference between (8) and (9) ties in with an interesting morphological fact about English, namely reduction of have to a, av and of after modals.10 (The spelling of for have can frequently be seen in student papers). The reduction of have after modals, as in (10), is different from the reduction elsewhere, e.g. as in (11). In (10), the verbal part of have is retained whereas in (11), the person marking is. Kayne (1997) considers the reduced form of a complementizer. I’ll argue that all forms are clitics in T indicating past tense that needs to have a modal join to it by head movement: (10) I shoulda seen him. (11) Ed’s seen him.

In accordance with e.g. Zwicky & Pullum (1983), but see also Kaisse (1983), I will argue that the auxiliary -a in (10) is a clitic, and needs to incorporate into

161


another head. The same holds for the variant forms. Zwicky & Pullum’s (1983: 503–4) arguments distinguishing clitics from agreement markers are well-known. They involve criteria such as the selection of a host (clitics attach freely), morphological and semantic idiosyncracies (less common for clitics), and attachment to material already containing clitics (clitics can attach to other clitics). All these point to -a being a clitic. For instance, clitics attach to personal pronouns, as in He’s gone there, to full nouns as in (11), relative pronouns, complementizers, and auxiliaries. They also attach to other clitics as in I’ll’av eaten by then. In Section 1.2, I mentioned that NegP is below TP and this fits with the contractions as well. If past av/a/of is in T and Neg below it, the modal on its way via Neg to T first picks up n’t and then av resulting in forms such as shouldn’av. Not can also be a specifier and is then not cliticized, as in coulda NOT gone. The use of contracted have, as in (10), starts around 1250, as I will show in Section 4 below, i.e. as soon as modals become complemented with have. This is probably evidence for modals already being in ASP, ‘causing’ the have to occupy T. 1.3.2 Statives A related argument that ASP is involved in the structure of deontic modals is that the Aktionsart of the verbs in the complement of deontic modals differs from that in the epistemic ones. Deontic modals take eventive11 verbs as in (12), as argued by e.g. Heine (1993), Barbiers (1995), and Abraham (1998; 2002). Stative complements, as in (13), are ungrammatical if the modal is deontic: (12) He must/may read that letter, i.e. Someone forces/allows him to read that letter (deontic reading and eventive verb). (13) *An orange must/may be healthy,12 i.e. Someone forces/allows an orange to be healthy (deontic reading and stative verb).

This suggests that eventive verbs are vP or VP and can therefore be the complement of a deontic in ASP, but that stative verbs are dominated by ASPP and can therefore not be the complement of a deontic in ASP. This is in fact what Diesing (1992:24–26) argues for individual level predicates, i.e. stative, and stage level predicates, i.e. eventives, respectively. She uses IP rather than ASPP but the result is the same. She shows that individual level predicates have their subjects in the Spec of IP but that stage level ones have them in the Spec of VP (now vP).


As mentioned, the reason that deontics occupy ASP may be that deontics originate historically as perfective, i.e. stative, verbs (can means ‘to know’, motan ‘to have measured’, and the non-preterite-present will is originally ‘to want’) and express that original unboundedness through being in ASP. So when modals occur in ASP, the main verb cannot be stative. With a stative verb, the unboundedness must be expressed by checking ASP or by have/be in ASP. Epistemic modals (in M) take either stative ASPP complements, as in (14), or eventive vP/VP ones, as in (15): (14) He must know that person, i.e. It is very likely/may be possible that he knows that person (epistemic reading and stative verb). (15) He might read that letter, i.e. It is likely/may be possible that he read that letter (epistemic reading and eventive verb).

One peculiarity involves the modal should. It is epistemic (historically perhaps the first to become epistemic, see (30) and (31) below) as well as deontic. However, its deontic use patterns with epistemics in preceding progressive be, as in (16), and a stative verb, as in (17): (16) You should be working harder. (17) You should know that answer.

There are a number of ways to account for this, none completely satisfactory. It could be that should despite its being deontic is situated in M, and behaves syntactically like an epistemic. Jamaican Creole uses reduced have after epistemic modals in a way similar to English. Durrleman (2000) argues that even though initially this have may have been perfective, it “did not retain its original perfective value” (p. 204) and no longer has an independent meaning. She provides a nice minimal pair of an epistemic with reduced have and a deontic one without, as in (18) and (19): (18) Jan mosa did haffi tell dem ‘John must have had to tell them.’ (19) Jan did mos haffi tell dem ‘John was obliged to tell them.’ (both from Durrleman 2000: 211; glosses adapted)

As will be shown in Section 2, there are many varieties of English with double modals, and (18) and (19) fit in that perfectly: the epistemic modal merged as


M but moved to T (picking up the clitic) and the deontic one merged as ASP. The point to be made here is that reduced have is very different from full have in a number of varieties of English. So far in this section, I have provided two arguments for the structures in (1) and (2). The first is that deontic modals resist having have/be follow them, expected if the modal and the auxiliaries are both in ASP. Epistemic modals do occur together with have/be since they are in a position different from ASP. A second argument supporting different positions for the two modals is that if the verb in the complement is stative, the modal has to be epistemic. An interesting observation is that have when preceded by the past tense deontics could, should, would and must, and the epistemic modals does not have a perfect interpretation (indicated in the table as ‘(=past)’. This is due, I argue, to -a being a T checking past tense features. Some other notable points: the behavior of should resembles that of an epistemic modal and may have to be generated in M; must, even though historically past, functions like a present tense; and shall occurs infrequently. In a table, this can be summarized as shown in Table 8.1.

2. Double modals In this section, I show that the varieties and stages of English that allow double modals provide further support for (1) and (2), because they are generally restricted to two modals, and because the order is always epistemic and then deontic. There is a lot of uncertainty as to the exact analysis, especially in combinations of double modals and have and I point those out without further using them for my analysis.

Table 8.1.Properties of English modals

Present deontic can, may, (shall), will,13 must Past deontic could, would Past deontic should Epistemic may, will, might, would, could, must, should

AUX have

AUX be

stative

–

–

–

+ (=past)

–

–

+ (=past)

+

+

+ (=past)

+

+


As is well-known, many varieties of English have double modals (e.g. Labov 1972, Boertien 1986, Di Paolo 1986, Brown 1991, to name but a few), as shown in (20), and earlier stages do as well, as in (21), from the year 1475: (20) He might can do it if he tried he might be able to do it if he tried (from Brown 1991: 76) (21) ye schall can lyve lyke a jentylman you shall be able to live like a gentleman (Paston Letters, #292, p. 486)

The pattern of double modals in (20) and (21) is quite wide-spread (and, according to di Paolo (1989: 196), easily acquired even as an adult) and presents some evidence for my analysis: (a) epistemic modals always precede deontic modals, as in (20), and (b) most dialects are restricted to two modals, as shown by the ungrammatical (22) for Ozark. Teresa Wells (p.c.) says (22) is “vaguely possible”, but with the meaning as in the gloss. Boertien (1986: 302) discusses triple modal constructions, as in (23), and suggests they have a structure where the first two modals form a unit: (22) You might could may go ‘someone is speculating that someone else might give you permission to go.’ (23) I might shouldn’t oughta. (from Boertien 1986: 302)

This pattern is more restricted than it is in languages such as German and Dutch where three or four modals (argued to be main verbs by many) occur regularly, as in (24), from Dutch: (24) Je zou hier eigenlijk moeten kunnen mogen kamperen you should here really must can may camp ‘You should really be able to be allowed to camp here.’

In standard English, deontic modals do not occur with perfect have but epistemic modals do. The same is not true in all English double modal varieties, as (25) from Scottish shows, which is unexpected given (1) and (2). Other varieties can also have a double have as in (26), making an analysis difficult: (25) He might no could have done it ‘It is possible that he was unable to do it.’ (from Brown 1991: 98) (26) I think that we should have ought’ve done that yesterday. (from Boertien 1986: 309)

Labov (1972:59) argues that in sentences such as (25), and (27) below, might (and must) “are functioning formally as adverbs, without a tense marker”. Di Paolo


(1989) presents an analysis where some double modals are one lexical item: (27) She still might don’t even like the thing. (from Labov 1972: 59)

If this is the case, these constructions present no evidence for the analysis in (1) and (2) above; (20) and (21) do. As to why (standard) English does not allow double modals, one explanation is that deontic modals, in double modal varieties, are either finite or nonfinite but are only finite in standard English, hence restricted. This restriction to finite form is part of the grammaticalization process of more reduced morphological shape. In double modal varieties, modals can be preceded by to if they are deontic (e.g. Brown 1991: 78; Visser 1963–73: 1751), but this too has long been lost in standard English. Structures (1) and (2) possibly account for the double modal varieties of English by making available two positions for modals, namely M and ASP. More work is needed, in particular to understand what the structures of (25) and (26) are. Having presented arguments that deontic modals in ModE are in ASP and that epistemic ones are in M, I will now provide two arguments for when they come to be situated there.

3. From V to ASP: Competition with ge In OE, modals function as main verbs, as well as deontic modals (see e.g. Lightfoot 1979; Traugott 1972, 1989; Roberts 1985; Goossens 1982, Denison 1993, Warner 1993, to name but a few). This is obvious from their use with DP objects and as participle, as in fifteenth century (28) and (29) respectively: (28) He can al langagis ‘He knows all languages.’ (Beryn 2662, from Visser 499) (29) yf by any meanes she had conde if by any means she had could ‘If she had been able to.’ (Blanchardyn 97/4, from Kellner 1890: liii)

Epistemic uses develop by Late OE and Early ME. Traugott (1989: 41–2) lists (30) and (31): (30) & to þam Pentecosten … wæs gesewen blod weallan of earþan. and at that Pentecost was seen blood well-up from earth


swa swa mænige sæden þe hit geseon sceoldan as many said that it seen should ‘At Pentecost, blood was seen to well up from the earth, as many said who supposedly saw it.’ (Anglo Saxon Chronicle E, anno 1100) (31) ic wat þ ðu geherdest oft reccan on ealdum leasu spellum þte Iob I know that you heard often tell of old false tales that Job Saturnes sunu sceolde bion se hehsta god ofer ealle oþre godu 7 he Saturn’s son should be the highest god over all other gods and he sceolde bion þæs heofenes sunu 7 sceolde ricsian on heofenu should be the heaven’s son and should rule in heavens ‘I know that you have often heard old fables be told that Job, Saturn’s son, should be the highest god over all other gods and that he should be heaven’s son and should rule the heavens.’ (Alfred, Boethius, Sedgefield p. 98.26–9)

Many have shown that modals can be used as main verbs in OE and ME and that they come to be restricted in the complement they take. OE modals appear non-finite in form, as in (29), and take DP complements, as in (28) above. This means at one point they are clear verbs. Rather than going through the same data again, I will focus on two points in this and the next section, not often mentioned in the literature: (a) in OE, many modals have a ge-prefixed verb following them, and (b) have after modals becomes an affix before other forms of have do, and starts to have another meaning, see Section 1.3.1. I claim these points show that the change is from V to ASP, and later to M. As mentioned, present tense deontic modals derive historically from perfective verb forms and the class is therefore called ‘preterite-present’. This suggests an ‘affinity’ with aspect, as I now show is the case. There is a frequent ge- prefix on the verb accompanying the (deontic) modal in OE. This shows that the modal is still a V with an IP or ASPP as complement or a v with an inner aspect as complement. Epistemic modals, as in (30), show the same but, as mentioned, they won’t be my focus. The use of ge- with epistemics is expected if they are in M, as in (2) above, even in this early stage. The link between deontic modals and ge- has been remarked upon in the older literature (see van Swaay 1901: 55–6; Lenz 1886: 18–20) but hasn’t figured prominently in the more recent literature. This ge- prefix is seen (e.g. Mustanoja 1960: 446) as indicating perfective aspect, and shows that OE modals are quite different from the ones discussed in e.g. (4) above since those cannot occur together with aspect. In Beowulf, the co-occurrence of a modal with an aspectual prefix is frequent, as in (32) and (33), with all modals except can:


(32) Swa sceal geong guma gode gewyrcean so shall young man good-dat accomplish ‘So should a young man through good works accomplish.’ (Beowulf 20) (33) þæt ic sænæssas geseon mihte that I sea-bluffs see might ‘So that I could see the cliffs.’ (Beowulf 571)

Other perfectivizing prefix forms are possible with OE modals as well, as in (34), but ge- is more frequent: (34) þone biscopdom forletan sceolde that diocese leave should ‘should leave that diocese.’ (Life of St Chad, p. 162)

The frequency suggests that early modals are verbs with complements that have ge- in the inner aspect, or E, indicating perfectivity. A possible structure is as in (35), with modals in v (since they become restricted to having VP complements) and ge- in (inner) aspect. This structure is similar to Abraham (2002) who argues for German that deontic modals are small clause structures in the VP: (35)

vP v′ v

EP E′

sceal E ge/for

VP V

As the role of ge- lessens (much has been written about this, e.g. Mossé 1938), the modal takes over the aspectual role and the inner aspect (E) is reanalyzed as outer ASP, as in (36):


(36)

ASPP ASP′ ASP

vP/VP

sceal

seon

Now the structure is clearly mono-clausal. The evidence for the reanalysis to outer aspect is that ge- is lost early with modals. I’ll now show a very different use of ge- with modals. Postma (2002) shows that there are three kinds of g(h)e in Middle Dutch, one of which occurs frequently after a modal, as in (37) and (38), and is argued to induce negative polarity on the verb: (37) Daer die vianden niet wel by en conden gecommen there those enemies niet well at neg could come ‘which the enemies couldn’t reach well.’ (Postma 2002: 210) (38) Hij en kan niet ver gezijn he neg can not far ge-be ‘He can’t be far.’ (Postma 2002: 212)

This ge- can occur on verbs such as zijn in (38), normally unable to have a prefix. In OE, this is very different: there are 5 instances of gebeon in the entire corpus of OE, but as Bosworth and Toller have it, most are past participles of ‘command’ and the past participle of beon ‘be’, as in (39), is very rare (only one instance): (39) þ he heafde gebeon on þes cynges swicdome and hit him on gefeaht that he had been on the king’s treason and it him through fight ‘that he had been in on the treason against the king, and maintained it against him through fighting.’ (Anglo Saxon Chronicle E anno 1096, Thorpe 1861: 362)

There seems not to be the frequent negative context as in Dutch either, so the use as in (37) and (38) does not exist in OE. The ge- affix discussed in this section can be selected by a modal and, as I’ll show in Section 4, by have as well. With the demise of ge-, modals become generated in the ASP phrase, and become incompatible with other aspectual information such as perfective have. In Chapter 10, I’ll return to the demise of ge- and suggest that before the complete collapse of ge-, it is reanalyzed as outer aspect.


4. Modals and have: Taking over the ASP position The use of auxiliary have and be with modals is late. The OED (entry for have II 26) gives twelfth century examples but the modals are either still main verbs or epistemic modals. As Traugott (1992: 200) puts it for OE: “[n]o two auxiliary verbs may occur in sequence except Pre-modal — Passive (e.g. might be destroyed)”. In fact, modals and have have similar complementation patterns, as I’ll now go into. In Section 4.1, I’ll demonstrate that ‘have’ is already an auxiliary in OE. Since I have argued that modals come to occupy outer ASP, I need to explain how that interacts with aspectual have. In Section 4.2, I argue that once the deontic modal is in ASP, have cannot be there at the same time as well and is reanalyzed as a clitic in T. 4.1 OE Since the use of auxiliaries in OE is more limited than it is in ModE, there are a number of linguists (e.g. Visser 2189; Mitchell I: 292) who assume have in constructions such as (40) is a main verb, namely a possessive, and that the past participle gefeormod is a modifier to the object unlyfigendes: (40) sona hæfde | unlyfigendes eal gefeormod soon had lifeless all eaten ‘As soon as he had eaten the lifeless one completely.’ (Beowulf 743–4)

Brinton (1988: 99ff) notes some problems with this traditional account of the development of the perfect in English, e.g. have with a participle is rare in OE with possessive meaning. In addition, the participle is only rarely inflected and DP-Participle or Participle-DP order seems to have no influence on the inflection. Also, the spread of the construction to intransitives is not evident since these already occur in OE. Brinton (1988: 102) concludes that ‘have’ and the participle are perfects and that there is no reason to believe have in (40) has possessive meaning.14 The instances of have in Beowulf also show that have is auxiliary-like. Mitchell’s scenario is unlikely for two reasons: (a) have also occurs with past participles that are intransitive, and (b) the ‘have’ that occurs with a lexical verb always occurs in a perfectivizing context. This, I argue, means have is in v, selecting a (lower) ASPP, as in (35) above (with an occasional object agreeing with the participle, as in (44) below).


Regarding (a), ‘have’ occurs with past participles that are intransitive, as in (41) to (42), and it would be hard to say that there is an object of ‘have’ that the participle modifies: (41) syððan hie togædre gegan hæfdon when they together gone had ‘as soon as they met.’ (Beowulf 2630) (42) we to symble geseten hæfdon we to meal seated had ‘We had sat down to the meal.’ (Beowulf 2104)

Regarding (b), in Beowulf, there are 81 occurrences of a form of ‘have’ and 47 of those occur in combination with a lexical verb. The lexical verb connected with ‘have’ always has a perfectivizing prefix such as ge- as in (43) and (44) (a total of 34 times), for- as in (45) (6 times), a- (3 times), be- (twice) and on(twice). Some are transitivizing. ‘Have’ is a main verb the remaining 34 times: (43) hu hit hringdene | … gebun hæfdon how it Danes … occupied had ‘How the Danes had occupied the Hall.’ (Beowulf 116–7) (44) Hæfde se goda Geata leoda | cempan gecorone had the good Goth people warriors chosen-p ‘(while) the good one chose warriors among the Geats.’ (Beowulf 205–6) (45) siþðan him scyppend forscrifen hæfde since him the creator banned had ‘Since the creator had banned him.’ (Beowulf 106)

Thus, the two criteria show that ‘have’, in connection with a participle, is no longer a main verb meaning ‘to hold’, but that it has auxiliary characteristics. Examining the 81 instances of have in Beowulf gives the following results: ‘have’ extends its meaning of physical possession, as in perhaps (40), occurs in a perfective context, as in (41) and (42), and has lost some of its phonological weight since it does not participate in alliteration. In OE, ‘have’ is an auxiliary (as well as a main verb), at least in Beowulf, and can be argued to be in inner aspect. When it starts to occur with deontic modals, it is no longer perfective. This shows that modals ‘push’ have out of ASP, which happens in ME. 4.2 ME In ME, ‘have’ is often reduced morphologically when it is a non-finite auxiliary

171


and occurs after a modal. The OED says that the reduction is “exceedingly frequent in 13–17th c.; in later times chiefly in representations of colloquial or familiar speech” (entry for a, verb). After modals, ‘have’ becomes a clitic and starts to occupy a functional category (presumably T, see 1.3.1) and the modal moves to T to join it. As a result, it has a simple past reading after modals. The reason for the change of have after modals is that both are competing for the same position. I’ll now show both the reduction and the modified interpretation in a number of fifteenth century and later texts. In the 15th century Paston Letters, the reduction is very frequent as in (46) to (50) and I will therefore take a closer look at this set of letters. As in ModE, the reduction only occurs with perfective have, not with main verbs, as (51) shows:15 (46) we haue not don owur duté, whyche was to haue seyn and ave waytyd up-on Šou ‘We haven’t done our duty, which was to have seen and have waited for you.’ (Paston Letters I, #399, anno 1481, Edmond Paston II) (47) … speed in that matter as i wolde ve hadde, and as i hope ye … Success in that matter as I would have had and as I hope you … (Paston Letters, #305, anno 1477, John Paston II) (48) it xuld a be seyd ‘It should have been said.’ (Paston Letters, #131, anno 1449, Margaret Paston) (49) Še wold a be plesyd ‘You would have been pleased.’ (Paston Letters, #176, anno 1464, Margaret Paston) (50) there xuld not a be do so mykele ‘There should not have been done so much.’ (Paston Letters, I, #205, anno 1469, Margaret Paston) (51) a vestment which they have for a prist a piece of clothing which they have for a priest (Paston Letters, #34 anno 1479)

Of the different letter writers in the Paston Letters, Margaret is the most frequent user of reduced a. Her 104 letters are in 29 different hands (Davis 1971: xxxvii). So dictating the letters might have made a difference. Others (some dictated as well) also use a reduced form. There is a wold not a in letter 23 by Agnes, a nede not a in letter 417 by Margery, a might not a in letter 88 by William II, and two forms of a in letters by Edmond. There are 7 instances of


wold a and 5 are by Margaret, one by Agnes and one by John III; of the 9 instances of xuld(e) a, all are in letters by Margaret. In Margaret’s letters, the reduced forms are only non-finite auxiliaries. She uses 631 instances of haue and 196 of have; 5 instances of a occur after an infinitival to; 40 after a modal as in (48) to (50) above or a modal with not/adverb; there are no cases of reduced third person hath forms; one ha occurs after a modal; 7 of an as in (52) but only before the participle had or before one starting in a vowel (had may have been produced without the initial consonant): (52) so þat he myth an had mony to… ‘so that he might have had money to …’ (Paston Letters, #128, anno 1448, Margaret)

In the Paston Letters, the prefix on the past participle has disappeared completely, as (53) shows, and if the modal is in ASP, have is not but is a clitic in T to which the modal moves. There are constructions where ‘have’ disappears even if it is finite, as in (54) and (55): (53) the sayd ryottys haue be done … ‘The above-mentioned riots have been done.’ (Paston Letters, #48, anno 1454, John I) (54) and for asmuche as we been enfourmed … ‘And for as much as we have been informed.’ (Paston Letters, #122A, anno 1467) (55) Še spoken to me of … ‘You (have) spoken to me of …’ (Paston Letters, #135, anno 1449, Margaret)

Sentences such as (56) to (58) show that ‘have’ combined with a past participle does not have a present relevance reading, as it does in ModE. These occur throughout the Paston Letters, and I have given them a modern gloss. They continue to be used until C19:16 (56) he hath acordyng to youre desyre spoken Yeluerton yesterday ‘He spoke to Y yesterday, as per your wish.’ (Paston Letters, #184, anno 1465, Margaret) (57) I haue send to Ser Thomas Howys yesterday … ‘I sent yesterday to Sir Thomas Howys.’ (Paston Letters, # 190, anno 1465, Margaret)

173


(58) I haue yesterday sent to Herry … ‘I sent to Harry yesterday …’ (Paston Letters, #352, anno 1472, John III)

In The Cely Letters, from a set of letters written by members of the Cely family, also in the 15th century, there are many instances of auxiliary a and an and from a variety of correspondents. As in the Paston Letters, reduction only occurs in non-finite instances after a modal: (59) and so myght Y a done syn I come vnto Calles ‘and so might I have done since I came to Calais.’ (Cely Letters, 22/39 George Cely, anno 1478) (60) Syr, in good fayth Y am sory therefor, for and Y had west that ye would a taked so sor Y would nat a wreten so vnto you, nat and I schuld a gette therbey xx nobelys … ‘Sir, in good faith, I am sorry for it, if I had known that you would have taken it so badly I would not have written to you that way … and I should have gotten 20 nobles thereby …’ (Cely Letters, 110/7–10 William Maryon, anno 1480) (61) he was purpossyd to a ben at thys Passe Martt, but … he myght nott, wharffor he prayth yow that the xl pownde sterlyng that he schulde a receyued off yow … that Še wyll kepe hit … ‘He was supposed to have been at this Easter Market, but … he wasn’t able. Therefore he asks you that the 40 pounds stirling that he should have received from you … that you will keep it …’ (Cely Letters 155/8–13 William Cely, anno 1482) (62) if I might a spoken wyth you I wold a boght hym of you, so þat Še wold a ben resvnabyll ‘If I might have spoken with you I would have bought it from you, so that you would have been fair.’ (Cely Letters, 166/3–5 William Adam, anno 1482) (63) þat he myght an sowlde all yowr woll ‘that he might have sold all your wool.’ (Cely Letters, 22/36 George Cely, anno 1478)

In the Helsinki Corpus, morphological reduction starts in ME2, i.e. in 1250, as in (64) and (65): (64) And ha men neuer ben sa bald ‘and men have never been so bold.’ (Cursor Mundi 3561)


(65) Deuel ha hold him ‘The devil has held him.’ (Capgrave’s Sermon 7537)

However, in some of these texts, both finite and non-finite auxiliaries are reduced. The earliest is ha from the Kentish Sermons; others occur in Man in the Moon, The Prose Psalter, and Margery Kempe. In the last period covered by HC (EMOD 3, i.e. up to 1710), there is one ha’, in Farquhar. Visser (1973: 2035) mentions instances from the 15th and 16th centuries, Mandeville, Capgrave, The Cely Papers and Shakespeare among others. The reduction found in these texts comes in the form of ha, a, an, ave, ve, and zero. They differ in what information is left out and unlike in ModE, there are a number of instances where person and number are not checked, as in (64) and (65) above. A reason for this might be the general loss of inflection that is occurring in Early ME (cf. also Trudgill 1997). Since the early 14th century Cursor Mundi has instances where finite auxiliaries are reduced as in (64), I will now take a closer look at this text. As mentioned in a note in Chapter 5, Cursor Mundi is a text of which several manuscripts exist (see note 52). Morris’s (1874–1893) edition has two northern and two Midlands versions but reductions are found mainly in the Cotton Vespasian A iii, which is a northern text from the early part of the 14th century. I have examined 12,000 (all written by the same scribe, cf. Hupe 1893) of the total of almost 30,000 lines. Most reduced forms in Cotton occur after modals as in (66) and then, as in the Paston Letters and the Cely Letters, the reduction is to a, i.e. a genuine clitic: (66) þat wald a don me o mi lijf that would have done me of my life ‘that would have taken my life.’ (Cursor Mundi, Cotton 5284)

There are also a few instances where ‘have’ is left out altogether after a modal (Cursor Mundi 7941 and 12088). The endings on the participle are there but the prefixes have disappeared. Subjunctives, as in (67), show reduction in one of the northern manuscripts: (67) þis ha faraon þe king … peace have pharaoh the King ‘May Pharaoh the King have peace.’ (Cursor Mundi, Cotton 5333)

The six instances where a finite verb in post-subject position is reduced are as in (68); these are all first and second person:17

175


(68) als yee ha said ‘as you have said.’ (Cursor Mundi, Cotton 5106)

Since inflection is lost in northern texts on a variety of verbs (especially first and second person), I ascribe the occurrence of (68) to that general tendency. In late 16th century, e.g. in Gammer Gurton’s Needle (1575) and in Shakespeare, endings may be reduced as well, as in (70) and (71). In the former, 8 instances of ha appear as in (69) but the situation is exactly as in Cursor Mundi. In Shakespeare’s First Folio Edition, there are 16 instances of ha(‘) and quite a number of have left out as in (72), even though (72) might be due to portraying a Northerner: (69) Where ha you ben fidging abrode. (Gammer 42) (70) Ophelia: So would I ha done by yonder Sunne. (Hamlet IV, v, 65) (71) Iustice: I know you ha’ pra-ctis’d. (2 Henry 4 II, i, 124) (72) Capt Iamy: I wad full faine heard some question tween you tway. (Henry V, III, ii, 127)

Unlike in the Paston Letters, there are finite auxiliaries that are reduced as in (71) as well as main verbs. They pattern as in other works, namely either the verb is preposed or the subject is a first or second person, and their form is ha. A surprising difference with the Paston Letters is that the reduced auxiliary is rare after a modal. Of the 9 reduced auxiliaries, 7 are finite as in (69). Hence, the reduction is of person and tense features. After the seventeenth century, reduced have becomes less common and is rarely left out due to prescriptive pressure. Instances from 17th century Bunyan (first edition 1666) are listed in (73) to (75), even though later editions change a to have. All of these occur after modals: (73) Chris. … I thought you would a come in by violent hand or a took the Kingdom by storm. Mer. Alas, to be in my Case, who that so was, could but a done so? …as I, that would not a knocked with all their might. (Bunyan, Pilgrim’s Progress ii, 203/191)18 (74) Mer. I might a had Husbands afore now. (Pilgrim’s Progress ii, 240/227) (75) Greath. Why what would, or could you a-done to a helped your self, if we indeed had been of that Company? Hon. Done! I would a fought as long as breath had been in me. (Pilgrim’s Progress ii, 260/247)


According to the OED, the usage has continued to the present (the OED lists many instances from Dreiser, Malamud, Naipaul, Irving) but was suppressed as English was more ‘standardized’. A structural representation for have is similar to that of the modal in (35) above, namely as in (76). The changes are also similar to those affecting the modal: the inner ASP is lost and reanalyzed as outer ASP, as in (77): (76)

vP v′ v

EP

have

(77)

E′ E

VP

ge

V

ASPP ASP′ ASP

vP/VP

have

V

The structure is now mono-clausal. Once the reanalysis of have to ASP takes place, it is in complementary distribution with the modal, as discussed before. In Chapter 11, I show that ge- is reanalyzed as outer aspect, ASP before being lost, the worsened ‘competition’ may have contributed to the demise of ge-. In conclusion, in the Paston Letters and the Cely Letters, only non-finite auxiliaries are reduced. In other texts, e.g. the Northern Cursor Mundi and Shakespeare, person and number features are deleted as well. This may be an indication of the transitionary period to checking in I, and not related to the modal. Regarding the function of have as ‘perfect’, the above shows that have indicates tense after modals (and in sentences such as (56) to (58)), not aspect. Strang (1970: 149) notes that “[t]he distinction between past and perfect was not yet so clear-cut as now. Compare Shakespeare’s the time has been, where … we should now prefer was”. Visser (1973: 2197) confirms this. It fits with what I argue in Chapter 10 that, in older English, tense is specially marked not aspect. This is a parametric change.


5. Late Merge and concluding remarks In conclusion, this chapter focuses on (a) the structure of modals in ModE, (b) the historical changes involving the deontic modal, and (c) those involving geand have. As to (a), I argue that in ModE, epistemic modals are base generated in M, deontic modals (and have and certain kinds of be) in ASP (or move to it), and light verbs (not discussed much here, but see Hale & Keyser 1993) in v. In accordance with Late Merge, the most grammaticalized, i.e. epistemics, are highest in the structure; then deontics; and then make. The structures proposed here make English similar to TMA languages such as those described by Bickerton and discussed in Chapter 7. I will leave for further research what the features are that are checked (uninterpretable or interpretable) and also whether C-T-M pattern as one unit, and ASP and V do as another. As to (b) and (c), I propose that both the deontic modal and (non-reduced) have are reanalyzed from v to (outer) ASP. The result is that they come to be in complementary distribution with each other (and ge-). Reduced have occurs with modals but is not the regular auxiliary. In this conclusion, I’d like to mention a curious construction that native speakers feel is not terribly ungrammatical, namely the one as in (78): (78) A hearer … would’ve not have captured the full meaning. (BNC J7X-167)

Most of these instances involve would and not, but it does indicate perhaps that the copy of have is not deleted because the higher one is too far. The modals involved here are irrealis and perhaps in M, not ASP. It remains for further research how to incorporate these. I have not examined the change of the epistemic modals from ASP to M. Evidence for this would have to be found in the complementation patterns, e.g. when they started to behave as in (6), (7), and (14) and (15) above.

Chapter 9

Perception verbs and ASPect1

In keeping with a cross-linguistic tendency for perception verbs (PVs) to allow reduced complements, I argue that (visual) PV complements in English involve an ASP(ect) Phrase. I saw him cross the street and I saw him crossing the street differ in interpretation but this is not only due to the bare infinitive and the present participle, but also to the category of the PV: ASP in the first sentence; V in the second. In OE, the situation is different, and this difference can be accounted for if PVs, such as saw, are always verbal rather than in ASP. Hence, sentences with PVCs are bi-clausal. The change from OE to ModE is from a lower (V) to a higher category (ASP), instigated by Late Merge, and to a monoclausal structure. The main differences between ModE on the one hand and languages such as OE and Dutch on the other emerge as (a) there are two kinds of ‘see’, one is verbal and the other is an evidential auxiliary, (b) the evidential only occurs in the past tense and occupies ASP, and (c) as argued by traditional grammarians, the changes are facilitated by the availability of -ing, and by the loss of aspectual markers. The structure I suggest reflects characteristics (a) and (b), making crucial use of the fact that in many languages evidentials and perfectives are related (see also Abraham 1998; 1999). The reason for this is that if an action is finished it must have happened. (c) will be discussed more in Chapter 10. The outline is as follows. In Section 1, I will examine the infinitival complement after PVs and aspectual restrictions on this complement in general. In Section 2, I’ll provide an analysis in terms of ModE PVs being in ASP. In Section 3, I will discuss the situation in OE and ME, and in Section 4, I’ll indicate the diachronic changes using Late Merge.

1.

PV complements and aspect

In this section, I briefly show that the complement to English PVs is typically either a bare infinitive, i.e. an infinitive without to, or a non-finite form ending


in -ing. The choice depends on aspectual considerations in ModE unlike Dutch, OE, and ME. Van Ek (1966: 86–9; 150–3) provides some complementation patterns and numbers indicating their frequencies in a million-word corpus of British English. These are given in (1). Van Ek has other complementation patterns, e.g. (2) and (3). The numbers for auditory PVs are given as well:2 (1) a. I saw him go: 52 I heard him go: 54 b. I saw him going: 101 I heard him going: 33 c. I saw him changed: 27 I heard him gone: 14 (2) Never tell him what she saw him to be. (George Eliot, Felix Holt I, Chapter 9, 172) (3) when she was seen to leave the lecture. (Nicholas Blake, Widow’s Cruise, 96 from van Ek, p. 155)

Constructions (1c), (2) and (3) seem archaic or even ungrammatical to native speakers of (American) English. I’ll return to these below in Section 3. As to the interpretation of the aspect in the (bare infinitival) complement, (4) and (5) show the well-known fact (e.g. Quirk et al. 1985: 1206) that ModE bare infinitives differ from those in other Germanic languages in that the event referred to by the infinitive in (4) must be completed; hence, (4b) is ungrammatical. In Dutch, on the other hand, the infinitive in (5a) can refer to the action in progress (or to the completed action), and hence, (5b) is grammatical: (4) a. I saw him cross the street. b. *I see him cross the street. (5) a.

Ik zag hem de straat oversteken I saw him the street cross ‘I saw him cross/crossing the street.’ b. Ik zie hem de straat oversteken I see him the street cross ‘I am seeing him cross the street.’

The two languages differ in that in English, but not in Dutch, the present tense is ill-formed, as (4b) and (5b) show. Since see is a state verb in (4b), it should not have to occur in the progressive, but note that the progressive is wellformed in (6) when see is an activity. This suggests that (4a) and (6) have a different structure, a fact I will come back to in the next section (Section 2.1): (6) I am seeing him cross the street.

Perception verbs and ASPect

Miller (2002: 256) provides a good summary of other aspectual differences between (1a) and (4a) on the one hand and (1b) and (5b) on the other, the former being a “genericized but telic event” whereas the latter is a “noncompletive, particularized event[] in progress that can have duration”. To indicate an incomplete action in English, a progressive verb is used in the complement, as in (7) and (8), not the bare infinitive, as in (4): (7) I see him crossing the street. (8) I saw her reading the book for hours.

With auditory PVs such as hear, not the focus of my study, instances such as (9) show that the PVC is not perfective and that the PV has not grammaticalized in the same way as see: (9) You’ll hear the President say at the very top of that agenda is … (CSE-WH97A)

Languages as diverse as Russian and Lele, a Chadic language, make similar morphological distinctions to indicate perfective, as in (10) and (12) respectively, or non-perfective complements, as in (11) and (13) (Buzarovska 2000 for Russian; Frajzyngier 1996: 278–9 for Lele): (10) Ja videla kak Bob pereshel ulicu I saw if Bob cross-pf street ‘I saw Bob cross the street.’ (11) Ja videla kak Bob perehodil ulicu I saw if Bob cross-impf street ‘I saw Bob crossing the street.’ (12) ng-gòl-ì wàl tù 1S-see-3 slaughter goat ‘I saw him slaughter a goat.’ (13) ng-gòl-í go jè wàl-dí kúlbà 1S-see-3 comp progr slaughter cow ‘I saw him slaughtering a cow.’

K. Hale (1991: 5–6) reports that, in Misumalpan, a group of languages spoken in Nicaragua and El Salvador, perception verbs pattern with aspectual verbs, such as ‘begin’ and ‘finish’, in selecting a participial complement, whereas other verbs select an infinitive. This suggests that perception verbs are somehow seen as belonging to aspectuals.

181


There are other indications that the infinitive in (4) really bears perfective aspect: (14) is not widely accepted by native speakers since the ‘for hours’ forces a durative reading, incompatible with the perfective, unlike its Dutch counterpart in (15): (14) ?I saw him read the book for hours. (15) Ik zag hem urenlang dat boek lezen I saw him hourslong that book read ‘I saw him read that book for hours on end.’

As in other constructions, a perfective is typically triggered with definite objects, and imperfective with indefinite ones. Therefore, (16) should be worse in English than (17). It is, but only very marginally so, however: (16) ?I saw him read books. (17) I saw him read the book.

2. PVs in ModE and Dutch: ASP and v There have been many analyses of PVCs (Akmajian 1977; Guasti 1993; van Gelderen 2001b, Felser 1999, Miller 2002, to name but a few). I will not review (all of) these. My account differs in that I argue that there are (at least) two kinds of PVCs since there are two kinds of perception verbs, and one of these is an evidential. Evidentiality and epistemic modality are connected in that in both belief in the truth of the statement is involved but in the former the source of the information is emphasized whereas in the latter the speaker’s attitude is. 2.1 Two kinds of see As is well-known, perception verbs such as see are typically stative (and that’s true for direct visual ones as well as for those of mental perception). Viberg (1983: 123) uses the term ‘experiencer-based’ for the stative ones. Because they are stative, the simple present is used in (18) rather than the progressive in (19): (18) I see (the) mountains. (19) ?I am seeing (the) mountains (as if for the first time).

As seen in (4) above, infinitival complements to these stative verbs are perfective. Not discussed in this connection in the literature, as far as I know, but


crucial for my analysis below, is the occurrence of (20) to (24), and (6) above, where the bare infinitive expresses duration, and perfectivity is not connected to the infinitive: (20) Seeing her swim is exciting. (21) I made them watch Michael swim (for hours). (22) Mary watched the boatman leave the house. (Gaskell, Mary Barton 31, from the OED) (23) Martin took it, feeling himself surrender. (from Visser, 2251) (24) We’d be hearing him holler for mercy. (from Scheffer 1975: 68)

So, when the perception verb is not a state, but an activity, or used in the progressive, as in (20) to (24), its bare infinitival complement need not be perfective (contra Giorgi & Pianesi 1997 who argue all infinitives are perfectives). I will argue that ‘see’ is ambiguous between a non-stative reading, as in (20), and a stative reading, as in (4a). In the latter case, I argue it is an evidential in ASP, and this is only the case in the past tense (cf. (4a) and (4b) and not with hear as (9) above shows). It is not unusual for verbs of perception to grammaticalize into evidentials. For instance, Gordon (1986: 75; 84) shows that in Maricopa, a Yuman language, see and hear can be either evidentials or full verbs with a sentential complement. The non-stative visual PVs in English, as in (20) to (22), are not typically see but watch, observe, look at, and perceive. I’ll elaborate on that in Section 2. Summarizing the current section up to now, there is some evidence that English (4) is different from Dutch (5) even though both involve bare infinitival complements: an exclusively perfective reading in (4), an incompatibility with duration adverbs in English (14), but not in Dutch (15), and a slightly marginal English (16). However, this is only the case if the bare infinitive is embedded under a stative perception verb, such as saw, as in (4a). The structure I suggest for English (4a), (7), and Dutch (5) is (25). Him and hem will move to Spec ASPP (for Case reasons presumably), and oversteken ‘cross’ will move to ASP in Dutch where it can check either perfective or imperfective aspect. In English, cross will only move if it is imperfective to ‘pick up’ -ing. This only happens if saw is present in a higher V (of another clause). If saw is present in ASP, as in (4a), it will check the perfective and the crossing of the street is at the same time as the seeing. So, saw in (4a) and -ing in (7) are in complementary distribution, the former checking perfective and the latter imperfective:

183


(25)

ASPP ASP¢ ASP

vP v¢ v

VP V¢ V

Ï saw ¸ Ì -ing ˝ Ó ˛

him/hem

...

cross/oversteken

While agreeing with Felser (1999) that PVCs involve an ASPP, I argue that the static PV is in ASP. Later on this section, I suggest a possible modification where saw starts out in v and then moves to ASP. It may be, as Ramchand (1997: 216) argues, that vP is really ASPP and can hence be deleted. I won’t go into that here. Assume that in (25) ASP indicates ±boundedness/perfectivity. Then ModE has two options for ASP: either perfective saw or imperfective -ing. If saw is stative, it originates in ASP or moves to it, as in (4a), and then cross cannot have an independent aspect but is dependent on ASP. This use of saw is evidential and it is not surprising that it occurs only in the past. Abraham (1998) argues that “evidentiality is … often triggered by the perfect or perfectiveness”. Comrie (1976: 108–110) argues that the perfect is typical for the inferential evidential, not the direct evidence one. I see no reason to restrict it that way, at least for English, and will assume that perfect can also be used for direct perception. In many languages, perception in the past is more grammaticalized than perception in the present. For instance, Turkish indicates evidentiality (direct vs. indirect perception) in the past tense (see Slobin and Aksu 1982: 188), and Buzarovska (p.c.) reports that in Greek and Macedonian PVs in the past tense, i.e. as in (4), have a special infinitival complementation that makes them more grammaticalized into evidentials than the present tense ones. Barnes (1984:259) shows that a verb with a visual evidential suffix is past unless specially marked.


The visual PV see can also be non-stative, and then it is a full verb higher in the tree with cross checking imperfective, as in (20), and possibly perfective. The tree would be as in (26), identical to (25) but with the higher VP showing. This is a bi-clausal structure and the two ASPect Phrases occur in separate clauses: ASPP

(26) ASP

VP V

ASPP ASP′ ASP

vP v′ v

VP V′ V

(-ing)

see

[±pf]

him

...

cross

In (26), the higher PV can of course be present tense, as in (7), since it is not checking ASP. Since -ing is not available in Dutch, ASP is ambiguous between ±perfective (‘pf ’ in (26)). Therefore, oversteken ‘cross’ in (5) can have independent aspect which it checks in ASP. Felser (1999: 205ff.) argues that Dutch aan ’t, in sentences such as (27), and German am is in an ASP projection as well. Intuitions about sentences such as (27) are that they are very marginal after perception verbs:3 (27) ??Ik zag hem ‘Harry Potter’ aan ’t lezen I saw him ‘Harry Potter on the reading ‘I saw him reading Harry Potter.’

It is interesting that historically many -ing forms in English derive from a preposition followed by a verb with -ing (cf. van Gelderen 1993: Chapter 8). Hence, (27) and the German counterpart with am would not be an unexpected development, predicted by (26).

185


The complement to a perception verb cannot be a stative, as in (28), or an individual level predicate, as (29) shows:4 (28) *I saw you be/being tall. (29) *I saw him know/knowing the answer.

As before in Chapter 8, I’ll use Diesing’s (1992: 24–26) account that argues that individual level predicates, i.e. statives, are IPs but that stage level predicates, i.e. eventives, are VPs (see also Rochette 1988, and Higginbotham 1983: 118 for a different account). If this is so, one can expect that IP complements such as in (28) will not occur since they do not ‘fit’ in (25) and (26). As expected, the internal aspect of the complement to the non-stative see need not remain nonstative either, as the grammaticality of (30) shows: (30) Seeing her be so healthy is a pleasure.

The structure of stative complements never allows auxiliary have or be (except passive)5 in either English or Dutch, as in (31) and (32) respectively: (31) *I saw him have crossed the street. (32) *Ik zag hem de straat zijn overgestoken I saw him the street be crossed

This is again explained by the structure: since perfect have and progressive be result in states (see Vlach 1981: 287 and Comrie 1976: 56), i.e. IPs, they cannot occur with perception verbs. Once have is used, as in (33), the structure changes into one where the -ing modifies the subject or object, and a comma intonation occurs between him and having: (33) I saw him having crossed the street.

Syntactically, this means that the complement in (4) and (7) is greatly reduced in structure, as shown in (25), not an IP or CP, but a vP in (4) and an ASPP in (7). The complement is reduced in Dutch as well, but can only be an ASPP in (5). I will not go into the structure of verbs such as watch, but there is evidence that their complement is a CP, not a reduced clause. As Kirsner (1977) has shown, the subject of the embedded infinitive cannot be passivized, as (34) shows, unlike the two kinds of see in (35) and (36): (34) *Nureyevi was watched ti to leap across the stage. (35) Nureyevi was seen ti to leap across the stage. (both from Kirsner, p. 174) (36) Hei was seen ti leaping across the stage.

Perception verbs and ASPect 187

If sentences such as (35) and (36) have a structure as in (26) above, it follows that an embedded DP can move to the subject of the passive of the higher clause since it would not be crossing a CP (and IP) boundary. Verbs such as watch have full-fletched CP complements and therefore lack passives. As is well-known, passives of PVCs are only grammatical with a to, as in (35), but not without, as in (37): (37) *He was seen leap across the stage.

Many accounts exist. For instance, Basilico (2003) accounts for why a bare infinitive cannot passivize by positing a TopicP with a pro in the case of ‘see’, and only the pro can be passivized. There is a much simpler answer, namely that the see in ASP cannot be passivized because it is too high up and too far from the V where (passive) voice is realized. The ungrammaticality of (37) therefore provides additional evidence for an analysis where the PV is in ASP. In 2.1, I have argued that visual PVs can be stative/evidential (in the case of the past tense) or non-stative in English. The structure for English statives is auxiliary-like, and I claim that saw occupies ASP. Below, I’ll show it starts in v and then moves to ASP. PVCs are reduced in English and Dutch, i.e. ASPP in Dutch and English non-statives and vP/VP for English statives, as in (4) above. 2.2 Additional evidence I now present two kinds of evidence in favor of (25), in addition to the complementation data just given in (28) to (36), and the lack of passives, as in (37). First, de Haan (1997: 5) argues that evidentials (the cases he discusses are different from ‘see’) cannot be in the scope of negation. If saw is an evidential when its complement has a perfective meaning, as I argue, the same incompatibility should hold and the evidential should not be able to be an evidential under negation. This is indeed so in (38) and (39), accepted by some native speakers. According to these native speakers, the interpretation is not exclusively that the crossing/drowning is finished. The negation forces the PV to be a regular verb:6 (38) I didn’t see him cross the street in a normal way, but I saw him crawl across. (39) I didn’t see her drown but someone else did and rescued her.

Other native speakers find that the bare infinitive is awkward and they would use an -ing complement only. For them, negatives cannot force evidentials to be


verbs when bare infinitives follow them. The data in (38) and (39) support the claim that saw followed by a bare infinitive with a perfective meaning, as in (4a), is an evidential since the aspectual interpretation shifts with a negative. A second piece of evidence in favor of (25) comes from an old and oftendebated problem, namely the different constituent structures of (4a), (5), and (7) above. Akmajian (1977) argues, on the basis of preposing and clefting, that the structures for (4) and (7) are quite different: in (4), the DP and infinitive are separate constituents; in (7), they are not. Thus, (40) is grammatical but (41) is not (with the intended structure of a subject the moon and a verb rise, and not of a compound moonrise): (40) It was [the moon rising over the mountain] that we saw. (41) *It was [the moon rise over the mountain] that we saw.7

Applying preposing to Dutch (42), the result is (43), where the infinitive patterns with the -ing in English (but see Koster 1987: 131 who considers similar constructions “highly ungrammatical”): (42) We zagen de maan door de bomen schijnen we saw the moon through the trees shine ‘We saw the moon shining through the trees.’ (43) [De maan door de bomen schijnen] is wat we gisteren zagen [the moon through the trees shine is what we yesterday saw

Using structure (25), with saw in ASP in ModE, the differences between (40) and (41) are readily accounted for. According to (25), when saw is in ASP, the subject we moves to Spec IP and its trace inside ASPP would not be bound when the ASPP preposes in (41). With -ing complements, as in (7), bare infinitives in (20) to (24), and in Dutch (5), the structure would be as in (44), i.e. with see less grammaticalized, we the subject of the higher clause, and the structure bi-clausal. In (44), rise moves to ASP and the moon to Spec ASP but we is never inside the ASPP and it can therefore propose, as in (40) (I have simplified the VP-shell for convenience):


(44)

VP V′ we

V

ASPP ASP′

saw ASP

VP V′

-ing

the moon

V

PP

rise

over the mountain

The difference between (25) with saw in ASP and (26) and (44) with saw in a higher clause accounts for the data in (40) and (41). In (40), the ASPP would move and the trace of the subject would be c-commanded by we; in (41), the trace of we would have preposed inside ASPP and would not be properly bound by we. Thus, the crucial difference is either having the subject of saw and saw inside or outside of ASPP. If sentences such as (20) to (24) are different from (4a), the preposing should be grammatical as well, and it is, as (45) and (46) show: (45) [The moon rise over the mountains] is what we missed seeing. (46) [The moon rise over the mountains] is what we wanted to see.

In (45) and (46), see is not stative/evidential and therefore, the complement can be preposed. We need to make sure that moon and rise are not construed as a compound noun but as the subject and verb.8 Thus, the difference between (41) on the one hand and (40), (45), and (46) on the other provides extra evidence for the correctness of (47) and (44). So far, I have given several pieces of evidence that stative saw in English in structures such as (4a) has a structure as in (25) above. It is possible to suggest a slight modification of (25), as I have mentioned in 2.1, and have saw start out in v and move to ASP, as in (47):


(47)

IP I′ I

ASPP ASP′ ASP

vP v′

[+pf]

v

VP

she

V′ saw him

V

DP

cross

the street

In (47), the subject would receive a theta-role, Experiencer, from saw in v, slightly more elegant than she receiving a theta-role from an element in ASP, as would be the case in (25). If (47) is correct over (25), saw is still merging relatively early, and (25) would be the next stage. What would the evidence be that English saw is base generated in ASP, as in (25): if it could only be complemented by a VP rather than a vP, saw would be in v not ASP. This is impossible to test since all light verbs can be complemented by another light vP, as in (48): (48) I made him make (her make) it roll down the hill.

For lack of empirical evidence, I will not decide between (25) and (47). (47) is an in-between stage with saw checking perfective features in ASP; (25) is the stage expected by Late Merge with saw base generated in ASP. 2.3 Other languages with a PV split Sentences (4a) and (7) above show a split in complement selection which is present in many languages, as shown in (10) to (13). Many people, e.g. Kuno (1973, Chapter 18), Dixon (1988: 38), have argued that PVs have complements different from other verbs. In Japanese, the complementizer is different; French


allows clitic climbing, indicating that there is a close connection between the PV and its complement; and English has to-less infinitives. There is also a split in whether PVs are used to express activity or experience. In English, watch and look are used for the former meaning whereas see is used mainly for the latter but sometimes for activity as well. In this subsection, which is mainly based on Viberg’s (1983) observations, I will look at this split a little more. It can be shown that in many languages, English included, stative or evidential PVs, such as saw in (4a), repeated here as (49), are lexically different from non-stative ones (activity based), such as see in (20), repeated here as (50), and look at (Viberg discusses a third kind but I will leave that out): (49) I saw him cross the street. (50) Seeing her swim is exciting.

Viberg provides data from a number of languages, but he is predominantly interested in the different lexical realizations, not what the nature of the difference is or what kind of complement there is. Certain languages form stative and non-stative variants through serial verbs (e.g. Vietnamese and Mandarin Chinese); others through compounds consisting of a noun and a (light) verb, as in Farsi. In the table Viberg provides (1983: 131), it is the activity verbs that have the compound form. The light verb that is included in the compound is typically kardan ‘do’ in Farsi, emphasizing the imperfectivity of the activity-based verb (even though in Farsi light verb constructions are typically more varied, e.g. harf zadan ‘speak’, literally ‘letter hit’, yod gereftan ‘learn’, literally ‘memory get’). PVCs in Farsi (cf. Lambton 1953: 155) are not infinitival, like English, but clausal, as in (51), even though (52) is very interesting with ura ‘him’ raised out of a finite clause: (51) didim ke inja hastand we-saw that here they-are ‘We saw they are here.’ (from Lambton, p. 155) (52) ura didam ke miraft him I-saw that is-going ‘I saw him going.’ (from Haim’s Larger Persian English Dictionary, s.v. didan)

So, from (51) and (52), it appears that didan ‘see’ in Farsi is not grammaticalized into an evidential. For Hindi (and the same holds for Urdu), Viberg lists dekhna (p. 133) as the equivalent for both ‘look at’ and ‘see’. However, even though dekhna ‘see’ can

191


be used as both (as didan can in Farsi, and see can marginally in English), there are many noun-verb compounds for the activity-based verb ‘see’, namely nazer kerna, malum kerna, deryaft kerna, and nagah kerna (see Sant Singh’s Practical Dictionary). The nouns that are part of the compound in Hindi/Urdu and Farsi are most often loanwords from Arabic and in the case of Hindi/Urdu from Farsi as well (see Platts’s Dictionary) whereas dekhna ‘see’ and karna ‘do’ have cognates in Sanskrit. In Hindi/Urdu, dekhna can be complemented by a present participle, as in (53), (comparable to the English -ing form) or by a past participle, as in (54) (comparable to the English bare infinitive): (53) meN ne use bezhte hue dekha I erg him sitting be see-past-masc ‘I saw him (in the act of) sitting down.’ (54) meN ne use bezhe hue dekha I erg him sat be see-past-masc ‘I saw him (in the state of) sitting down.’ (both from Barker II: 35, with N indicating that the preceding vowel is nasal, and z that the alveolar stop is retroflex) In informal speech (Anju Kuriakose p.c.), an infinitive is used as well, as in (55), but intuitions differ as to the exact (aspectual) interpretation, i.e. whether (55) has an interpretation similar to (53) or to (54): (55) meN ne use jane dekha I erg him go-inf see-past ‘I see him go/going.’

The conclusion is that the simple verb (see, didan, dekhna) is often used for experiencer based, i.e. stative, meanings. Since a PV in Farsi and Hindi/Urdu has a clausal complement, it has not grammaticalized as in English. The reason for this is that the difference between experiencer and activity based ‘see’ is expressed in another way, namely through compounds. Compounds are most often used for activity based meanings with the light verb emphasizing the imperfectivity. I now turn to the development of words such as see, watch, look at in English. If ModE saw is indeed grammaticalized into an evidential marker, as I have claimed in (25), it may be the case that the non-stative PVs show some lexical differences, as they do in Hindi/Urdu and Farsi. As is clear from reflecting on (20) to (24), the preferred non-evidential PV is watch or look at, not see. In fact, (19) above and (56) are somewhat marked:


(56) ?I am seeing that problem for the first time.

According to the OED, watch (or rather its unpalatalized form wake) means ‘be awake’ and ‘remain/keep awake’ in OE. By 1200, it acquires the meaning of ‘be vigilant’, and by 1600 or so, it acquires the modern meaning of observing someone, as in (57): (57) Ill watch Titania, when she is asleepe, and drop the liquor of it in her eyes. (Shakespeare, Midsummer Night’s Dream II, i, 177)

Observe and perceive are straightforward loans, the former being a late 14th century loan with the initial meaning of ‘obey, follow’ and the latter being an early 14th century loan. Peer at, glance, stare are all quite specialized forms of non-stative PVs, and in OE, look at means ‘direct one’s sight to’, according to the OED. Thus, in OE, the general perception verb for both stative and nonstative perception is see. The addition and changes of perception verbs, I claim, makes the grammaticalization of see as a perception verb possible. In conclusion to Section 2, I have argued that ModE has two kinds of ‘see’, one moves to an ASP as in (47), and one is a verb, as in (26). I have also provided evidence that other languages distinguish between two kinds of PVs, a stative or evidential one and an activity one. I’ll now turn to OE and show that the evidential use is not there yet, and that older English patterns with the other Germanic languages.

3. OE and ME I’ll now show that the situation in OE and ME is comparable to that of the Germanic languages other than ModE, discussed in Section 1. The reason for going into this is to see how saw has grammaticalized. In OE and ME, the bare infinitive occurs in the PVC, as in (58) and (59), with an -an ending. A present participle also occurs with an -inde/-ende ending, as in (60) and (61), or -ande and later in -ing: (58) Ic seah turf tredan I saw earth tread ‘I saw earth being walked on.’ (Riddle 14, from Callaway 1913: 35) (59) ða of wealle geseah weard Scildinga … beran ofer then on wall saw guardian of-Shieldings being-carried over


bolcan beorhte randas gangplank bright shields ‘When from the walls, the watchmen of the Shieldings saw shields glittering on the gangplank.’ (Beowulf 229–31, from Callaway 1913: 34) (60) se hælend … seah hia hremende 7 uoepende ‘The savior saw her weeping and weeping.’ (Lindisfarne Gospel, John 11, 33) (61) He seye … a grom cominde ‘he sees a man coming.’ (Guy of Warwick 5799, from Visser, p. 2344)

Interestingly, there are more instances of geseon and gehieran in Callaway than of seon and hieran (namely 7, 20, 2, and 10 examples respectively). This indicates the PV is connected to inner aspect in E(xtent), as in (62): (62)

EP EP′ E ge

VP V

...

seon

beran

Later, as ge- is lost and E is reanalyzed as ASP, the PV can move there but it ‘competes’ with to, -ing, and the past participle, as I’ll come back to below. Many people argue that the present participle, as in (60) and (61), is not “a native idiom” in OE but appears in texts that are translated from Latin (see e.g. Callaway 1913:225–30; Miller 2002:271ff.). Sentence (60) is indeed an interlinear gloss. Sentences such as these would have the same analysis as (7) in ModE with see having a sentential complement, and whether or not they are native or not is not really important for that argument. Some people, e.g. Miller (2002), have argued that the difference in aspect between constructions such as (4) and (7) is already present in Late OE. Mossé (1962) and others argue there isn’t much difference in meaning between the infinitive and the participle; they are competitors. I will only look at ME but argue that the difference wasn’t there, and hence was very unlikely in OE as well. See Zeitlin (1908: 72) for a nice list of examples of both. Unlike their ModE counterparts, the ME bare infinitive constructions from Chaucer in (63) to (66) have imperfective interpretations. For instance, in (63),


to and fro indicates duration, and (65) and (66) could not be translated into ModE without an -ing. This makes them similar to the present participle complements of ModE (7): (63) The fairnesse of that lady that I see | Yond in the gardyn romen to and fro | Is cause of … ‘The beauty of that lady that I see roaming in that garden is the cause of …’ (Chaucer, Knight’s Tale 40.1098–9, from Kerkhof 1966: 55) (64) The deeth he feeleth thurgh his herte smyte ‘The death he feels cutting through his heart.’ (Chaucer, Knight’s Tale 42.1220) (65) I sawgh hyr daunce so comlily | Carole and synge so swetely | Laughe and pleye so womanly ‘I saw her dancing so becomingly, dancing and singing so sweetly, laughing and playing so womanly.’ (Chaucer, Book of the Duchess 340.848–50) (66) But … in hande I saw hym holde | Two firy dartes … ‘But in his hand I saw him holding two fiery arrows …’ (Chaucer, Legend of Good Women 594.166–7)

There is also an -ynge after perception verbs, as in (67): (67) And saw his barge saylynge in the se ‘And saw his barge sailing in the sea.’ (Chaucer, Legend of Good Women 624.2196)

Miller (2002: 265f.), based on arguments from Fischer (1995) and Kerkhof (1966), argues that in Chaucer the aspectual difference is as in ModE. Based on (63) to (66), this is unlikely. Miller does have an interesting set of examples that shows there is some sense for the difference, even in OE, as in (68). He argues that in (68) the standing is not being perceived but the ascending is: (68) ða geseah he on swefne standan ane hlæddre fra eorþan to heofonan then saw he in dream stand a ladder from earth to heaven 7 Godes englas upp stigende and God’s angels up ascending ‘Then he saw in a dream a ladder stretch from earth to heaven and God’s angels climbing up.’ (Genesis 28.12, from Miller 2002: 263)

I will argue that even though there is in older English some evidence that the infinitive differs from the participle, it is not expressed in the grammar.


In More’s English (Visser 1946–52), from the beginning of the 16th century, some infinitives have a perfective meaning resembling ModE, as in (69) with it, with hear, but not all do. Sentence (70) emphasizes the action through the adverb so faste and ModE would use running. This text has many -ing forms too (see Visser 1952: 810): (69) She hard him boast it ‘She heard him boast about it.’ (Apology 489 E 8) (70) The fox … saw him run so faste ‘The fox saw him running so fast.’ (Richard 71 C 1, from Visser, 1952: 761–2)

In the First Folio Edition (1623) of Shakespeare’s works, there are only two -ing complements, after an approximate 2000 forms of ‘see’. One occurs after see and one after saw, namely (71) and (72), but many bare infinitives occur, as in (73) to (75), with imperfective meaning. There are also some infinitives with to, as in (76), and many past participles, as in (77): (71) who you saw sitting by me on the Turph. (As You Like It III, IV, 52) (72) may you see it comming. (Macbeth V, v, 37) (73) to see thee weare thy heart in a scarfe. (As You Like It V, ii, 23) (74) when she sees me worke. (Tempest III, i, 12) (75) When shall you see me write a thing in rime. (Love’s Labor’s Lost IV, iii, 181) (76) I saw her corall lips to moue. (Shrew I i 175) (77) To see me thus transformed to a boy. (Merchant II vi 39)

Just to get some sense for the auditory PVs, there are 1200 or so forms of ‘hear’ in the First Folio edition. Instances of bare infinitival complements after ‘hear’, as in (78) to (81), are very frequent: (78) and another Storme brewing, I heare it sing ith’ winde. (Tempest II, ii, 20) (79) harke, do you not heare the people crie Troylus? (Troilus & Cressida I, ii, 244) (80) I heard the Owle schreame, and the Crickets cry. (Macbeth II, ii, 16) (81) Me thought I heard a voyce cry, Sleep no more. (Macbeth II, ii, 35)

As with ‘see’, these are different from ModE since the bare infinitive occurs after a present tense form in (78) and (79) and because the infinitive need not be


perfective. Five occur with -ing complements, two in Hamlet, as in (82) and (83), two in King Lear, and one in First Henry IV. Four of these have the same verb in the complement. (Note the variable spelling): (82) I heare him comming. (Hamlet III, i, 55) (83) Withdraw, I heare him coming. (Hamlet III, iv, 7)

Thus, as late as Shakespeare, the infinitive continues to be used as a non-perfective to a PV and after a present tense form of the verb see (and hear). It is also more frequent than the -ing complement, unlike the ModE data given in (1). Participles, as in (84), (88), and (89), and to-infinitives, as in (85) to (87), occur as well. These were briefly mentioned in Section 1 (examples (1c), (2) and (3)) as being archaic in ModE: (84) You heare all these matters deni’d. (Merry Wives I, i, 193) (85) Who heard me to denie it? (Comedy of Errors V, i, 26)

Both these constructions are frequent from later ME on, as in (86) to (89), and seem especially frequent in the 16th and 17th centuries: (86) Whanne God had seen the erthe to be corrupt. (Wyclif, Genesis VI 12) (87) You dwell, (said he,) in the City of Destruction, … I see it to be so. (Bunyan, Pilgrim’s Progress 12/11) (88) And tolde hym al, as ye han herd me sayd ‘And told him all as you have heard me say.’ (Chaucer, Franklin’s Tale 188.1547, from Visser 1952: 761) (89) Whan Troilus hadde herd Pandare assented | To ben his help ‘When Troilus had heard Pandare assent to be of help.’ (Chaucer, Troilus 487.1009–10, from Visser 1952: 894)

In the early 15th century writings by Pecock (cf. Zickner 1900: 67), see is complemented by a bare infinitive only twice but by to four times and forto once. So, at the time the infinitival ending is disappearing, the to-infinitive expresses durativity for a while. Bunyan’s seventeenth century English has both the reduced auxiliary, discussed in Chapter 8, quite a lot of progressives (Brusendorff 1930: 243) indicative of an ASP, and constructions such as (87). The latter construction also occurs with modals only if the modal is very far removed from the infinitive (Visser 1952: 590; 620). The past participle expresses perfective aspect. So rather than be limited to one interpretation, ASP after PVs is broader. In OE, the prefixes on the verbs determine perfectivity. As the


prefixes disappear, -ing (and to) are reanalyzed as imperfective and for a while the past participle, as in (84) and (88), is a perfective. At what point does the ModE situation, with a structure as in (25) or (47) arise? Is it a relatively sudden change or do old forms peter out? To answer that question, I have looked at later texts, an 18th century and a 19th century one. It is sometimes difficult to interpret whether the complement is perfective or not but it is not difficult to see if the present tense of the PV is used, as is no longer grammatical (see (4b) above). The 18th century text (from 1710), A Treatise Concerning the Principles of Human Knowledge by George Berkeley contains quite a number of instances such as (90) and (91): (90) But when we see things go on in the ordinary course, they do not excite in us any reflection. (Luce & Jessop II, Principles 57) (91) And I would fain see any one explain any the meanest phenomenon in Nature by it. (Principles 72)

Berkeley’s language is ‘modern’ enough not to have needed a modernized edition. However, I think instead of the infinitive a present participle would have been used. The picture is very different in the 19th century novel, Emma by Jane Austen. There is only one bare infinitival complement after see, as in (92), but here see is infinitival not present tense, so it is not like (4b). There are many such infinitives after saw, as is the case in ModE, as in (93). The -ing complements all forms of ‘see’, as shown in (94) and (95). There is also an infinitive with to, as in (96): (92) She was delighted to see her father look comfortable. (Emma, I, Chapter 3) (93) saw her go away in the evening attended by … (Emma, I, Chapter 1) (94) saw Frank Churchill looking intently across the room at Miss Fairfax. (Emma, II, Chapter 8) (95) I cannot see you acting wrong. (Emma, III, Chapter 7) (96) and it was not long before he saw it to be Dixon. (Emma, III, Chapter 6)

Thus, by Jane Austen’s time, apart from (96), the situation is ‘modern’ with respect to (4) and (7) above. In Chapter 10, I show there is another change occuring at exactly the same time. There I argue that the unmarked setting of ASP changes from non-bounded to bounded. That resetting would have resulted in (25) or (47), with saw in ASP checking perfective/boundedness.


4. Conclusion In this chapter, I have argued that the ModE perception verbs that have a (perfective) bare infinitival as complement are evidential and move to ASP from v, as in (47), or are base generated in ASP, as in (25). This analysis accounts for certain preposing facts, for the lack of passives, as well as for the aspectual interpretation of (4). The situation in older English is different and points towards see being a verb, often with an inner aspectual marker, as in (62). The reanalysis that takes place is again one where the grammaticalizing element moves higher in the tree. It is similar to that discussed in the previous chapter since the inner aspect is replaced by an outer aspect. In addition to the bare infinitival PVC, there is a complement with an ing-affixed verb. The structure of this is bi-clausal, one where the perception verb is verbal. Thus, visual PVs are of two kinds, activity and experience, in accordance with Viberg (1983). The former is less grammaticalized and is usually not expressed by see but by verbs such as watch and look. The two kinds occupy different positions and allow different interpretations. With activity-based see, the complement is typically not perfective, but marked by -ing. Assuming that -ing occupies ASP (or is checked in ASP), one could argue it is in complementary distribution with saw. If saw occupies ASP, the verb in its complement is perfective since it is dependent on saw. In sentences such as (4), the verb is always past tense, and I am calling this use evidential. Apart from accounting for the perfectivity constraint, other advantages of the proposed structures in (25) or (47) are an explanation for the different constituent structure of (40) and (41), and the difference between (4) versus (20) to (24), both in terms of perfectivity and constituent structure. Sentences such as (4a) are grammaticalized: saw behaves like an auxiliary. In Dutch and English up to the 18th century, zag/saw is not grammaticalized (as far as it is in ModE). I argue the gradual changes in aspect marking show that perception verbs undergo grammaticalization from main verb to evidential modal and that the change involves a reanalysis of V to ASP. However, it continues to be possible to use the visual perception verb in its non stative use. ASP has been available for a long time for saw to move to (to check perfective) but not until the 19th century did that happen.

Chapter 10

Aspect The Tense Aspect Parameter and inner to outer aspect1

In the last two chapters, I have examined some verbal elements that change to become aspectual markers in accordance with Late Merge. This current chapter provides a general picture of the changes in aspect in English and of the interpretation of the present tense that takes place in the 19th century. I argue that the setting of a parameter is switched from having aspect as unmarked to tense as unmarked. The implication is that aspect is replaced by tense. This change is fast, as expected of parameters. The chapter also examines changes in aspectual prefixes, such as ge- and for-. The loss of these shows that inner aspect is replaced by outer aspect in the form of have and -ing. The inner aspect in turn is ‘replenished’ by new phrasal verbs. The outline is as follows. In Section 1, I will briefly describe the data on the use of the simple present and progressive in a variety of Germanic languages and provide an account for the differences. Then, in Section 2, building on Chapter 8, I’ll look at the changes involving the demise of the aspectual prefix and the introduction of the progressive in the history of English, and in Section 3, at another aspectual marker, namely do. In Section 4, I will examine the development of inner aspect, and in Section 5, Giorgi & Pianesi’s (1997) theory, since they provide a competing theory to account for the changes. The Appendix formulates an account of English aspect in a Reichenbachian framework.

1.

The simple present and the progressive: The Tense Aspect Parameter

In 1.1, I review the present tense system of ModE. It differs from other languages in having to mark the simple present as progressive in aspect. Cowper’s (1999) theory presents a way to account for this. In 1.2, I compare ModE to other Germanic languages and show that, even though the latter do not have to mark the progressive, they can do so.


1.1 ModE As is well-known, in ModE, it is impossible to have (1). The simple present of activity (e.g. run) and accomplishment (e.g. eat an apple) verbs, i.e. durative verbs, results in a habitual reading, incompatible with a time adverbial such as now. Such a reading is not the result for state (e.g. see, know) and achievement (e.g. reach the top) verbs, as (2a) and (2b) show respectively: (1) a. *I run right now. b. *I eat an apple right now. (2) a. I know it right now. b. I reach the top right at this moment.

For durative verbs, the progressive is used, as in (3), which is ungrammatical for non-durative verbs, as (4) shows (achievements can be made into accomplishments by using an ing): (3) I am running/eating right now. (4) *I am knowing it right now.

Some recent accounts, for instance, Ramchand (1997) and Cowper (1999), have provided analyses using ASP for present tense constructions. Cowper (1999: 218), using the Davidsonian event argument, argues that “languages choose either MOMENT (perfective) or INTERVAL (imperfective) as the unmarked representation of events … In English the unmarked value of e is MOMENT, while in French it is INTERVAL. While English has inflectional morphology making sentences imperfective, French has inflectional morphology making sentences perfective”. In addition, Cowper needs a (universal) principle excluding two simultaneous temporal points, and a discourse anchor which is a point/moment. In French, the simple present, as in (5), denotes an interval and since the event takes place at the same time as the moment of discourse, the constraint is not violated. In the English translation in (6), the simple present is a moment and so is the discourse. This is excluded by the constraint: (5) Elle écrit une lettre. (6) *She writes a letter.

With special morphology, a marked form is possible, i.e. -ing in English indicates that e is interval. In Cowper’s model stative sentences lack e. Her model builds on the work of Smith (1991) and van Voorst (1987). What I am most interested in are the parametric differences between languages. In ModE,

Aspect 203

tense is the default. This is done by considering points in time basic and needing to mark periods of time with special morphology. French, Dutch, German, and English before the 19th century, in contract, are aspect based and their default is a period of time, as we’ll see in the next subsection. 1.2 The progressive in other Germanic languages If English has moment as unmarked, the other Germanic languages have interval as the unmarked option, as in French, since the simple present is grammatical. This is shown for OE in (7), for ME in (8), for Early Modern in (9), for Dutch in (10), and its German counterpart in (11): (7) nu ic arisu cwið drihten now I rise said lord ‘Now I rise up said the lord.’ (Vespasian Psalter 11.6, Visser 663) (8) What do ye, maister Nicholay? what do you master Nicholay ‘What are you doing, master Nicholay?’ (Chaucer, Miller’s Tale 71.3437) (9) What say you, Scarlet and John? ‘What are you saying, Scarlet and John?’ (Shakespeare, Merry Wives I, i, 155) (10) Wat doe je? Ik eet een appel what do you I eat an apple ‘What are you doing? I am eating an apple.’ (11) Was machst du? Ich esse einen Apfel.

In accordance with Cowper’s system, there would have to be ways in these languages to mark perfectivity (inner aspect), and there are, e.g. by verbal prefixes such as in- and door- in Dutch. There are, however, also special ways to mark continuous action (outer aspect) by means of a preposition and a (durative) infinitive, as in (12), from Dutch, (13) and (14) from OE, (15) from ME, and (16) from 17th century English. These markings are optional since the unmarked case is interval. As in ModE, in Dutch, as (17) shows, and in older English, none of these involve states: (12) Ik ben een cactus aan het planten I am a cactus on the plant ‘I am planting a cactus.’


(13) Exorciste beoð on getacnunge Cristes gespellan Exorciste is on teaching Christ’s story ‘Exorciste is teaching the story of Christ.’ (De Eccles. Gradibus) (14) on feohtende wæron oþ niht on fighting were until night ‘(they) were fighting until night.’ (Anglo Saxon Chronicle C, D, E, 871 Thorpe 1861: 138–9) (15) þær he wes an slæting there he was on hunting ‘he was hunting there.’ (Layamon, Caligula 6139) (16) I am upon writing a little treatise (Pepys’s Diary 31 Dec 1666, from Visser 1998) (17) *Ik ben dat antwoord aan het weten I am that answer on the know ‘I know that answer.’

In addition to the forms with a preposition, there is, in OE, a form in -ende (or -ande or -inde depending on the dialect) that resembles the modern construction in (3) above, namely (18) and (19), and in ME, one in -yng/-ing, as in (20). However, many of these, e.g. (19), cannot be translated with an -ing in ModE or an aan het construction in Dutch: (18) ac se æglæca ehtende wæs but the monster pursuing was ‘but the monster was chasing.’ (Beowulf 159) (19) þe þer were wuniende who there were living ‘who were living there.’ (Lambeth Homilies 41, from Mossé 1938, I: 81) (20) We han ben waitynge al this fourtenyght ‘We have been waiting these two weeks.’ (Chaucer, Knight’s Tale 38.929)

It is often argued that in OE these participles are adjectives (cf. Mossé 1938, I: 3) and that the constructions involve copulas followed by adjectival forms. The form in -ing in (20) is a ME innovation, but whether it is a direct continuation of the -ende form is contested (see Mossé 1938: II, 36 and Jespersen 1940: 415). In Chaucer, it is not that frequent after a form of to be (it is frequent as a verbal noun and adjective). Some evidence that the two constructions are connected is the fact that in many texts, both forms occur, seemingly ambiguous:

Aspect 205

(21) Where þe dragun was wonande where the dragon was living ‘where the dragon lived.’ (Handlyng Synne 1760) (22) Wher þe old man wonyng was where the old man living was ‘where the old man lived.’ (Handlyng Synne 8504)

However, the OE -ing/ende form occurs mainly with verbs of dwelling, as in (21), and movement, as in (18) (see also Pessels 1896; Raith 1951: iii) and I therefore agree with Jespersen’s (1931: 168) claim that the ME -ing is not a continuation of the OE one. There is no end to this debate, especially in the linguistic literature of the early 20th century. In Late ME, examples such as (23) show that the form with the -ing ending is (still) an adjective since the construction needs a Case marking of before the object. This preposition becomes necessary after adjectives (and nouns) cease assigning inherent Case: (23) thanne thei ben not scorninge of God but worschipyng then they are not scorning of God but worshipping ‘then they don’t scorn God but worship him.’ (Wyclif, Sermon, from Visser 2002)

This option is utilized only after 1400 and is rare. Visser (p. 2002) says that Wyclif is the first to have of before the object but gives examples of the participle not followed by a preposition. The construction becomes extinct at the beginning of the 20th century (Visser 1993ff). In the 15th century Paston Letters, the special progressive is relatively rare (around 20 perhaps in a corpus of over 250,000 words): (24) þer ys comyng, …, more than a thowsand there is coming more than a thousand ‘More than a thousand are coming.’ (Paston Letters #187, anno 1465) (25) where the seid felechep is abydung ‘Where the above-mentioned fellowship abides.’ (Paston Letters #40, anno 1452) (26) syche mony that he is owyng ‘such money that he owes.’ (Paston Letters #336, anno 1469)

These occur with the same verbs as in OE and most of those would not get -ing in ModE, e.g. (25), indicating that it is not the same construction, in accordance with Jespersen (1931) as mentioned above.


The present is typically expressed as in (27) to (29), with the latter possibly being fossilized: (27) I send you ‘I am sending you.’ (Paston Letters #3, anno 1425) (28) I make þis day a new apelle ‘I am making a new appeal today.’ (Paston Letters #4, anno 1426) (29) I recomaunde me to you ‘I commend myself to you.’ (Paston Letters #3, frequent formula in letters, anno 1425)

By the time of Thomas More, i.e. the early part of the 16th century, the progressive is “employed rather sparingly” (Visser 1946: 248). In Shakespeare, -ing is used on occasion, e.g. in (30), cf. also Franz (1909: 514). Visser (p. 662) says, about both More and Shakespeare that “at the time the choice between the two possibilities did not yet depend on any fixed principle”. The simple present is used frequently, as in (31) and (32): (30) Now she’s going to my wife. (Shakespeare, Merry Wives III, ii, 36) (31) Whether go you. … To see your wife ‘Where are you going.’ (Shakespeare, Merry Wives II, ii, 10) (32) But what saies shee to mee? (Shakespeare, Merry Wives II, ii, 75)

In Shakespeare, there are a number of cases where the -ing progressive is used with stative verbs such as live, as in (33), but not with know and see. So, it seems to me that -ing is definitely used by Shakespeare as a progressive, as in (33): (33) To see what Friends are liuing, who are dead (1 Henry 4 V, iv, 165)

So, OE, ME, German, and Dutch have no obligatory progressive, unlike ModE. Thus, Dutch (34a) and (34b) are variants, unlike English (35a) and (35b), their literal translations: (34) a.

Ik lees nu I read now ‘I am reading now.’ b. Ik ben nu aan het lezen I am now on the read ‘I am reading now.’

(35) a. *I read now. b. I am reading now.

Aspect 207

The pattern in (35ab) starts in the nineteenth century and is due to a resetting of the Tense-aspect Parameter. 1.3 From aspect to tense The simple present in older English, as in (31) and (32), is so much more common than the progressive that they can’t really be compared. So when does the Modern situation arise? The construction in (30) above increases especially in the 19th and 20th centuries, see Scheffer (1975: 110ff.), and this is the point at which interval changes to moment. Checking certain simple present and progressive forms (eat/is eating; run/is running) in Jane Austen’s early 19th century Emma, Persuasion, and Pride and Prejudice shows that by this time the use is modern, with the simple present being used for habitual aspect, as in (36) and (37), i.e. moment, and the progressive for interval, as in (38) to (40): (36) I dare not let my mother know how little she eats. (Emma II, ch 9) (37) whenever she comes to Highbury. (Emma I, ch 10) (38) he is writing about it now. (Persuasion ch 23) (39) At this moment … Mr Elton is shewing your picture … (Emma I, ch 7) (40) If you are looking for my master … he is walking towards the little copse. (Pride & Prejudice II, ch 7)

Additional evidence for the form in -ing being optional early on can be found in 17th and 18th century (pedagogical) grammars. Only the ones from the middle of the 18th century include the -ing form as a regular present. Wharton (1654) does not allude to a special present tense form ending in -ing, except in one sentence: “A Participle of the Present tens signifieth the time present, and endeth in ing; as loving, teaching…” (p. 54). Miege (1688), who was born in France, comments (p. 67) “[l]astly, ’tis to be observed, that the English has a peculiar Way of using the Verb to be, with a Participle of the Present Tense. As, I am writing for I write…”. Lane (1700) comments: “[t]he Auxiliaries of the Present Tense, are, do, dost, …; am, art, …; as I do call …; I am calling” (p. 44). Duncan (1731) and Fisher (1750) clearly imply that the -ing is a present tense; Greenwood (1711) seems to suggest the use is optional. As late as 1771, Fenning writes: “the present tense represents the action as now doing, without any other limitation; as I sup, that is, I am now at supper” (1771: 74), but the example is as in Greenwood, so may have been copied and no longer be correct in the late 18th century.


Considering this, one can say that the unmarked aspect (encoded in the speech situation, see the Appendix) changes to moment in the 19th century. This is a fast, parametric change, that can be formulated as in (41), where [−bound] represents the interval and [+bound] the moment: (41) Tense Aspect Parameter Change ASP » ASP [−bound] [+bound]

2. Changes in ASP Building on the structural changes that modals and PVs are argued to undergo in the previous two chapters, I now go into more detail on how the inner aspect (expressed in OE and early ME by ge- and for- etc) is replaced by outer aspect around C15 (with have for the perfective and optionally -ing for the imperfective). Mossé (1938, II, 2ff.), based on Streitberg (1891), attributes the popularity of (3) to the demise of the aspectual system occurring from the 13th century on. In OE and Germanic, the simple verb is durative, whereas verbs with prefixes (he calls them ‘préverbes’) are not. The same is true in Slavic languages (cf. Comrie 1976 [1987]). To give a Dutch example, slapen ‘sleep’ is durative, whereas inslapen ‘fall asleep2’ is not. Mossé’s formulation is similar to Cowper’s that I adopt above in which the imperfective is unmarked in Germanic. So, the prefixes are perfectivizing, and have is not generally used for perfective in OE, but a prefix is. Brinton (1988: 202 ff.) argues that OE verbal prefixes indicate telicity, but that ge- has become “seriously over-extended” (p. 212) by ME. In Gothic, the prefix is on occasion an independent morpheme (see Lenz 1886: 11) separated from the verb by a question particle or clitic. Mustanoja (1960: 446) writes that ge- remains productive as a perfectivizer “down to the 13th century”. Its disappearance is due to Norse influence and occurs first in the North according to Mustanoja. Chaucer only uses a limited palatalized version in the poetry. For retains its productivity up to the end of the ME period. The verbal prefixes are replaced by particles and phrasal verbs extend their domain in ME and become less marked (Brinton 1988: 226). There are many different opinions regarding ge-. Hiltunen (1983: 56–6) says that ge- is often meaningless in OE. If ge- is indeed meaningless, the construction using -ing cannot be its direct replacement because of the time gap, another indication that (22) is not the source for (3). Around 1400, ge- is lost

Aspect 209

completely as a marker of perfective. Is there a replacement, as Traugott (1972: 91) suggests? I will argue that there is a replacement and that as a result of the demise of ge-, the ASP position is rather ‘crowded’, to use a structuralist term. This is mentioned in Chapters 8 and 9. In ME and later, the past participle ending, as in (42) and (43), infinitival to, as in (44) and (45), and increasingly the present participle ending in -ing express aspect, each with a different interpretation (these sentences are repeated from Chapter 9, Section 3): (42) And tolde hym al, as ye han herd me sayd ‘And told him all as you have heard me say.’ (Chaucer, Franklin’s Tale 188.1547) (43) Whan Troilus hadde herd Pandare assented to ben his help ‘When T had heard P agree to be of help.’ (Chaucer, Troilus 487.1009–10) (44) Whanne God had seen the erthe to be corrupt.3 (Wyclif, Genesis VI 12) (45) I see it to be so. (Bunyan, Pilgrim’s Progress 12/11)

The latter is a frequent construction from later ME on, especially in the 16th and 17th centuries. According to Zickner (1900: 67), 15th century Reginald Pecock has see complemented by a bare infinitive only twice, as in (46), but by to four times and forto once. However, checking the text (Babington 1860), all cases of (for)to infinitives are either as in (47) or (48), i.e. not direct complements or passives: (46) alle men whiche schulen se him go or meete with him ‘all men that should see him go or meet with him.’ (Repressor, 238.1) (47) and thei schulden se and knowe weel hem silf to be fonnys and foolis ‘and they should see and know well themselves to be fools and fools.’ (Repressor 129.6) (48) tho chapitres schulen be seen forto not repugne bitwixe hem silf ‘Those chapters should be seen as inconsistent between them.’ (Repressor 408.22)

The role of to and the past participle shows that rather than being limited to one interpretation, ASP is broader, as also explained in the previous chapter. In OE, the prefixes on the verbs determine inner aspect or perfectivity. As the prefixes disappear, -ing, as in (20), and to, as in (44) and (45), are seen as imperfective and for a while the past participle, as in (42) and (43), is a perfective. In Chapters 8 and 9, I have argued that this shift involves a reanalysis of the


inner to the outer aspect (see Section 4 also). This is represented by the shift from (49) to (50). Note that the label for inner aspect is E(xtent), see Chapter 7 and Section 4 below: (49)

vP v′ EP

v′

E′ E for/ge

(50)

VP V

...

ASPP ASP¢ ASP

vP v¢ v

Ï to/ing ¸ Ì have/modal ˝ Ó ˛

VP V

...

This shift takes place at the end of the ME period. It is not until the 19th century that the other shift takes place, namely from interval to moment. So, two major aspectual changes take place. By the 19th century, the pattern in (1) and (3) of obligatory -ing is fixed (see Section 1) as well as that of a perfective interpretation of the infinitive after PVs (see Chapter 9). These changes are the result of the parameter switch from interval to moment. The earlier changes, loss of ge-, optional -ing, to, changes in phrasal verbs, and other aspectuals (see next section) result from the reanalysis of the lower (inner) aspect to a higher one, as in (49) to (50), with different words ‘competing’ for the same position. This is due to Late Merge. A third change involves the shift of a past tense reading of have (as in (56) to (58) of Chapter 8) to one specially marked for unbounded aspect.

Aspect

3. Other aspectual markers: do? So far, we have seen modals and PVs becoming located in ASP, ge- disappearing, and -ing being introduced. To after the PV might be in ASP as well. In this (short) section, I examine the role of do. There has been no shortage of work on the auxiliary do, but not as an aspect marker. Do, as is well-known, grammaticalizes from a full verb, to possibly a causative verb, as in (51), to a marker of negation, interrogatives, ellipsis, and emphasis: (51) preyng you þat ye wole do them spede them praying you that you will cause them speed them(selves) ‘asking you to cause them to hasten.’ (Paston Letters, from Denison 1993: 257)

The basic change is thus from V to (causative) v to I. As Denison (1993: 281) argues: “[t]he new DO + infinitive construction could be seen as an experimental form of Aktionsart marking”, i.e. a perfective marker. In the Early ModE examples from the HC, there is a clear habitual aspect, as in (52) and (53), both from the middle of the 16th century: (52) for all men doo thynke that they have well deserued ‘because all men think that they have deserved well.’ (Colville’s Boethius, p. 110) (53) it may well appeare, that the braine doth gouerne the tongue ‘It may well appear that the brain governs the tongue.’ (Ascham’s Scholemaster, p. 185)

Around 1700, however, an interesting switch occurs. Contemporary grammars provide evidence for this possible change. Miege (1688), as quoted in the previous section, gives as the present tense I love or I do love. Lane (1700), as mentioned, says: “[t]he Auxiliaries of the Present Tense, are, do, dost, …; am, art, …; as I do call …; I am calling” (p. 44). Duncan (1731: 22) provides for the present tense the following: I carry, or I do carry, or I am carrying. Lowth (1762: 40) remarks: “Do and did mark the action itself, or the time of it, with greater distinction”. This seems to suggest that around 1700, -ing and do are in complementary distribution, with -ing more common. I will argue that do, like -ing, indicates interval, but only at the time that it switches, which seems to be a little after 1700, if the grammars are right. In George Berkeley’s 1710 Treatise Concerning the Principles of Human Knowledge, habitual aspect, or generic statements still occur with do, as in (54):

211


(54) a body moving in a place which is in motion doth participate the motion of its place. (Luce & Jessop, II, Principles 111)

By the time of Jane Austen, auxiliary do is used in its ModE forms. It indicates questions, as in (55), negatives, and emphatics: (55) Why do you smile? (Emma I, ch 8)

Do expresses aspect until the beginning of the 18th century, as in (54), and is then in ASP, like have and others. Now, I’ll examine some other characteristics of inner aspect.

4. Inner aspect: The E(xtent) Phrase In Chapter 7, I mentioned that phrasal verbs in ModE provide some evidence for inner aspect because particles occupy that inner aspect position. There, I also introduced the term E(xtent) Phrase to indicate boundedness and definiteness. English phrasal verb particles are similar to particles in languages such as Dutch where eten ‘eat’ is imperfective but opeten ‘eat up’, with a prepositional prefix, is perfective (cf. also Streitberg 1891; Abraham 1996ab; Brinton 1988; Raith 1951, to name but a few). In OE, “[t]he perfective aspect” is “often indicated by means of verbal prefixes” (Mustanoja 1960: 446; see also Quirk & Wrenn 1955: 114ff). For instance, þurh- ‘through’, of-, and to-, as in þurhbrecan ‘break through’, ofsceotan ‘shoot off’, tobrecan ‘break up’, render an imperfective verb perfective by specifying the goal.4 In OE, ge- expresses perfectivity without have, e.g. both (56) and (57) are well-formed. The prefix foris also a transitivizer in OE and Early ME as in (58). These prefixes are in E, responsible for the relationship between perfective and the presence of objects:5 (56) Sona þæt gesawon soon that saw ‘As soon as they saw that.’ (Beowulf 1591) (57) ac hie hæfdon gefrunen þæt … but they had found-out that ‘But they had heard that …’ (Beowulf 694) (58) ac he sigewæpnum forsworen hæfde but he victorious-weapons cast-a-spell-on had ‘But he had put a spell on the weapons.’ (Beowulf 804)

Aspect

In this section, I examine inner aspect and changes from inner to outer aspect. First (in 4.1), I’ll look at the ModE situation where what are originally adverbs are used to indicate aspect in the E position. This is in accordance with Late Merge. Then (in 4.2), I’ll take a look at Dutch, which shows a fascinating complementarity between certain aspectual prefixes and the perfective prefix ge-. I suggest a Spec to Head change may be involved in certain changes. The Dutch data contrast interestingly with those from older periods of English (as shown in 4.3). Finally (in 4.4), I discuss a number of the properties of the E position. 4.1 English Chapter 7 mentions the reason for having a structural inner aspect, namely to express the relationship that exists in many languages between perfectivity and a particular Case. Kiparsky (1996) shows this for Finnish and Abraham (1996a) for Old High German. This relationship would be accounted for by a special functional category that, apart from containing aspect, is responsible for Case checking. Just as a tensed verb (in most languages) makes nominative Case checking possible, a perfective verb would enable certain objective Case checking. Sanz (1996) links perfectivity and transitivity by assuming, based on Tenny (1987), the feature [measure] to occupy a head that is checked by an object DP. Hence, ‘I ate’ is not perfective but ‘I ate the apple’ is. I modify Sanz’s tree in (59), where the apple moves to the Spec of EP, and results in ‘I ate the apple up’. In case of an indefinite object, as in ‘I ate up an apple’ movement typically does not occur: (59)

vP v′ v

EP

ate

E′ E

VP V′

(up) the apple

V

...

213


In (59), because an E head with perfective features is present, the verb licenses an object. Often, speakers will emphasize perfectivity by adding a preposition as in (61) and (62) below. In ModE, eat is imperfective whereas (the phrasal verb) eat up is not. In (59), this relationship is expressed by having the prefix occupy a functional head responsible for completed aspect. Particles are grammaticalized adverbs, and hence the change is from (60a) to (60b): (60) a.

vP v′ v

VP V′

turn

V

A off

b.

vP v′ v

EP Ε′

turn

E

VP

off

Faarlund (1990:187) looks at a similar correspondence between prepositions in Old Norse and particles in Modern Norwegian. Blom (2003), as we’ll see in the next subsection, does so for preverbs in the history of Dutch. Van Kemenade & Los (2002) do so for Gothic and OE. None of them puts it in this tree form, however. Adverbs, as in (61) and (62), are new potential sources for changing towards inner aspect: (61) ASU Librarian: We received it in for you. (62) Do you mind if I open me eyes up slowly. (Onslow in ‘Keeping up appearances’)

Aspect

Other examples are collect it in and change it up. This means speakers add a new (resultative) particle in A, as in (63), comparable to (60a) above, because there is no such verb such as upopen in English: (63)

vP v′ v

VP

receive

V′ it

V

A in

The particle in turn can grammaticalize, and be reanalyzed as an inner aspect, as in (59) and (60b), in accordance with Late Merge. Sentences such as (64) would be the result since indefinite objects such as a book would remain in the VP but definites such as it would move to the Spec of EP: (64) a. I received in a book. b. I received it in.

4.2 Dutch verbal prefixes The representation of aspect is still quite controversial. The analysis I present here no doubt will be as well. I will now give an analysis for Dutch and then relate it to the situation in OE and ME. In Dutch, as is well-known from the grammar books, there are two kinds of prefixes: separable and inseparable. Many of the separable ones have spatial meaning and are more prepositional. Instances of separable ones with their verbs are opzoeken ‘look up’, uitzoeken ‘figure out’, uitlezen ‘finish reading’, voorlezen ‘read to someone’, and of the inseparable ones doorzoeken ‘search’, voorspellen ‘predict’, doorgronden ‘understand’. When the verb moves (to second position) in main clauses, it leaves the separable prefix, as in (65), but takes the inseparable prefix along, as in (66): (65) Ik zoek dit uit I search this out ‘I am getting to the bottom of this.’

215


(66) Ik doorzoek de tuin I through-search the garden ‘I am searching the garden.’

A second difference is that the separable prefixes are compatible with ge-, as in (67), but the inseparable prefixes are not, as (68a) shows, and a ge-less form is needed as in (68b): (67) Ik heb dit uitgezocht I have this out-searched ‘I have gotten to the bottom of this.’ (68) a. *Ik heb dit doorgezocht I have this through-searched ‘I have searched this.’ b. Ik heb dit doorzocht

This means ge- and the inseparable prefix are in complementary distribution, both in inner aspect, I argue. Assuming structures as in Kayne (1994) and Zwart (1997) for (surface) SOV languages, (67) can be derived as follows. Uit ‘out’ is in the Spec of EP and the verb moves to E, as in (69): (69)

vP v¢

ik v

EP E¢

uit E ge

VP V¢

dit V

A

zocht

Evidence for having uit in a specifier position is that the verb skips the ‘particle’ in (65) and this is only possible if the latter is in the specifier position. A second piece of evidence is (70), where the ‘particle’ is modified, again only possible if it is a specifier:

Aspect 217

(70) Ik ben die trap recht opgelopen I am those stairs straight up-walked ‘I walked straight up those stairs.’

The same modification is not possible with inseparable prefixes. In the case of inseparable prefixes, the tree is similar but with the prefix in the head E, as in (71). This particle is incompatible with ge-, which is also in E: (71)

vP v¢ ik

v

EP E¢ E

VP

door

V¢ dit

V

A

zocht

The verb moves to E in the case of (68b), but if there is no auxiliary, as in (66), the verb and prefix move to the second position together since the verb cannot skip the particle in head position. Blom (2003) and Blom & Booij (2003) provide a grammaticalization schema and argue that certain separable prefixes become inseparable. For instance, the separable over in (72) from Middle Dutch becomes inseparable in Modern Dutch (73) and is in Modern Dutch incompatible with ge-, as (74) shows: (72) mer tis een flaute die hem over ghecomen is but it-is a swoon that him over come is ‘But it is a swoon that has happened to him.’ (from Blom 2003: 33) (73) Hem overkwam een flauwte him happened a fainting-fit ‘He fainted.’ (74) *Hem is een flauwte overgekomen him is a fainting-spell happened ‘He fainted.’


The change from adverb to particle in E is expected under Late Merge. The change from separable to inseparable is also expected since it is from specifier to head. 4.3 OE prefixes Throughout the history of English, ‘préverbes’, as Mossé calls them, such as aon ahefe and aris in (75), disappear. They are replaced by other adverbs such as in (61) above. Even (75), where the prefixes still occur, has two upp forms that strengthen the aspectual meaning that is perhaps no longer as strong in a-: (75) Ahefe hine þone upp & upp aris lift him then up and up get ‘Lift him up and get up then.’ (Quadrupedibus, p. 9; HC OE3)

Van Kemenade & Los (2002: 20–21) see the a in ut-a-leoran ‘flee away’, ut-asellan ‘grant outright’, and ut-a-tynan ‘exclude’ as a weakened form of ut. With upp, it is not an actual doubling as with ut (Los p.c.), but I’ll argue it is reinforcing the original prefix. OE has prefixes that are comparable to Dutch inseparable ones with (inner aspect), e.g. a-, as in (76), and separable ones such as upp in the specifier of EP or in a higher (outer) ASPP: (76) Hu lange sceal min feond beon uppahafen ofer me ‘How long shall my enemy be elevated over me.’ (Paris Psalter, p. 19; HC OE3)

Unlike in Dutch, it seems as if the inseparable prefixes may have ge- preceding them, as in (77): (77) 7 he geandbidode þone frofer þe behaten wæs þam and he waited the consolation which promised was the folce israhel people Israel ‘And he was waiting for the consolation which had been promised to the people of Israel.’ (Alfric Homilies I, p. 249)

As Los (p.c.) points out, if the prefix is phonologically strong, as and is in (77), the verb derives from a noun, and Bosworth & Toller’s dictionary lists a noun anbid ‘waiting’. Even though there are many instances of geanbidian as well with the weaker prefix an-, the -ian ending suggests a causative, as in ‘make wait’, with a structure, as in (78):

Aspect 219

(78)

vP v¢ v

VP

ian

V¢ V

N anbid

Having an- be part of the noun does not tell us where ge- is located. It could be in either inner or outer aspect. Other forms that have ge- precede a prefix are geunrotsad ‘saddened’, geanlæht ‘united’, geuntrumod ‘weakened’, but these are formed from adjectives (rot ‘glad’, ænlic ‘only’, trum ‘strong’) and hence the un/æn- is part of the adjective, not situated in inner aspect. They have a derivation as in (78), with the adjective in place of the noun, also moving to v. Having discarded geanbidian and geuntrumod as not being likely to show where ge- is located, I’ll turn to verbs in Bosworth & Toller that are not causative. As far as I can determine, these are geanbyrdan ‘resist’ and geanmetan ‘encourage’; geandettan ‘confess’ and geforwearþan ‘perish’ may be derived from nouns, even though they lack causative markers. Examples are given in (79) and (80): (79) Gif he on ðone geonbyrde ðe hine slog if he on him resisted who him beat ‘If he strove against him who slew him.’ (Laws 76, Bosworth & Toller 425) (80) he him ða to fultume com 7 hiene swiðe geanmette he him then to help came and him quickly encouraged ‘He then came to his help and encouraged him quickly.’ (Orosius 76.17–8)

An analysis for these would have ge- in the outer aspect and the prefix in the inner aspect. There are not many of these and most instances of ge- preceding a prefix are found in Alfredian writings. There is no ge- preceding an or un prefixes in the poetic corpus of Beowulf, The Exeter Book, Vercelli, or in the Junius Manuscript. If (79) and (80) show that ge- is changing from E to ASP,


this is unlike Dutch where ge- can only be in E. The change to ASP predicts the occurrence of (79), (80), and possibly (77), completely ungrammatical in Dutch, as well as instances where ge- is (still) in E, as in (76) above and (81). Pre-950 (HC OE1–2), there are 17 instances of ofaslegen ‘slain’, as in (81), but only two in HC OE3. This means inner aspect is disappearing: (81) Her Offa Miercna cyning het Æþelbryhte rex in-this-year Offa Mercian king commanded Æthelbryht king þ heafod ofaslean the head struck-off ‘In this year, Offa, the Mercian King, commanded the head of King Æthelbryht to be slain off.’ (Anglo Saxon Chronicle A, anno 792, Thorpe p. 98)

Note that ofaslean occurs in all versions of the Chronicle except the Peterborough one, which is northern and has lost the prefix a-. In OE3, there are 3 instances of ge- preceding anbidian, whereas there are none in OE1–2. Like Modern Dutch, OE has inseparable prefixes such as a-. They eventually weaken and are replaced by new adverbs, as in (75). OE is different from Dutch in that ge- (infrequently) precedes an aspectual prefix, a sign of grammaticalization from E to ASP. 4.4 E markers In this short section, I look at some other instances of E. In Spanish, as argued by Sanz (1996), Bonneau et al. (1994), and others, there is a telic clitic se as in (82): (82) Pedro se comió tres manzanas Pedro himself ate three apples ‘Pedro ate three apples.’ (Sanz’s sentence)

This clitic would be in E marking completion. In ME, there is a period in which third person pronouns indicate telicity as in (83) to (85). It is never a frequent construction (fewer than ten occurrences in Chaucer) but it could be indicative of an overt telicity marker in E: (83) And with glad herte he wente hym hoom ful soone and with glad heart he went him home very soon ‘And with a happy heart he went home immediately.’ (Chaucer, Knight’s Tale 55.2270)

Aspect 221

(84) And to his bed he wente hym hastily and to his bed he went him hastily ‘and he went hastily to bed.’ (Chaucer, Merchant’s Tale 160.1779) (85) Forð him gon ride. Arður þe riche away him began ride Arthur the great ‘Arthur the great began to ride away.’ (Layamon, Caligula 9939)

Similar markers occur in many other languages, e.g. Frajzyngier (1997) mentions (86), from Mina, a Chadic language, even though he argues the (possessive) pronoun is a marker of affectedness of the subject rather than of telicity: (86) í-tsù t‘tàn á wt’ t‘tàn 3.p-go 3.p prep village 3.p ‘They went home.’

In Section 4, I have provided some more reasons for assuming inner aspect (EP). First, I show that ModE makes use of inner aspect for phrasal verbs. Then, I examine Dutch separable and inseparable prefixes in relation to ge- and show the inseparable prefixes are in complementary distribution with ge-. The difference between separable and inseparable prefixes in Dutch is that the former are in the specifier of the EP and the latter in the head. Some change from separable to inseparable, i.e. from Spec to Head, as expected. In English, both ge- and prefix heads in E disappear by the end of the ME period and are replaced by new adverbs as well as by markers such as have and -ing. These move up the tree, as predicted, throughout the history of English. Lastly, I provide some instances of other markers of E.

5. Giorgi & Pianesi: The demise of the infinitival ending and aspect Giorgi & Pianesi (1997, hence G&P) argue that bare (eventive or durative) infinitives without -e(n), as in English, are perfective but that infinitives with -e(n), as in German and Dutch, are not.6 According to them, this explains the data discussed in Chapter 9, PV complements in English being perfective but not in Dutch and German, as well as the data in (1) and (3) of this chapter, the simple present in English not being used to express present tense. The presence of [+perf], implying boundedness, on eventive forms is only compatible with the progressive which is also bounded (G&P, p. 169). Stative verbs such as know and see are not associated with [+perf] since, like habituals, they are associated with a generic operator.


I show that there are several problems with Giorgi & Pianesi’s analysis. First, there are languages such as Afrikaans where infinitives do not have an ending but where no specially marked progressive is used. Second, the infinitival ending in English is lost several centuries before the bare verb becomes perfective, as defined in Giorgi & Pianesi, and before the simple present ceases to be used. Third, as shown in Chapter 9, eventive (bare) infinitives are not always interpreted as perfective in ModE. They only are as complements to stative perception verbs, not as complements to non-stative perception verbs. Starting with the first, Modern Afrikaans presents a counterexample to the link between infinitival endings and a simple present being grammatical. It is a language without infinitival endings (e.g. te drink ‘to drink’, te se ‘to see’, te kom ‘to come’) but one in which eventives can be in the present tense, as in (87): (87) Nou dans die poppe now are-dancing the dolls ‘Now the fat is in the fire.’ (Ponelis 1991: 187; Paul Roberge p.c.)

As to the second objection, in OE, as is well-known, infinitives have endings, but these drop off in ME. In Chaucer, who writes at the end of the fourteenth century, both -e and -en endings occur, as they do in Modern Dutch where the -en ending is mainly pronounced -e. The ending is presumably pronounced in Chaucer since smyte in (88) rhymes with Arcite: (88) How greet a sorwe suffreth now Arcite! how great a sorrow suffers now Arcite The deeth he feeleth thurgh his herte smyte the death he feels through his heart strike ‘Arcite is suffering great sorrow and feels death striking his heart.’ (Chaucer, Knight’s Tale 42.1219–20)

Minkova (1991), citing a number of other scholars, argues that the ending is disappearing in the North from 1100 on and that “[b]y 1400 final unstressed -e had been abandoned in all parts of the country” (p. 30). Görlach (1990: 47) says that from 1400 onwards, the -e spelling is “arbitrary and optional” (see also Moore & Marckwardt 1951). Southworth (1947: 925) estimates that even in Chaucer the final infinitival -e is not pronounced in 82% of the cases. After Chaucer, the infinitival ending changes, and this is when we would expect an increase of forms in -ing, and a loss of (88). This is not the case and problematic for G&P. In the 15th century Paston Letters, the infinitival ending -en is fairly rare: apart from ben ‘to be’, there are perhaps 20 in a large corpus.

Aspect 223

However, -ing is not popular, as shown in Section 1. The Paston Letters also show that, after the endings are lost, no immediate changes with PVCs occur. In More’s English (Visser 1946–52), from the beginning of the 16th century, i.e. a century or so after the loss of the infinitival ending, some infinitives have an imperfective meaning. Sentence (89), for instance, repeated from Chapter 9, emphasizes the action through the adverb and ModE would use running. This text has many -ing forms too (see Visser 1952: 810): (89) The fox … saw him run so faste (Richard 71 C 1, Visser, 1952: 762)

Thus, even though the infinitive has no ending, it can be used as a non-perfective. An Early ModE grammarian, Mulcaster, who writes an Elementarie in 1582, divides final -e into “soundeth or … silent” (p. 111). The first category includes me, see, we, agree, yee, and e in Latin words, but the section is very short; the silent -e section is much longer (and talks about nouns as well as verbs). Silent -e is said to have an effect on the length of the vowel preceding it, as in made, cure, and is used in many other situations, e.g. in cause, excuse, deceiue, loue, moue. Thus, Mulcaster’s description shows that Elizabethan English infinitival endings are not pronounced differently from ModE endings. Franz (1909: 21) says the infinitival -e is used “ziemlich prinziplos” in Shakespeare’s time, but -en is never used. This loss of -e(n), however, does not seem to force an increase in the use of -ing, since very few complements such as (90) and (91) occur, as mentioned in the previous chapter. (92) to (94) are more likely: (90) Who you saw sitting by me on the Turph. (As You Like It III, IV, 52) (91) may you see it comming. (Macbeth V, v, 37) (92) to see thee weare thy heart in a scarfe. (As You Like It V, ii, 23) (93) when she sees me worke. (Tempest III, i, 12) (94) When shall you see me write. (Loves Labor Lost IV, iii, 181)

Thus, in Shakespeare, there is no connection between the loss of -en and the perfective interpretation of the bare verb: the infinitive has lost its ending but it continues to be used as a non-perfective. In conclusion to Section 5, G&P account for the differences between Dutch (10) and ModE (1) by assigning [+perf] to the English eventive verb. This explanation encounters empirical problems. Even as late as Shakespeare’s time, long after the disappearance of -en, (92) to (94) above are grammatical. I therefore suggest that this problem is independent of the infinitival ending, but depends on what is in ASP. The loss of -en cannot be shown to coincide with


the bare infinitive becoming perfective. It is more likely that the infinitive remains ambiguous until gradually -ing is reanalyzed as imperfective marker. The progressive is available in OE, i.e. is not introduced with the loss of -en, and there is even no sudden increase of constructions such as (90) and (91).

6. Conclusion and further research In the previous chapters, I argued that modals and perception verbs undergo grammaticalization from main verb to ASP. Intertwined with this change is the gradual introduction of the ASP marker -ing which is examined in this chapter. In an attempt to provide an account why (1) is ungrammatical in ModE but not in Dutch, OE, and ME, I argue that the basic parameter setting changes, and one framework to express this in is Cowper’s but even in Streitberg the point is clear. The typological change is one of marking aspect to marking tense, as formulated in the Tense Aspect Parameter in (41). Cowper’s account, unlike G&P’s, does not give an independent reason why a language would have one choice unmarked over the other. I assume Cowper’s theory over G&P’s, since the latter’s account of the infinitival ending causing the difference cannot be maintained, as shown in the previous section. Table 10.1.Changes involving ASP and E

a. b. c. d. e. f. g. h.

change

century

ge-/for- Æ Ø modal Æ ASP have Æ a after modals to/do/ing in ASP A/P Æ E I eat now Æ * saw Æ ASP have seen yesterday Æ *

C12–C14 C15 C15 C14–C17 continual C19 C19 C19

Table 10.1 summarizes the changes discussed in this and the previous two chapters: a loss of inner aspect in (a); the introduction of outer aspect in (b) and (d), due to Late Merge; the contraction of have in (c), being the result of the competition in (b) and (d); the continual replenishment of inner aspect by adverbs and prepositions in (e); the change from interval to moment in (f) (independent of the other changes); and the Late Merge of saw in (g). The change in (h) is discussed in Chapter 8 and Section 2 above.

Aspect 225

Appendix Many linguists have used Reichenbach’s (1948) system to account for the tense system of languages (e.g. Hornstein 1990; Adger 1996). In this appendix, I’ll relate Cowper’s aspectual system as used above to Reichenbach’s tense one. The main reason that I find the Tense-aspect Parameter attractive is that it accounts for the differences between ModE and other languages in an intuitive way. Where does moment reside in English: C or T or ASP? I think this is where Reichenbach’s system of S(peech), E(vent), not to be confused with Extent, and R(eference) helps. If Reichenbach’s S equals Cowper’s moment, and if S is at the same point as the Event, then the present tense needing a special marker follows immediately and so does the non-moment interpretation of the simple past. First, I’ll outline Reichenbach’s system. Reichenbach (1947: 290ff) presents a schema where E is the point of the Event, S the point of Speech, and R is the Reference Point. A simple present would have all three connected, as in (95). In a simple past, as in (96), both the Event and the Reference points are in the past: (95) E,S,R (96) E,R__S

A present perfect7 has an event in the past, but with the reference at the time of speech as in (97). In a past perfect, the event precedes the reference point which in turn precedes the time of speech as in (98): (97) E___S,R (98) E__R__S

The future tenses are as follows, simple future in (99) and future perfect in (100): (99) S___R,E (100) S___E___R

Reichenbach indicates aspect by means of a modification of the Event representation. Crucial for my analysis is that only the Speech time has moment (or interval depending on the language). This predicts that in simple pasts the event could actually be an interval, since the event is not at S, but not in simple presents. This is a correct prediction, as (101) and (102) show: (101) Yesterday, I ate for hours (at that fake Roman feast). (102) *Today, I eat for hours. (except as a future)


The interpretation of a present perfect, as in (103), is also typically interval, and the same holds for future perfect, as in (104): (103) I have worked on this for ages. (104) He will have worked for six hours.

The -ing is optional, if the actual action is emphasized but this is more like the ME situation for the simple present. If S is the only one marked for moment, this fits. A structural representation using S(peech), R(eference), and E(vent) could be (105): (105)

CP C¢ C [pres]

IP They

I¢ I [pres] (ha)ve

ASPP ASP¢ ASP t

VP V¢ V gone

S

R

...

E

In (105), (ha)ve originates in ASPP (except when have follows modals) and then moves to I to check its present tense features. The VP has variable tense features depending on the time of the event, here past.8 Since the time point of I is the same as in C, there is present relevance even though the event happened in the past, indicated by the past participle that moves to ASP. The tense of S is situated in C, adapting an idea from Enç (1987: 643) that anchoring is a necessity and that “[i]f Comp does not have a governing category, it is anchored

Aspect 227

if and only if it denotes the speech time”. Hornstein (1990: 14) argues likewise that S “is a deictic element anchored within the discourse situation, often to the moment of Speech”. Thus, have does not check the past tense features in VP. Have, after checking its present tense features against I (again the features in I are uninterpretable but those of have are interpretable), moves to C covertly to check the tense features there. Have is ‘taken out’ of the lexicon as an element with present features that are checked in I; gone has past features and checks those in VP. In cases where have checks other features in addition to present, it is specified in the lexicon differently (e.g. the finite has needs to check person and number). I will account for the relation of R to the other tenses by arguing that, in simple tenses, the tense in I is the same as that of VP. The verb has interpretable tense features and checks the tense of I. Thus, the structure of a simple past is as in (109): (106) __S,R,E__ (107) E,R__S (108) S__R,E (109)

CP C¢ C [pres]

IP She

I¢ I

VP

[past]

V went [past]

S

R

E

Chapter 11

Late Merge Heads to higher Heads

Chapter 6 shows that all lexical and grammatical categories participate in crossclausal grammaticalization by becoming complementizers and creating CPs. This chapter shows the same for the creation of IPs and VP shells. It lists all the possibilities where grammaticalization affects V, E, ASP, M, and T. Givón (1979: 218) claims that “the only universal source for tense-aspect affixes … is via the reanalysis of main verbs” (quoted in Heine & Reh 1984: 114). This is too narrow. The real difference between CP-grammaticalization and IP/VP grammaticalization is that with the former the specifier is active. Hence, phrases (such as pronouns and wh-phrases) participate, as well as heads. With IP and VP, mainly heads are involved. Exceptions are particles going from Spec to Head of E, discussed in the previous chapter. I continue to talk about I(nflection) when I speak in general, but will show this position is in fact divided into T, M, and ASP in those cases where there is structural evidence. This is in accordance with what I argue in Chapter 7. I won’t go into the change from V to v (see Chapter 8) and will continue to assume that there is no separate AGR(eement) position. As in Chapter 6, this chapter is divided into two sections. Section 1 examines the changes from lexical to grammatical categories and Section 2 the ones from one kind of grammatical category to another kind of grammatical category.

1.

From lexical to grammatical

1.1 N (to A/P) to I A direct change from N to I has not been attested as far as I know. An adverb or preposition deriving from a noun may grammaticalize into an auxiliary, however. This is possibly the case with the Bantu language Kikuyu where thutha ‘back’ becomes the adverb ‘afterwards’ (see H&K 2002: 49). This adverb may


subsequently become a tense marker. Other similar examples in H&K come from Egyptian and Bambara. 1.2 V to I Two of the changes from V to I have been talked about in Chapters 8 and 9. Other examples abound and the change from V to I is the most common instance of Late Merge. I will just give two additional examples, one from Spanish and one from Lhasa. In Spanish, a participle preceded by venir ‘come’, ir ‘go’, andar ‘walk’, as in (1), and seguir ‘follow’, as in (2), expresses progressive aspect (Armitage & Meiden 1979:381). The different auxiliaries result in slightly different meanings: (1) Anda buscando flores [he] walks around looking for flowers ‘He is looking for flowers.’ (2) seguimos viajando hacia el sur ‘[we] continued traveling towards the south.’ (from Armitage et al. 382)

Myhill (1988: 230; 238) shows that clitic climbing is very common with andar when it is a progressive, indicating it is no longer a main verb (see also the discussion in Chapter 7). In 71% of the sentences where andar is progressive, clitic climbing takes place; with other progressives, this percentage is higher, i.e. with estar, ir, and venir, it is 89%, 86%, and 83% respectively. Myhill (1992: 218) gives the percentages of clitic climbing for ir as 86% when it is a progressive, 76% when future, and 29% when it is a motion verb. He doesn’t list seguir. So, with the progressive, there is no intervening IP and ir etc. are auxiliaries in a mono-clausal structure. There is a possible case of V becoming ASP. Hopper & Traugott (1993) mention verb-serialization in Lhasa, a Tibeto-Burman language where one verb becomes the verbal inflection on the other verb (data from DeLancey 1991). The verb-serialization construction is shown in (3) where two actions occur. Once the non-finite verb phyin ‘go’ loses its inflection, tshar ‘finish’ becomes an auxiliary, as in (4). The final stage is for the ‘auxiliary’ to become an ending on the verb, as in (5): (3) Khos phyin-byas tshar-ba red he went-nf finish perf ‘He went and finished it.’ (from Hopper & Traugott 1993: 200)

Late Merge

(4) Khos phyin tshar-ba red he went finish perf ‘He has gone.’ (also from Hopper & Traugott) (5) Nga krom-la phyin-tshar I store went-perf ‘I’ve gone to the store.’ (again from Hopper & Traugott)

Hopper & Traugott argue (1993: 200) that verb-chaining is different from the change in English from V to AUX since the AUX is not inflection. The change of V to inflection can, however, be seen as a change to a higher position, from V to ASP. I’ll now turn to another source for the ‘creation’ of I-elements. 1.3 A to I: Adverbs of time as T, of mood as M, and of manner as ASP In Chapter 10, I have considered the case of adverbs replenishing inner aspect, as a change from A to E. I won’t go into that more here. Instead, I’ll first briefly discuss cases from the literature of temporal adverbs changing to I (in 1.3.1). Then (in 1.3.2), I’ll provide some background on Mood Particles (or MPs) and show that MPs are polysemous with adverbs. The respective order of MPs suggests they are in T, M, and ASP positions in, for instance, contemporary Dutch. Later (in 1.3.3), it is shown that OE has MPs but that, by ME, they cease to occur. I argue that, due to the grammaticalization of the auxiliaries in ME, the MPs are lost in that language and hence the German, Dutch, and Swedish MPs given below are almost untranslatable. This means that MPs and auxiliaries are at a certain point ‘competing’ for the same position in ME and that the adverbs probably occupy I, and more specifically T, M, and ASP. Most of the conclusions in this section are based on van Gelderen (2001c) where many other examples are given. 1.3.1 Temporal adverbs There are languages that lack tense markers (one famous case being Hopi, see Malotki 1983) and here temporal adverbs play a large role and could be argued to have become tense markers. Heine & Reh (1984: 120) mention the case of ‘then’ in Bari, a Nilotic language, which comes to mark future as it moves from clause-initial position to one immediately preceding the verb. Lotuko, another Nilotic language, uses adverbs as tense markers (Heine & Reh 1984: 121) as do West African languages, such as Klao, a Kru language. Baka, a Niger-Congo

231


language, uses the word for ‘yesterday’ to mark past, and so do River Cess Bassa and Neyo, the latter two Kru languages (H&K 315). 1.3.2 Germanic MPs In most Germanic languages except ModE, the area between the CP and the VP can have MPs such as doch, immer in German, toch, noch in Dutch, and väl, ju in Swedish (cf. Abraham 1990; 1991ab; 1995, Altmann 1976, Hentschel 1986, Helbig 1988, J. Jacobs 1983, König 1991, Weydt 1979). What I’ll argue is that these MPs in all Germanic languages derive from adverbs of time and manner (not controversial given they can still function that way), in accordance with Late Merge. An example of a sentence with MPs is given in (6) from Dutch: (6) dat ik het hem dan toch maar weer eens een keer uitgelegd heb that I it him then yet but again once a time explained have ‘that I have once again explained it to him.’

The particles dan, toch, maar, weer, eens on their own (e.g. sentence-initially) can be used as adverbs (in the case of dan, toch, weer, eens) and as conjunctions (in the case of maar ‘but’). As adverbs and complementizers, they are stressed. In (6), the MPs have lost much of their original meaning. They occur in the ‘middle’, in the area between CP and VP: more precisely, the area between the definite DPs on the left and the indefinite DPs on the right, as in (6) and (7). Abraham (1991: 244) calls this the area between thematic and rhematic. The ungrammaticality of (8), a variant of (6), shows that MPs are not allowed in the definite domain that het and hem are located in. Sentence (9) shows that MPs are not allowed in the (indefinite) VP-domain either where een boek is assumed to stay in the VP (in these sentences, dan and toch are the clear MPs): (7) dat ik dan toch weer eens een boek heb gelezen that I then yet again once a book read have ‘that I’ve managed to read a book again.’ (8) *dat ik het dan toch weer eens hem een keer uitgelegd heb that I it then yet again once him a time explained have (9) *dat ik een boek dan toch weer eens gelezen heb that I a book then yet again once read have

Another characteristic of MPs is the number of such elements, i.e. the possibility of stacking. In Dutch many are possible, e.g. sequences of seven MPs are possible, as (10) shows, even though, according to van der Wouden (1999: 294), six are quite ‘spectacular’ already:

Late Merge 233

(10) Doe dat dan nu toch maar weer eens even over nieuw do that then now yet but again once just again new ‘Do that again.’

The order is roughly that of tense, then mood, and then aspect. Thus, dan and nu are also time adverbials; toch and maar can be seen as mood markers; and weer, eens and even also function as aspectual adverbials), and other orders are ungrammatical, as (11), a rearrangement of the MPs from (10), shows: (11) *Doe dat even eens weer maar toch nu dan over nieuw do that just once again but yet now then again new

The sequence in (10) might initially indicate that MPs are possibly situated in functional categories, as in Cinque (1999). However, most have become units, e.g. dan nu, toch maar, and weer ‘ns even. These units are compatible with the three positions I have argued for in ModE. In ModE, whenever two or more particles occur together, they tend to become one phrase as well, e.g. already, also, all right, well now, although, as well, even though. In short, languages such as Dutch have MPs that occur in a very definite TMA order and that are polysemous with temporal, modal, and aspectual adverbs. The argument that MPs find their origin in Adverbs, and hence are a result of Late Merge, is that the older Germanic languages do not have MPs to the same extent. For instance, Old High German (cf. Hentschel 1986) and Middle Dutch (cf. Vismans 1994) have similar particles, and so do OE and ME, but they are much more restricted. Vismans (1994: 102ff.) shows that in Middle Dutch (up to 1500), toch/doch occurs as MP in all texts examined. (This is not unexpected since Gothic already seems to have had thau, see Hentschel 1986). Dan occurs only in 3 texts and eens in one. In the early 17th century, ook and nou are added, and the number of MPs gradually increases. In the next section, I show that Dutch-like MPs occur in OE but in a much more restricted way. They never have the chance to develop the way those in the other Germanic languages do as I will also show. 1.3.3 OE I will provide some instances of MPs in OE, which as in Middle Dutch and Old High German are in the developing stages. I then show that the development of MPs in English was stopped and that this coincides with the loss of V2. This fits with Abraham’s (1991) observations that it is not the verb-final aspect but the V2 aspect of the Mittelfeld that is relevant to licensing MPs. The loss of MPs


also coincides with the complementizer becoming optional in the subordinate clause (see van Gelderen 1993), another indication of the C-position ‘losing influence’. Then, as the verb ceases to move to C and the I position becomes more important (argued to occur around Chaucer’s time in e.g. van Gelderen 1993; 1997a), MPs in English become restricted to the CP periphery. Though is a cognate of Dutch/German toch/doch. The OED lists some instances of ‘adversative particle’1 though from OE on. For instance, (12) and (13), provided by the OED, and (14) and (16) from the HC. Notice the multiple MPs in (16): (12) Ne maŠon þis þeah ealle men don not might this though all men do ‘Not al men might do this though.’ (Blickling Homilies 37, from 971) (13) ac hi þah ledað to deðe on ende but they though lead to death in end ‘but they, however, lead to death in the end.’ (Lambeth Homilies 119, from 1175) (14) hie þeh Philippus besirede mid his lotwrencum they though Philippus deceived through their cunning ‘They deceived Philippus through cunning though.’ (Alfred, Orosius, 64.18–9) (16) Philippuses yfel mehte þeh þagiet be sumum dæle gemetlic þyncan Philippus’s evil could though then-yet to some extent sufficient seem ‘Philippus’s evil might seem sufficient already.’ (Alfred, Orosius 66.6–7)

In ME, the MP use of though dies out. The Caligula version of Layamon’s Brut from the 13th century has about 35 instances (that I could find) of variants of though (þah 10 times; þeh 22 times, and the rest þeah, þeih, þaih). These all introduce clauses. In late ME, e.g. in Chaucer, though is never an MP (even though other MPs occur). It occurs 378 times and only functions as a clause introducer, often followed by that or preceded by and, but, for, as, all, and a few times by ne or eke. There are 90 instances of a variant of although, used the same as though. In ModE though is still used on the periphery of the clause, as in ME. An instance from Mark Twain’s Tom Sawyer is (17): (17) Well, I’ll just bet I will, though. (Tom Sawyer, Chapter 7)

Although is limited to complementizer in that text, and occurs only six times. Showing there is a possible correlation between the loss of MPs and the loss of a Mittelfeld does not account for this relation. It is possible to argue that the

Late Merge 235

loss of V2 frees up the CP, and makes it possible for MPs to be in CP. This cannot be correct since the structure of the CP is quite elaborate in ME (see Chapter 3 above), more so than in later stages of the language (when the language becomes more IP-oriented, see van Gelderen 1993; 2000). Another explanation is to build on the insight that stacking gives us, namely that they show a TMA order, as in (10) above, and that their loss coincides with the grammaticalization of auxiliaries. Abraham (1995; p.c) argues there are at least three functional MoodPs above the VP as well as TP and ASPP to accommodate MPs in German. As is well-known, main verbs start to grammaticalize into auxiliaries (see Lightfoot 1979, Traugott 1972 for changes in auxiliaries) in ME. The language learner confronted with these new auxiliaries places them in positions such as TP and ASPP. This banishes MPs to the periphery, namely CP, or the right periphery, as in (17). The problems with this account are theoretical. If MPs are heads (and they look like heads), V-to-C movement should be blocked in Modern German and Dutch but obviously is not since verbs move to the second position. This can be solved if the MP heads are (optional) adjuncts and do not count as barriers.2 If MPs are heads (that do not form barriers for head movement), the difference between English and Dutch is that the former uses CP more for expressions of mood because the head positions of the TMA categories are used by auxiliaries (sequences such as may have been being occur in English but not Dutch). Some evidence for this account can be found in Chaucer. Around the time of Chaucer, modals, perfect and passive auxiliaries start combining. Their combination with MPs is interesting, for instance, in that the original manner adverb ek/eke ‘also’ always follows the perfect and modal auxiliaries. It never precedes them, as expected if MPs occupy specific FCs, and manner is aspectual and close to the V: (18) I have ek understonde | How ye ne do but holden me in honde I have also understood how you not do but hold me in your hand ‘I have also understood how you deceive me.’ (Troilus and Criseyde 581.1614–5) (19) And for the harm that myghte ek fallen moore ‘and for the harm that might also descend more.’ (Troilus and Criseyde 495.454) (20) Oure maydens shal eke pluk hym so ‘Our maidens will also pluck him so much.’ (Romaunt of the Rose 749.5989)

So, eke might be in ASP in (18) to (20).


The temporal adverb now mainly precedes the auxiliary in Chaucer. For instance, there are 40 instances of ‘now wol I/ye’ but only 3 of ‘I wol now’. Looking through the 1411 instances of now, this ratio seems representative. This indicates that unlike ek/eke which is a manner adverb, now may be being pushed out since tense has to be filled. This needs some further investigation. The changes in the MPs in Germanic can be seen as changes from Adverb to TMA, i.e. instances of Late Merge. The reason MPs are lost in English is possibly through ‘competition’ with the auxiliaries that are also occupying TMA positions. 1.4 P to higher positions Other changes from P to I have also been talked about frequently. I’ll first talk about to and where it will be discussed further. Then, I’ll talk about another P becoming an outer aspect marker, and finally I’ll briefly mention a P possibly changing to a passive marker, i.e. the head in a Voice Phrase, not talked about much in this book, but still in accordance with Late Merge. The most famous example in English of P to I is of course to, going from P to I, or more precisely to ASP and to M (never to T though) and finally to C (see Chapter 6 and Section 2 below). The evidence for to being in ASP for a short period is the appearance of sentences such as (21), repeated from the previous chapter: (21) Whanne God had seen the erthe to be corrupt. (Wyclif, Genesis VI 12)

The evidence that to is in M is related to to also moving to C, and hence will be discussed in Section 2. I’ll first discuss another example of P to I. During one period of English, prepositions such as (up)on, at are used to indicate progressive aspect, as in (22): (22) I am upon writing a little treatise to present to the Duke. (Pepys’s Diary 31 Dec 1666, from Visser 1998)

These prepositions of location are often used for this purpose, e.g. Egyptian, as in (23), Dutch, as in (24), and German, as in (25): (23) hr shrt Rtnw on/after overthrowing Retjnu ‘on overthrowing Retjnu.’ (Gardiner 1988: 128, orthography simplified) (24) Ik ben aan het lezen I am on the read ‘I am reading.’

Late Merge 237

(25) Ich bin am lesen I am on-the read ‘I am reading.’

H&K give numerous examples from other languages. For instance, the preposition ‘in’ becomes a progressive marker in Lamang, a Chadic language, Vai, a Mande language, and Lezgian, a Caucasian language (pp. 178–9). Other languages they mention (2002: 202) as having locative markers becoming progressive are Imonda, Irish, and Chinese. There is of course also the use of Dutch te, as in (26), and similarly with German zu, and earlier English to, where the infinitival marker has a passive meaning. This means te is in Voice: (26) Dit is niet te eten this is not to eat ‘This can’t be eaten.’

So, the originally prepositional infinitival marker has many grammatical uses: ASP, Voice, M, and C.

2. From grammatical to grammatical: From (P to) ASP to M to C C(omplementizer)s do not change to I, and this is of course expected given Late Merge. The only change that is possible is from ‘I to I’, i.e. from ASP to M to T, and from I to C. In Chapter 2, I have given perhaps the most frequent example of a D(eterminer) becoming an I-element, namely a pronoun in D becoming agreement in T.3 This occurs in many languages, and is due both to Head Preference Principle (a phrasal subject becoming a head) and Late Merge (the subject merging late). In this section (in 2.1 and 2.2), I’ll show that to undergoes a change from P to I and from I to C. I will refer to the to that is in the expanded I position, as in (27), as the lower to and the to that is in C, as in (28), as the higher to: (27) It would be nice for them not to have to go there. (28) It would be nice to not have to go there.

I also consider the interaction with not, since not, as seen in Chapter 2, is changing from specifier to head. If not is a head, however, it stops to from moving to C. So, this situation is interesting since the two economy principles


go against each other, i.e. they bleed each other, to use early terminology. I discuss that interaction in Section 2.3. Section 2.4 addresses other possible positions of to, and Section 2.5 is a conclusion. 2.1 The lower to As is well-known from early work (e.g. Akmajian et al. 1979), modals, do, and infinitival to are in complementary distribution, and are all seen as occupying T (AUX or I[nflection] in that work). The proposals regarding modals in Chapter 8 do not change anything, since modals have to move from either ASP or M to T and would therefore not be able to occur together with to. In the early framework, there are only two functional categories, namely I and C, and arguments that to (and modals) are in I are therefore simple: (a) They are not part of the VP since VP-deletion and VP-preposing occur, leaving both modals and to, as in (29) and (30): (29) He wants to read and she wants to. (30) He can read and she can.

To, as in (31), but modals as well, can be split off from the VP by an adverb, assumed to be outside VP. Hence, modals and to are both in I: (31) I’ve had to actually drag it out of you. (BNC-KRP 278)

This split infinitival construction is of course prescriptively ‘incorrect’ and I’ll come back to that as an external reason that impedes the economy principle. (b) Another reason modals and infinitival to are seen as occupying the same functional category is that they are in complementary distribution, as in (32): (32) *I want to can do it.

Both modals and to indicate future (e.g. in headlines such as Bishop to go to jail) and unreal actions, so they should both be in the same position. Using these arguments, one of the positions that to can occupy is I. (c) To is not in C in (33) since an overt complementizer occurs together with to, as in (33): (33) I expected for them to do their homework.

Of course, there is a to that does occupy C, which will be the topic of Section 2.2. Thus, in a framework with just I and C, to in (29), (31), and (33) is in I. However, in the late 1980s, the I is split, and more positions become available.

Late Merge 239

Lower to can now be either ASP, M, or, T. ASP can be ruled out since to can precede aspectual material in English, as in (34), and hence T or M are left as possibilities: (34) It is a pleasure not to be going there.

Historically, the ASP position was not always excluded (see also Abraham 2002 for a lower position for Dutch and German ‘to’). Only after 1500 does to start to occur with aspectual auxiliaries. There is some additional evidence that to is in ASP in earlier stages, namely the occurrence of (35), repeated from Chapter 9, where to indicates progressive aspect: (35) I saw her corall lips to moue. (Shakespeare, Shrew I, i, 75)

Having eliminated ASP for ModE, there are two arguments in favor of to being in M, as in (38), over it being in T. These arguments involve the position of negative not below the T and also the higher to in C. The first reason is that sentences such as (36) are typical, not those as in (37). This is expected if to is in M, following not: (36) it would be inconsistent for Thatcher not to do this. (BNC-AA9 753) (37) %… for Thatcher to not do this.

A second reason, mentioned in Chapter 6 (Section 2.3), is that sequences such as to not to occur regularly (in Cobuild six times, in the BNC seven times, and in the CSE twice). These are cases where to is moving to the higher position but where its copy is not deleted. Cases of to to not do not occur in the BNC, Cobuild, CSE, and HC. If to were ever in T, the latter sequence would be expected. In (38) both possible positions for to are indicated. If the copy (in M) does not delete, both will be present:


(38) CP C¢ C

TP T¢

to

NegP

T

Neg¢ Neg

MP M¢

not M

ASPP ASP¢

to ASP

vP4

In (38), I have included T to clarify the possibilities but the fact that to to not never occurs points to to not stopping in T on its way to C. Hence, the correct structure is (39) without the TP (see also Stowell 1982):

Late Merge 241

(39)

CP C¢ C

NegP

to

Neg¢ Neg

MP M¢

not

M

ASPP ASP¢

to ASP

vP

Unlike modals and auxiliaries, to doesn’t cliticize to the negation. This means the negation has to be a specifier of NegP in (39) since otherwise it would block movement of the head to to C. Many people have commented on the differences in to: to in complements to control verbs, such as try, is irrealis, but it is dependent on the tense of the main verb with verbs such as believe. Stowell (1982) attributes it to the presence or absence of C. IJbema (2002: 105ff) argues for Dutch that there is a realis te in T (with an independent tense that is not unrealized or future) and an irrealis one in M. The latter is the case, I argue, in English, where to is not in T. Concluding 2.1, I have shown that lower to, as in (33), is in M. I’ll turn to higher to now. 2.2 Higher to There are, as noticed in Beukema & den Dikken (1989) and Bernstein (1994),5 reasons to assume a second position for to. Using data from corpora, I provide additional arguments that there are two positions, and that the ‘high’ position of to is C. As outlined in Chapter 3, in ModE, that and finite for are in Force not in Fin. The infinitival complementizer for, as in (38) of that chapter, cannot occur with a topic, and I argued that non-finite clauses do not project a split CP. This eliminates a number of possibilities for the position of to. If for is


present, I will argue below, to is not in C. Now, I’ll provide some examples of the to not construction. Since a number of linguists (e.g. Bernstein 1994; Kayne 2000) argue that to not, i.e. higher to, involves phrasal negation, I’ll provide some instances that show it doesn’t always (see also Section 2.4). Sequences of to not occur as complements to adjectives, as in (40), to nouns, as in (41), to verbs, as in (42), as adverbial clauses, as in (43): (40) It would be unrealistic to not expect to pay higher royalties. (BNC-CSS 245) (41) the freedom to not attend lessons. (BNC-AHG 1258) (42) He professes to not be ready for that. (BNC-CGB 1649) (43) to train the dog to not be afraid of people. (BNC-K54 6582)

Almost all of these have not to counterparts; some constitute minimal pairs: (44) It would be unrealistic not to show them to be human. (BNC-CBF 14312) (45) one of the best advertisements not to have an animal. (BNC-HV0 52) (46) He professes not to want the job. (BNC-ABJ 970) (47) We’ll train you how not to ‘blow it’. (BNC-CFV 2052)

Sometimes the reason for the choice of one over the other is structural, e.g. in (47), the C (or Spec CP) is occupied by how, blocking movement of to to C. There are a number of arguments that to in (40) to (43) is in C. One of the most convincing ones is mentioned in relation to (36) above, namely that to not does not occur with the complementizer for (or how in (47). For instance, in the BNC, (48) occurs 233 times (with pronominal and two-word subjects) but for followed by to not occurs only once in a passage with a number of peculiarities, namely (49): (48) she prefers for me not to stay on the phone for very long she does. (BNC-KPY 150) (49) it was an hour late, which it was late enough for us to not to know if it was coming, so we had to go and get another PA, otherwise we’d have gone to get, there was no other option. (BNC-J92 528)

In the (smaller) American corpus (CSE), there are ten sentences with for followed by a subject followed by not to, as in (50):

Late Merge 243

(50) if you as a parent choose for your child not to participate, you may opt out. (CSE-COMR6B97)

In this corpus, there are, however, also four apparent counterexamples. (51) is an instance of an ECM verb and it is often argued that the object moves higher to check Case (in AGRo). So (51) is not decisive. Checking want in the BNC, it occurs 26 times in a construction as (51) but always with not to. So, (51) may be an exception. The other three possible counterexamples are as in (52), and involve not only/just … but constructions where not is part of a VP conjunction, and phrasal negation is indeed at stake: (51) They don’t want them to not get them before their science and social studies. (CSE-COMR6A97) (52) You can look for the president to not only talk about legislative initiatives … but also to tie those to … (CSE-WH94)

So to when it precedes not is (generally) in complementary distribution with for. This means that in that position, it is in C. A second argument that the to of to not is in C involves verbs such as seem. It is known from independent tests that seem does not have a CP complement (raising and lack of an overt complementizer). If these constructions are C-less, they should not have the to not sequence, and this is indeed the case. For instance, in the BNC, forms of seem with not to, as in (53), occur frequently (namely 249 times), whereas there is one hesitant to not, as in (54): (53) At first, the darkening official mood seems not to have troubled Prokofiev. (BNC-ABJ 524) (54) pay schemes … seem to not y’know deal with that very effectively. (BNC-JTO 120)

Verbs other than seem are less clear. For instance, with complements to try and want, it is often assumed that there is an empty complementizer in sentences such as (55). If that is the case, does to occupy this C? The numbers are as follows: 782 forms of try have not to, and only 3 have to not, of which one is (56) with a double to: (55) You tried to go. (56) I mean I don’t try to not to take a lot of notice of it. (BNC-G5R 32)

This suggests that the C in complements to control verbs is an empty C, as argued e.g. in van Gelderen (1993: Chapter 5), indicating a future/unrealized


tense. Unfortunately, control verbs such as try cannot be compared with ECM-verbs, generally believed not to have a C, such as believe, guess, suspect, imagine, know, since none of the latter occurs as an ECM verb in the BNC or CSE in the relevant sense. Native speakers prefer (57) over (58), but that may be because of prescriptive reasons: (57) I believe him not to be a good guy. (58) %I believe him to not be a good guy.

To not occurs in a purposive context, e.g. as a complement to verbs such as hope, and as purposive clause in (59): (59) Luckily we are playing well enough to not be under much pressure at all lately. (BNC J1G)

A third argument that to of to not is in C involves the fact that sentences often end in not to, as in (60), but never in to not, as in (61). This can be explained if to is in C in (61): (60) because they desperately wanted not to realize it. (BNC-A69 1473) (61) *because they desperately wanted to not realize it.

The notion of ‘phase’, as in Chomsky (2001a) is helpful in explaining this since CP and vP are phases that can be deleted but TP (and ASPP) cannot. This means that in (60) either CP or vP is deleted. CP is unlikely since some remnant of the embedded clause remains. Therefore, vP is deleted and to is in T, or in M, or in ASP. The reason (61) is ungrammatical is that to is in C, and ASPP or TP is deleted. In the next section, I will examine which position the lower to is in. A last argument, mentioned already, that there are two positions is that in a number of modern corpora, (62) and (64) occur. This is less an argument about where to is than that there are 2 tos and that movement takes place: (62) — as a request to not to — (CSE-WH97A) (63) … is to try to not necessarily set those problems up. (CSE-COMM797) (64) This is to try to not to overturn the … (CSE-WH97A)

In (62) to (64), assuming the copy theory of movement as in e.g. Chomsky (1995; 2001ab), the copy of to is not deleted (cf. (38)). This construction is relatively frequent, e.g. in the BNC, it occurs 27 times, and native speakers find it reasonably acceptable. It is too consistent to be a ‘performance’ error that is independent of the actual structure.

Late Merge 245

2.3 Adverbs vs. Negation: Specifier vs. head Data from different corpora show that the split infinitive is relatively infrequent with negatives. For instance, in the BNC (the written and spoken part), there are 17381 instances of not to and 93 of to not, which means .54% is split. In just the spoken BNC, these figures are 1164 and 43, which is 3.6% split. In the (American) CSE, there are 381 instances of not preceding infinitival to, as in (65), and 59 of to not, as in (66), indicating that 13.4% of negative infinitives are split: (65) I was trying not to be specific. (CSE-COMM697) (66) … they chose to not use calculators. (CSE-COMM8A97)

These differences between spoken and written and between American and British point towards a prescriptive rule being at work. The relative numbers of (65) and (66) are very different for non-negative adverbs, e.g. always, actually. The typical pattern is the split infinitive with many, as in (67), and not the non-split one, as in (68): (67) You’re going to always need more. (CSE-COMR797) (68) So our view is always to find outstanding individuals. (CSE-WH96A)

There are some exceptions, as shown in Table 11.1. Lower adverbs such as completely are expected to always show a split infinitive, since to is in M and the adverb closer to VP. This is indeed the case, and whether or not to moves to C wouldn’t make a difference. The higher adverbs would be above M and would be expected not to be split. However, TP-adverbs such as probably when split show that to has moved to C (see Table 11.1). In the case of to not sequences, there is evidence that to is in C, and that’s true for other split constructions with TP-adverbs as well. In the entire BNC, for does not occur with to probably/to allegedly. As in the case of not, constructions with double to are found, namely (69) to (72): (69) the thing to do would be to probably to separate the six … (BNC-JS8 368) (70) To possibly to help to get those people around. (Cobuild) (71) Okay time to perhaps to relax. (BNC-F88 295) (72) I should be able to just to deal with this. (linguist in e-mail message)

Again, as with the to not to, these ‘errors’ occur in a specific pattern and are therefore not ‘just’ performance errors, but indicate a copy is not deleted.


Table 11.1.Split vs. non-split infinitives in spoken corpora

CP-adverbs frankly TP probably perhaps possibly v/VP always completely

BNC (spoken)

Cobuild

CSE

split

non-spl

split

non-spl

split

non-spl

1

0

0

0

0

3

5 37 4

17 45 5

19 17 4

17 37 3

8 6 1

0 5 1

35 3

27 0

40+ 6

18 0

8 1

3 0

Comparing the movement of to to C in negatives to that in non-negatives shows that the latter is more frequent. There is, however, much variation in the corpora: the BNC and Cobuild selections being only British spoken and the CSE American spoken. For instance, in the case of probably, to moves to C in 22.7% of the cases in the BNC, but in the Cobuild, this is almost 50%, and in the CSE a 100%. This may also have to do with not every speaker placing the adverbs as high. In conclusion, not to is preferred over to not but with non-negative TP-adverbs the pattern is different. The VP adverbs are expected to be split from to, and the CP-ones expected not to be split (and this is so in the case of CSE frankly, but not the BNC). The data predicted under Late Merge would be to always have to in C. In 2.5, I’ll discuss why this is not the case, but first I look at some other accounts. 2.4 Other accounts Kayne (2000: 301–3) provides an analysis of sentences such as (65) and (66) by using a difference in meaning. According to him, the split infinitive in (66) involves constituent negation, whereas the non-split one in (65) involves sentential negation (Bernstein 1994: 17; 20 argues the same). The derivations of (66) would be as in (73), and that of (65) as in (74), where WP and ZP are arbitrary projections: (73)

WP[

[trying tj]i tok ZP[ [not be specific]j tk [ti]].

(74)

WP[

[trying not tj]i tok ZP[ [be specific]j tk [ti]].

Late Merge 247

The relevant part of the derivation merges trying not be specific and to. After this, either not be specific, as in (73), or be specific, as in (74), moves to the Spec of to. Then, to moves to the head of WP and the remaining VP to the Spec of WP. The difference between the sentences is whether or not not is stranded. Even though Kayne’s account explains the VP-deletion data in (60) and (61) above, there are two problems with this account. One is that Kayne predicts a sharp difference between (75) and (76) on the one hand and (77) on the other. Only (77) is an impossible derivation since both the sentential and phrasal not would be preposed with promised: (75) John promised to not not do it. (76) John promised not to not do it. (77) John promised not not to do it. (Kayne 2000: 303)

With the right intonation, some native speakers accept all but some accept none. The BNC and Cobuild only have instances of (76) and in most of them there is doubling to not to not to, as in (78): (78) The aim is not to not to gain weight. (BNC-FL6 74)

The second problem is that multiple to, as in (62) to (64) above, cannot be derived using (73) or (74). In (79) and (8), which parallel (73) and (74) respectively, the copies of to are left but will never result in a to not to sequence: (79)

WP[

[try tj]i to ZP[ [not overturn the decision]j to [ti]].

(80)

WP[

[try not tj]i to ZP[ [overturn the decision]j to [ti]].

Kayne’s framework provides a way to first merge the lexical material and after that the functional material. Theoretically, this is very desirable, but does not account for (62) to (64) and (75) to (78). 2.5 Conclusion The conclusion reached in this section is that the behavior of not and TP-adverbs in relation with to shows that to moves from M to C on occasion. It can only do so when the C is present and not filled by for or another element. When C is empty, however, it is not clear why to moves on one occasion but not on the other, as in (81) and (82) respectively, and other minimal pairs listed above: (81) I’ll be trying to perhaps get invited to the meeting. (BNC-KS1 236) (82) I’d just like perhaps to seek a little clarification from Mr Curtis. (BNC-HVJ 169)


We’d expect Late Merge to take place to avoid having to merge as well as move. The problem for the ‘merge over move’ principle is to explain why to would not always be in the higher position. There are two possible reasons. One is that the rule against splitting infinitives is still very strong in standard English (see Fowler 1926 [1950]: 558; Quirk & Greenbaum 1973: 312). This is also evident from the difference between CSE and BNC. The second is that typically not is a head (see Chapter 2) and that in order to allow to to move across not, as in (40), not is forced to be a specifier. Unlike modals and auxiliaries, to doesn’t cliticize to the negation. This means the negation is a specifier of NegP in (40) to (43), but possibly a head in (44) to (47), the way it is in the case of modals, blocking movement to C. This ambiguity of not’s status still occurs in ModE since nothing else is in the specifier position (due to the powerful, prescriptive prohibition on double negation). The situation sketched above can be seen in terms of grammaticalization of an element from a lower to a higher functional category. In present-day English, both M and C are used but in earlier forms to is ‘lower’ (probably in ASP), as are the modals. It is expected that further changes will continue the development and have to become a C element.

3. Conclusion In this chapter, I have provided lists of the categories, both lexical and grammatical, that are the source for I-elements. The most frequent of these is the v/V becoming ASP and later M or T. Other frequent grammaticalizations are from A and P to ASP or M or T. D and N are infrequent sources. Their semantic features do not lend themselves to being reanalyzed as grammatical ones. Table 11.2 summarizes the possible changes, from this chapter and from Chapter 6, leading to C and to I.

Table 11.2.Grammaticalization as Late Merge (C and I can be split; % = unattested)

N> V> A> P>

I

C

% modal/PV MP/‘then’ on/aan/hr

till/and ‘say’ so/yet for/like

D> I> C>

I

C

% modal %

? to that (Fin to Force)

Part IV

Chapter 12

The Layer Parameter and Pronominal Argument Languages

The main thesis of this book is that principles of economy explain certain cases of (cross-clausal) grammaticalization. The present chapter is independent of this thesis. It tries to get into the ‘genius’ of a particular language, as Sapir puts it, or the ‘macroparameters’, as Baker does. It does, however, argue that the impetus of a macroparametric shift can be grammaticalization, which in turn is due to Economy. Based on the work by Jelinek (1984), Willie (1991), and Baker (2003a), I’ll argue that languages such as OE are partial Pronominal Argument Languages (hence PALs) and that the paratactic nature, as expressed by a certain setting of the Layer Parameter, follows. OE has many characteristics of a PAL: pro-drop, argumental verbal agreement, minimal embedding, and no special reflexives. It loses these features by ME. In Section 1, I’ll revisit the Layer Parameter from Chapter 1. In Section 2, I outline the typical characteristics of a PAL, and in Section 3, I show that OE has many of these traits, and, in particular, that PAL is connected to the paratactic nature of OE. In Section 4, I’ll briefly examine the switch from OE to ME to ModE and connect that to observations I have made regarding pronouns in Chapter 2. Section 5 is a conclusion.

1.

The Layer Parameter

In Chapter 3, it is shown that ‘older’ languages make less use of embedding, and that their topics are less integrated into the sentence structure. This means they make less use of a CP and can be described in terms of a macroparameter as mentioned in Chapter 1 and repeated here as (1): (1) The Layer Parameter CP/IP Italian, Modern English, Modern Greek VP Chinese, Chantyal, Nunggubuyu, Old English, Navajo


Languages such as Italian and ModE have been shown in Chapter 3 (and other chapters for English) to make extensive use of the CP and IP. Modern Greek does too if one considers the data by Poletto and others presented in Chapter 3 (Sections 1.3 and 2.4). Languages such as Chantyal and Nunggubuyu have been described as virtually without CP. Thus Noonan (1996) sees Chantyal, a TibetoBurman language, as not having finite subordination. Nunggubuyu, as argued in Heath (1984), is even more extreme in not having any infinitival or nominalized complements either. The reason I add (Mandarin) Chinese to (1) is that (except in questions and some conditionals) the VP layer, with v and inner aspect, is quite important as shown in Chapter 7. Traditionally, Chinese is not seen as having tense or agreement, so the IP layer may be lacking completely.1 The data from Navajo will be discussed below. In work such as Cinque (1999), all languages have the same array of functional categories. What I hope to have shown above is that that is not the case. What I’ll try to show in this chapter is that some languages are more ‘lexically-oriented’, i.e. towards the VP, and others more ‘grammaticallyoriented’, i.e. towards the IP. The CP in many languages is necessary to accommodate questions, topics, and other discourse information, but is not integrated into the grammatical system. I’ll now turn towards polysynthetic languages and show that they typically have no embedding, so the Layer Parameter is in fact a derivative parameter.

2. Characteristics of PAL/Polysynthetic Languages In the 1970s and 1980s, non-configurational languages were ‘discovered’ and became a challenge for generative linguistics. K. Hale’s (1976; 1982; 1983a) work on Australian languages shows that certain languages do not seem to have a hierarchical phrase structure, as in for instance English, hence the term nonconfigurational, and need not express arguments through overt nominals. K. Hale (1983a) argues that there is a Lexical Structure (LS) and Phrase Structure (PS), and that in non-configurational languages, the projection principle holds only for the former, so theta-marking is done in the Lexical Structure only. There is, however, still a certain amount of hierarchy in non-configurational languages, e.g. subjects and objects behave differently. Jelinek (1984) in dealing with Warlpiri argues that languages have either lexical or pronominal arguments. In non-configurational languages, “[c]litic Pronouns [are] Verbal Arguments” (1984: 43). Navajo is one such language, as are the Salish languages.

The Layer Parameter and Pronominal Argument Languages 253

Jelinek’s version of this difference/parameter is (2), which unlike Hale’s formulation, does not rely on the difference between LS and PS: (2) Configurationality Parameter a. In a configurational language, object nominals are properly governed by the verb. b. In a […] non-configurational language, nominals are not verbal arguments, but are optional adjuncts to the clitic pronouns that serve as verbal arguments. (Jelinek 1984: 73)

Making a similar point, Baker (1995; 2001) proposes the following macroparameter: (3) The Polysynthesis Parameter Verbs must include some expression of each of the main participants in the event described by the verb (the subject, object, and indirect object). (Baker 2001: 111)

Polysynthetic is used for a language that is “more than ordinarily synthetic” (Sapir 1921: 128). Initially, non-configurational languages are defined as having free word order (e.g. Hale) but then the emphasis shifts away from that (e.g. Navajo has relatively strict word order, and languages with free word order such as German can be accounted for through scrambling). I’ll use both PAL and polysynthetic. Jelinek (1984) notes that in pro-drop languages such as Spanish subject pronouns can be argued to be PAs. This, I will show in Section 2, is true in OE as well. Other polysynthetic languages vary where the incorporation of objects is concerned. For instance, among the Uto-Aztecan Cupan languages, Cahuilla marks subjects and objects on its verbs but Luiseño doesn’t mark objects (see R. Jacobs 1975: 105). A noteworthy characteristic of PALs is the absence of specifiers in the core domain. A PAL, as it were, is in complete accordance with the ‘Head-over-Spec’ Principle. Partial PALs allow more use of specifiers as arguments. Characteristics of PAL languages, as in Baker and Jelinek, are (a) optionality of nominals (DPs as well as independent pronouns), as in (4), for Navajo, (b) sentences with more than one nominal are rare, (c) nominals as in (5) are adjuncts, sometimes with a different case system (e.g. Jelinek 1989), (d) absence of anaphors and non-referential quantified DPs, (e) minimal embedding:


(4) Nanishté na-ni-sh-té around-you-I-carry.impf ‘I am carrying you around.’ (5) (Diné bizaad) yíníshta’ (Navajo language 1-study ‘I am studying Navajo.’

The optionality of nominals, as in (4), follows if they are adjuncts, and so does their being specially Case marked in some languages. Regarding (d), Baker (1995: 49f.) makes the point that anaphors such as ‘himself ’ would be adjuncts and hence outside the c-command domain of the real subject. As to (e), Hale (1989) notes that (non-)configurationality is confined to constructions, not languages, and notes that sentential complements such as (6) and (7) in Navajo have to be configurational, even though Navajo as a whole is non-configurational. These constructions occur infrequently and are still paratactic (e.g. notice the embedded first person): (6) Shi-zhé’é kinla’nígóó deesháál nízin my-father Flagstaff-to 1-will.go 3-want ‘My father wants to go to Flagstaff.’ (from K. Hale 1989: 300) (7) doogáál ní 3-arrive 3-said (disjoint reference) ‘He said that he arrived.’ (from Willie 1991: 143)

Baker (1995: Chapter 10) and Mithun (1984) say that polysynthetic languages avoid embedded arguments, rather than saying that configurationality is linked to constructions. Constructions such as (6) are rare in Navajo; the preferred embedding strategy being nominalization, as in (8) for instance: (8) honeesná-nígíí yoodlá 3.win-nom 3.believe (free reference) ‘He believes he won’ or ‘he believes the winner.’ (Willie 1991: 178)

In Section 1, I argue that certain languages are more ‘VP-prominent’, or VP-oriented. The actual structure for Navajo is not just an expanded VP-shell but most likely has inner and outer aspect and subject and object agreement. A possible tree for (4) looks like (9), with only heads:


ASPP

(9) na

vP sh

v¢ v

VP ni

V té

So the main characteristic of a PAL is its VP dominance, and that is expressed in e.g. K. Hale’s (1983a) LS, but also in (9).2 OE (and German, see e.g. Haider 1989) have been described as languages in which arguments are marked morphologically and this is partly lexical, depending on the nature of the verb.

3. OE as (partial) PAL In this section, I’ll show that OE is like a regular pro-drop language such as Spanish in terms of the subject being optional, so that it is a limited PAL in that respect. A number of other features are shared with regular PA languages. Faarlund (1990: 82ff.) examines Old Norse as a non-configurational language that has pro-drop as well, but I will just consider OE. OE verbal subject agreement is argumental. Evidence for this can be found in pro-drop, as in (10) and (11), and overt verbal agreement, e.g. the -est ending in (10): (10) ær ðon ðe hona creawa ðriga mec onsæcest before that that rooster crows thrice me-acc deny-2S ‘You will deny me three times before the rooster crows.’ (Lindisfarne Gospel, Matthew 26.75) (11) þæt healreced hatan wolde | medoærn micel men gewyrcean that palace command would meadhall large men to-build ‘that he would order his men to build a big hall, a big meadhall.’ (Beowulf 68–9)

Instances of (10) and (11) are especially numerous with third person singular and plural. This use occurs up to the mid 13th century. Berndt (1956) estimates


that in some OE texts only 20% of subjects are overt (see also van Gelderen 2000: Chapter 4). OE verbal agreement distinguishes person and number separately. In addition, subjects in the traditional sense are often optional, as mentioned in Chapter 5 (Section 4), and a topic appears, as in (12) and (13), from the 15th century, without a clearly integrated structure: (12) As for þe toþer.tway enemyes. wich ben … seruauntes to hem. […]. mowe sone be ouer come. whan here lordis and maystris ben ouercome ‘As for the other two enemies, which are servants to them … [They] must soon be defeated, when their lords and masters are defeated.’ (The Tree of xii frutes 149.11–14) (13) As for the secunde þinge wiche longith to a religious tree þat is plantid in religioun: is watering ‘As for the second thing which pertains to a religious tree that is planted in religion is watering.’ (Idem 5.8)

Baker (1995: 53) shows that polysynthetic languages such as Mohawk do not have object reflexives. The absence of reflexive pronouns in OE is well-known (e.g. Faltz 1985; van Gelderen 2000); simple pronouns, as in (14), function reflexively instead: (14) Ic on earde bad | … ne me swor fela I on earth was-around not me-dat swore wrong ‘I was around on earth … I never perjured myself.’ (Beowulf 2736–8)

Quantifiers or, as Lightfoot (1979) calls them, pre-quantifiers, are quite complex in OE and ME. They are inflected as adjectives; many have an adjectival meaning, e.g. eall means ‘complete’; and some have pronominal functions. This shows they are more referential in keeping with what is known about quantifiers in polysynthetic languages. If subject pronouns are adjuncts in OE (and the agreement is the real argument), how are subjects Case marked? It can be argued that there is a split Case system for subjects and objects, with objects being assigned inherent Case (by the V, Adj, or P), and subjects default nominative Case. Thus, in (15), repeated from Chapter 8, the object him is dative because the verb forscrifan ‘proscribe’ assigns a Goal theta-role. Inherent Case depends on the theta-role and can be genitive, dative, or accusative. As a result, (transformational) passives do not occur in OE: (15) siþðan him scyppend forscrifen hæfde since him-dat creator banned had ‘since the creator had banned him.’ (Beowulf 106)


In OE, subjects are adjuncts with default nominative and objects have inherent Case. Thus, OE Case is interpretable (in terms of Chomsky 1995) and so is agreement. The loss of inherent Case around 1200 triggers checking of uninterpretable features in functional categories and a change in the Layer Parameter. Checking of uninterpretable can be done in two ways: Head–Head and Spec– Head. The former is more economical but only possible for heads, i.e. pronouns. This, however, means that ModE pronouns are very similar to agreement markers in polysynthetic languages. Up to this point, I have assumed that objects are argumental. The reason there can’t be object reflexives is because the subject is pronominally marked on the verb and too low to c-command the reflexive object. Is there any evidence that the object is pronominally marked on the verb as well? There is minimal object agreement in OE on past participles, but mostly in number (see e.g. (44) of Chapter 8), never in person. This is not enough to identify the argument. Object-drop appears in rare cases (cf. van Gelderen 2000:147–9; Visser 97–188), but this is not comparable in numbers to subject drop. There is one possible object marking on the verb, namely the intransitive-transitive alternations, as in (16), and one could argue that the prefix is object agreement or E, identifying an object. I will leave that question for further research: (16) gan ‘to go’ restan ‘to rest’ winnan ‘to labor, toil’ (from Visser 126)

gegan ‘to overrun, subdue’ gerestan ‘to give rest’ gewinnan ‘to gain, conquer’

I’ll now turn to the feature of PALs that is most connected to the rest of the book, namely the lack of embedding (and of the absence of more than one auxiliary).3 In PALs, clausal arguments are unexpected since it would be difficult to represent them on the verb. As has been shown in Chapters 3, 4, 5, and 6, older languages have a more paratactic structure, e.g. as in ME (17), repeated from Chapter 3. (The punctuation of (17) has been checked against the facsimile of the manuscript): (17) An preost was on leoden. laŠamon wes ihoten. he wes leouenaðes sone. liðe him beo drihten. he wonede at ernleŠe. at æðelen are chirechen A priest was among people. Layamon was called. He was Liefnoth’s son. kind him be God. He lived at Areley. at lovely a church ‘There was a priest living here, called Layamon. He was the son of Liefnoth, may God be him kind. He lived at Areley, at a lovely church.’ (Layamon, Caligula 1–3)


Vachek (1961: 38) compares an Alfredian OE translation to the Latin original and finds many more verbs and independent clauses in the OE, which fits with the general paratactic nature. A feature of PALs that hasn’t received (as) much attention is the lexical incorporation of aspectual information, such as ‘around’ in (4) and the prefixes in (16). In Chapter 11, it was shown that inner aspect is typical for OE. Again, the VP is accorded more prominence than the grammatical structure. In this section, I have indicated that a number of the key characteristics of PA languages also hold for OE. A feature OE does not have is noun-incorporation, but a number of core PALs do not either (e.g. Navajo), so I won’t regard that further. In Chapter 1 and Section 1 of this chapter, I mentioned the Layer Parameter, but if the above is correct, the choice of focusing on the VP really follows from the PA Parameter. The use of aspectual prefixes (represented in VP), pro-drop, and verbal agreement also does. The big question is what causes the shift towards ME. I will argue it is the Economy Principles that bring about cross-clausal grammaticalization. Spec to Head and Late Merge ‘cause’ changes in clausal structure such that the paratactic nature is lost.

4. The switch to Middle and Modern English The changes that take place between OE and ME/ModE change a synthetic language into an analytical one. Verbal agreement and pro-drop are lost; topics are more tightly integrated; functional categories are introduced responsible for the checking of the, now, structural inflections; reflexives are introduced; and CPs are used more generally. Approximate dates for a few of these changes are as in (18) (see van Gelderen 2000 for more): (18) a.

loss of agreement and pro-drop: 950–1230 (the dates of the Lindisfarne Gospels and Katherine respectively) b. topic-incorporation: 15th century (e.g. Paston Letters) c. reflexives: 15th century (e.g. Paston Letters) d. the increased use of CP and the introduction of a split CP: late 15th century (e.g. York Plays for the split CP).

If we see the loss of agreement and pro-drop as an indication of the loss of PA-hood, the PA-parameter is reset earlier than the Layer Parameter, as mentioned before. The Layer Parameter may be a description of some of the ramifications of the PA-Parameter, i.e. not an independent one. This is formulated in (19):


(19) PALs and Layers In PA Languages, the VP Layer is more prominent than the CP one.

As argued in Chapter 2, ModE is moving towards a stage where pronouns count as agreement, i.e. again towards a PAL where subject arguments are concerned. There are, however, puzzling developments. For instance, Weber (1934: 103) argues English can no longer be called Indo-European in type (he calls it Bau ‘build’). His arguments mainly involve the isolating/analytic character of English simple sentences. Vachek (1961) focuses on other aspects of the “analytical trend”, namely the use of infinitives and participles (-ing as well as -ed), by comparing a synthetic language such as Czech with English. In a Czech translation of an English work, Vachek (1961: 33) finds only 75 non-finite ‘condensers’ where the English original has 168. These developments go against those of the pronoun being incorporated. I will leave the typological description of ModE for future research.

5. Conclusions In this chapter, I have argued that OE is a partial PA language, since it has (subject) pro-drop and verbal agreement. The latter two are effectively lost by 1200, and other changes are taking place: reflexives are introduced (possibly because the subject argument now c-commands the reflexive) and CPs are used more and are expanded through Spec-to-Head and Late Merge. The latter changes take place over a longer period but find their culmination in the 15th century. In PALs, the VP is more prominent and when Late OE and ME lose PA-hood, the Layer Parameter is reset as well to having IP and CP more prominent.

Chapter 13

Conclusion

The main focus of this book has been to account for cross-clausal grammaticalization in terms of two Economy Principles. In this concluding chapter, I will list the main results. In Section 1.1, I’ll summarize the cases that show that Economy Principles account for grammaticalization processes. Several grammarexternal processes may interfere, however. In Section 1.2, I will present a list of some of the individual changes talked about in the book. In Section 1.3, I’ll discuss the implications of this book for the analysis of clauses. Section 2 clarifies some apparent counterexamples to the theory of changes articulated here. Section 3 discusses how changes in principles are different from changes in parameters, and returns to the definition of these two terms.

1.


1.1 Head Preference, Late Merge, innovation, and prescriptivism In the derivation of a sentence, a number of lexical (and grammatical) items are selected and merged into a structure of the type shown in Chapter 2. After the merging of the items, some move to check features. I have argued that two Economy Principles help in the merge process, one is that it is better to merge and move as a head than as a phrase, and the second is that if an element has to move, it is better to wait than to merge early and have to move several times. These are synchronic processes, at work in the production of utterances by native speakers, and in Chapter 2, I have provided several arguments that this is indeed the case. In the remainder of the book, I have shown that these Principles are also at work in language change. The examples of the Head Preference Principle in Chapter 4 involve the incorporation of one CP into another, i.e. two independent clauses become connected. A demonstrative in one clause becomes a relative pronoun in the specifier of another CP and the dependence of one clause on the other is the result. Adverbs such as oþ ‘until’ undergo the same process from adverb in one


clause to marker of dependence in another clause. Not much attention has been paid to these specifier to head changes in the grammaticalization literature, even though both negation (Jespersen’s Cycle) and articles go through this. In both the grammaticalization and generative literature, more attention has been paid to head to head changes, for instance, prepositions becoming complementizers. These changes I argue (but see below for some possible counterexamples) always involve changes to a higher position in the tree structure. Often, these higher positions are ones that ultimately the grammaticalizing element would have had to move to. These changes can be argued to be brought about by Late Merge and affect both CP grammaticalization (in fact the introduction of the split CP is ‘helped’ by the change of the complementizer from Fin to Force), as well as IP grammaticalization. Instances of the latter are modal verbs changing to ASP, perception verbs becoming ASP, and infinitival to changing from P to ASP to M and then later to C. In Chapters 6 and 11, I systematically go through all possibilities, i.e. the lexical categories N, V, P, and A to the functional categories C and I; and the functional categories D, C, and I to higher functional categories C and I (where C and I can be expanded). If the Economy Principles ran their course unimpeded, the array of lexical items would soon be depleted and all specifiers would be heads. This is where Jespersen’s ‘tug-of-war’ between ease, i.e. Economy Principles (internal change), and innovation and resistance to change (external change) becomes relevant. The innovations that introduce new specifiers are the new negative phrases such as nothing (Chapter 2) and new wh-relatives (Chapter 4). New heads continue to emerge as well, for instance, prepositions such as till, for, like and to grammaticalize to complementizers (Chapters 5, 6, and 11) and new adverbs mark aspect (Chapter 10). The prescriptive rules impeding internal change involve the ban against the split infinitive (stopping to from moving higher, Chapters 6 and 11), multiple negation as an affirmative (stopping a second negative from agreeing in Spec NegP, Chapter 2), like as being restricted to prepositional use (stopping Late Merge, Chapter 6), and not ending a sentence with a preposition (favoring a wh- over a that-relative and thereby stopping Spec to Head, Chapters 2 and 4). 1.2 Individual changes I now provide a list of the major changes, first from Spec to Head and from Head to Head, then those involving the incorporation of a phrase as a Spec, and finally those involving a parametric change.

Conclusion 263

Spec to Head: a. Pronouns > agreement = head of IP (Chapter 2) b. Neg Words > negation = head of NegP (Chapter 2) c. Demonstratives > complementizers = head of CP (Chapter 4) d. AP and PP > complementizers = head of CP (Chapters 4 and 6) e. Dutch particles > aspect = head of EP (Chapter 10) f. AP and PP > outer ASP = head of ASPP (Chapter 10) Head to Head: a. For, from P > C (Chapters 2 and 5) b. That, from Fin > Force (Chapters 4 and 5) c. Till, from P > Force (Chapter 5) d. Like, from P > C (Chapter 6) e. To: P > ASP > M > C (Chapters 6, 10, and 11) f. Modals and do: v > ASP (Chapter 8) g. Have: v > ASP (Chapter 8) ASP > T (-a-) (Chapter 8) h. Saw: v > ASP (Chapter 9) i. Ge E > ASP > Ø (Chapter 10) j. Adverb > ASP (Chapter 10) Becoming Spec: a. Demonstrative > Spec CP (Chapter 4) b. Topic > Spec CP/TopP (Chapter 5) Parametric and Macroparametric: a. Tense Aspect Parameter: reset in C19 b. PA and Layer Parameter: reset in ME 1.3 Clauses and layers In this book, clauses have been considered to consist of three layers, CP, IP, and VP. The Layer Parameter (as discussed in Chapters 1, 3, and 12) expresses the fact that not all languages make use of the layers in the same way: CPs can be split or not, IPs contain T, M, and ASP positions (at least in English), and VP shells may have inner aspect. Non-finite complement clauses are typically not split since the matrix verb selects the features on its sister and Force would block that selection (Chapter 5, Section 2).


In Chapter 1, the direction of cross-clausal grammaticalization has been sketched that is observed cross-linguistically; for instance, certain full clauses tend to be reduced to infinitives. Both Economy Principles predict the tightening of the connection between two clauses, through the change from verb to auxiliary and through the introduction of new complementizers. Possible counterexamples are discussed in Section 2 below. In Chapter 3, I have provided a classification of verbal complements based on the work of Kiparsky & Kiparsky and Hooper & Thompson. I argue that assertive and non-assertive verbs have different complements: assertives just have a complementizer in Force, whereas non-assertives have mood features, in addition to a complementizer (see Chapter 3, Section 4.3). The mood features in C make the complement less transparent. Hence embedded topicalizations, raising, and ECM are harder with non-assertives. When the complements undergo grammaticalization, the result is different as well. In the case of assertives, there is less to reduce, so the typical reduction is an infinitive; in the case of non-assertives, the reduction can be a subjunctive or a for-to infinitive. In Chapter 7, I have examined the IP and VP and how they differ. Because the changes of a head to a higher head are slow, the changes from V to I are also, i.e. from bi-clausal to mono-clausal structures, and there are therefore inbetween stages. In He managed to go there, manage might be an auxiliary or a main verb. I use proximity to to and obligatory control as diagnostics of auxiliarihood. Chapters 8 and 9 show how clause boundaries are lost when V changes to I.

2. Apparent counterexamples to uni-directionality There are inevitable counterexamples to any proposal, and in this section I’ll try to account for some. In 2.1, I will first look at a few heads that seem to go from C to P and heads that become specifiers when a pronoun becomes a full noun. These are violations of Late Merge and of the Head Preference Principle respectively. I’ll also examine some work by Klamer and H&K that seems to suggest violations of Late Merge. Then, in 2.2, I look at changes involving subjunctives. These present a possible challenge to the clausal stages mentioned in Chapter 1, but are not a direct challenge to the Economy Principles. Here, too, I suggest the violation is only apparent.

Conclusion 265

2.1 High heads to low heads and heads to specifiers? As discussed in van Gelderen (1997a), man seems to change from noun to pronoun to noun and for from P to C to P. These are counterexamples to grammaticalization going from head to specifier and from high (Late Merge) to low in the tree structure. The reason for these apparent reversals, I will argue, is that in many changes of say P to C, the P-use remains, and can always be ‘activated’ again. The change of man/men seems to go from lexical noun to indefinite pronoun and then back to noun since third person singular features are gained and lost again. In texts from the early 13th century, me(n) is third person singular, as in (1), hence a pronoun, whereas it is variable in Chaucer’s late 14th century variety, as the plural shows in (2) (for more see van Gelderen 1997a: Chapter 9) and is therefore a full DP with an empty D: (1) for men hit saide wel i-war because men it said-S everywhere ‘because it was said everywhere.’ (Layamon, Otho 6869) (2) Thurgh which men myghten any light discerne through which men might-p any light discern ‘Through which one might see light.’ (Chaucer, Knight’s Tale 52.1989)

Perhaps the most famous instance of a reversal is English for. As a preposition of location, it initially assigns Case to a DP, and loses this Case assigning property as it becomes a complementizer around 1200, as in (3), discussed in Chapter 2. It, however, starts to be complemented by DPs again around Chaucer’s time, as in (4). This might look as if for changes from a higher C to a lower P: (3) þet is umbe forte leaden that is about to lead ‘which is designing to lead.’ (Hali Meidhad, Bodley 4/24) (4) preyen for folk to avauncen hem ‘plead for people to promote them.’ (Chaucer, Parson’s Tale 315.785)

Rather than consider these counterexamples, I argue that the ‘reversal’ with both men and for is possible because the older meanings remain in the lexicon, e.g. even today for is also still a preposition and man a noun. Apart from English, there are a number of languages where a change from head to head seems to have gone in the opposite direction from that expected. For instance, in Ewe, Haitian, and Lingala, the verb ‘come from’ seems to have


evolved into a preposition ‘from’, according to H&K (2002: 71). In Pilara and Lahu, the verb ‘come’ has become a benefactive (H&K 73). I have not looked into these further, but they may again be cases of ‘reactivation’. Many consider serial verbs to be different from auxiliary-verb sequences since one of the ‘verbs’ in a serial verb construction may be homophonic with a preposition, e.g. the Mandarin Chinese co-verb construction as in (5) and (6): (5) women chang dao Niuyue qu we often go/to New York go ‘We often go to New York.’ (from Li & Thompson 1981:363, tones left out) (6) ta gei Lisi mai shu he give Lisi buy book ‘He bought Lisi a book.’

On their own, dao and gei can be used as verbs, meaning ‘to arrive’ and ‘to give’, but they are used here together with other verbs and sound to English ears more like prepositions. Li & Thompson (1974: 274–77) provide a list of 54 of these co-verbs, “prepositions which are still in the process of changing from earlier verbs” (p. 274). As a result of this, some of these are more verbal than others. If one thinks of preverbs such as dao as ASP or v, there is no violation of Late Merge since the P does merge higher. An example that is not quite a counterexample yet but that may become one is that presented in Klamer (2002). She mentions the case of wàngu ‘use’ in Kambera, a Malayo-Polynesian language, where she sees the following stages: full verb, compound verb, preposition indicating instrument, clause linker, raising verb. The stage from V to P would be unexpected. It is, however, quite possible that the examples of P are ambiguous, as in (7): (7) ku-palu wà-nja ài da ahu-mu nyumu I-hit wangu-3p.dat wood art dog-poss you ‘I hit your dogs with a stick.’ (Klamer 2002: 16, glosses slightly modified)

Klamer shows that the reduced form wà, as in (7), still has a verbal ending agreeing with a definite object and that therefore it is not clearly prepositional. There are, however, cases where wàngu is ambiguous. With indefinite objects, verbs are not marked with the clitic of the indefinite object and hence in (8), wàngu could be reanalyzed as a P: (8) Ngangu kokuru juaa ngangu wàngu tolung eat coconut only eat with meat ‘eat coconut only, with meat.’ (Klamer 2002: 17)

Conclusion 267

Even though this could happen, it hasn’t yet. The stage where wàngu moves towards I, as in (9), would also be unexpected if we only took into account one clause. Considering the structure as biclausal, wàngu can be seen as the V of a higher clause or possibly the T (the status of pa is unclear),1 so it is merged higher: (9) talanga la andaka nyungga hi na-wàngu pa-urang while loc road I conj 3-wangu subord-rain ‘While I was on the road, it began to rain.’ (Klamer 2002: 23)

In short, the counterexamples to Spec to Head and to Late Merge that I have discussed above are not real ones. 2.2 Subjunctives and infinitives In (2) of Chapter 1, a scenario of cross-clausal grammaticalization is given which is uni-directional. Many of the changes are explained in terms of Economy, but not all have been discussed. One of these is the subjunctive. Counterexamples to the uni-directionality of the changes would be (i) subjunctives changing into indicatives and (ii) infinitives changing into subjunctives or indicatives. I have not really talked about subjunctives (except briefly in Chapter 3) since changes in them do not involve the Economy Principles. Here I will argue regarding (i) that the subjunctive is not replaced by the indicative in the history of English, contrary to common views (see Visser 788), but rather by an infinitival so that most change is going from (2d) to (2e), and regarding (ii) that, when the change from infinitive to inflected verb happens, e.g. in Greek, it is really a loss of inflection that results in (2c), independent of the stages in (2) of Chapter 1. I’ll say something in general first. Subjunctives are ‘weakened’ CPs. This means they never show a split CP and have a dependent mood. The mood can be dependent on the matrix verbs, as in (10), or be dependent on a sentential negation or question, as in (11), from French: (10) I insist that he go. (11) Je ne croit pas qu’il aille là I not believe not that-he go-subj there ‘I don’t believe he’ll go there.’

Quer (1998: 32), following unpublished work by Stowell, calls (10) and (11) intensional and polarity subjunctives respectively. Many languages show no


overt subjunctive and can still be argued to have it since the complements are more transparent, as in (12) from Serbo-Croatian, where the clitic ga optionally moves to the higher clause (see Chapter 7) and where negative polarity items in the ‘subjunctive’ are licensed by a negative in the matrix: (12) Milan želi da ga vidi Milan wants that him sees ‘Milan wants to see him.’ (from Progovac 1993: 119)

Uchibori (2000) shows that certain kinds of subjunctives in Japanese allow A-movement and accounts for this in terms of Chomsky’s phase theory: defective tense in certain subjunctives results in a non-phase and hence allows extraction. Compared to older varieties of English and German, ModE makes little use of subjunctives. Fowler (1926 [1960]: 574) says that “it is moribund” and most modern ones “are either deliberate revivals by poets … or antiquated survivals as in pretentious journalism”. This loss of the subjunctive in English seems to be a change from (2d) to (2c), and hence a counterexample to (2) in Chapter 1. I will argue this is not correct and that the change is from subjunctive to infinitive. OE has a specially marked form for the subjunctive, but this is replaced by the past or should in Late OE. It is not clear that the subjunctive is ever obligatory (see Chapter 3). Many Renaissance grammarians do not describe the ‘subjunctive’, as in (10) above, as a subjunctive but as a bare form in discussing we shall overtake him though he run: “may we not suppose that the word run is the radical form (which answers to the infinitive mood in other languages)” (Priestley 1761 from Visser 788). Consider the history of typically subjunctive clauses. For instance, the earliest instances of the verb require, a French loan word, are from the fourteenth century (according to Visser and the OED). Require is a verb that traditionally takes a subjunctive complement, as in (13) and (14): (13) requeir … That Šhe pres you ‘require that she press you.’ (1375 Barbour) (14) I require yow here, þat Še me telle with trawþe ‘I require that you tell me the truth.’ (from the OED entry)

The complements are subjunctives. The first infinitival complement in Visser (and the OED) is later than the subjunctive one, namely c1420, as in (15). The earliest in the HC are 15th century Malory (Morte d’Arthur) and Caxton:

Conclusion 269

(15) I the requere … Nat to arecte to resumpcioun ‘I require you not to lift up declaring (something) void.’ (Lydgate, Troy I, 2303, from Visser 2286)

So, these facts show that stage (d) becomes (e) and not the other way around. In a modern spoken corpus of American English (CSE), require has 8 clear subjunctives, 1 indicative, 47 infinitives (one of which with the complementizer for), and an -ing complement. This shows that the indicative is not a real replacement at all, and that the infinitive has taken over from the subjunctive. Another instance is important, which is complemented by a visible subjunctive, as in (16), in 25 instances, and by an indicative clause, as in (17), in 10 instances in the same corpus. The high number of subjunctives in this corpus is due to its relatively formal nature. The subjunctive is still taught and therefore still used in formal spoken English: (16) it’s very important that it represent a statement … (CSE-COMM597) (17) it is important that people are aware … (CSE-COMM 697)

It is hard to compare this use to the use of a for-infinitive, as in (18), which occurs 77 times, or a to-infinitive, as in (19), occurring 176 times, since subjunctives versus indicatives are only clearly marked when the verb is be, with negatives, and with third person singular verbs:2 (18) it’s important for those differences to be bridged. (CSE-WH97B) (19) But it is probably important to really deal with it. (CSE-COMR6A97)

There is a total of 118 instances of that clauses following important, but only in 35 cases is it clear whether the verb that follows is indicative (in 8.5%) or subjunctive (in 21%). The trend is clear, however, the infinitive is a more likely ‘replacement’ for the subjunctive than the indicative is. The complement to each verb and adjective has to be looked at separately since they all differ. As mentioned in Chapter 3, the history of important is interesting in that its use as a predicative adjective with a complement is very late. It too is a loanword from French but quite a late one, namely late 16th century (the verb import is from the early 16th century), and is initially only used attributively. Even 18th and 19th century texts, such as (George) Berkeley, (David) Hume, (Emily) Bronte, and (Jane) Austen only contain instances of attributive use, so the subjunctive use, as in (20), repeated from Chapter 3, must have arisen late. Some early American texts (e.g. Samuel Adams’s 18th century writings) just have subjunctives, as in (20), not yet infinitival complements, and therefore the direction from Chapter 2 is adhered to:


(20) Some of our military gentlemen have, I fear, disgraced us; it is then important that every anecdote that concerns a man of real merit among them, and such I know there are, be improved, as far as decency will admit of it, to their advantage and to the honor of a colony, which, for its zeal in the great cause, well as its sufferings, deserves so much of America. (Samuel Adams, letter to E Gerry 1775, Cushing ed.)

Turning to Greek, a very well-known counterexample to (2) of Chapter 1 is the loss of the Balkan infinitive (as described e.g. in Joseph 1983). In several South Slavic languages, as well as in Greek and Romanian, the earlier infinitive, as in (21a), is replaced by an inflected form, as in (21b): (21) a.

Persai gar theon nomizousi einai pur persians for god-acc think-3p be-inf fire-acc ‘For the Persians think that fire is a god.’ (Herodotus 3.16, from Joseph 1983: 39, but without accent marks) b. nomizo oti/pos i fotia ine theos think-1S that the fire is god-nom ‘I think that fire is a god.’ (also from Joseph)

In (21b), there is a choice between a slightly more formal oti and a more colloquial pos. These follow verbs such as ‘think’ and ‘say’. Verbs of volition, as in (22b), are followed by na with an inflected verb in Modern Greek but also derive from an infinitive as in (22a): (22) a.

pant’ ethelo: domenai all want-1S give-inf ‘I am willing to give all back.’ (Iliad 7.364, from Joseph 1983: 40) b. thelo na tin do want na her-acc see-1S ‘I want to see her.’ (again from Joseph)

Joseph is not concerned with grammaticalization in his 1983 work but does wonder how this trend from infinitive to inflected clause occurs. His cautious suggestion is that “losing the category infinitive would be a first step towards doing away with all finite forms and categories” (p. 214; see also Joseph 1985: 390). This would mean that the inflected infinitive is not really comparable to the infinitives mentioned in (2) of Chapter 1. Banfi (1990) argues that the infinitive in most Balkan languages is still alive and well and only in some dialects (due to sociolinguistic reasons) moribund. Jakab (2001) provides a monoclausal analysis for subjunctive complements in Serbo-Croatian and this

Conclusion 271

would solve this problem, since the subjunctive would be a more grammaticalized form than the infinitive. In short, there are a number of exceptions to the general direction of change sketched in (2) of Chapter 1, namely subjunctives and infinitives becoming indicatives and subjunctives respectively. In Section 2.2 of this chapter I show this sequence of events is not correct and they in fact follow the direction outlined in Chapter 1. There are also a few counterexamples to Late Merge, discussed in Section 2.1. Most of these can be explained by the old, nongrammaticalized, form never having ‘left’ the grammar.

3. Principles and Parameters In this section, I’ll compare changes in parameters, which are typically fast, to changes involving principles. The changes I discuss in this book mainly involve principles. Economy Principles are absolute and apply whenever possible. There is no definite evidence that pronouns are less often coordinated, hence more often heads, than they were 400 years ago. This is expected if principles are invariable. The slight changes are due to the changing status of pronouns, from lexical to grammatical. This book also discusses two parameters, although very different ones. A ‘traditional’ parameter is the Tense Aspect Parameter, discussed in Chapter 10, which is reset from having aspect as unmarked to having tense as the default in the nineteenth century. I also use the Layer Parameter but, as mentioned before, I employ this term in the sense of Baker (1995). Macroparameters do not change overnight, unlike the parameter discussed in Chapter 10. In this section, I will discuss an additional ‘traditional’ parameter and contrast the speed between change due to parameters and that due to principles. Lightfoot (1999) argues that changes due to the resetting of parameters are fast, ‘catastrophic’ in Lightfoot’s terms. I will briefly provide some data about the loss of morphological Case, a parameter change (in the traditional sense) from morphological Case to checking Case in a functional category such as I(nflection), see also Kiparsky (1995) and Lightfoot (1999). The first person pronoun is the first to lose the special morphological Case. Since the change is fast, corpora data cannot ‘predict’ it, and the changes are indeed parameter switches. I then compare this to changes in pronominal phrase or head status. In Old English, nouns and pronouns can be marked genitive, dative, or accusative depending on the character of the governing verb, as explained in


Section 3 of the previous chapter. This Case is not related to position but to theta-marking. The change in Case is related to the change from a PAL to a non-PAL. If agreement marking in Old English actually represents an argument, as in PALs, checking in a higher functional category is not necessary. As endings disappear, checking becomes necessary and overt arguments occur. I now show what the data in the HC say about the Case change. As shown in Table 13.1 below, between OE1–2 and OE3 of the HC, i.e. around 1000, the relationship between the specially marked accusative mec and the dative/accusative me remains stable. An example of both is given in (23) from the Vespasian Psalter: (23) ða ðe swencað mec … monge arisað wið me that that oppress me-acc many rise with me-dat ‘that oppress me … many rise with me.’ (Vespasian Psalter, 3.1, Kuhn edition)

By OE4, again shown in Table 13.1, i.e. from 1050 on, mec has disappeared, however. So, even though the ratio between the two forms stays the same between OE1 and OE3, after that, suddenly, one form disappears. This means that the change is quick and the corpus does not help predict the change. From the examples of change presented in this book, it becomes obvious that change due to Principles is not as ‘catastrophic’ as that shown in Table 13.1. For instance, the changes involving relatives occur over a fairly long period of time, and, as I show now, pronouns become heads relatively slowly. In the HC EMOD period (from 1500 to 1700), pronouns are used more often in coordination, indicating a more phrasal use, than in ModE but the change is slow, as Table 13.2 shows, and the supposed difference may be due to the corpora used. There is no spoken corpus available for older stages of English and I use HC, a mixture of formal and less formal texts, and the spoken BNC. I differentiate between nominative and accusative pronouns since the checking may be different and hence the need to be heads. Table 13.1.First person changes in Case in HC OE1–2 (–950)

OE3 (950–1050)

OE4 (1050–1150)

ME1 (1150)

me mec

597 90 (=13%)

1282 194 (=13%)

234 0

669 0

total

687

1476

234

669

Conclusion 273

Table 13.2.First person singular coordinated pronouns in HC and spoken BNC

HC BNC

coordinated/total nominative

coordinated/total accusative

coordinated/total all

45/8166 (= 0.55%) 656/304612 (= 0.2%)

34/1446 (=2.35%) 492/27703 (=1.8%)

79/9612 (= 0.82%) 1148/332315 (= 0.34%)

Comparing the HC and the spoken BNC reveals that there is a change towards less coordination where pronouns are concerned. This is accounted for by the Head Preference Principle (or Spec to Head). The difference between BNC and HC is statistically significant (at p < 0.001, X-square 39.776). The reason I am hesitating to call it a real shift is that the earlier and later texts are not really comparable. The CSE, discussed in Chapter 2 (Section 1.1), is a slightly more formal corpus and falls in between HC and BNC. A shift in language-invariant principles would not be expected either unless the character of the pronouns changes to agreement marker. This may be happening since there is a difference in cliticization, as seen in Chapter 2 Table 2.2, between first person on the one hand and second and third on the other. That is confirmed in Table 13.3 for first and third person. In this last section of the book, I have compared a change in a traditional parameter, namely the structural Case parameter, to a change brought about by the Spec to Head Principle. The speed is radically different since principles apply when they structurally can. Chomsky (2000: 102; 142, n. 30), discussing cases where the less economical operation Move is chosen over Merge, suggests thinking about “a proper interpretation of the qualification ‘when possible’” (2000: 142). I will leave this for further investigation. Principles don’t change whereas parameters can be reset. The speed of the Tense-aspect Parameter switch discussed in Chapter 10 is also fast, as expected. I have also mentioned

Table 13.3.Coordination and person in the CSE-FAC uncoordinated

coordinated

total

first third

3024 227

20 (=.66%) 4 (=1.73%)

3044 231

total

3251

24 (=.73%)

3275


that my use of the term parameter in the Layer Parameter is as in Baker (1995; 2001). The Layer Parameter has been used more as a descriptive tool to illuminate cross-linguistic differences than as a theoretical entity.

Notes

Notes to Chapter 1 1. Cf. Noonan (1996). 2. Cf. Heath (1984). 3. For instance, Stockwell & Minkova (1991: 370ff.) remind us of Kellner’s (1892) four-way distinction where I just have one stage. Kuno (1973: chap 17) provides detailed arguments for the degrees of coordination/subordination in Japanese. Givón (1995: 125–6) provides a continuum for clause integration of verbs of truth and action that is also a lot more finegrained. For clauses connected through lack of intonation break, see e.g. Mithun (1988). C. Lehmann’s (1988: 183–4) scale of ‘hierarchical downgrading’, another term for clausal grammaticalization, has eight stages of which the first seven are adverbial/modifying, as in (i), the translation of one of his sentences: i.

The sun was hot when they arrived in Bamako.

Only the eighth is argumental. 4. In Lakoff (1968: Chapter 3), there is evidence that Latin changes from (f) to (c). I will not go into that here since the sociolinguistic background of that change is complex. 5. In van Gelderen (1997a: Chapter 10), two other cases are mentioned: grammaticalization of a preposition to a (coordinating) conjunction, in the come and see construction and with which functions at one point as a coordinator but no longer does in modern English. 6. Mithun (1988: 351ff.) notes that certain languages develop coordination late. 7. As is common in the generative literature, I capitalize Case when it indicates a nominal function. 8. H. Koopman (1996; 2000) argues that Kayne’s (1994) LCA predicts that the specifier and head cannot both be filled. With more functional projections, this is easier to implement since the Spec and head that look as if they are in the same phrasal projection could be assigned to different functional projections. However, this ‘doubly-filled phrase filter’ would not explain the directional changes observed in this book whereas making a crucial distinction between Specs and heads does. 9. Chomsky (2001b: 7–8) reformulates the notions of merge and move as external and internal merge respectively. “Argument structure is associated with external merge (base structure); everything else with internal merge (derived structure)” (p. 8). The latter leaves a copy in place, but is otherwise similar to merge. Both, according to Chomsky, are free, unlike in the older system where move is less economical. I will continue to assume that


internal merge involves more steps and is less economical. In the same paper, Chomsky also starts using ‘late merge’ for adjuncts, which is completely different from my use. 10. Diessel & Tomasello (2001), based on de Villiers (1999), argue that in a biclausal construction, such as (i), the child only sees the embedded verb as relevant to the action. The higher verb is deontic, epistemic, or evidential: i.

I know this piece go [Adam 2;6, p. 116]

This may show Late Merge. Once the child learns that know, and see and think among others, has more meaning it is merged differently.

Notes to Chapter 2 1. Montgomery & Bailey (1991) report on the use of a ‘non-standard in which’, that is used in formal writing as a conjunction and relativizer to ensure non-stranding of the preposition. 2. This observation is made by Hiltunen (1998: 436). 3. First and second person pronouns have more nominal characteristics, hence unlucky we/us is possible, and we linguists. Lyons (1999: 27; 141) argues that first and second person pronouns are Ds and that the absence of third person they linguists is due to the being the third person D rather than he/she/they. 4. If they occur, they are usually non-restrictive, as in (i), or the pronoun is a generic, as in (ii). There are many archaic sounding ones (‘blessed are they that …’): (i) But if he, who had undoubtedly been always so much the most in love of the two, were to be returning with the same warmth of sentiment which he had taken away, it would be very distressing. (Jane Austen, Emma, Vol 3, chap 1) (ii) He who believes in separate and innumerable acts of creation will say, that … (Charles Darwin, Origin of Species, chap 6) Pronouns in (i) and (ii) are forced to be phrases, but, as we’ll see below, pronouns are generally not modified this way. 5. The numbers for coordinated nouns and pronouns include ‘and X’ as well as ‘X and’. I have not taken into account instances where ‘and’ functions as an adverb introducing new sentences. 6. There is an exception to this. Some pronouns cannot be optionally modified, as in the case of all-‘float’. Assuming that in the base structure all precedes the DP, as in QP[all DP], movement should either be possible as a QP in (i) or as a DP, as in (ii). As (ib) shows, with pronouns, only DP movement is possible: (i) a. All these painters are happy. b. *All they are happy. (ii) a. b.

These painters (all) are (all) happy. They (all) are (all) happy.

Notes 277

Only with non-pronominal DPs is the movement of QP possible. So, this is a case of having to move obligatorily as a head. 7. In earlier English (see Newmeyer 1998: 270–1), the second person pronoun cliticizes to the verb if the latter is in C, as in questions such as What wiltow seyn of this (HC-ME3). I won’t go into this. There is perhaps something special with second person since in Dutch verbal inflection is reduced in exactly those circumstances. In many of the modern Germanic languages (e.g. Dutch and Swedish), there is a restriction for many speakers not to separate subject pronouns from C, as in (i) for Dutch and (ii) for Swedish (see Holmberg 1991: 166): (i) Heeft mogelijk *ze/Monica het niet gelezen? has possibly *she it not read ‘Could she/Monica possibly not have read that.’ (ii) Har möjligen *hon/Monica inte sett boken? has possibly *she/Monica not seen the-book ‘Could she/Monica possibly not have seen the book.’ The reason may be that these pronouns need to incorporate. I will just restrict myself to English. 8. These are my own judgments, checked with other fluent speakers of Dutch and English. 9. I heard this attributed to Henk van Riemsdijk who informs me that it is not his work. My apologies for not being able to track the reference. 10. Neurolinguistic work on the difference between pronouns and lexical NPs is not decisive. Even though Kean’s (1979) work would suggest that if pronouns incorporate they would be harder for Broca’s aphasics, this is not borne out in work such as Friederici et al. (1991). However, a lot of the neurolinguistic research on pronouns is done on understanding coreference, not on agreement. 11. However, in Korean, double negation is also possible with both Spec and Head filled, as in (i), but unlike Early English, the result is positive: (i) Mia-nun hakkyo-ey an ka-ci an h-ass-ta Mia-nom school-to not go-inf not do-past-decl ‘It is not the case that Mia didn’t go to school.’ (Sohn 1999: 393, glosses adapted) 12. Their example is Chinese, where they argue ne in (i) is the wh-particle. This particle is optional and therefore this might not be the best example: (i) Hufei chi-le sheme (ne) Hufei eat-asp what part ‘What did Hufei eat?’ 13. Also note the additional me in (42), which could be a copy.


Notes to Chapter 3 1. Doherty (2000: Chapter 2) provides some evidence that complementizer-less embedded clauses are IPs rather than CPs. I won’t go into that here. This position seems to me, however, incorrect for infinitival clauses that can have for, as in (i): (i) They want her to leave. 2. As is well-known, the Mainland Scandinavian languages allow sentences such as (i) and hence their C is nominal: (i) Jag tackade honom för att han hade hjälpt mig I thanked him for that he had helped me ‘I thanked him for having helped me.’ (adapted from Wessén 1968) Starting in the 12th century, prepositions can be followed by that-clauses in English. I discuss this briefly in Chapter 3 (Section 5). 3. Some people (e.g. Varlokosta 1994: 44) argue that subjects and adjuncts can be extracted from non-factives, but not from factives. I think that distinction is harder to make, especially in the case of subjects, because (i) seems fine with either a factive or non-factive: (i) Who do you regret/believe t read Moby Dick? 4. This could mean that Swedish has no expanded CP, but there is work by e.g. Santelmann (1995) showing that children acquiring questions and topicalization have slightly different verb-movement patterns, indicating that the CP is possibly split. 5. Notice that, unlike topicalization, left dislocation is only possible in a main clause as in (i) and not in a subordinate as in (ii): (i) Those books, I read them. (ii) *I know that those books I read them. 6. As a reviewer points out, adverbs such as frankly do not occur very often with other adverbs either. 7. In an earlier article (van Gelderen 2001d), I used native speaker judgments and found that sentences such as (i) and (ii) were unacceptable: (i) *I know that frankly she left (ii) *I know frankly that she left. 8. In the BNC, there are 3 instances of for followed by unfortunately, but these all involve finite for. 9. See also Hoekstra & Zwart (1994: 198) for similar sentences. 10. W. Koopman (1997; 1998) has shown that the kind of initial element influences the ratio of Verb-movement. In sentences with preposed objects, authors such as Alfred invert the subject half the time (I am only taking into account full NP subjects since pronominal ones may be clitics). Sentences with adverbs such as witodlice ‘truly’ show less inversion than those that can be seen as sentence connectors such as þa. In sentences with initial þa, the verb is usually in second position, indicating that þa is integrated in the structure of the CP.

Notes 279

Roberts’ solution is certainly more elegant than that e.g. in Pintzuk (1991; 1999: 85) with both the topic and the subject in Spec IP, adjoining heads and maximal projection. 11. A bridge verb, as in Erteschik-Shir (1973), is a term that is used when extraction out of it is possible and is mostly synonymous with non-factive but also includes a manner-ofspeaking verb such as whisper, mumble, shout, mutter. With manner-of-speaking verbs, Stowell (1981: 404) argues that the emphasis is on the main verb and hence the complement is adjunct-like and does not receive a theta-role, hence no that-deletion and no extraction. Looking through the BNC, however, sentences such as (i) occur frequently without that. So, in order to emphasize the complement, it needs to be fronted: (i) ‘It’s spookier than ever in here’, Mary whispered ‘It’s even worse when you see the headless horseman’, Billy replied with a mischievous grin. (BNC — B3J 2271) Non-bridge verbs typically do not allow wh-extraction, as in (ii): (ii) *Who did Bill mutter t was playing too much poker. I won’t go into bridge and non-bridge-verbs, such as whisper. There is also a link to possibilities of that-deletion. Chomsky (1977: 85) says that it is not clear what allows a verb to be a bridge. Erteschik’s (1973) explanation has to do with ‘focussability’, but in essence says if the main verb draws too much attention to itself, the subordinate is background information, and does not allow extraction. 12. As Seuren (1999: 146) points out, know is not a syntactically factive verb. He suggests it may have lexically incorporated the fact. 13. There is some debate in the literature about this, see the review in Maki, Kaiser, & Ochi (1998: 291). Japanese counterparts are clear, in contrast: (i) a.

John-wa kono hon-wa Mary-ga yonda to sinziteiru John-top this book-top Mary-nom read comp believe ‘John believes that this book Mary read.’ b. *John-wa kono hon-wa Mary-ga yonda no-o kookaisiteiru John-top this book-top Mary-nom read comp-acc regret ‘John regrets that this book Mary read.’ (from Maki, Kaiser, & Ochi 1998: 295–6; Note the difference in Case marking too)

14. The Dutch translation of ‘the fact that’ het feit dat is very rare in sentences comparable to (86a). ANS does not mention this at all, and searching the Dutch newspapers in Lexis Nexis for the past 10 years, I find over 40 after betreuren ‘regret’ but they all seem quite formal/ archaic: (i) Het algemeen bestuur van de Gemeenschappelijke Regeling … betreurt het feit dat de centrumgemeente Hulst … ‘the central administration of the … regrets the fact that the community of Hulst …’ (De Stem, 3 February 1994) 15. Reis (1977: 217ff.) entertains the possibility that assertion equals Focus and non-assertion (presupposition in her terms) equals Topic, but concludes they don’t. 16. I ignore adjectival predicates such as be possible.


17. There are some instances of þæt followed by þe but these will be shown to be a Specifier followed by a Head in the next chapter. 18. However, the earliest example from the OED is from 1175 (Lambeth Homilies) with after that. 19. M. Hale (1987) argues that there is wh-movement to C in Sanskrit, but it isn’t clear to me how one would decide that the position to which the wh-element moves isn’t the focus position since certainly full phrases occur in it.

Notes to Chapter 4 1. Jespersen (1926: 168) is clear that that is not a pronoun and refers to an article by Kruisinga in 1924 who argues that that is a conjunction. There is a debate about this in nongenerative circles, e.g. Quirk and Poutsma say it is a pronoun. 2. I think the construction is widespread. For instance, Falk (1989) reports Cape Breton Island English, in Canada, of the 1970s to include: (i) … where I am standing at. (ii) … when you were hired on. 3. Examples were pointed out by Johanna Wood. 4. As Stein (1998: 68) has shown, there is in the 15th century a “landslide” use of which. This is mainly used in non-restrictive clauses and with inanimate antecedents. 5. This occurs at the end of a line and may therefore be written closer together. 6. I do not have access to facsimiles where these examples occur, but one of the pages of Orosius is reproduced in Bately and þat te is more separate than hatte, a very similar word: (i) þæm iglande þæt te Gades hatte to-the island that Gades was-called ‘to the island which was called G.’ 7. Cf. Ukaji (1997), an article which I have only recently come across, for more examples of whether. 8. Another reason could be ‘analogy’ with the other wh-complementizers, such as in I don’t know what they saw.

Notes to Chapter 5 1. The material in this chapter is based on work in van Gelderen (2001d). 2. In the West-Saxon versions, that is used. 3. While that occurs once, and when that three times.

Notes 281

4. There are a number of versions of Cursor Mundi. Morris’s (1874–1893) edition has 2 Northern and 2 Midlands versions. The Cotton Vespas. A iii is a northern text (e.g. retention of -ande and of -a-; no palatalization; the use of qu- for wh-) from the early part of the 14th century and is written in three different hands. 5. Til is replaced by at in Fairfax and by to in the other two versions. 6. I am leaving ai out but it would be in the Spec of a second TopP. 7. As Fischer (1989) points out, the introduction of these constructions cannot be directly related to the demise of the subjectless for to, since both occur in the same text. 8. The exception is the semi-auxiliary gan, as in (i). I will not go into that here: (i) The markys cam and gan hire for to calle ‘The marquis came and began to call her.’ (Chaucer, Clerk’s Tale 141.289) 9. Supposedly, this is based on M. Hale (1987), but as far as I can see M. Hale (1987) assumes a CP with wh-movement, and M. Hale (1996: 177) has a CP in addition to a Topic and Focus Phrase. 10. One of W. Koopman’s aims is to test the claim that OE pronouns are clitics. This point is not relevant to my purposes.

Notes to Chapter 6 1. Fowler (1926 [1950]: 325ff.) is not too clear in the following excerpt but is not happy with the use of like except as preposition. He writes: It will be best to dispose first of what is, if it is a misuse at all, the most flagrant & easily recognizable misuse of like. A sentence from Darwin quoted in the OED [Oxford English Dictionary] contains it in a short & unmistakable form: Unfortunately few have observed like you have done. Every illiterate person uses this construction daily; it is the established way of putting the thing among all who have not been taught to avoid it … in good writing this particular like is very rare. 2. Kellner (1892: 59) has more examples, one from OE.

Notes to Chapter 7 1. Interestingly, these are not the only reflexives, and often a distinction is made between simple reflexives (zibun, caki, ziji, sich, se) and complex (taziji, herself, si stesso, elle même): the simple ones are less restricted. Sometimes languages have two kinds of reflexives (e.g. Chinese, Italian, French). In Chinese, the complex reflexive is taziji ‘himself/herself ’, woziji ‘myself ’ etc. The minimal domain of these is (a), as (i) shows: (i) *wo bu xu [erzi ma woziji] I not permit [son swear myself ‘I won’t permit my son to swear at me.’


2. I have only examined one element between the verb and to. In the case of EC-verbs, this is typically an adverb whereas in the case of PC-verbs, it is a DP or pronominal object. The figures would probably be much higher for the PC-verbs if I took into account more than one word interruptions. 3. There are 10 instances of wantta which I included. 4. It is also significant between hate and want, however, and hate and need most likely form an in-between class.

Notes to Chapter 8 1. Palmer (1986: 12) excludes ‘dynamic’ modals such as can, but Jespersen (1924) includes them. 2. Most speakers do not like (i), and hence the original deontic meaning is no longer present: (i) Might I smoke here? Abraham (2002: 20–1) shows that, certainly in American English, most modals are used epistemically, and that the deontic meaning is altogether lost. This is expected considering Late Merge. 3. Cormack & Smith (2002) also argue that there are two positions, one above and one below the Negation. They do not see the split as epistemic versus deontic, however. 4. Abraham (1999) argues that epistemic modals in German are in T, the same way I do, but that deontic modals occur in between ASP and V. This might be the case in German and Dutch, where sentences such as (i) occur: (i) Hij heeft het boek moeten lezen (voor school) ‘He has been obliged to read the book (by school).’ If that order is correct for German and Dutch, the order of the functional categories varies cross-linguistically. 5. For Wurmbrand, deontic ones are raising as well. 6. Axel (2001) shows that, in Old High German, modals have no epistemic meaning but are nevertheless raising verbs. This is problematic to an approach such as mine, since structure and meaning do not go together then. 7. As a reviewer points out, progressives and perfectives are very different and obviously not base generated in the same position. A perfect and a progressive auxiliary also occur, be it infrequently, in the same construction, as in (i), again indicating they are not in the same position: (i) It seemed to me that you have been saying something a little different … In the British National Corpus, there are close to half a million instances of have and over 70,000 instances of have been, but the be is not very often progressive. The problem is not directly relevant to my point since following modals, have is never perfect. 8. Must is considered by many as a present tense modal (e.g. Quirk et al. 1985).

Notes 283

9. The same occurs when have follows to, as in (i). To, like the modal, is in M with possibly no T in the sentence, but not one that have can relate to for a present relevance interpretation: (i) They believe to have seen him yesterday. 10. Swedish and Norwegian have complete reduction of ha ‘have’ after past tense modals. Platzack (1986) and more recently Julien (2000) have provided accounts for this. I think the reduction in English is different since the perfect meaning is lost. I won’t go into this further. 11. Eventive, as used here, includes Vendler’s activity, accomplishment, and achievement. Stative is Vendler’s state. 12. There are exceptions to this, as a reviewer points out. For instance, (i) is grammatical: (i) A car must be safe. Be safe in (i) is also meant to convey ‘be made safe’, so perhaps that’s the reason why. 13. After future will, the behavior of -ing is interesting, but won’t be investigated here. Quirk & Greenbaum (1973: 49f) note that it has two main functions, namely ‘future-as-matter-ofcourse’, as in (i), and politeness, as in (ii): (i) a. b.

He’ll be doing his best. He’ll do his best.

(ii) a. b.

When will you be coming? When will you come? (both from Quirk et al. 1972: 49)

14. The OED lists ‘have’ as an auxiliary and claims that its use was “already frequent in OE” (a) with transitive verbs; (b) in eME, with verbs of action without an object; and (c) extended to other intransitives. “Verbs of motion and position long retained the earlier use of the auxiliary be”. The earliest example of (a) dates from 832; of (b) from 1175; and of (c) from 1205. 15. There are other texts, e.g. the northern Cursor Mundi and Shakespeare, where person and number features are deleted, but this is due to agreement reduction in certain positions such as fronting. 16. Van der Wurff (1994: 296) points out that the frequency of these examples is low. 17. There is even one (non-finite) main verb reduced as in (i). If the verb is past tense, it is never reduced: (i) Quar sal he euer ha medicine ‘Where shall he ever have medicine.’ (Cursor Mundi, Cotton 9458) The reason (i) occurs may be a confusion as to forms. Quite close to (i) in the text (ii) occurs, where the infinitival form is has (but haue in the other versions): (ii) Als feird to has sli sisteres fre as afraid to have such sisters noble ‘As if afraid to have such noble sisters.’ (Cursor Mundi, 9532) 18. The two sets of page numbers refer to Wharey (1928) and its revision by Sharrock (1960) respectively.


Notes to Chapter 9 1. This chapter was originally presented as a paper during a conference at Groningen University in 1999 and a preliminary version has appeared as van Gelderen (2001b). 2. For reasons that aren’t so clear, the use of (1a), at least with visual perception, seems to be in decline with respect to (1b). This is the reverse of the Old English situation. The data in part of the CSE confirm that the evidential use of see/saw is quite rare. Out of 217 instances of see only five have bare infinitival complements and none are visual, as (i) shows. Of the 26 instances of saw, none have a bare infinitival complement. This may mean a reversal of earlier changes: (i) a model for how you’d like to see a university operate. 3. Werner Abraham (p.c.) reports that (27) is grammatical for him in German if the direct object is left out. 4. Ikawa (1999) shows that Japanese PVCs can be both stage- and individual level predicates. Interestingly, the subjects of PVCs have nominative Case in Japanese, perhaps indicating PVCs are IPs (this is not Ikawa’s conclusion though). 5. Van Ek (1966: 151) quotes Jespersen (1940: 155) as giving unattested examples such as (i): (i) I saw him be assaulted. Native speakers tend to make be into being. 6. Dik & Hengeveld (1991: 241) show that a PVC as in (i) cannot be negated. Native speakers do not seem to be too clear in their judgments, however: (i) *He saw the girl not cry/crying. 7. Sentences such as (i) and (ii) occur, but here the perception verb is not stative: (i) … and [see him win] I will. (ii) … and [hear those dogs yapping] I did. 8. James Berry (p.c.) points out that less ambiguous sentences, such as (i), are more awkward, though still grammatical: (i) [The dog jump through the hoop] is what we missed seeing.

Notes to Chapter 10 1. A number of arguments in this chapter are adapted, sometimes quite radically, from van Gelderen (1997b). 2. There is another meaning, i.e. ‘die’. 3. This construction also occurs with modals only if the modal is very far removed from the infinitive (Visser 1952: 590; 620). 4. Cf. Brinton’s Appendix B for a summary of the different aspectual meanings of the prefixes.

Notes 285

5. That would presumably mean the Case is structural whereas in other work, and in Chapter 12, I argue that Case in Old English is inherent, dependent on the theta-structure of the verb. I will not go into this. 6. Hoekstra & Hyams (1998) present an interesting argument that Dutch children use the -en on eventive verbs when wanting to express obligation or volition. This had already been noticed in van Ginneken (1917). This suggests a connection with either ASP or M. 7. As Comrie (1976) notes ‘perfective aspect’ is a bad term for the English perfect since ‘present relevance’ is involved in the latter but not the former. 8. This accounts for why, except in the case of irregular verbs such as go, the inflection of simple past and past participle is the same. It also accounts for why one often hears I have went and I was bit.

Notes to Chapter 11 1. Chen (2000: 104) stresses its origin may not be adversative, but additive. There is also the frequent conjunctive function. 2. Grohmann (2003) formulates a proposal on anti-locality that would also account for this. Alternatively, Chomsky (2001b) argues adverbials enter the derivation late. 3. Botne (1995) argues that the origin of the evidential in Lega, a Bantu language spoken in Zaire, is a pronoun. Botne also discusses possible alternative sources that are verbal which he argues against. The data don’t seem clear enough to me, however. 4. Note that Baltin (1995) provides a good argument that the PRO subject stays internal to the VP (based on the position of quantifiers). 5. The problem with Beukema & den Dikken’s account is that the discrepancy between not and other adverbs is not explained, since other adverbs presumably adjoin to AGR. Also, the behavior of not as a clitic in finite clauses, such as (i) would be unexplained, since it wouldn’t be adjoined to AGR: (i) … that I didn’t want food. Bernstein (1994) suggests to (obligatorily) moves to C, but this runs into problems accounting for the not to sequence.

Notes to Chapter 12 1. However, Chiu (1993: 74–80) argues convincingly that words such as keneng ‘likely’, which have sometimes been called S-adverbs in Chinese, are really raising verbs, and in my terminology epistemic modals. She does not provide a structural analysis but they would readily fit as I. A number of interesting parallels with English occur. They cannot co-occur with aspectual le as in (i) and can be sentence-initial as in (ii):


(i) *Lisi keneng le du nei ben shu Lisi likely asp read that cl book (ii) Keneng Lisi du le nei ben shu likely Lisi read asp that cl book ‘Lisi likely read that book.’ (from Chiu 1993: 75) The aspectual marker le must follow a verbal element and the main verb (du in (ii)) moves to ASP. The reason keneng cannot move to join with le is because it is in a position higher than ASP, namely I, and cannot move via ASP. 2. Note that e.g. K. Hale (2003) has more ‘modern’ structures and basically accepts the PA-hypothesis. 3. Early on, K. Hale noted that PALs, such as Warlpiri and O’odham, often have an auxiliary. OE shows the same but I won’t go into that here.

Notes to Chapter 13 1. The evidence in Klamer (1998: 190ff; 338) suggests that pa may be a v changing to C. 2. I have only looked at that/for immediately following important. There are many pseudoclefts that I haven’t considered, but the ratio of subjunctive/indicative/infinitive would be expected to be the same. I have made sure that all indicative forms could indeed be subjunctive, i.e. are ‘unreal’.

References Primary texts, dictionaries, and corpora d’Ardenne, S. 1977. The Katherine Group. Paris: Les Belles Lettres. Barron, W. & S. Weinberg 1995. Layamon’s Brut. Harlow: Longman. Babington, Churchill 1860. The Repressor. London: Longman. Bartlett, John 1894. A New and Complete Concordance or Verbal Index to Words, Phrases, & Passages in the Dramatic Works of Shakespeare. London: MacMillan. Bately, Janet 1980. The Old English Orosius. Oxford: Oxford University Press [EETS S. S. 6]. Beadle, Richard 1982. The York Plays. London: Arnold. Benson, Larry 1987. The Riverside Chaucer. Boston: Houghton Mifflin Company. British National Corpus (BNC). Available online at thetis.bl.uk. [100 million words of British English, 10% spoken]. Brook, G. & R. Leslie 1963. Layamon: Brut. Oxford: Oxford University Press [EETS 250]. Clark Hall, J. R. 1894 [1960]. A Concise Anglo-Saxon Dictionary. Toronto: University of Toronto Press [reprint]. Clemoes, Peter 1997. Aelfric’s Catholic Homilies. Oxford: Oxford University Press [EETS SS 17]. Cobuild Corpus. Available at titania.cobuild.collins.co.uk. [45 million word corpus of British and American written and spoken English]. Corpus of Spoken Professional American English (CSE). Not available online, but see www.athel.com. [2 million words; divided in COMM, FACMT, WH] Davis, Norman 1971. Paston Letters and Papers of the Fifteenth Century, Part I. Oxford: Clarendon. Gardner, John 1965. The Complete Works of the Gawain Poet. Chicago: University of Chicago Press. Godden, Malcolm 1979. Ælfric’s Catholic Homilies, the second series. London: Oxford University Press. [EETS SS 5]. Hanham, Alison 1975. The Cely Letters. London: Oxford University Press. Haim, S. 1997. Farhang Jama’ Farsi — Angresi [The Larger Farsi — English Dictionary]. Tehran. Hecht, Hans 1965. Übersetzung der Dialoge Gregors des Grossen. Darmstadt: Wissenschaftliche Buchgesellschaft. Helsinki Corpus (HC), available through the OTA, and the Penn-Helsinki Parsed Corpus. [available at www.ling.upenn.edu/mideng, OE1–2 before 950; OE3 950–1050; OE4 1050–1150; ME1 1150–1250; ME2 1250–1350; ME3 1350–1420; ME4 1420–1500; EMOD1 1500–1570; EMOD2 1570–1640; EMOD3 1640–1710]. Kellner, Leon 1890. Caxton’s Blanchardyn and Eglantine. London (EETS). Klaeber, Fr. 1922 [1941]. Beowulf. Boston: Heath & Co.


Kökeritz, Helge 1954. Mr. William Shakespeares Comedies, Histories and Tragedies. Facsimile edition of the First Folio of 1623. New Haven: Yale University Press. Krapp, G. P. 1931. The Junius Manuscript. New York: Columbia University Press. Krishna, Valerie 1976. The Alliterative Morte Arthure. New York: Burt Franklin & Co. Kuhn, S. M. 1965. The Vespasian Psalter. Ann Arbor: University of Michigan Press. Luce, A. & T. Jessop 1949. The Works of George Berkeley II. London: Nelson & Son. Morris, Richard 1873. Trinity Homilies. OE Homilies of the 12th Century. London, EETS. Morris, Richard 1874–1893. Cursor Mundi, 7 Parts. Trübner & Co. Mulcaster, Richard 1582. Elementarie. Menston: Scolar Press [1970 reprint]. Oxford English Dictionary (OED) 1933. Oxford: Oxford University Press, and OED online. Platts, John 1884/1982. A Dictionary of Urdu, Classical Hindi, and English. Oxford: Oxford University Press. Practical Dictionary Anglo-Urdu. 1987. Delhi: J. S. Sant Singh Publishers [17th edition]. Schaefer, K. G. 1972. An Edition of Five Old English Homilies. Columbia Diss. Sedgefield, Walter 1899. King Alfred’s Old English Version of Boethius. Oxford: Clarendon. Sharrock, Roger 1960. John Bunyan’s The Pilgrim’s Progress. Second Edition of Wharey 1928. Oxford: Clarendon. Sweet, Henry 1871 [1934]. King Alfred’s West-Saxon Version of Gregory’s Pastoral Care. London: Oxford University Press. Thorpe, Benjamin 1861. Anglo-Saxon Chronicle I and II. London: Longman. Vaissier, Johannes 1960. A deuout treatyse called the tree & xii. frutes of the holy goost. Groningen: Wolters. Wharey, James Blanton 1928. John Bunyan’s The Pilgrim’s Progress. Oxford: Clarendon. Zupitza, Julius 1959. Beowulf, reproduced in facsimile [2nd edition, EETS 245].

Linguistic References Abbott, E. A. 1872. A Shakespearean Grammar. London: MacMillan & Co. Abraham, Werner 1990. “Modalverben in der Germania”. Akten des VIII. Internationalen Germanisten-Kongresses. Abraham, Werner (ed.) 1991a. Discourse Particles. Amsterdam: John Benjamins. Abraham, Werner 1991b. “The Grammaticalization of German Modal Particles”. In: Elizabeth Traugott & Bernd Heine (eds), Approaches to Grammaticalization II, 331–380. Amsterdam: John Benjamins. Abraham, Werner 1996a. “Case, Aspect, and Referentiality”. In: Ans van Kemenade & Nigel Vincent (eds), Parameters of Morphosyntactic Change, 29–61. Cambridge: Cambridge University Press. Abraham, Werner 1996b. “Kasus, Aspekt und nominale Referenz”. Linguistische Berichte Sonderheft 7: 22–70. Abraham, Werner 1997. “The Base Structure of the German Clause”. In: Werner Abraham and Elly van Gelderen (eds), German: Syntactic Problems, 11–42. Tübingen: Niemeyer.

References 289

Abraham, Werner 1998. “The Morphological and Semantic Classification of ‘Evidentials’ and Modal Verbs in German: the perfect(ive) catalyst”. Groninger Arbeiten zur germanistischen Linguistik 42: 192–206. Abraham, Werner 1999. “Evidentials”. International Conference on Historical Linguistics paper. Abraham, Werner 2002. “Modal Verbs: Epistemics in German and English”. In: Sjef Barbiers et al. (eds), Modality and its Interaction with the Verbal System, 19–50. Amsterdam: John Benjamins. Abraham, Werner 2004. “The Grammaticalization of the Infinitival Preposition: towards a theory of ‘grammaticalizing reanalysis’”. Journal of Comparative Germanic Linguistics 7.2: 111–170. Adger, David 1996. “Aspect, Agreement, Measure phrases”. In: Robert Borsley & Ian Roberts (eds), The Syntax of the Celtic Languages, 200–222. Cambridge: Cambridge University Press. Akmajian, Adrian, Susan Steele & Tom Wasow 1979. “The Category of AUX in Universal Grammar”. Linguistic Inquiry 10.1: 1–64. Akmajian, Adrian 1977. “The Complement Structure of Perception Verbs in an Autonomous Syntax Framework”. In: Peter Culicover, Thomas Wasow & Adrian Akmajian (eds), Formal Syntax, 427–460. New York: Academic Press. Algemene Nederlandse Spraakkunst (ANS). 1984. G. Geerts et al., eds. Groningen: Wolters Noordhoff. Allen, Cynthia 1977. Topics in Diachronic English Syntax. UMass Diss. Altmann, Hans 1976. Die Gradpartikeln des Deutschen. Tübingen: Niemeyer. Armitage, Richard & Walter Meiden 1979. Beginning Spanish. Boston: Houghton Mifflin Company. Anderwald, Lieselotte 2002. Negation in Non-Standard British English. London: Routledge. Authier, J. Marc 1992. “Iterated CPs and Embedded Topicalization”. Linguistic Inquiry 23.2: 329–336. Avram, Larisa 2000. “Comparative Remarks on the Syntax and Semantics of a putea”. University of Bucharest ms. Axel, Katrin 2001. “Althochdeutsche Modalverben als Anhebungsverben”. In: Reimar Müller and Marga Reis (eds), Modalität und Modalverben im Deutschen, 37–60. Hamburg: H. Buske. Bacquet, P. 1962. La structure de la phrase verbale à l’époque Alfrédienne. Paris: Les Belles Lettres. Baker, Mark 1995. The Polysynthesis Parameter. Oxford: Oxford University Press. Baker, Mark 2001. The Atoms of Language. New York: Basic Books. Baker, Mark 2003a. “Agreement, Dislocation, and Partial Configurationality”. In: Andrew Carnie et al. (eds), Formal Approaches to Function in Grammar, 107–132. Amsterdam: John Benjamins. Baker, Mark 2003b. Lexical Categories. Cambridge: Cambridge University Press. Bakker, Peter, Marike Post, & Hein van der Voort 1995. “TMA Particles and Auxiliaries”. In: Jacques Arends, Pieter Muysken, & Norval Smith (eds), Pidgins and Creoles, 247–258. Amsterdam: John Benjamins. Baltin, Mark 1995. “Floating Quantifiers, PRO, and Predication”. Linguistic Inquiry 26.2: 199–248.


Banfi, Emanuele 1990. “The Infinitive in South-east European Languages”. In: Johannes Bechert et al. (eds), Towards a Typology of European Languages, 165–183. Berlin: Mouton. Barbiers, Sjef 1995. The Syntax of Interpretation. Leiden Diss. Barbiers, Sjef 2002. “Remnant Stranding and the Theory of Movement”. In: Artemis Alexiadou et al. (eds), Dimensions of Movement, 47–67. Amsterdam: John Benjamins. Barker, Muhammad et al. 1967. Spoken Urdu, 2 vols. Ithaca [1975 edition]. Barnes, Janet 1984. “Evidentials in the Tuyuca Language”. International Journal of American Linguistics 50.3: 255–271. Basilico, David 2003. “The Topic of Small Clauses”. Linguistic Inquiry 34: 1–35. Bean, Marian 1983. The Development of Word Order Patterns in Old English. London: Croom Helm. Bech, G. 1955/1983. Studien über das deutsche Verbum Infinitum. Tübingen: Niemeyer. Bentahila, Abdelali & Eirlys Davies 1983. “The syntax of Arabic-French code-switching”. Lingua 59: 301–30. Benveniste, Emile 1968. “Mutations of Linguistic Categories”. In: Winfred Lehmann & Yakov Malkiel (eds), Directions for Historical Linguistics, 85–94. Austin: University of Texas Press. Bergs, Alexander 2002. “Grammaticalizing Relative who”. NRG2 talk. Berndt, Rolf 1956. Form und Funktion des Verbums im nördlichen Spätaltenglischen. Halle: Max Niemeyer. Bernstein, Michael 1994. “Some Notes on Infinitives and Negation in English”. Cornell Working Papers in Linguistics 12: 1–25. Besten, Hans den 1983. “On the Interaction of Root Transformations and Lexical Deletive Rules”. In: Werner Abraham (ed.), On the Formal Syntax of the Westgermania, 47–131. Amsterdam: John Benjamins. Beukema, Frits & Marcel den Dikken 1989. “The Position of the Infinitival Marker in the Germanic Languages”. In: Dany Jaspers et al. (eds), Sentential Complementation and the Lexicon, 57–75. Dordrecht: Foris. Bickerton, Derek 1981. Roots of Language. Ann Arbor: Karoma. Bickerton, Derek 1984. “The Language Bioprogram Hypothesis”. The Behavioral and Brain Sciences 7: 173–188. Bisang, Walter 1992. Das Verb im Chinesischen, Hmong, Vietnamesischen, Thai und Khmer. Tübingen: Gunter Narr Verlag. Blom, Corrien 2003. “The Diachrony of Complex Predicates in Dutch”. GLAC talk. Blom, Corrien & Geert Booij 2003. “The Diachrony of Complex Predicates in Dutch”. Acta Linguistica Hungarica: 61–91. Boeckx, Cedric 2003. Islands and Chains: Resumption as Stranding. Amsterdam: John Benjamins. Boertien, Harmon 1986. “Constituent Structure of Double Modals”. In: Michael Montgomery and Guy Bailey (eds), Language Variety in the South, 294–318. University of Alabama Press. Bonneau, José, J. Bruhn-Garavito & Alan Libert 1994. “Feature Checking and Spanish SE”. WECOL 24 Proceedings. Botne, Robert 1995. “The Pronominal Origin of an Evidential”. Diachronica 12.2: 201–221.

References 291

Brinton, Laurel. 1988. The Development of English Aspectual Systems. Cambridge: Cambridge University Press. Brown, Keith 1991. “Double Modals in Hawick Scots”. In: Peter Trudgill & J. K. Chambers (eds), Dialects of English, 74–103. London: Longman. Brunner, Karl 1962. Die englische Sprache II. Tübingen: Niemeyer. Brusendorff, Aage 1930. “The Relative Aspect of the Verb in English”. A Grammatical Miscellany Offered to Otto Jespersen on his Seventieth Birthday, 225–247. Copenhagen: Levin & Munksgaard. Bullokar, William 1586. Bref Grammar. In: J. R. Turner (ed), The Works of William Bullokar II. Leeds 1970. Buzarovska, Eleni 2000. “On Complementation of English and Russian Verbs of Perception”. Ms. Bybee, Joan 1985. Morphology. Amsterdam: John Benjamins. Bybee, Joan, Revere Perkins & William Pagliuca 1994. The Evolution of Grammar. Chicago: University of Chicago Press. Callaway, Morgan 1913. The Infinitive in Anglo-Saxon. Washington D. C.: Carnegie Institution. Campbell, Alistair 1959. Old English Grammar. Oxford: Oxford University Press. Cardinaletti, Anna & Michael Starke 1996. “Deficient Pronouns: a View from Germanic”. In: Höskuldur Thráinsson et al. (eds.), Studies in Comparative Syntax II, 21–65. Dordrecht: Kluwer. Carstensen, B. 1959. Studien zur Syntax des Nomens, Pronomens und der Negation in den Paston Letters. Bochum-Langendreer. Chapman, Carol 1995. “A Subject-Verb Agreement Hierarchy”. Historical Linguistics 2: 35–44. Chen, Guohua 2000. “The Grammaticalization of Concessive Markers”. In: Olga Fischer et al. (eds), Pathways of Change, 85–110. Amsterdam: John Benjamins. Chiu, Bonnie Hui-Chun 1993. The Inflectional Structure of Mandarin Chinese. UCLA Diss. Choi, Ji-Young 1999. The Acquisition of English Reflexive Pronouns by Adult Korean Speakers. ASU MA. Chomsky, Noam 1955/1979. The Logical Structure of Linguistic Theory. Chicago: University of Chicago Press. Chomsky, Noam 1957. Syntactic Structures. Den Haag: Mouton. Chomsky, Noam 1962. “Discussion”. In: Archibald Hill (ed.), Third Texas Conference on Problems of Linguistic Analysis. Austin: University of Texas. Chomsky, Noam 1977. “On Wh-movement”. In: Peter Culicover, Thomas Wasow & Adrian Akmajian (eds), Formal Syntax, 71–132. New York: Academic Press. Chomsky, Noam 1986. Barriers. Cambridge: MIT Press. Chomsky, Noam 1989. “Some Notes on Economy and Derivation”. MIT Working Papers in Linguistics 10; also appears as Chapter 2 in Chomsky (1995). Chomsky, Noam 1995. The Minimalist Program. Cambridge: MIT Press. Chomsky, Noam 2001a. “Derivation by Phase”. In: Michael Kenstowicz (ed.), Ken Hale: A Life in Language, 1–52. Cambridge: MIT Press. Chomsky, Noam 2001b. “Beyond Explanatory Adequacy”. MIT ms.


Cinque, Guglielmo 1993. “A Null Theory of Phrase and Compound Stress”. Linguistic Inquiry 24: 239–97. Cinque, Guglielmo 1999. Adverbs and Functional Heads. Oxford: Oxford University Press. Cinque, Guglielmo 2001. “The Status of Mobile Suffixes”. In: Walter Bisang (ed.), Aspects of Typology and Universals, 13–19. Berlin: Akademie Verlag. Cinque, Guglielmo (ed.) 2002. Functional Structure in DP and IP. New York: Oxford University Press. Coar, Thomas 1796. A Grammar of the English Tongue. London. Comrie, Bernard 1976. Aspect. Cambridge: Cambridge University Press. Condoravdi, Cleo & Paul Kiparsky 2001. “Clitics and Clause Structure”. Journal of Greek Linguistics 2: 1–39. Cormack, Annabel & Neil Smith 2002. “Modals and Negation in English”. In: Sjef Barbiers et al. (eds), Modality and its Interaction with the Verbal System, 133–163. Amsterdam: John Benjamins. Corver, Norbert & Henk van Riemsdijk (eds) 2001. Semi-lexical Categories. Berlin: Mouton. Cowling, George 1915. The Dialect of Hackness. Cambridge: Cambridge University Press. Cowper, Elizabeth 1999. “Grammatical Aspect in English”. The Linguistic Review 16: 205–226. Craenenbroeck, Jeroen van 2002. “Van as a Marker of Dissociation”. In: Jan-Wouter Zwart & Werner Abraham (eds), Studies in Comparative Germanic Syntax, 41–67. Amsterdam: John Benjamins. Dam, Johannes van 1957. The Causal Clause and Causal Prepositions in Early Old English Prose. Groningen: Wolters. Dekeyser, Xavier 1986. “Relative Markers in the Peterborough Chronicle: 1070–1154”. Folia Linguistica Historica 7.1: 93–105. Dekeyser, Xavier 1996. “WH- and That: two competing strategies in the history of English relative clause formation”. Leuvense Bijdragen 85.3–4: 293–302. Delancey, Scott 1991. “The Origins of Verb Serialization in Modern Tibetan”. Studies in Language 57: 1–23. Denison, David 1986. “On Word Order in OE”. Dutch Quarterly Review of Anglo-American Letters 16: 277–295. Denison, David 1993. English Historical Syntax. London: Longman. Deutscher, Guy 2000. Syntactic Change in Akkadian. Oxford: Oxford University Press. Di Paolo, Marianna 1986. A Study of Double Modals in Texas English. University of Texas Diss. Di Paolo, Marianna 1989. “Double Modals as Single Lexical Items”. American Speech 64: 195–224. Dik, Simon & Kees Hengeveld 1991. “The Hierarchical Structure of the Clause and the Typology of Perception-verb Complements”. Linguistics 29: 231–259. Diesing, Molly 1992. Indefinites. Cambridge: MIT Press. Diessel, Holger 1999. Demonstratives. Amsterdam: John Benjamins. Diessel, Holger & Michael Tomasello 2001. “The Acquisition of Finite Complement Clauses”. Cognitive Linguistics 12.2: 97–141. Dikken, Marcel den 2002. “On the Syntax of WH-movement”. CUNY ms.

References 293

Dikken, Marcel den 2003. “On the Morpho-Syntax of wh-movement”. In: Cedric Boeckx and Kleanthes Grohmann (eds), Multiple Wh-fronting, 77–98. Amsterdam: John Benjamins. Disterheft, Dorothy 1980. The Syntactic Development of the Infinitive in Indo-European. Columbus: Slavica. Dixon, R. M. 1988. A Grammar of Boumaa Fijian. Chicago: Chicago University Press. Doherty, Cathal 2000. Clauses without ‘that’. New York: Garland Publishing. Duncan, Daniel 1731. A New English Grammar. Menston: Scolar Press [1967 reprint]. Durrleman, Stephanie 2000. “The Architecture of the Clause in Jamaican Creole”. Generative Grammar in Geneva 1: 189–240. Eckardt, Eugen 1875. Über die Syntax des gotischen Relativpronomens. Halle. Enç, Mürvet 1987. “Anchoring Conditions for Tense”. Linguistic Inquiry 18: 633–57. Ek, J. van 1966. Four Complementary Structures of Predication in Contemporary British English. Groningen: Wolters. Emonds, Joe 1985. A Unified Theory of Syntactic Categories. Dordrecht: Foris. Emonds, Joe 2001. “The Flat Structure of Semi-lexical Heads”. In: Norbert Corver & Henk van Riemsdijk (eds), Semi-lexical Categories, 23–66. Berlin: Mouton. Erteschik-Shir, Nomi 1973. On the Nature of Island Constraints. MIT Diss. Evers, Arnold 1990. “The Infinitival Prefix ‘zu’ as INFL”. In: Günther Grewendorf and Wolfgang Sternefeld (eds), Scrambling and Barriers, 217–238. Amsterdam: John Benjamins. Eythórsson, Thórhallur 1996. “Functional Categories, Cliticization, and Verb Movement in the Early Germanic Languages”. Studies in Comparative Germanic Syntax II, 109–139. Dordrecht: Kluwer. Eythórsson, Thórhallur 2003. “Looking at the Overlooked”. DGfS talk. Faarlund, Jan Terje 1990. Syntactic Change: Towards a Theory of Historical Syntax. Berlin: Mouton de Gruyter. Falk, Lilian 1989. “Regional Usage in the English of Cape Breton Island”. Atlantic Provinces Linguistic Association Proceedings. Faltz, Leonard 1985. Reflexivization: A Study in Universal Syntax. New York: Garland. Farkas, Donka 1992. “On the Semantics of Subjunctive Complements”. In: Paul Hirschbühler & Konrad Koerner (eds), Romance Languages and Modern Linguistic Theory, 69–104. Amsterdam: John Benjamins. Felser, Claudia 1999. Verbal Complement Clauses. Amsterdam: John Benjamins. Fenning, Daniel 1771. A New Grammar of the English Language. Menston: Scolar Press reprint. Fischer, Olga 1995. “The Distinction between to and Bare Infinitival Complements in Late Middle English”. Diachronica 12: 1–30. Fischer, Olga 1998. “On Negative Raising in the History of English”. In: Wim van der Wurff et al. (eds), Negation in the History of English, 55–100. Berlin: Mouton de Gruyter. Fischer, Olga et al. (eds) 2000. The Syntax of Early English. Cambridge: Cambridge University Press. Fisher, Ann 1750. A New Grammar. Menston: Scolar Press reprint. Fowler, H. W. 1926 [1950]. A Dictionary of Modern English Usage. Oxford: Clarendon. Frajzyngier, Zygmunt 1995. “A Functional Theory of Complementizers”. In: Joan Bybee & Suzanne Fleischman (eds), Modality in Grammar and Discourse, 473–502. Amsterdam: John Benjamins.


Frajzyngier, Zygmunt 1996. Grammaticalization of the Complex Sentence. Amsterdam: John Benjamins. Frajzyngier, Zygmunt 1997. “Domains of Affected Subject and Coreferentiality”, Ms University of Colorado. Franz, W. 1909. Shakespeare-Grammatik, 2nd edition. Heidelberg: Carl Winter. Friederici, Angela Weissenborn, Jurgen & Michele Kail 1991. “Pronoun Comprehension in Apahasia”. Brain and Language 41: 289–310. Gardiner, Alan 1957. Egyptian Grammar. Oxford: Griffith Institute [Third Revised Edition 1988]. Gelderen, Elly van 1993. The Rise of Functional Categories. Amsterdam: John Benjamins. Gelderen, Elly van 1997a. Verbal Agreement and the Grammar behind its ‘Breakdown’. Tübingen: Niemeyer. Gelderen, Elly van 1997b. “Structures of Tense and Aspect”. Linguistic Analysis 27:3–4. Gelderen, Elly van 1998. “For to in the History of English”. American Journal of Germanic Language and Literature 10.1: 45–72. Gelderen, Elly van 2000. A History of English Reflexive Pronouns. Amsterdam: John Benjamins. Gelderen, Elly van 2001a. “Phrases, Heads, Grammaticalization, and Economy”. International Conference on Historical Linguistics paper, appears as van Gelderen 2004. Gelderen, Elly van 2001b. “Evidentials and Aspect”. GAGL 44: 54–74. Gelderen, Elly van 2001c. “The Syntax of Mood Particles in the History of English”. Folia Linguistica Historica 22: 301–331. Gelderen, Elly van 2001d. “The Force of ForceP in English”. Southwest Journal of Linguistics 20.2: 107–120. Gelderen, Elly van 2003. “ASP(ect) in English Modal Complements”. Studia Linguistica 57.1: 27–44. Gelderen, Elly van 2004. “Specifiers, Heads, Grammaticalization, and Economy”. Journal of Comparative Germanic Linguistics 7.1: 59–98. Gestel, Frank, Jan Nijen Twilhaar, Tineke Rinkel, & Fred Weerman 1992. Oude Zinnen. Leiden: Martinus Nijhoff. Ginneken, Jac. van 1917. De Roman van een Kleuter. Nijmegen: Malmberg. Giorgi, Alessandra & Fabio Pianesi 1997. Tense and Aspect. Oxford: Oxford University Press. Givón, Talmy 1979. “Language Typology in Africa”. Journal of African Languages and Linguistics 1: 199–224. Givón, Talmy 1991. “The Evolution of Dependent Clause Morpho-syntax in Biblical Hebrew”. In: Elizabeth Traugott & Bernd Heine (eds), Approaches to Grammaticalization II, 257–310. Amsterdam: John Benjamins. Givón, Talmy 1995. Functionalism and Grammar. Amsterdam: John Benjamins. Goldman, Robert & Sally Sutherland 1987. Devavanipravesika [2nd edition]. Berkeley: University of California. Gonda, Jan 1971. Old Indian. Leiden: Brill. Goossens, Louis 1982. “On the Development of the Modals and the Epistemic Function in English”. International Conference on Historical Linguistics 5 talk. Gordon, E. V. 1927. An Introduction to Old Norse. Oxford: Clarendon Press.

References 295

Gordon, Lynn 1986. “The Development of Evidentials in Maricopa”. In: Wally Chafe & Johanna Nichols (eds), Evidentiality, 75–88. New York: Academic Press. Görlach, Manfred 1990. Introduction to Early Modern English. Cambridge: Cambridge University Press. Greenwood, James 1711. An Essay towards a Practical English Grammar. Menston: Scolar Press reprint. Grohmann, Kleanthes 2003. Prolific Domains. Amsterdam: John Benjamins. Grossmann, Heinrich 1906. Das angelsächsische Relativ. Berlin Diss. Guasti, Maria Teresa 1993. Causative and Perception Verbs. Torino: Rosenberg & Sellier. Haan, Ferdinand de 1997. “Evidentiality in Dutch”. University of New Mexico ms. Haddad, Ghassan & Michael Kenstowicz 1980. “A Note on the Parallels between the Definite Articles and the Relative Clause Marker”. Studies in the Linguistic Sciences 10.2: 141–147. Haeberli, Eric. 2000. “Adjuncts and the Syntax of Subjects in Old and Middle English”. In: Susan Pintzuk et al. (eds), Diachronic Syntax, 109–131. Oxford: Oxford University Press. Haider, Hubert 1986. “V-Second in German”. In: Hubert Haider & Martin Prinzholm (eds), Verb Second Phenomena in Germanic Languages, 49–75. Dordrecht: Foris. Haider, Hubert 1989. “[Theta]-tracking Systems”. In: László Marácz & Pieter Muysken (eds), Configurationality: The typology of asymmetries, 185–206. Dordrecht: Foris. Haiman, John 1974. Targets and Syntactic Change. Den Haag: Mouton. Haiman, John & Sandra Thompson 1984. “Subordination in Univeral Grammar”. BLS 10: 510–523. Haiman, John & Sandra Thompson (eds) 1988. Clause Combining in Grammar and Discourse. Amsterdam: John Benjamins. Hale, Ken 1976. “The Adjoined Relative Clause in Australia”. In: R. M. W. Dixon (ed.), Grammatical Categories in Australian Languages. Humanities Press. Hale, Ken 1982. “Preliminary Remarks on Configurationality”. NELS Proceedings 12: 86–96. Hale, Ken 1983a. “Warlpiri and the Grammar of Non-configurational Languages”. Natural Language and Linguistic Theory 1: 5–47. Hale, Ken 1983b. “Papago (k)c”. International Journal of American Linguistics 49.3: 299–327. Hale, Ken 1989. “On Non-configurational Structures”. In: László Marácz & Pieter Muysken (eds), Configurationality: The typology of asymmetries, 293–300. Dordrecht: Foris. Hale, Ken 1991. “Misumalpan Verb Sequencing Constructions”. In: Claire Lefebvre (ed.), Serial Verbs, 1–35. Amsterdam: John Benjamins. Hale, Ken 2003. “On the Significance of Eloise Jelinek’s Pronominal Argument Hypothesis”. In: Andrew Carnie et al. (eds), Formal Approaches to Function in Grammar, 11–43. Amsterdam: John Benjamins. Hale, Ken & Jay Keyser 1993. “On Argument Structure and the Lexical Expression of Syntactic Relations”. In: Ken Hale & Samuel Keyser (eds), The View from Building 20, 53–109. Cambridge: MIT Press. Hale, Ken & Jay Keyser 2002. Prolegomenon to a Theory of Argument Structure. MIT Press. Hale, Mark 1987. “Notes on Wackernagel’s Law in the Language of the Rigveda”. In: Calvert Watkins (ed.), Studies in Memory of Warren Cowgill, 38–50. Berlin: Mouton de Gruyter. Hale, Mark 1996. “Clitic Placement in the Language of the Rigveda”. In: Aaron Halpern & Arnold Zwicky (eds), Approaching Second, 165–197. Stanford: CSLI Publications.


Haller, Beat & John Watters 1984. “Topic in Zulgo”. Studies in African Linguistics 15.1: 27–46. Harris, Alice & Lyle Campbell 1995. Historical Syntax in Cross-linguistic Perspective. Cambridge: Cambridge University Press. Hashimoto, Anne 1971. Mandarin Syntactic Structures. Unicorn 5. Haspelmath, Martin 1989. “From Purposive to Infinitive”. Folia Linguistica Historica 9: 287–310. Haspelmath, Martin 1997. “Why is the Change from Lexical to Functional Categories Irreversible?” International Conference on Historical Linguistics paper. Heath, Jeffrey 1984. Functional Grammar of Nunggubuyu. Canberra: Australian Institute of Aboriginal Studies. Heine, Bernd 1993. Auxiliaries. Oxford: Oxford University Press. Heine, Bernd & Mechthild Reh 1984. Grammaticalization and Reanalysis in African Languages. Hamburg: Buske Verlag. Heine, Bernd & Ulrike Claudi 1986. On the Rise of Grammatical Categories. Berlin: Dietrich Reimer. Heine, Bernd, Ulrike Claudi & Friederike Hünnemeyer 1991. Grammaticalization. Chicago: Chicago University Press. Heine, Bernd & Tania Kuteva 2002. World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. Heine, Bernd & Derek Nurse (eds) 2000. African Languages. Cambridge: Cambridge University Press. Helbig, Gerhard 1988. Lexikon deutscher Partikeln. Leipzig: VEB Verlag. Henry, Alison 1995. Belfast English and Standard English. New York: Oxford University Press. Hentschel, Elke 1986. Funktion und Geschichte deutscher Partikeln: Ja, doch, halt, und eben. Tübingen: Niemeyer. Higginbotham, James 1983. “The Logic of Perceptual Reports”. Journal of Philosophy 80.2: 100–127. Higginbotham, James 1985. “On Semantics”. Linguistic Inquiry 16: 547–593. Hiltunen, Risto 1983. The Decline of the Prefixes and the Beginnings of the English Phrasal Verb. Turku: Turun Yliopisto. Hiltunen, Risto 1998. “Phrasal Verbs in English Grammar Books before 1800”. In: Mats Rydén et al. (eds), Bamberger Beiträge zur Englischen Sprachwissenschaft, 433–445. Frankfurt: Peter Lang. Hinterhölzl, Roland 1999. Restructuring Infinitives and the Theory of Complementation. USC Diss. Ho, Li-Ting 1999. Reflexives in Chinese and English. ASU MA. Hock, Hans 1991. Principles of Historical Linguistics [second edition]. Berlin: Mouton de Gruyter. Hoekstra, Eric 1993. “Dialectal Variation inside CP as Parametric Variation”. Linguistische Berichte 5. Hoekstra, Eric & Jan-Wouter Zwart 1994. “De Struktuur van de CP”. Spektator 23.3: 191–212. Hoekstra, Teun 1995. “De Status van van”. Tabu 25: 160–164.

References 297

Hoekstra, Teun & Nina Hyams 1998. “Aspects of Root Infinitives”. Lingua 106: 81–112. Holm, John 1989. Pidgins and Creoles II. Cambridge: Cambridge University Press. Holmberg, Anders 1991. “Distribution of Scandinavian Weak Pronouns”. In: Henk van Riemsdijk & Luigi Rizzi (eds), Clitics and their Hosts. Eurotyp Working Papers II, Tilburg University. Hooper, Joan & Sandra Thompson 1973. “On the Applicability of Root Transformations”. Linguistic Inquiry 4.4: 465–497. Hopper, Mike 1987. “Emergent Grammar”. Berkeley Linguistic Society Proceedings 13: 139–157. Hopper, Mike & Elizabeth Traugott 1993. Grammaticalization. Cambridge: Cambridge University Press. Horn, Laurence 1989. A Natural History of Negation. Stanford: CSLI. [2001 reprint]. Hornstein, Norbert 1990. As Time Goes By. Cambridge: MIT Press. Hupe, H. 1893. “On the Filiation and the Text of the mss of the Middle English Poem Cursor Mundi”. In: Richard Morris (ed.), Cursor Mundi, 7. Trübner & Co. IJbema, Aniek 2002. Grammaticalization and Infinitival Complements in Dutch. Leiden Diss. Ikawa, Hisako 1999. Events and Anaphoric Processes. University of Arizona Diss. Jacobs, Joachim 1983. Fokus und Skalen. Tübingen: Niemeyer. Jacobs, Roderick 1975. Syntactic Change: a Cupan (Uto-Aztecan) Case Study. Berkeley: University of California Press. Jakab, Edit 2001. The Modality of Non-finite clauses in Slavic and Finno-Ugric. Princeton Diss. Jake, Janice 1994. “Intrasentential Codeswitching and Pronouns: on the categorial status of functional elements”. Linguistics 32: 271–298. Jelinek, Eloise 1984. “Empty Categories, Case and Configurationality”. Natural Language and Linguistic Theory 2: 39–76. Jelinek, Eloise 1989. “The Case Split and Pronominal Arguments in Choctaw”. In: László Marácz & Pieter Muysken, Configurationality: The typology of asymmetries, 117–141. Dordrecht: Foris. Jelinek, Eloise 1997. “Functional Projections and Clitics”. University of Arizona ms. Jelinek, Eloise 1998. “Voice and transitivity as Functional Projections in Yaqui”. In: Miriam Butt & Wilhelm Geuder (eds), The Projection of Arguments, 195–224. Stanford: CSLI. Jelinek, Eloise 2001 “Pronouns and Argument Hierarchies”. Paper presented at the Workshop on Agreement in Argument Structure, Utrecht. Jespersen, Otto 1917. Negation in English and Other Languages. Copenhagen: A. F. Høst. Jespersen, Otto 1921. Language. London: Allen & Unwin. Jespersen, Otto 1924. The Philosophy of Language. London: Allen & Unwin. Jespersen, Otto 1926. Modern English Grammar III. London: Allen & Unwin [1974 reprint]. Jespersen, Otto 1931. Modern English Grammar IV. London: Allen & Unwin [1970 reprint]. Jespersen, Otto 1940. Modern English Grammar V. London: Allen & Unwin [1970 reprint]. Josefsson, Gunlög 2000. “Scandinavian Pronouns and Object Shift”. In: Henk van Riemsdijk (ed.), Clitics in the languages of Europe, 731–757. Berlin: Mouton de Gruyter.


Josefsson, Gunlög & Gisela Håkansson 2000. “The PP-CP Parallelism Hypothesis and Language Acquisition”. In: Susan Powers and Cornelia Hamann (eds), The Acquisition of Scrambling and Cliticization, 397–422. Dordrecht: Kluwer. Jucker, Andreas 1991. “Between Hypotaxis and Parataxis. Clauses of Reason in the Ancrene Wisse”. In: Dieter Kastovsky (ed.), Historical English Syntax: 203–220. Berlin: Mouton de Gruyter. Julien, Marit 2000. “Optional ha in Swedish and Norwegian”. Working Papers in Scandinavian Syntax 66: 33–74. Julien, Marit 2001. “The Syntax of Complex Tenses”. The Linguistic Review: 125–167. Kaisse, Ellen 1983. “The Syntax of Auxiliary Reduction in English”. Language 59: 73–89. Kayne, Richard 1994. The Antisymmetry of Syntax. Cambridge: MIT Press. Kayne, Richard 1997. “The English Complementizer of”. Journal of Comparative Germanic Linguistics 1: 43–54. Kayne, Richard 1999. “Prepositions as Attractors”. Probus 11: 39–73. Kayne, Richard 2000. Parameters and Universals. Oxford: Oxford University Press. Kayne, Richard 2002. “On some Prepositions that look DP-internal”. Ms. Kean, Marie-Louise 1979. “Agrammatism”. Cognition 7: 69–84. Keenan, Ed & Bernard Comrie 1977. “Noun Phrase Accessibility and Universal Grammar”. Linguistic Inquiry 8: 63–99. Kellner, Leon 1892. Historical Outlines of English Syntax. London: MacMillan. Kemenade, Ans van 1997. “V2 and Embedded Topicalization in Old and Middle English”. In: Ans van Kemenade & Nigel Vincent (eds), Parameters of Morphosyntactic Change, 326–352. Cambridge: Cambridge University Press. Kemenade, Ans van & Bettelou Los 2002. “Particles and Prefixes in Dutch and English”. Ms. Kerkhof, J. 1966. Studies in the Language of Geoffrey Chaucer. Leiden: Universitaire Pers. Kiparsky, Paul 1995. “Indo-European Origins of Germanic Syntax”. In: Adrian Battye and Ian Roberts (eds.), Clause Structure and Language Change, 140–169. Oxford: Oxford University Press. Kiparsky, Paul 1996. “Partitive Case and Aspect”, ms Stanford. Kiparsky, Paul & Carol Kiparsky 1970. “Fact”. In: Manfred Bierwisch & Karl Heidolph (eds), Progress in Linguistics, 143–173. Den Haag: Mouton. Kirsner, Robert 1977. “On the Passive of Sensory Verb Complement Sentences”, Linguistic Inquiry 8.1: 173–9. Kirszner, Laurie & Stephen Mandell 1992. The Holt Handbook [third edition]. Fort Worth: Harcourt Brace Jovanovich. Kiss, Katalin (ed.) 1995. Discourse Configurational Languages. Oxford: Oxford University Press. Kivimaa, Kirsti 1966. þe and þat as Clause Connectives in Early Middle English with Especial Consideration of the Emergence of the Pleonastic þat. Helsinki: Societas Scientiarum Fennica. Klamer, Marian 1998. A Grammar of Kambera. Berlin: Mouton de Gruyter. Klamer, Marian 2002. “Multi-Categorial Items as Underspecified Lexical Entries: the case of Kambera wàngu”. ms Leiden University. Kock, Ernst Albin 1897. The English Relative Pronouns: a critical essay. Lund: Hjalmar Möller. König, Ekkehard 1991. The Meaning of Focus Particles. London: Croom Helm.

References 299

Koopman, Hilda 1996. “The Spec Head Configuration”. UCLA Working Paper in Linguistics 1. Koopman, Hilda 2000. The Syntax of Specifiers and Heads. London: Routledge. Koopman, Willem 1997. “Topicalization in Old English and its Effects”. In: Raymond Hickey (ed.), Language History and Linguistic Remodelling, 307–321. Berlin: Mouton de Gruyter. Koopman, Willem 1998. “Inversion after Single and Multiple Topics in Old English”. In: Jacek Fisiak & Marcin Krygier (eds), Advances in Historical Linguistics, 135–150. Berlin: Mouton de Gruyter. Koster, Jan 1987. Domains and Dynasties. Dordrecht: Foris. Kroch, Anthony & Anne Taylor 1997. “Verb Movement in Old and Middle English”. In: Ans van Kemenade & Nigel Vincent (eds), Parameters of Morphosyntactic Change, 297–325. Cambridge: Cambridge University Press. Krug, Manfred 2000. Emerging English Modals. Berlin: Mouton de Gruyter. Kuno, Susumu 1973. The Structure of the Japanese Language. Cambridge: MIT Press. Kwee, John 1968. Indonesian. Hodder & Stoughton. Kytö, Merja & Matti Rissanen 1988. “The Helsinki Corpus of English Texts”. In: Merja Kytö, Ossi Ihalainen, & Matti Rissanen (eds), Corpus Linguistics, hard and soft, 169–70. Amsterdam: Rodopi. Labov, William 1972. Language in the Inner City. Philadelphia: University of Pennsylvania Press. Lakoff, Robin 1968. Abstract Syntax and Latin Complementation. Cambridge: MIT Press. Lambrecht, Knut 1981. Topic, Antitopic, and Verb Agreement in Non Standard French. Amsterdam: John Benjamins. Lambton, Ann 1953. A Persian Grammar. Cambridge: Cambridge University Press [1990 reprint]. Landau, Idan 2000. Elements of Control. Dordrecht: Kluwer. Lane, Archibald 1700. A Key to the Art of Letters. Menston: Scolar Press [1969 reprint]. Larson, Richard 1988. “On the Double Object Construction”. Linguistic Inquiry 19.3: 335–391. Lasnik, Howard & Mamoru Saito 1992. Move [Alpha]. Cambridge: MIT Press. Lehmann, Christian 1988. “Towards a Typology of Clause Linkage”. In: John Haiman & Sandra Thompson, 181–225. Amsterdam: John Benjamins. Lehmann, Winfred 1978. “Topic to Subject in Indo-European”. In: Charles Li (ed.), Subject and Topic, 447–456. New York: Academic Press. Lehmann, Winfred 1980. “The Reconstruction of Non-simple Sentences in Proto-IndoEuropean”. In: Paolo Ramat (ed.), Linguistic Reconstruction and Indo-European Syntax, 113–144. Amsterdam: John Benjamins. Lehmann, Winfred 1993. Theoretical Bases of Indo-European Linguistics. London: Routledge. Leiss, Elisabeth 2000. Artikel und Aspect. Berlin: Walter de Gruyter. Lenerz, J. 1985. “Diachronic Syntax: Verb Position and COMP in German”. In: Jindrich Toman (ed.), Studies in German Grammar, 103–132. Dordrecht: Foris. Lenz, Philipp 1886. Der Syntactische Gebrauch der Partikel GE in den Werken Alfred des Grossen. Darmstadt: G. Otto. Li, Charles & Sandra Thompson 1974. “Co-verbs in Mandarin Chinese”. Journal of Chinese Linguistics 2: 257–278.


Li, Charles & Sandra Thompson 1981. Mandarin Chinese: A Functional Reference Grammar. Berkeley: University of California Press. Lightfoot, David 1979. Principles of Diachronic Syntax. Cambridge: Cambridge University Press. Lightfoot, David 1999. The Development of Language. Oxford: Blackwell. Lin, Tzong-Hong 2001. Light Verb Syntax and the Theory of Phrase Structure. UC Irvine Diss. Lockwood, William 1968. Historical German Syntax. Oxford: Clarendon. Longobardi, Giuseppe 1994. “Reference and Proper Names”. Linguistic Inquiry 25: 609–665. Lord, Carol 1993. Historical Change in Serial Verb Constructions. Amsterdam: John Benjamins. Lowth, Robert 1762. A Short Introduction to English Grammar. Menston: Scolar Press. Luraghi, Silvia 1990. Old Hittite Sentence Structure. London: Routledge. Lyons, Christopher 1999. Definiteness. Cambridge: Cambridge University Press. Lyons, John 1977. Semantics, Vol 2. Cambridge: Cambridge University Press. MacSwan, Jeff 1999. A Minimalist Approach to Intrasentential Code Switching. New York: Garland. Mahajan, Anoop 2000. “DP Scrambling and VP Scrambling”. Talk, University of Arizona. Maki, Hideki, Lizanne Kaiser, & Masao Ochi 1998. “Embedded Topicalization in English and Japanese”. WECOL Proceedings 10: 289–304. Malotki, Ekkehart 1983. Hopi Time. Berlin: Mouton. Manzini, Rita & Ken Wexler 1987. “Parameters, Binding Theory, and Learnability”. Linguistic Inquiry 18.3: 413–444. Marácz, László & Pieter Muysken (eds) 1989. Configurationality. Dordrecht: Foris. Marácz, László 1990. “Some Notes on Long Movement in Hungarian”. Groningen ms. Massam, Diane 1985. Case Theory and the Projection Principle. MIT Diss. Mathesius, Vilém 1975. A Functional Analysis of Present Day English. Den Haag: Mouton. Matisoff, James 1991. “Areal and Universal Dimensions of Grammaticalization in Lahu”. In: Elizabeth Traugott & Bernd Heine (eds), Approaches to Grammaticalization II, 383–453. Amsterdam: John Benjamins. Mätzner, Eduard 1864. Englische Grammatik. Berlin: Weidman. McCloskey, Jim 1991. “Verb Fronting, Verb Second and the Left Edge of IP in Irish”. Talk, Stuttgart workshop on Comparative Germanic Syntax. Meillet, Antoine 1937. Introduction a l’étude comparative des langues indo-européennes. Paris: Hachette. Meinunger, André 1995. Discourse Dependent DP (De)placement. Potsdam Diss. Meinunger, André 2002. “Verb Position, Verbal Mood, and the Anchoring (Potential) of Sentences”. Ms. Miege, Guy 1688. The English Grammar. Menston: Scolar Press [1969 reprint]. Miller, Gary 2002. Nonfinite Structures in Theory and Change. Oxford: Oxford University Press. Miller, Jim 1993. “The Grammar of Scottish English”. In: James Milroy & Lesley Milroy (eds), Real English, 99–138. London: Longman. Minkova, Donka 1991. The History of Final Vowels in English: The Sound of Muting. Berlin: Mouton de Gruyter.

References 301

Mitchell, Bruce 1985. Old English Grammar I and II. Oxford: Oxford University Press. Mithun, Marianne 1984. “How to Avoid Subordination”. Berkeley Linguistic Society Proceedings 10: 493–509. Mithun, Marianne 1988. “The Grammaticalization of Coordination”. In: John Haiman & Sandra Thompson (eds), Clause Combining in Grammar and Discourse, 331–359. Amsterdam: John Benjamins. Montgomery, Michael & Guy Bailey 1991. “In Which: A new form in written English”. American Speech 66.2: 147–163. Moore, Samuel & Albert Marckwardt 1951. Historical Outlines of English Sounds and Inflections. Ann Arbor: George Wahr. Mossé, Fernand 1938. Histoire de la Forme périphrastique être + participe présent en germanique. Paris: Klincksieck. Mossé, Fernand 1962. Manuel de l’anglais du Moyen Age II. Paris: Aubier. Munn, Alan 1992. “A Null Operator Analysis of Parasitc Gaps”. The Linguistic Review 9.1: 1–26. Mustanoja, Tauno 1960. A Middle English Syntax. Helsinki. Myhill, John 1988. “Variation in Spanish Clitic Climbing”. Georgetown University Round Table: 227–250. Myhill, John 1992. Typological Discourse Analysis. Oxford: Blackwell. Newmeyer, Frederick 1998. Language Form and Language Function. Cambridge: MIT Press. Noonan, Michael 1996. “Complementation”. In: Timothy Shopen (ed.), Language Typology and Syntactic Description II [2nd edition], 42–140. Cambridge University Press. Noonan, Michael 1997. “Versatile Nominalizations”. In: Joan Bybee et al. (eds), Essays on Language Function and Language Type, 373–394. Amsterdam: John Benjamins. Nortier, Jacomine 1990. Dutch-Moroccan Arabic Code Switching. Dordrecht: Foris. Oku, Satoshi 2001. “A Minimalist Theory of LF Copy”. In: Galina Alexandrova et al. (eds), The Minimalist Parameter, 281–94. Amsterdam: John Benjamins. O’Neil, Wayne 1977. “Clause Adjunction in Old English”. General Linguistics 17: 199–211. Ordonez, Francisco & Esthela Trevino 1999. “Left Dislocated Subjects and the Pro-drop Parameter”. Lingua 107: 39–68. Österman, Aune 2001. “Where your treasure is, there is your heart”: A Corpus-based Study of THERE Compounds and THERE/WHERE Subordinators in the History of English. Helsinki: Société Néophilologique. Ouhalla, Jamal 1990. “Sentential Negation, Relativised Minimality and the Aspectual Status of Auxiliaries”. The Linguistic Review 7: 183–231. Palmer, F. 1986. Mood and Modality. Cambridge: Cambridge University Press. Park, Kabyong 1992. Light Verb Constructions in Korean and Japanese. UNC Chapel Hill Diss. Payne, Doris 1985. Aspects of the Grammar of Yagua. UCLA Diss. Pessels, Constance 1896. The Present and Past Periphrastic Tenses in Anglo-Saxon. Trubner: Strassburg. [Johns Hopkins diss] Picallo, Carmen 1984. Opaque Domains. CUNY Diss. Picallo, Carmen 1990. “Modal Verbs in Catalan”. Natural Language and Linguistic Theory 8: 285–312.


Pierce, Amy 1994. “On the Differing Status of Subject Pronouns in French and English Child Language”. In: Barbara Lust et al. (eds), Syntactic Theory and First Language Acquisition II, 319–333. Hillsdale: Erlbaum. Pintzuk, Susan 1999. Phrase Structures in Competition. New York: Garland [1991 Diss]. Plag, Ingo 1998. “The Syntax of some Locative Expressions in Sranan”. Journal of Pidgin and Creole Languages 13.2: 335–353. Platzack, Christer 1986. “COMP, INFL, and Germanic Word Order”. In: Lars Hellan & Kristi Koch Christensen (eds), Topics in Scandinavian Syntax, 185–234. Dordrecht: Reidel. Platzack, Christer 1998. Svenskans inre grammatik — det minimalistiska programmet. Lund: Studentliteratur. Poletto, Cecelia 2000. The Higher Functional Field. Oxford: Oxford University Press. Poletto, Cecelia 2001. “Complementizer Deletion and Verb Movement in Standard Italian”. In: Guglielmo Cinque and Giampaoli Salvi (eds), Current Studies in Italian Syntax, 265–286. Amsterdam: Elsevier. Pollock, J-Y. 1989. “Verb Movement, UG and the Structure of IP”. Linguistic Inquiry 20: 365–424. Ponelis, F. A. 1991. Afrikaanse Sintaksis. Pretoria: van Schaik. Postma, Gertjan 2002. “Negative Polarity and Modality in Middle Dutch ghe-particle Constructions”. In: Sjef Barbiers et al. (eds), Modality and its Interaction with the Verbal System, 205–244. Amsterdam: John Benjamins. Poussa, Patricia (ed.) 2002. Relativization on the North Sea Littoral. München: Lincom Europa. Priestley, J. 1761. Rudiments of Grammar. London. Progovac, Ljiljana 1993. “Locality and Subjunctive-like Complements in Serbo-Croatian”. Journal of Slavic Linguistics 1.1: 116–144. Quer, Josep 1998. Mood at the Interface. The Hague: Holland Academic Graphics. Quirk, Randolph et al. 1973. A Grammar of Contemporary English. London: Longman. Quirk, Randolph et al. 1985. Comprehensive Grammar of the English Language. London: Longman. Quirk, Randolph & C. Wrenn 1955. An Old English Grammar. London: Methuen. Radford, Andrew 1997. Syntactic Theory and the Structure of English. Cambridge: Cambridge University Press. Raith, J. 1951. Untersuchungen zum englischen Aspekt. Munchen: Max Hueber. Ramchand, Gillian 1997. Aspect and Predication. Oxford: Clarendon Press. Reichenbach, H. 1947. Elements of Symbolic Logic. New York: MacMillan. Reinhart, Tanya 1983. Anaphora and Semantic Interpretation. Chicago: Chicago University Press. Reis, Marga 1977. Präsuppositionen und Syntax. Tübingen: Niemeyer. Reis, Marga 1999. “Coherent and Incoherent Infinitives”. DGfS talk. Riemsdijk, Henk van 2000. Clitics in the Languages of Europe. Berlin: Mouton. Rizzi, Luigi 1978. “A Restructuring Rule in Italian Syntax”. In: S. J. Keyser (ed.), Recent Transformational Studies in European languages, 113–158. Cambridge: MIT Press. Rizzi, Luigi 1990. Relativized Minimality. Cambridge: MIT Press.

References 303

Rizzi, Luigi 1994. “Early Null Subjects and Root Null Subjects”. In: Teun Hoekstra & Bonnie Schwarz (eds), Language Acquisition Studies in Generative Grammar, 151–176. Amsterdam: John Benjamins. Rizzi, Luigi 1997. “The Fine Structure of the Left Periphery”. In: Liliane Haegeman (ed.), Elements of Grammar, 281–337. Dordrecht: Kluwer. Rizzi, Luigi 2001. “On the Position ‘Int(errogative)’ in the Left Periphery of the Clause”. In: Guglielmo Cinque et al. (eds), Current Studies in Italian Syntax, 287–296. Amsterdam: Elsevier. Rizzi, Luigi 2002. “Cartography and some Asymmetries in the Theory of Locality”. Comparative Germanic Syntax Workshop talk. Roberts, Ian 1985. “Agreement Parameters and the Development of English Modal Auxiliaries”. Natural Language and Linguistic Theory 3: 21–58. Roberts, Ian 1993. Verbs and Diachronic Syntax. Dordrecht: Kluwer. Roberts, Ian 1996. “Remarks on the Old English C-System and the Diachrony of V-2”. Linguistische Berichte. Sonderheft: 154–168. Roberts, Ian & Anna Roussou 1999. “A Formal Approach to ‘Grammaticalization’”. Linguistics 37: 1011–1041. Roberts, Ian & Anna Roussou 2003. Syntactic Change. Cambridge: Cambridge University Press. Rochette, Anne 1988. Semantic and Syntactic Aspects of Romance Sentential Complementation. MIT Diss. Romaine, Suzanne 1982. Socio-historical Linguistics. Cambridge: Cambridge University Press. Rosenbaum, Peter 1967. The Grammar of English Predicate Complement Constructions. Cambridge: MIT Press. Rothstein, Susan 1983. The Syntactic Forms of Predication. MIT Diss. Roussou, Anna 2000. “On the Left Periphery”. Journal of Greek Linguistics 1: 65–94. Russom, Geoffrey 1976. “A Syntactic Key to a Number of Pearl-group Cruxes”. Journal of English Linguistics 10: 21–29. Rydén, Mats 1983. “The Emergence of who as Relativizer”. Studia Linguistica 37: 126–134. Rynell, Alarik 1952. Parataxis and Hypotaxis as a Criterion of Syntax and Style. Lund: Gleerup. Safir, Ken 1985. Syntactic Chains. Cambridge: Cambridge University Press. Santelmann, Lynn 1995. “Topicalization, CP and Licensing in the Acquisition of Swedish”. Proceedings of Boston University Conference on Language Development 19: 499–510. Sanz, Montserrat 1996. Telicity, Objects and the Mapping onto Predicate Types. Rochester Diss. Saxena, Anju 1988. “On Syntactic Convergence: the case of the verb ‘say’ in Tibeto-Burman”. Berkeley Linguistic Society, Parasession on Grammaticalization: 375–388. Scheffer, Johannes 1975. The Progressive in English. Amsterdam: North Holland Publishing. Schwenter, Scott & Elizabeth Traugott 1995. “The Semantic and Pragmatic Development of Substitutive Complex Prepositions in English”. In: Andreas Jucker (ed.), Historical Pragmatics, 243–73. Amsterdam: John Benjamins. Seuren, Pieter 1999. “Factivity”. In: Keith Brown & Jim Miller (eds), Concise Encyclopedia of Grammatical Categories, 145–6. Oxford: Pergamon.


Shi, Yuzhi 2002. The Establishment of Modern Chinese Grammar. Amsterdam: John Benjamins. Shopen, Timothy (ed.) 1985. Language Typology and Syntactic Description II. Cambridge: Cambridge University Press. Simpson, Andrew & Xiu-Zhi Zoe Wu 2002. “Agreement Shells and Focus”. Language 78.2: 287–313. Simpson, Andrew & Zoe Wu n.d. “IP-raising, Tone Sandhi, and the Creation of Particles”. Slobin, Dan & Ayhan Aksu 1982. “Tense, Aspect, and Modality in the Use of the Turkish Evidential”. In: Paul Hopper (ed.), Tense-Aspect: between Semantics and Pragmatics, 185–200. Amsterdam: John Benjamins. Smith, Carlota 1991. The Parameter of Aspect. Dordrecht: Kluwer. Snyder, William 2000. “An Experimental Investigation of Syntactic Satiation Effects”. Linguistic Inquiry 31.1: 575–82. Snyder, William 2001. “On the Nature of Syntactic Variation: evidence from complex predicates and complex word-formation”. Language 77.2: 324–342. Sobin, Nicholas 1987. “The Variable Status of COMP-Trace Phenomena”. Natural Language and Linguistic Theory 5: 33–60. Sohn, Ho-min 1999. The Korean Language. Cambridge: Cambridge University Press. Southworth, James 1947. “Chaucer’s Final -e in Rhyme”. Proceedings Modern Language Association 42: 910–935. Speas, Margaret 1990. Phrase Structure in Natural Language. Dordrecht: Kluwer. Stein, Dieter 1998. “Relative Sentences in Late Middle English”. In: Dieter Stein & Rosanna Sornicola (eds), The Virtues of Language, 67–77. Amsterdam: John Benjamins. Stockwell, Robert & Donka Minkova 1990. “Verb Phrase Conjunction in Old English”. In: Henning Andersen & Konrad Koerner (eds), Historical Linguistics 1987, 499–515. Amsterdam: John Benjamins. Stockwell, Robert & Donka Minkova 1991. “Subordination and Word Order Change in the History of English”. In: Dieter Kastovsky (ed.), Historical English Syntax, 367–408. Berlin: Mouton de Gruyter. Stoelke, H. 1916. Die Inkongruenz zwischen Subjekt und Prädikat im Englischen und in den verwandten Sprachen. Marburg Diss. Stowell, Tim 1981. Origins of Phrase Structure. MIT Diss. Stowell, Tim 1982. “The Tense of Infinitives”. Linquistic Inquiry 13: 560–70. Strang, Barbara 1970. A History of English. London: Methuen. Streitberg, W. 1891. “Perfective und imperfective Actionsart im Germanischen”. Beiträge zur Geschichte der deutschen Sprache und Literatur 15: 70–177. Stroik, Tom 1991. “Expletives Revisited”. Linguistic Analysis 21: 23–33. Sun, Chaofen 1996. Word-order change and Grammaticalization in the History of Chinese. Stanford University Press. Swaay, H. van 1901. Het prefix ga-, gi-, ge-, zijn geschiedenis, en zijn invloed op de ‘Actionsart’. Utrecht: Kemink & Zoon. Swadesh, Morris 1946. “Chitimacha”. In: Harry Hoijer et al. (eds), Linguistic Structures of Native America, 312–336. New York: Viking Fund Publications in Anthropology. Swan, Michael 1980. Practical English Usage. Oxford: Oxford University Press. Sybesma, Rint 1999. The Mandarin VP. Dordrecht: Kluwer.

References 305

Tait, Mary, Cynthia Thompson, & Kirrie Ballard 1995. “Subject-object asymmetries in agrammatic comprehension of four types of wh-question”. Brain & Language 51: 77–79. Taraldsen, Tarald 1985. “On Verb-second and the Functional Content”. In: Hubert Haider & Martin Prinzhorn (eds), Verb-Second Phenomena in Germanic Languages, 7–25. Dordrecht: Foris. Taraldsen, Tarald 1992. “Agreement as Pronoun Incorporation”. GLOW talk. Tenny, Carol 1987. Grammaticalizing Aspect and Affectedness. MIT Diss. Thráinsson, Höskuldur 1979. On Complementation in Icelandic. New York: Garland. Thráinsson, Höskuldur & Sten Vikner 1995. “Modals and Double Modals in the Scandinavian Languages”. Working Papers in Scandinavian Syntax 55: 51–88. Traugott, Elizabeth 1972. The History of English Syntax. New York. Traugott, Elizabeth 1989. “On the Rise of Epistemic Meanings in English: An example of subjectification in semantic change”. Language 65.1: 31–55. Traugott, Elizabeth 1992. “Syntax”. In: Richard Hogg (ed.), The Cambridge History of the English Language, 168–289. Cambridge: Cambridge University Press. Traugott, Elizabeth & Bernd Heine 1991. Grammaticalization. Amsterdam: John Benjamins. Travis, Lisa 2000. “The L-Syntax/S-Syntax Boundary”. Formal Issues in Austronesian Linguistics, 167–194. Dordrecht: Kluwer. Uchibori, Asako 2000. The Syntax of Subjunctive Complements in Japanese. University of Connecticut Diss. Ukaji, Masatomo 1997. “A History of whether”. In: Masatomo Ukaji et al. (eds), Studies in English Linguistics, 1236–1261. Tokyo: Tashukan. Underhill, Robert 1988. “Like is, like, Focus”. American Speech 63: 234–246. Vachek, Josef 1961. “Some Less Familiar Aspects of the Analytical Trend of English”. Brno Studies in English 3: 9–71. Varlokosta, Spyridoula 1994. Issues on Modern Greek Complementation. University of Maryland Diss. Veenstra, Tonjes 1996. Serial Verbs in Saramaccan. Dordrecht: ICG. Vennemann, Theo 1984. “Verb-second, Verb Late, and the Brace Construction”. In: Jacek Fisiak (ed.), Historical Syntax, 627–636. Berlin: Mouton de Gruyter. Viberg, Åke 1983. “The Verbs of Perception: a typological study”. Linguistics 21: 123–162. Villiers, Jill de 1999. “Language and Theory of Mind”. In: S. Baron-Cohen et al. (eds), Understanding Other Minds, 83–123. Oxford: Oxford University Press. Vikner, Sten 1995. Verb Movement and Expletive Subjects in the Germanic Languages. Oxford: Oxford University Press. Vincent, Nigel 1999. “The Evolution of C-structure”. Linguistics 37: 1111–1153. Vismans, Roel 1994. Modal Particles in Dutch Directives: A Study in Functional Grammar. Amsterdam: IFOTT. Visser, F. Th. 1946–52. A Syntax of the English Language of St Thomas More. Leuven: Librairie Universitaire. Visser, F. Th. 1963–1973. An Historical Syntax of the English Language. Leiden: Brill. Vlach, Frank 1981. “The Semantics of the Progressive”. In: Philip Tedeschi and Annie Zaenen (eds), Syntax and Semantics: Tense and Aspect, 272–292. New York: Academic Press. Voorst, Jan van 1987. “The Structure of Aspect”. Canadian Journal of Linguistics 32: 261–275.


Wal, Marijke van der 2002. “Relativization in the History of Dutch”. In: Patricia Poussa (ed.), Relativization on the North Sea Littoral, 27–35. München: Lincom Europa. Warner, Anthony 1982. Complementation in Middle English and the Methodology of Historical Syntax. University Park: The Pennsylvania State University Press. Warner, Anthony 1993. English Auxiliaries. Cambridge: Cambridge University Press. Wasow, Tom 2002. Postverbal Behavior. Stanford: CSLI. Watters, John 2000. “Syntax”. In: Bernd Heine & Derek Nurse (eds), African Languages, 194–230. Cambridge: Cambridge University Press. Weber, Georg 1934. Der Bau der englischen Sprache. Leipzig: Mayer & Müller. Wessén, Elias 1968 [1970]. Vårt svenska språk. Stockholm: Almqvist & Wiksell. Weydt, Harald (ed.) 1979. Die Partikeln der deutschen Sprache. Berlin. Wharton, Jeremiah 1654. The English Grammar. Menston: Scolar Press [1970 reprint]. Windisch, Ernst 1869. “Untersuchungen über den Ursprung des Relativpronomens in den indogermanischen Sprachen”. Curtius Studien II: 201–419. Willie, Mary 1991. Navajo Pronouns and Obviation. University of Arizona Diss. Wilson, L. 1906. Chaucer’s Relative Constructions. Philological Club of the University of North Carolina. Wolfram, Walt, Kirk Hazen, and Natalie Schilling-Estes 1999. Dialect Change and Maintenance on the Outer Banks. Tuscaloosa: University of Alabama Press. Wood, Johanna 2003. Definiteness and Number. Arizona State University Diss. Wouden, Ton van der 1999. “Kleine Rotwoordjes: Smeermiddelen van de taal”. Onze Taal 68.11: 292–295. Wu, Xiu-Zhi Zoe 2000. Grammaticalization and the Development of Functional Categories. University of Southern California Diss. Wülfing, E. 1894–1901. Die Syntax in den Werken Alfreds des Grossen I and II. Bonn: Hanstein. Wurff, Wim van der 1994. “Some Observations on the Nature and Historical Development of the English Perfect”. In: R. Boogaart & J. Noordegraaf (eds), Nauwe Betrekkingen, 291–299. Münster: Amsterdam/Nodus Publikationen. Wurmbrand, Susi 2001. Infinitives. Berlin: Mouton de Gruyter. Zeitlin, Jacob 1908. The ACI and Some Kindred Constructions in English. Columbia Diss. Zickner, B. 1900. Syntax und Stil in Pecock’s Repressor. Berlin: Greifswald. Zou, Ke 1995. “A Minimalist Approach to the Chinese ba-Construction”. West Coast Conference on Linguistics Proceedings. Fresno: University of California Press. Zwart, Jan-Wouter 1997. Morpho-syntax of Verb Movement. Dordrecht: Kluwer. Zwicky, Arnold & Geoffrey Pullum 1983. “Cliticization vs. Inflection: English n’t”. Language 59: 502–513.

Index

Some entries are not exhaustive. For example, the OED is only listed when relevant to the argument developed in the text, not for every citation. Some entries, e.g. ‘OE’, ‘ModE’, ‘Economy’, for, to, ‘grammaticalization’ and ‘Late Merge’, are not listed at all. The general table of contents should be used here.

Reference Index A Abbott, E. A. Abraham, Werner Adger, David Aksu, Ayhan Akmajian, Adrian Algemene Nederlandse Spraakkunst (ANS) Allen, Cynthia Altmann, Hans d’Ardenne, S. Armitage, Richard Anderwald, Lieselotte Authier, J. Marc Avram, Larisa Axel, Katrin B Babington, Churchill Bacquet, P. Bailey, Guy Baker, Mark Bakker, Peter Ballard, Kirrie Baltin, Mark Banfi, Emanuele Barbiers, Sjef Barker, Muhammad

Barnes, Janet Basilico, David Bean, Marian Bech, G. Bentahila, Abdelali Benveniste, Emile Bergs, Alexander Berndt, Rolf Bernstein, Michael Besten, Hans den Beukema, Frits Bickerton, Derek Bisang, Walter Blom, Corrien Boeckx, Cedric Boertien, Harmon Bonneau, Jose Booij, Geert Bosworth, Joseph Botne, Robert Brinton, Laurel British National Corpus (BNC) Brown, Keith Bruhn-Garavito, J. Brunner, Karl Brusendorff, Aage Bullokar, William


Buzarovska, Eleni Bybee, Joan C Callaway, Morgan Campbell, Alistair Campbell, Lyle Cardinaletti, Anna Carstensen, B. Chapman, Carol Chen, Guohua Chiu, Bonnie Hui-Chun Choi, Ji-Young Chomsky, Noam Cinque, Guglielmo Claudi, Ulrike Coar, Thomas Cobuild Corpus Comrie, Bernard Condoravdi, Cleo Cormack, Annabel Corpus of Spoken Professional American English (CSE) Cowling, George Cowper, Elizabeth Craenenbroeck, Jeroen van D Dam, Johannes van Davies, Eirlys Davis, Norman Dekeyser, Xavier Delancey, Scott Denison, David Deutscher, Guy Di Paolo, Marianna Dik, Simon Diesing, Molly Diessel, Holger Dikken, Marcel den Disterheft, Dorothy Dixon, R. M. Doherty, Cathal Duncan, Daniel Durrleman, Stephanie

E Eckardt, Eugen Enç, Mürvet Ek, J. van Emonds, Joe Erteschik-Shir, Nomi Evers, Arnold Eythórsson, Thórhallur F Faarlund, Jan Terje Falk, Lilian Faltz, Leonard Farkas, Donka Felser, Claudia Fenning, Daniel Fischer, Olga Fisher, Ann Fowler, H. W. Frajzyngier, Zygmunt Franz, W. Friederici, Angela G Gardiner, Alan Gelderen, Elly van Gestel, Frank Ginneken, Jac. van Giorgi, Alessandra Givón, Talmy Goldman, Robert Gonda, Jan Goossens, Louis Gordon, E. V. Gordon, Lynn Görlach, Manfred Greenwood, James Grohmann, Kleanthes Grossmann, Heinrich Guasti, Maria Teresa H Haan, Ferdinand Haddad, Ghassan Haeberli, Eric

Index 309

Haider, Hubert Haiman, John Hale, Ken Hale, Mark Haller, Beat Harris, Alice Hashimoto, Anne Haspelmath, Martin Heath, Jeffrey Heine, Bernd Hengeveld, Kees Haim, S. Håkansson, Gisela Hazen, Kirk Helbig, Gerhard Helsinki Corpus (HC) Henry, Alison Hentschel, Elke Higginbotham, James Hiltunen, Risto Hinterhölzl, Roland Ho, Li-Ting Hock, Hans Hoekstra, Eric Hoekstra, Teun Holm, John Holmberg, Anders Hooper, Joan Hopper, Mike Horn, Laurence Hornstein, Norbert Hünnemeyer, Friederike Hupe, H. Hyams, Nina I IJbema, Aniek Ikawa, Hisako J Jacobs, Joachim Jacobs, Roderick Jakab, Edit Jake, Janice Jelinek, Eloise

Jespersen, Otto Josefsson, Gunlög Jucker, Andreas Julien, Marit K Kail, Michele Kaisse, Ellen Kaiser, Lizanne Kayne, Richard Kean, Marie-Louise Keenan, Ed Kellner, Leon Kemenade, Ans van Kenstowicz, Michael Kerkhof, J. Keyser, S. J. Kiparsky, Paul Kiparsky, Carol Kirsner, Robert Kirszner, Laurie Kiss, Katalin Kivimaa, Kirsti Klaeber, Fr. Klamer, Marian Kock, Ernst Albin König, Ekkehard Koopman, Hilda Koopman, Willem Koster, Jan Kroch, Anthony Krug, Manfred Kuno, Susumu Kuteva, Tania Kwee, John Kytö, Merja L Labov, William Lakoff, Robin Lambrecht, Knut Lambton, Ann Landau, Idan Lane, Archibald Larson, Richard


Lasnik, Howard Lehmann, Christian Lehmann, Winfred Leiss, Elisabeth Lenerz, J. Lenz, Philipp Li, Charles Libert, Alan Lightfoot, David Lin, Tzong-Hong Lockwood, William Longobardi, Giuseppe Lord, Carol Los, Bettelou Lowth, Robert Luraghi, Silvia Lyons, Christopher Lyons, John M MacSwan, Jeff Mahajan, Anoop Maki, Hideki Malotki, Ekkehart Mandell, Stephen Manzini, Rita Marácz, László Marckwardt, Albert Massam, Diane Mathesius, Vilém Matisoff, James Mätzner, Eduard McCloskey, Jim Meiden, Walter Meillet, Antoine Meinunger, André Miege, Guy Miller, Gary Miller, Jim Minkova, Donka Mitchell, Bruce Mithun, Marianne Montgomery, Michael Moore, Samuel Morris, Richard

Mossé, Fernand Mulcaster, Richard Mustanoja, Tauno Muysken, Pieter Myhill, John N Newmeyer, Frederick Nijen Twilhaar, Jan Noonan, Michael Nortier, Jacomine Nurse, Derek O Ochi, Masao Oku, Satoshi O’Neil, Wayne Ordonez, Francisco Österman, Aune Ouhalla, Jamal Oxford English Dictionary (OED) P Pagliuca, William Palmer, F. Park, Kabyong Payne, Doris Perkins, Revere Pessels, Constance Pianesi, Fabio Picallo, Carmen Pierce, Amy Pintzuk, Susan Plag, Ingo Platts, John Platzack, Christer Poletto, Cecelia Pollock, J-Y. Ponelis, F. A. Post, Marike Postma, Gertjan Poussa, Patricia Priestley, J. Progovac, Ljiljana Pullum, Geoffrey

Index

Q Quer, Josep Quirk, Randolph R Radford, Andrew Raith, J. Ramchand, Gillian Reh, Mechthild Reichenbach, H. Reinhart, Tanya Reis, Marga Rinkel, Tineke Rissanen, Matti Rizzi, Luigi Roberts, Ian Rochette, Anne Romaine, Suzanne Rosenbaum, Peter Rothstein, Susan Roussou, Anna Russom, Geoffrey Rydén, Mats Rynell, Alarik S Safir, Ken Saito, Mamoru Santelmann, Lynn Sanz, Montserrat Saxena, Anju Scheffer, Johannes Schilling-Estes, Natalie Schwenter, Scott Seuren, Pieter Shi, Yuzhi Shopen, Timothy Simpson, Andrew Slobin, Dan Smith, Carlota Smith, Neil Snyder, William Sobin, Nicholas Sohn, Ho-min Southworth, James

Speas, Margaret Starke, Michael Steele, Susan Stein, Dieter Stockwell, Robert Stoelke, H. Stowell, Tim Strang, Barbara Streitberg, W. Stroik, Tom Sun, Chaofen Sutherland, Sally Swaay, H. van Swadesh, Morris Swan, Michael Sybesma, Rint T Tait, Mary Taraldsen, Tarald Taylor, Anne Tenny, Carol Thompson, Cynthia Thompson, Sandra Thráinsson, Höskuldur Toller, T. Northcote Tomasello, Michael Traugott, Elizabeth Travis, Lisa Trevino, Esthela U Uchibori, Asako Ukaji, Masatomo Underhill, Robert V Vachek, Josef Varlokosta, Spyridoula Veenstra, Tonjes Vennemann, Theo Viberg, Åke Villiers, Jill de Vikner, Sten Vincent, Nigel

311

312


Vismans, Roel Visser, F. Th. Vlach, Frank Voorst, Jan van Voort, Hein van der W Wal, Marijke van der Warner, Anthony Wasow, Tom Watters, John Weber, Georg Weerman, Fred Weissenborn, Jurgen Wessén, Elias Wexler, Ken Weydt, Harald Wharton, Jeremiah Windisch, Ernst

Willie, Mary Wilson, L. Wolfram, Walt Wood, Johanna Wouden, Ton van der Wrenn, C Wu, Xiu-Zhi Zoe Wülfing, E. Wurff, Wim van der Wurmbrand, Susi Z Zeitlin, Jacob Zickner, B. Zou, Ke Zupitza, Julius Zwart, Jan-Wouter Zwicky, Arnold

Language Index A Afrikaans Akkadian American English Ani Arabic (see also Moroccan) B Baka Bambara Bari Belfast English Bulgarian C Cahuilla Cape Breton Island English Catalan Chantyal Chinese modern

old Chitimacha Czech D Danish Diegueño Dutch modern middle E Egyptian Ewe F Farsi Finnish French

Index

G Ga German, modern old high Gokana Gothic Greek, modern ancient H Haitian Hebrew Hindi Hittite Hopi I Icelandic Igbo Ik Imonda Indonesian Irish Italian J Jamaican Creole Japanese K Kambera Kikuyu Klao Korean L Lahu Lakota Lamang Latin Lega Lezgian Lhasa Lingala Lotuko

Luiseño M Macedonian Malagasy Maricopa Mina Misumalpan Mohawk Moroccan Arabic N Nawdm Nama Navajo Neyo North Country English Norwegian Nunggubuyu O Old Norse O’odham Ozark English P Pilara R River Cess Bassa Romanian S Salish Sango Sanskrit Saramaccan Scottish English Serbo-Croatian Spanish Sranan Susu Swahili Swedish

313


T Tagalog Thai Tibeto-Burman Turkish Twi Tzotzil

V Vai Vietnamese W Warlpiri

X !Xun

Y Yagua Yorkshire English

U Urdu

Z Zulgo

Topic Index A aan het absolutive accusative activity adjective adjunct adverbial adverb agent AGR agreement Aktionsart all-float alliteration anaphor (see also reflexive) anterior appositive argument article as for ASP aspect assertive auxiliary auxiliation

B barrier bi-clausal binding theory borrowing bound bridge verb C cartographic Case inherent morphological Case Resistance Principle ‘catastrophic’ change causative child language clause-union clitic clitic climbing code switching colloquial comparative competence complementizer complex NP constraint configurational

Index

continuous control coordination coordinator corpora C(P) CPadverb deletion grammaticalization layer movement oriented cyclic change D dative definiteness demonstrative deontic modal discourse dislocation (see left dislocation) distance to to do double complementizer double modal doubly filled comp filter E E (see also Extent) ‘ease’ principle edge ECM E-language embedded topic embedding emotive emphatic epistemic adverb modal evaluative adverb eventive evidential exhaustive control

experience expressivity external Extent (see also E) extraction F factive feature categorial semantic uninterpretable finite first person focus force formal fronting Full Interpretation functional category functionalism G ge generative genitive goal gradual change H habitual have head movement constraint head preference hypotactic I I(P) I-language imperative imperfective incorporation noun verb topic

315


indicative individual level predicate infinitival infinitive marker ing inherent Case inner aspect layer innovation inseparable prefix integrated interlacing internal interpretable feature interrogative interval IPadverb deletion grammaticalization oriented irrealis island it-pronominalization J Jespersen’s Cycle L layer parameter left-dislocation lexicon LF like location M macroparameter markedness merge minimalism modal modal particle

modification moment mono-clausal mood move movement MP multiple negation N negation negative raising neurolinguistic nominative non-assertive non-emotive non-factive non-stative noun nuclear O object object agreement opaque orthographic outer aspect layer P parameter paratactic partial control participle particle partitive passive passivization perception verb (PV) verb complement (PVC) perfective performance periphery

Index

PF phase pied piping polysynthetic possessive pragmatic predicate prefix preposition prescriptive preterite-present principle pro-drop progressive pronominal argument parameter language pronoun psycholinguistic purposive Q QP question quantifier quotative R raising realis reanalysis reduced have reflexive relative clause relativized minimality remnant resetting restrictive restructuring resultative resumptive root infinitive runic S

same subject Satzverknüpfung scrambling second person semantic semi-factive separable prefix serial verb serialization specificity speech act adverb split CP infinitive IP stage level predicate statistics stative stranding strong pronoun subject subject doubling subjunctive subordination subordinator T T(P), see IP adverb telic telicity temporal tense Tense Aspect Parameter terminative that that-deletion the fact that theta-role third person till TMA TOP topic topic incorporation

317

318


topicalization transitive transparent U unaccusative unergative uni-directional uninterpretable feature unisentiation V V-2 V-to-C verbchaining final movement second (see V2) third vocative Voice volition(al) VPadverb

complement coordination deletion oriented preposing VP-shell W wanna weak pronoun whclause Criterion element embedding extraction form movement preference pronoun relative question whether word order

In the series Linguistik Aktuell/Linguistics Today the following titles have been published thus far or are scheduled for publication: 1

2 3

4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29

KLAPPENBACH, Ruth (1911–1977): Studien zur Modernen Deutschen Lexikographie. Auswahl aus den Lexikographischen Arbeiten von Ruth Klappenbach, erweitert um drei Beiträge von Helene Malige-Klappenbach. (Written in German). 1980. xxiii, 313 pp. EHLICH, Konrad and Jürgen REHBEIN: Augenkommunikation. Methodenreflexion und Beispielanalyse. 1982. viii, 150 pp. With many photographic ills. ABRAHAM, Werner (ed.): On the Formal Syntax of the Westgermania. Papers from the 3rd Groningen Grammar Talks (3e Groninger Grammatikgespräche), Groningen, January 1981. 1983. vi, 242 pp. ABRAHAM, Werner and Sjaak De MEIJ (eds.): Topic, Focus and Configurationality. Papers from the 6th Groningen Grammar Talks, Groningen, 1984. 1986. v, 349 pp. GREWENDORF, Günther and Wolfgang STERNEFELD (eds.): Scrambling and Barriers. 1990. vi, 442 pp. BHATT, Christa, Elisabeth LÖBEL and Claudia Maria SCHMIDT (eds.): Syntactic Phrase Structure Phenomena in Noun Phrases and Sentences. 1989. ix, 187 pp. ÅFARLÍ, Tor A.: The Syntax of Norwegian Passive Constructions. 1992. xii, 177 pp. FANSELOW, Gisbert (ed.): The Parametrization of Universal Grammar. 1993. xvii, 232 pp. GELDEREN, Elly van: The Rise of Functional Categories. 1993. x, 224 pp. CINQUE, Guglielmo and Giuliana GIUSTI (eds.): Advances in Roumanian Linguistics. 1995. xi, 172 pp. LUTZ, Uli and Jürgen PAFEL (eds.): On Extraction and Extraposition in German. 1996. xii, 315 pp. ABRAHAM, Werner, Samuel David EPSTEIN, Höskuldur THRÁINSSON and C. Jan-Wouter ZWART (eds.): Minimal Ideas. Syntactic studies in the minimalist framework. 1996. xii, 364 pp. ALEXIADOU, Artemis and T. Alan HALL (eds.): Studies on Universal Grammar and Typological Variation. 1997. viii, 252 pp. ANAGNOSTOPOULOU, Elena, Henk van RIEMSDIJK and Frans ZWARTS (eds.): Materials on Left Dislocation. 1997. viii, 349 pp. ROHRBACHER, Bernhard Wolfgang: Morphology-Driven Syntax. A theory of V to I raising and prodrop. 1999. viii, 296 pp. LIU, Feng-Hsi: Scope and Specificity. 1997. viii, 187 pp. BEERMAN, Dorothee, David LEBLANC and Henk van RIEMSDIJK (eds.): Rightward Movement. 1997. vi, 410 pp. ALEXIADOU, Artemis: Adverb Placement. A case study in antisymmetric syntax. 1997. x, 256 pp. JOSEFSSON, Gunlög: Minimal Words in a Minimal Syntax. Word formation in Swedish. 1998. ix, 199 pp. LAENZLINGER, Christopher: Comparative Studies in Word Order Variation. Adverbs, pronouns, and clause structure in Romance and Germanic. 1998. x, 371 pp. KLEIN, Henny: Adverbs of Degree in Dutch and Related Languages. 1998. x, 232 pp. ALEXIADOU, Artemis and Chris WILDER (eds.): Possessors, Predicates and Movement in the Determiner Phrase. 1998. vi, 388 pp. GIANNAKIDOU, Anastasia: Polarity Sensitivity as (Non)Veridical Dependency. 1998. xvi, 282 pp. REBUSCHI, Georges and Laurice TULLER (eds.): The Grammar of Focus. 1999. vi, 366 pp. FELSER, Claudia: Verbal Complement Clauses. A minimalist study of direct perception constructions. 1999. xiv, 278 pp. ACKEMA, Peter: Issues in Morphosyntax. 1999. viii, 310 pp. RŮŽIČKA, Rudolf: Control in Grammar and Pragmatics. A cross-linguistic study. 1999. x, 206 pp. HERMANS, Ben and Marc van OOSTENDORP (eds.): The Derivational Residue in Phonological Optimality Theory. 2000. viii, 322 pp. 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. xv, 474 pp. + index. 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. x, 332 pp. + index. Expected Fall 2004 71 GELDEREN, Elly van: Grammaticalization as Economy. xv, 320 pp. + index. Expected Fall 2004

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

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)

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

Quantifier Scope in German (Linguistik Aktuell Linguistics Today)

Islands and Chains: Resumption As Stranding (Linguistik Aktuell Linguistics Today, LA 63)

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)

Theoretical Approaches to Universals (Linguistik Aktuell Linguistics Today)

The Copy Theory of Movement (Linguistik Aktuell Linguistics Today)

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

Diachronic Clues to Synchronic Grammar (Linguistik Aktuell Linguistics Today)

The Structure of Stative Verbs (Linguistik Aktuell Linguistics Today)

The Grammar of Focus (Linguistik Aktuell Linguistics Today)

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

Diachronic Clues to Synchronic Grammar (Linguistik Aktuell Linguistics Today)

Movement Theory of Control (Linguistik Aktuell Linguistics Today)

The Grammar of Focus (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)

Continuity and Change in Grammar (Linguistik Aktuell Linguistics Today)

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

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

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)

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

Quantifier Scope in German (Linguistik Aktuell Linguistics Today)

Islands and Chains: Resumption As Stranding (Linguistik Aktuell Linguistics Today, LA 63)

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)

Theoretical Approaches to Universals (Linguistik Aktuell Linguistics Today)

The Copy Theory of Movement (Linguistik Aktuell Linguistics Today)

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

Diachronic Clues to Synchronic Grammar (Linguistik Aktuell Linguistics Today)

The Structure of Stative Verbs (Linguistik Aktuell Linguistics Today)

The Grammar of Focus (Linguistik Aktuell Linguistics Today)

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

Diachronic Clues to Synchronic Grammar (Linguistik Aktuell Linguistics Today)

Movement Theory of Control (Linguistik Aktuell Linguistics Today)

The Grammar of Focus (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)

Continuity and Change in Grammar (Linguistik Aktuell Linguistics Today)

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

Recommend Documents