Dynamic Two-Place Indirect Verbs in French: a Synchronic and Diachronic Study in Variation and Change of Valence

Dynamic Two-place Indirect Verbs in French: A Synchronic and Diachronic Study in Variation and Change of Valence

Michelle Ann Troberg

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Graduate Department of French University of Toronto

Library and Archives Bibliothèque et Canada Archives Canada

Published Heritage Direction du Branch Patrimoine de l’édition

395 Wellington Street 395, rue Wellington Ottawa ON K1A 0N4 Ottawa ON K1A 0N4 Canada Canada

Your file Votre référence ISBN: 978-0-494-58105-6 Our file Notre référence ISBN: 978-0-494-58105-6

NOTICE: AVIS:

The author has granted a non- L’auteur a accordé une licence non exclusive exclusive license allowing Library and permettant à la Bibliothèque et Archives Archives Canada to reproduce, Canada de reproduire, publier, archiver, publish, archive, preserve, conserve, sauvegarder, conserver, transmettre au public communicate to the public by par télécommunication ou par l’Internet, prêter, telecommunication or on the Internet, distribuer et vendre des thèses partout dans le loan, distribute and sell theses monde, à des fins commerciales ou autres, sur worldwide, for commercial or non- support microforme, papier, électronique et/ou commercial purposes, in microform, autres formats. paper, electronic and/or any other formats. . The author retains copyright L’auteur conserve la propriété du droit d’auteur ownership and moral rights in this et des droits moraux qui protège cette thèse. Ni thesis. Neither the thesis nor la thèse ni des extraits substantiels de celle-ci substantial extracts from it may be ne doivent être imprimés ou autrement printed or otherwise reproduced reproduits sans son autorisation. without the author’s permission.

In compliance with the Canadian Conformément à la loi canadienne sur la Privacy Act some supporting forms protection de la vie privée, quelques may have been removed from this formulaires secondaires ont été enlevés de thesis. cette thèse.

While these forms may be included Bien que ces formulaires aient inclus dans in the document page count, their la pagination, il n’y aura aucun contenu removal does not represent any loss manquant. of content from the thesis.

Dynamic Two-place Indirect Verbs in French: A Synchronic and Diachronic Study in Variation and Change of Valence

Michelle Ann Troberg

Doctor of Philosophy

Graduate Department of French University of Toronto

2008 Abstract

This dissertation provides an account of an often-noted change in the history of French: the shift in the expression of the internal argument of a small class of dynamic two-place verbs best represented by aider ‘to help’ from “dative”, i.e., as an indirect object with the preposition à, to

“accusative”, i.e., as a direct object with no preposition. The change does not appear to be correlated with a change in the meaning of the verbs. Traditional commentators have viewed it as random, affecting only a few lexical items, rather than systematic. One of the central results of this thesis is that the valency change affects a class of some twenty verbs at approximately the same period and that it follows the same time course. Moreover, three properties distinguish this class of verbs from all others taking indirect objects in French: following current ideas about the syntactic manifestation of verbs and their arguments, they have a non relational argument structure, they do not possess lexical directionality, and they select for first or third order entities.

These facts suggest that a structural change underlies the change in the realization of the internal argument. Adopting Lightfoot’s (1999, 2006) “cue-based” approach to language change, it is proposed that the valency change is a result of the loss of a functional item encoding directionality. Directionality is a derived property in Medieval French, available in particular to prepositions. It is demonstrated that when à was able to encode direction, first and third order ii indirect objects were licensed in a broader range of contexts, namely, with aider-type verbs. The loss of this functional item is also correlated with several other structural changes that occurred in the 16th and 17th century.

iii

Acknowledgments

The research for this thesis was funded by the Social Sciences and Humanities Research Council of Canada, the Ontario Graduate Scholarship Program, the Scace Graduate Fellowship, the Trentwith and Galipeau French Linguistic Fellowship, and the Department of French at the University of Toronto. This study profited from the opportunity to participate at two international conferences both supported by the Department of French at the University of Toronto: Diachro III: Évolutions en français in Paris and the NORMS Workshop on Argument Structure in Lund. A number of research assistantships with Yves Roberge enabled me to pursue our collaborative work on indirect objects, the results of which are central to the present study.

My deepest gratitude is to my director, Yves Roberge, who provided me with the ideal intellectual climate in which to engage in my research and who allowed me the freedom to work at my own rhythm. I am extremely fortunate to be among the many influenced by Yves’ ideas, by his approach to educating, by his generosity and by his joie de vivre. Yves is also responsible, perhaps unwittingly, for my choice in undertaking a thesis in diachronic linguistics; my single greatest inspiration was Marianne Adams’ 1986 doctoral thesis, a copy of which he lent to me in my master’s year.

It is with much warmth that I thank Brian Merrilees. To him, I owe my interest in Medieval French and my continued contact with it via the precious Aalma manuscripts. I also thank him for once very wisely reminding me that, even in cases of regular change, “chaque mot a son histoire”. I am grateful to both he and his wife, Pat Merrilees, for they have offered me support on every possible level.

I extend my deep appreciation to France Martineau of the University of Ottawa, who agreed to sit on my thesis committee, whose work in diachronic linguistics has been a model to me, and whose insightful questions and wise advice have very much influenced the shape of this thesis.

I am grateful to my external examiner, Ian Roberts of the University of Cambridge. His thorough assessment has improved my work and his comments and suggestions have taken me beyond this study and have me eager to develop related topics and take on new ones.

I am also very appreciative of my colleagues and informants David Fournier, Krystina Karenova, Caroline Lebrec, David Martin, Vincent Masse and Antje Ziethen.

I thank my parents, Judy Troberg and Harvey Troberg, for their steadfast support when I returned to graduate school. By example, they have conveyed to me the importance of the journey over the destination. This perspective has allowed me to embrace the research process and to understand that a doctoral thesis is just one step along an ever engaging and generous road of learning.

Finally, without my wonderful partner Nicholas, this thesis would surely never have come to fruition. He set aside countless hours of his own research time to take care of the daily logistics of a home and family so that I could continue working. And, as a counterpoint to research and writing, our little Xavier has provided the simple and timeless joy of play. We sometimes wondered just how this thesis would ever get finished with a new baby in our lives, but we now realize that he was instrumental to its completion and to our relative sanity. This thesis is dedicated to Nicholas and Xavier.

Table of Contents

ACKNOWLEDGMENTS...... IV

TABLE OF CONTENTS...... VI

CHAPTER 1 INTRODUCTION ...... 1

1.1 LANGUAGE CHANGE ...... 8 1.1.1 Theoretical assumptions...... 8 1.1.2 Methodological problems particular to historical linguistics...... 12 1.1.3 The value of language change to a theory of grammar ...... 12 1.2 ARGUMENT STRUCTURE ...... 14 1.2.1 Basic assumptions ...... 14 1.2.2 Events and argument structure ...... 17 1.2.3 External subjects, internal objects ...... 18 1.2.4 Event introducers: flavours of v...... 19 1.2.5 Roots and Conflation...... 20 1.3 THEMATIC INDIRECT OBJECTS IN FRENCH: A TYPOLOGY...... 21 1.3.1 Three-place verbs of transfer and comparison...... 26 1.3.2 Change of location verbs...... 28 1.3.3 Locative and relation verbs ...... 30 1.3.4 Psychological verbs...... 33 1.3.5 Verbs of interaction and reflection ...... 38 1.3.6 Thematic indirect objects are prepositional phrases ...... 40 1.4 CHAPTER SUMMARY...... 49

CHAPTER 2 THE VERBS...... 50

2.1 VERBS UNDER CONSIDERATION...... 52 2.1.1 Aider ...... 53 2.1.2 Applaudir...... 55 2.1.3 Assister ...... 57 2.1.4 Commander...... 59 2.1.5 Congratuler...... 62 2.1.6 Contrarier...... 62 2.1.7 Contredire...... 64 2.1.8 Dominer...... 66 2.1.9 Empêcher...... 68 2.1.10 Ennuyer...... 70 2.1.11 Épargner...... 71

2.1.12 Étudier ...... 73 2.1.13 Favoriser...... 75 2.1.14 Insulter...... 76 2.1.15 Offenser ...... 78 2.1.16 Persuader...... 79 2.1.17 Prier...... 81 2.1.18 Secourir ...... 82 2.1.19 Servir ...... 84 2.1.20 Supplier...... 87 2.2 EXCLUSIONS ...... 88 2.2.1 Insufficient evidence...... 89 2.2.2 Loss of the IO but no replacement form ...... 96 2.2.3 Stative verbs...... 100 2.2.4 Opposite change...... 101 2.3 CHAPTER SUMMARY...... 109

CHAPTER 3 QUANTITATIVE STUDY...... 111

3.1 CORPUS AND METHODOLOGY ...... 113 3.1.1 Corpus ...... 113 3.1.2 Methodology ...... 114 3.2 EVOLUTION OF INDIVIDUAL VERBS ...... 117 3.2.1 Aider ...... 117 3.2.2 Applaudir...... 120 3.2.3 Assister ...... 122 3.2.4 Commander...... 123 3.2.5 Contrarier...... 126 3.2.6 Contredire...... 128 3.2.7 Ennuyer...... 130 3.2.8 Favoriser ...... 131 3.2.9 Insulter...... 132 3.3 SUMMARY AND ANALYSIS ...... 134 3.3.1 The S-shaped curve ...... 134 3.3.2 Initial period of stability...... 135 3.3.3 Actuation...... 142 3.3.4 Diffusion ...... 144 3.3.5 The dative clitic...... 146 3.3.6 Animacy ...... 147 3.4 GENERAL EVOLUTION OF REMAINING VERBS...... 148 3.5 CHAPTER SUMMARY ...... 149 vii

CHAPTER 4 QUALITATIVE STUDY ...... 151

4.1 AIDER-TYPE VERBS ARE NCTVS ...... 152 4.1.1 Non core transitive verbs...... 152 4.1.1.1 Semantic distinction between CTVs and NCTVs...... 153 4.1.1.2 Syntactic distinction between NCTVs and CTVs...... 157 4.1.1.3 Typological distinction between NCTVs and CTVs...... 158 4.1.2 The argument structure of NCTVs ...... 162 4.2 STRUCTURAL AND SEMANTIC DISTINCTIONS...... 164 4.2.1 Structural distinction...... 164 4.2.1.1 Structural vs. inherent indirect objects...... 169 4.2.1.2 Other proposals for the structure of verbs like aider...... 171 4.2.2 Directionality...... 176 4.2.3 First, second and third order entities...... 181 4.2.4 Generalisation...... 186 4.3 CHAPTER SUMMARY ...... 188

CHAPTER 5 CAUSE AND EFFECT ...... 189

5.1 PREVIOUS ACCOUNTS ...... 190 5.1.1. Analogical change...... 190 5.1.2 IO/DO ambiguity and random evolution...... 192 5.1.3 Grammaticalization of the S-V-IO construction...... 194 5.1.4 Functional reanalysis...... 195 5.1.5 Loss of inherent dative case...... 200 5.2 THE MEANING OF À ...... 201 5.3 DIRECTIONALITY AND À...... 209 5.3.1 Directionality as a derived property ...... 213 5.4 A CUE-BASED APPROACH TO CHANGE ...... 215 5.4.1 Directional verb particles...... 216 5.4.2 Preverbs...... 220 5.4.3 Verb restructuring...... 223 5.5 THE PERSISTENCE OF LUI AIDER AND LUI PRIER ...... 226 5.6 CHAPTER SUMMARY...... 227

CONCLUDING REMARKS ...... 229

REFERENCES...... 231

Scientific works...... 231 Corpus...... 240 Electronic text-based corpora ...... 240 Metalinguistic references...... 240

viii 1

Chapter 1 Introduction

This dissertation is about the connection between regularity in form-meaning mapping and regularity in language change. Verbs can be grouped according to common meanings and common syntactic behaviour. When these regularities are captured formally, then a view of valency change emerges which is more systematic than previously thought. Change is contingent on argument structure and the language specific grammatical items associated with it. This is shown to be the case in French where a group of verbs is set apart from others taking an indirect object when the preposition à ceases to encode directionality.

Our study has two goals. On the one hand, we wish to present a comprehensive diachronic study of a valency change in a group of dynamic two-place French verbs. In Medieval French, they occur with an indirect object, i.e., with the preposition à, but in Modern French, they generally only occur with a direct object in Modern French. This group includes the following verbs:

aider ‘help’ épargner ‘spare, save’ applaudir ‘applaud’ étudier ‘study’ assister ‘assist’ favoriser ‘show favour, support’ commander ‘command, have authority’ insulter ‘insult’ congratuler ‘congratulate’ offenser ‘offend’ contrarier ‘annoy, thwart’ persuader ‘persuade’ contredire ‘contradict, refute’ prier ‘pray, beg’ dominer ‘rule, command, dominate’ secourir ‘aid, assist’ empêcher ‘trouble, hinder’ servir ‘serve’ ennuyer ‘annoy, bore, put out’ supplier ‘beseech’

The central claim is that the systematic change in the expression of the object of these verbs is the result of a common underlying cause. We propose that the local cause of change is related to a structural change affecting the possible interpretations that the preposition à could have, namely, that it could no longer encoded directionality from the 16th century onward. Our hypothesis is supported by a quantitative and qualitative analysis of the verbs and strong correlations with other structural changes that occurred during the same period.

On the other hand, we wish to advance a formal analysis of this class of verbs, linking their argument structure representation to the variability of the morphosyntactic expression of the object, attested both synchronically, diachronically and crosslinguistically. Our examination of the French facts reveals strict licensing requirements on dative objects and serves to explain why the loss of the directional interpretation of the preposition à would entail the replacement of some indirect objects by an analogous direct object, but not of others.

Note that the term valence is meant to refer to the number of arguments with which a verb can combine to make a syntactically well-formed sentence. It also refers to the category or grammatical role of these constituents. Intransitive verbs (marcher, éternuer) are monovalent, or one-place verbs whose sole argument is the subject. Verbs such as aider, détruire, ressembler and parler are bivalent, or two-place verbs, denoting events involving a subject and an object. Ditransitive verbs such as donner, voler, and comparer are trivalent, or three place-verbs, taking a subject, a direct object, and an indirect object. The valency change studied here is concerned with a change in the category of the object of two-place verbs.

Consider the verb aider ‘to help’, for example. In Old and Middle French, aider was most often followed by an indirect object (henceforth IO) headed by the preposition à (1a), commutable with a pronoun of the type lui/leur (1b) when referring to animate objects, or y (1c) in the case of inanimates.1 Although less frequent, direct objects were also used with aider, (1d) and (1e). Furthermore, it was not uncommon for an author to alternate between a direct and an indirect object within the same text, illustrated by the examples in (1a) and (1d) taken from Villehardouin’s early 13th century chronicle of the Fourth Crusade.

1 “Indirect object” in French is understood here to be a PP constituent headed by à and which is a core internal argument to the verb. It is not associated with any one particular syntactic position or with any one specific meaning, i.e., we do not adopt a unified analysis of this grammatical function. Similarly, “direct object” is taken to mean the NP or DP complement of a verb introduced without the intermediary of a preposition. It is not associated with any one particular position or meaning.

(1) a. Et Nostre Sires demonstra ben ke il voloit aider à nostre gent.

‘And our lord indeed showed that he was willing to help our people.’

(Villehardouin, De la conqueste de Constantinople, c.1207, §566; Nyrop

1930:186)

b. les Romains ne leur aideroient point en leurs afferes

‘the Romans didn’t help them at all in their engagements’

(Simon de Phares, Recueil des ... astrologues, c.1494-1498, f° 93 v°;

DMF1)

c. la chrestienté enseigne que la foy et le salut est par l'ouye, et que la veuë y

nuict plus qu' elle n'y ayde.

‘Christianity teaches that faith and salvation are attained through the ability to

hear, and that sight is more detrimental than it is helpful to these ends.’

(Charron, De la Sagesse, Trois Livres, 1601, p.89; Frantext, ARTFL)

d. Si les envoia en la terre de Romenie por aidier cels d’Andrenople.

‘He sent them into Byzantium to help the people of Adrianopolis.’

(Villehardouin, De la conqueste de Constantinople, c.1207, §404; Ménard

1976:124)

e. Or puet dire a son Dieu que il le viengne aidier

‘He is able to ask his God to come and help him’

(Jehan de Saint-Quentin, early 14th c., 128; Herslund 1980:52)

In contrast, in Modern Standard French, aider is always followed by a direct object (henceforth DO) when it is animate (2a).2 The inanimate complement of aider is also expressed as a DO (2b), although some action nouns can appear as IOs (2c).

(2) a. Mon fils a aidé Jean/ l’a aidé dans ses travaux.

‘My son helped Jean/him with his work.’

b. si cet ordre, […] a un but, c'est d'aider le libre progrès

‘… if this order, […] has a goal, it is to facilitate free progress’

(J. Michelet, Le Peuple, 1846, p.359; TLFi)

c. Ces mesures pourront aider au rétablissement de l’économie.

‘These measures will help towards the recovery of the economy.’

(Le Petit Robert)

Crucially, the valency change does not appear to be correlated with a change in the meaning of the verb and thus with any obvious change in the selectional restrictions imposed on the internal argument. This tells us that the change in the expression of the verbal complement does not have to be accompanied by changes in meaning and that such change can be structural. The valency change in question is schematized in (3). In Medieval French, the verbs occur with an indirect object and may also occur with a direct object which may or may not have the same semantic role as the IO. In Modern French, the verb occurs with a direct object, which is otherwise equivalent to the former indirect object. The DO is thus a continuation of the IO.

(3) V-IO1/DO1/2 → V-DO1

Table 1 offers a sample of the various types of verbs that take indirect objects in French. Only a small group undergo the valency change described in (3): those highlighted in grey. These verbs are restricted to dynamic two-place verbs of interaction and reflection.3

2 In certain varieties of French (e.g., Franco-provençal, Swiss French and Canadian French), aider still allows an animate IO. This study is limited to Standard French uses, but a brief discussion in section 5.6 addresses the persistence of animate IOs in these varieties of French.

Table 1: Verbs taking indirect objects in French

VERBS OF CHANGE OF LOCATIVE & PSYCH VERBS VERBS OF TRANSFER & LOCATION RELATION INTERACTION COMPARISON VERBS VERBS & REFLECTION attribuer aller confiner agréer céder ‘attribute’ ‘go’ ‘adjoin’ ‘suit’ ‘yield’ apporter arriver demeurer convenir coopérer ‘bring’ ‘arrive’ ‘remain’ ‘suit’ ‘cooperate’ donner descendre habiter démanger obéir ‘give’ ‘go down’ ‘live in’ ‘itch’ ‘obey’ lancer entrer manquer importer parler ‘throw’ ‘enter’ ‘lack’ ‘matter’ ‘speak/talk’ rendre monter résider manquer remédier ‘give back’ ‘go up’ ‘reside’ ‘miss’ ‘remedy’ arracher partir vivre nuire penser ‘pull off’ ‘leave’ ‘live’ ‘harm/bother’ ‘think’ retirer parvenir appartenir plaire réfléchir ‘take away’ ‘reach’ ‘belong’ ‘appeal’ ‘reflect’ voler rester correspondre peser aider ‘steal’ ‘stay’ ‘correspond’ ‘weigh’ ‘help’ annoncer retourner ressembler profiter applaudir ‘announce’ ‘return’ ‘resemble’ ‘benefit’ ‘applaud’ dire sortir succéder répugner dominer ‘say’ ‘go out’ ‘succeed’ ‘disgust’ ‘dominate’ comparer tomber préexister suffire empêcher ‘compare’ ‘fall’ ‘pre-exist’ ‘suffice’ ‘hinder’ préférer venir survivre tarder servir ‘prefer’ ‘come’ ‘survive’ ‘be impatient’ ‘serve’

The shift from the possibility of taking an IO to the more or less strict use of a DO poses an interesting problem both diachronically and synchronically. From a diachronic point of view, there is a purely descriptive interest in establishing whether or not these two-place verbs change as a group, as part of a unified and regular change. Two pieces of evidence strongly support the idea that they underwent a systematic change. The first is provided by our quantitative study,

3 Some of the verbs in table 1 show slight alternations with a direct object at earlier periods in the history of French, i.e., ressembler, survivre, convenir, démanger, importer, manquer, nuire, plaire, obéir. These cases are not, however, examples of the type of change described in (3). They are discussed, among others, in chapter two, section 2.2.4

6 based on literary texts, which shows that the actuation of change is limited to a 150-year period between 1500 and 1650. The second is clear syntactic and semantic constraints on the change. Structurally, the valency change is restricted to dynamic two-place verbs whose IO is the sole internal argument. Semantically, the verbs are nondirectional and take first, second or third order entities as objects.

There is further interest in pursuing the cause of change, since this sheds light on the nature of the system undergoing the change. Various dramatic structural changes affecting the expression of verbs and their arguments occur during the 16th and 17th century: the loss of productive preverbs, the verb particle system, verb restructuring and the possibility of IOs with verbs like aider. Based on previously published work and original observations and analysis, we propose a chain of events in chapter five that involves a shift in the way French encodes directionality. This shift involves the loss of a functional item which encodes direction. No longer a derived property, directionality is subsequently lexicalized as a property of individual lexical items, namely verbs and prepositions. This has consequences for the verb particle system, the expression of manner of motion events, verb restructuring and inherent dative case with verbs like aider. The preposition à, now purely locative, is shown to have strong selectional restrictions on the kind of object with which it can occur, excluding first and third order objects in non relational, adirectional contexts. This explains why IOs are not licit with aider in Modern French, but were perfectly acceptable in these contexts when à was able to encode directionality.

From a synchronic point of view, we take stock of all thematic indirect objects in French in order to understand why only aider-type verbs undergo a valency change. A comprehensive typology of IOs in French is proposed based on common meaning and syntactic behaviour. From a licensing point of view, IOs are either structurally or semantically licensed. The majority of verbs that occur with IOs fall into the former group. Structurally licensed IOs are generated within a relational structure that presupposes an IO; the expression of these arguments shows little evidence of variation or change in the history of French. In this respect, they contrast with the smaller group of semantically licensed IOs occuring with the aider-type verbs which show significant synchronic, diachronic and crosslinguistic variation. These verbs are derived, we claim, in the same basic way as transitive verbs like manger ‘eat’, voir ‘see’, etc, the category of the complement being structurally underspecified. In line with Levin’s (1999) account of non core transitive verbs and independent cross-linguistic accounts of “helping” verbs, the goal-like

7 object complement of aider is most often expressed as an oblique or dative-marked object, but in the absence of an appropriate marker, these objects are realized in the accusative case.

The valency change in French gives us the perfect testing ground for identifying why the IO is no longer a legitimate form of the object of these kinds of verbs. Our analysis shows that the expression of the internal argument as an IO depends crucially on the interaction between the degree of abstraction of the object (following Lyons’ 1977 typology), structural vs. semantic licensing and the presence or absence of lexical directionality. This is confirmed by an examination of productive IO/DO alternations in French.

The dissertation is organized as follows. Chapter one introduces the two core phenomena of language change and argument structure with a detailed analysis of indirect objects in Modern French. Chapter two presents evidence that the same change in argument realization has affected a class of twenty verbs without affecting their meaning. Other verbs that have undergone variation and/or change of valence are shown to be peripheral to the change that concerns us here. Chapter three presents a quantitative analysis of nine of these twenty verbs, showing in detail how complement structures of these verbs change following the S-shaped curve, well- known from other work on language change, and how the changes take place during roughly the same period (starting in Pre-classical French). Chapter four provides a qualitative analysis of the verbs that underwent the valency change, demonstrating the three ways in which the change is constrained. First, only non relational dynamic two-place verbs undergo a valency change. Second, only verbs whose meaning is unspecified for directionality are susceptible to the change. Third, the change only affects verbs that occur with (animate) first order and (inanimate) third order indirect objects; (inanimate) second order entities tend to resist the change. This further supports our hypothesis of a unified change and identifies the properties connected to the local cause of change. Chapter five looks at possible causes and effects of the change. The principle claim is that directionality was once a derived property in French which had a number of structural consequences, one of which being a broader set of contexts permitting inherent dative objects. Chapter six concludes.

The present chapter introduces the basic assumptions on which our study is based. Section 1.1 outlines our assumptions about language change within a generative grammar framework. A theoretical overview is presented in section 1.1.1; section 1.1.2 addresses the main

8 methodological problems associated with historical linguistics; and section 1.1.3 underscores the value of language change within a generative grammar framework. Section 1.2 presents our assumptions about argument structure and argument realization. The theoretical approach is outlined in sections 1.2.1 to 1.2.5, covering the basic mechanisms and elements involved in the derivation of a verb phrase. Section 1.3 details our typology of verbs which occur with indirect objects in French. Formal representations of the various types of verbs are proposed, crucial to our analysis in chapters four and five. Section 1.4 concludes.

1.1 Language change

This section frames critical concepts such as “language”, “grammar”, and “syntax” within a Minimalist perspective so that our assumptions about what “language change” is are made as explicit as possible.

1.1.1 Theoretical assumptions

A fundamental distinction in a generative grammar approach to language change is the difference between one’s biological grammar and language.4 The former is understood to be the abstract system that develops in an individual according to genetic principles and to the linguistic environment to which he or she is exposed as a child. It characterizes an individual’s linguistic knowledge and, in conjunction with other cognitive functions, generates grammatical linguistic forms in the mature speaker. Language, on the other hand, is something external to individuals, a social object consisting of norms and conventions. A language like French, English, or Swahili is, in a very general sense, the non biological product of a community of speakers. The generative grammar enterprise takes an individual’s internal grammar as its object of study. As it is for synchronic linguistics, our approach to diachronic linguistics also takes the grammar to be its proper object of study.

Within a research programme concerned with the grammar of individual speakers, “change” can only mean the “set of differences” (Hale 1998:3) between the grammar characterized by mature

4 Chomsky (1986) refers to these two concepts as I-language (internal language) and E-language (external language).

9 speakers and that of a newly acquired grammar. Grammatical change therefore happens at the time of acquisition and has no temporal value. Following Lightfoot (1979, 1991, 1999, 2006), grammatical change is based on two fundamental assumptions: first, each child possesses an inherent language faculty comprised of the computational principles applicable to any language (universal grammar, or UG). This linguistic knowledge comes for free; it is not acquired in the course of a child’s development. Second, particular grammars develop in children based on what they hear (the primary linguistic data, or PLD) in conjunction with the innate computational system. Accordingly, children have no direct access to the grammar of any other individual and certainly not to an idealized grammar of preceding generations. Rather, they develop their own grammars independently and inductively. The grammar of a speaker of one generation

(Grammar1) becomes the PLD for a speaker of the next and is thus completely discontinuous from the new grammar (Grammar2):

(4) Primary data1 → Grammar1

Primary data2 → Grammar2 (Adams 1987:4)

Since no two children are exposed to the exact same PLD, grammars vary slightly from one speaker to another in a linguistic community just as a child’s grammar differs from that of an adult. Any number of prosodic, phonetic, cultural or pragmatic factors, including innovation, can affect the utterances produced by adults and thus the PLD to which child learners are exposed. Linguistic environments are rarely ever stable and input data for one generation will not be the same as that for the next. In this sense, grammatical change is inevitable. Grammatical change becomes interesting when the PLD shifts in some critical way so that learners exposed to the new PLD form grammars which end up generating structures not produced by the grammars of preceding generations. What begins as the reanalysis of one grammatical hypothesis for another, is then accurately transferred to subsequent generations of speakers. Language change is the diffusion of grammatical change and in contrast to grammatical change, it has temporal properties.

Grammar1

 Reanalysis = Grammatical change Grammar2  Diffusion = Language change

Grammar2' Grammar2'' Grammar 2'''

Figure 1 : Grammatical vs. language change

Grammatical change is therefore not only contingent on particular shifts in the linguistic environment, but also on the way in which children acquire linguistic knowledge. This latter point, the way in which PLD determines an abstract grammatical system, involves the identification of the cue, or trigger for a particular structural phenomenon along with the role of frequency in acquisition. What constitutes a cue and how many times a child must hear it before the underlying structure associated with it is acquired are ongoing questions in the field. The historical linguist can only suggest the most likely cue associated with a specific set of structures based on careful examinations of the data and correlations with other changes.

The way in which syntactic change is conceived within the Minimalist Program (Chomsky 1995, 2000, 2001) is very specific, and follows from the basic assumption of an inherent computational system and the aspects of language that are relevant in the acquisition process, explicitly stated in Chomsky (1995:169-170).

The standard idealized model of language acquisition takes the initial state S0 to be a function mapping experience (primary linguistic data, PLD) to a language. UG is

concerned with the invariant principles of S0 and the range of permissible variation. Variation must be determined by what is “visible” to the child acquiring language, that is, by the PLD. It is not surprising, then, to find a degree of variation in the PF component, and in aspects of the lexicon: Saussurean arbitrariness (association of concepts with phonological matrices), properties of grammatical formatives (inflection, etc.), and readily detectable properties that hold of lexical items generally (e.g., the head parameter). Variation in the overt syntax or LF component would be more problematic, since evidence could only be quite indirect. A narrow conjecture

is that there is no such variation: beyond PF options and lexical arbitrariness (which I henceforth ignore), variation is limited to nonsubstantive parts of the lexicon and general properties of lexical items. If so, there is only one computational system and one lexicon, apart from this limited kind of variety. Let us tentatively adopt that assumption – extreme, perhaps, but it seems not implausible – as another element of the Minimalist Program.

Longobardi (200:277-8) interprets these basic Minimalist notions, specifying that “syntax, by itself, is diachronically completely inert.” What this means is that syntactic change is construed as a change in some property of a particular lexical or functional item which has structural repercussions on the set of possible utterances. The locus of syntactic change is the abstract features of individual lexical items, the elements that ultimately dictate syntactic realization. In Longobardi’s words, “syntactic change should not arise, unless it can be shown to be caused – that is, to be a well-motivated consequence of other types of change (phonological changes and semantic changes, including the appearance/disappearance of whole lexical items) or, recursively, of other syntactic changes” (Longobardi 2001:278).

Lightfoot sums up grammatical change as internal, local and abstract. By internal it is meant that it occurs in the biological grammar of individuals, distinct from language as a social construct. By local, he means that principles of history or principles of UG are not explanatory forces of language change. Language changes because cues in PLD change. By abstract, it is meant that the locus of change is found in the abstract grammatical system itself – in lexical features. Grammatical change describes what changed; this is the primary focus of our analysis. On the other hand, shifts in the linguistic environment which affect the PLD for subsequent generations are external, by which we mean that they occur in language, the social phenomenon. They tell us about why a grammatical change arises and why it happens when it does.

This Minimalist, cue-based approach to language variation and change guides our examination of valency change in French. We wish to identify the property that changed in the abstract grammatical system of French speakers which had as an effect a widespread valency change in a well defined group of two-place verbs. By linking the valency change to a specific shift in the linguistic experience of a particular generation, we can explain why the change occurred, and why it did at a particular period in time and no other.

1.1.2 Methodological problems particular to historical linguistics

A central methodological problem to most historical study, in linguistics as in other domains, resides in the limited amount of available data. For linguistics in particular, experiments cannot be undertaken in order to verify hypotheses or to broaden the range of data. Information obtained by eliciting acceptability judgements from informants, a method central to synchronic grammatical analysis within a generative framework, is also unavailable.

The difficulties that diachronic syntax faces are therefore significant, but not impossible to overcome if one makes two reasonable assumptions, as outlined in Kroch (1989:200). The first assumption, the uniformitarian principle, is not limited to historical linguistics, and was originally advanced in order to account for the geological evolution of the Earth. It allows for the analysis of past phenomena by using knowledge of the present. For the historical linguist, it means that the general principles that hold for living languages should equally hold for languages that are no longer spoken. Principles believed to be universal such as Merge, Binary Branching, and the Single Complement Hypothesis thus hold equally for Modern French as they do for Old French, Latin, and Proto-Indo-European. The second assumption pertains to what constitutes an unacceptable utterance in an obsolete language. For simple and relatively common constructions, if a certain type of construction does not appear in a large corpus, it is reasonable to assume that it is not grammatical in the language of that corpus. This last assumption can be problematic in so far as absences in a corpus depend on a great number of things, such as the well-known differences between written and oral grammars, corpora limited to specific contexts, etc. However, given that the present study examines common verb-complement constructions over a large and varied corpus, this assumption does not weaken the analysis.

1.1.3 The value of language change to a theory of grammar

Faced with the natural limitations imposed on historical linguistics, it seems reasonable to ask what real interest it holds in a theory of grammar when synchronic studies of living languages and studies in acquisition already provide extremely rich data to work with. Yet, given that the very nature of language acquisition necessarily entails grammatical change, it is important to be able to identify what changes and why. Otherwise, structural change can be attributed to any number of external changes or predetermined principles which are entirely removed from the abstract notion of language, the mechanism which generates utterances to begin with (see

Lightfoot 1999, ch.2, for an in depth discussion of this topic). In Hale’s words, “explanatory adequacy requires us to account for all acquisition – this includes the acquisition of grammars that diverge from their sources (i.e., change) as well as that of fully convergent grammars.” (Hale 1998:1). Kroch (1989:200) reminds us that historical linguistics presents a type of information that is necessarily absent from synchronic data and thus offers a unique contribution to a general theory of grammar.

[...] perturbing a complex system and observing its subsequent evolution is often an excellent way of inferring internal structure. In addition, since the features of any language at a given point in time are the result of a complex interweaving of general principles of language and particular historical developments, knowledge of the historical process by which a language has reached a given state may be important to the proper assignment of responsibility to historical and general factors and so to the proper formulation of linguistic theory.

Recent approaches to language change can advance fundamental notions within a generative theory of grammar. Lightfoot’s (1999) development of a cue-based theory of language change, for instance, identifies the conditions under which structural changes take place which in turn sheds light on the conditions of language acquisition. Likewise, Roberts and Roussou’s (2003) work on grammaticalization provides a new perspective on the nature of functional categories. By looking at the diachronic development of functional elements, they motivate an account of the long-standing issue of non random pathways of language change. This is significant within a theory of language which does not predict the well attested clustering of certain combinations of parameters.

This dissertation hopes to contribute to a theory of argument structure through a better understanding of thematic indirect objects in French. The history of French is particularly instructive in this regard because the indirect object of a discrete class of verbs was replaced by a direct object without a change in meaning. The change reveals a dependency between argument structure, directionality, and the degree of abstraction of the object referent. This correlation describes the distribution of many IOs in French, not at all obvious from a synchronic point of view. In addition, the valency change implies conceptual and structural similarities between two- place verbs like aider, transitive in Modern French, and two-place indirect verbs such as

14 contribuer, obvier, remédier, etc. Again, this kind of insight is simply not available to synchronic studies. From this privileged point of view then, we are able to propose new ways of considering two-place verbs in French, both semantically and syntactically. From a typological view, our analysis is able to given an account of the necessary conditions that permit Old French to be classified among the satellite-framed languages and why Modern French is considered to be a verb-framed language.

1.2 Argument structure

This section defines our basic assumptions about argument structure as event structure. The elements that constitute a verb phrase and how they combine are described in some detail, since we claim that valency change is contingent on event structure and the language specific grammatical items associated with it.

1.2.1 Basic assumptions

Indirect objects are no longer licit with a group of two-place verbs in French that includes aider, applaudir, commander, contrarier, etc. In order to account for the fact that the change was restricted to a specific group of verbs, we examine how these verbs differ from others. To do so, a well articulated theory of how words come together to form meaningful sentences must be invoked. This section reviews our basic assumptions about the lexicon-syntax interface, the components of a verb phrase and how they combine to form structures.

As outlined earlier, our formal analysis assumes a Minimalist approach to grammar, where the grammar is considered to be purely computational. The most basic elements that enter the computation to form structures are referred to here as listemes. They are situated in the lexicon, a kind of reservoir of vocabulary items and can be either lexical items which express real-world concepts (livre, amour, marche, dans, etc.), or functional items which express grammatical notions such as tense, cause, number, etc. Each listeme exists as an entry in the mental lexicon and can be understood as the arbitrary pairing of sound and meaning. This is the most basic information that distinguishes a particular listeme from others.

The computational system is the mechanism that combines the various elements of a sentence to create grammatical utterances. From a generativist point of view, this aspect of the language faculty comes for free; it is innate. Sometimes referred to as the syntax, the computational

15 system is invariable, so that when one speaks of syntactic change, one is not speaking of a change to the system itself, but rather to a lexical property that determines some specific syntactic realization.

The most fundamental operation in the computation is Merge as defined in Chomsky (1995:226):

“The simplest syntactic operation takes a pair of syntactic objects (SOi, SOj) and replaces them by a new combined syntactic object SOij.” This operation is responsible for the recursive property of natural languages. By recursive, it is meant that syntactic operations can apply to their own output so that in principle, an infinite array of different sentences can be created by simply embedding structures within structures. Elements merge to form larger constituents according to the general principles of Bare Phrase Structure and constraints such as Binary Branching (Kayne 1984) and the Single Complement Hypothesis (Larson 1988). This is to say that structures are hierarchical and a grammatical phrase (XP) can have at most one specifier (ZP) and one complement (YP).

(5) XP 2 ZP X’ 2 X YP

Following Chomsky (2001), there are two types of Merge: external Merge, which is ‘argument’ merge broadly construed, and internal Merge, which is the association, higher in the structure, of copies of elements already present in the derivation (formerly Move).

The derivation of a sentence is complete when it yields a representation that will serve as input at the interface to other relevant parts of the cognitive system, namely PF (phonological form) and LF (logical form). In the Minimalist Program, the principle of Full Interpretation requires that every derivation “check” all uninterpretable features of its elements. Only these derivations converge and are “legitimate objects” at the interface.

A verb phrase is thus built by the successive merger of listemes, yet natural language does not allow for an unlimited number of different kinds of verb phrases. In fact, a single verb has no more than two internal arguments. Moreover, verbal arguments correspond to a limited number of general semantic roles (agent, experiencer, goal, theme, etc.) which correlate with the

16 grammatical function of subject in a hierarchical fashion. Hale and Keyser (1993, 2002) attribute these restrictions to the limited structural relations determined by Binary Branching and the Single Complement Hypothesis, rather than stipulating them as independent principles of grammar (i.e., the Theta Criterion, thematic hierarchies, Baker’s UTAH, etc.). Their theory is that the syntax determines possible argument structures and the correct interpretation of verbal arguments since argument structures themselves are associated with particular meanings.

While Hale and Keyser consider argument structure (l-syntax) to be distinct from overt syntax, many neo-constructionists find the distinction unnecessary.5 We follow the latter line of thinking and in so doing, attribute very little syntactic information to a lexical entry and no computational power to the lexicon. We take the point of view that a lexical entry does not contain categorial or syntactic linking information associated with insertion frames, for instance. It has been shown on empirical grounds that lexicalist models are too constraining; they cannot, in any reasonably economical way, account for the natural flexibility of words which can be easily used in novel and creative contexts, dramatically illustrated, for instance, in Borer (2005):

(6) a. The factory horns sirened throughout the raid.

b. The factory horns sirened midday and everyone broke for lunch.

c. The police car sirened the Porsche to a stop.

d. The police car sirened up to the accident site.

e. The police car sirened the daylights out of me.

In (6), the word siren has been coerced into the function of a verb in five different syntactic environments. Examples like these argue in favour of the idea that the categorial and syntactic possibilities of a word are a function of the computation and distinct from the invariable and idiosyncratic information contained in a lexical entry. The idea that a word is derived from an

5 Harley (2002b, 2004) demonstrates that interpreting Hale and Keyser’s l-syntax as syntactic as opposed to lexical is not only desirable, but also Minimalist in spirit. It eliminates theta theory as an independent component of the grammar; it solves the potential problem of distinguishing between unaccusative and unergative verbs in bare phrase structure; and their conflation mechanism motivates previously unexplained incorporation and head to head movement phenomena.

17 acategorial root is crucial to our approach to the derivation of verb phrases (see section 1.3., and psych verbs in particular).

Borer further argues that the isomorphism inherent in projecting syntactic information into the syntactic component is fundamentally redundant – contrary not only to the basic working hypothesis of generative grammar that the language faculty is non redundant, but also to a parsimonious theory of grammar. Our study is therefore associated with a neo-constructionist model of grammar in that words do not build complex meanings, but rather the computational component of language builds structures which have themselves, very specific meanings.

1.2.2 Events and argument structure

Much recent research has converged on the idea of “event” as the organizing principle around which verb meaning is constructed and verb phrases built. In this regard, argument structures are built on the basis of event types; verbs describe events in which arguments are participants and of which adjuncts are modifiers. In (7a), for instance, the subject Marie is understood as the agent of a simple touching event. The direct object le vase is understood as the participant that is more or less affected by the event. Sentence (7b) expresses a complex event. The subject Jean is the agent of a causing event, while the verb casser and the noun phrase le vase describe the change of state in the vase. In both sentences, the adjunct ce matin situates the event in time.

(7) a. Marie a touché le vase (ce matin).

‘Marie touched the vase this morning.’

b. Jean a cassé le vase (ce matin).

‘Jean broke the vase this morning.’

We adopt a syntactic approach to argument and event structure in the spirit of Hale and Keyser’s research project, developed in, among others, Cuervo (2003), Harley (1995, 2002b), Harley and Folli (2005), Marantz (1997), and Pylkkänen (2002).6 Within this framework, syntactic elements

6 In many respects, this approach applies predicate decomposition as developed in the work of Levin and Rappaport (Hovav) for instance, to the syntax proper.

18 correspond to elements of event structure. Functional heads and DPs for instance, are interpreted compositionally at LF as event predicates and participants respectively. Their interpretation is thus derived from their inherent meanings in conjunction with the structural relationship that holds between them.

The set of possible argument structures is limited to a relatively small number and is determined by the stipulation that there are only two kinds of syntactic relations (head-complement and head-specifier), and by the inventory of functional heads that introduce event types. Arguments are licensed as participants of an event based on these two basic syntactic relations and the type of functional head that introduces the event.

1.2.3 External subjects, internal objects

The strict compositionality of objects and their verbs is demonstrated in Marantz (1984) in which he shows how the interpretation of a verb phrase can vary dramatically depending on the kind of object that is used:

(8) kill a bug = cause the bug to die

kill a conversation = cause the conversation to end

kill an evening = while away the time span of the evening

kill a bottle = empty the bottle

kill an audience = entertain the audience to an extreme degree

Marantz points out that objects contribute to idiosyncratic interpretations of the event in ways that subjects do not. The asymmetry between the subject and the object is defined by Marantz as an asymmetry in the licensing of subjects and objects. Objects are arguments of the verb and as such are licensed within the domain of the verb phrase (internally). This is the domain in which the meaning of the event is derived compositionally. On the other hand, subjects are not arguments of the verb but of the verb-object predicate. They are thus licensed outside of the verb phrase (externally).

This idea is formalized in Kratzer (1996), where she proposes that certain subjects are syntactically and semantically licensed by the head of a Voice phrase which relates the subject

19 participant with the event denoted by the verb phrase. The Voice phrase takes the verb-object predicate as its complement from which the specific inerpretation of the subject is derived. With an activity, the subject is generally interpreted as an agent (John danced all night); with a caused event, the subject is interpreted as a causer (The wind/John opened the door).

(9) Voice P 3 ‘subject’ 3 Voice vP

Only external subjects are generated in the specifier of Voice. “Internal” subjects, those that are merged as objects of the verbal predicate – like the subjects of unaccusative and copular verbs – are not generated in Voice. The valency change discussed in this study concerns the objects of activity verbs like aider, contrarier, ennuyer, etc. These verbs take agentive/causer subjects which we assume to be base generated in the specifier of Voice. Since the valency change is isolated to the domain of the vP, the presence of an external subject often remains implicit in our discussion.

1.2.4 Event introducers: flavours of v

We further assume that little v is the functional element that introduces an event and defines the domain of the verb phrase. Following Harley (2002b), there are three types of little v which correspond to three basic event types:7

“Agentive” little v: this functional head may or may not have a phonological realization and may or may not be affixal. It can be glossed as MAKE, DO, or CAUSE depending on whether a Thing, an Event, or a State is brought into existence. VoiceP takes this little v as its complement, relating an agentive or causer external argument to the event. “Agentive” little v also possesses an accusative case feature, assuming Burzio’s Generalization (Burzio 1986) that only verbs that take an external argument assign accusative case. For the derivation of the French verbs examined here, agentive little v has no phonological realization.

7 Variations on these three event types have been proposed in the literature; see Cuervo (2003), for example. The ontological types of event introducers as assumed here are inspired by the Vendler-Dowty aspectual verb classes (activity, accomplishment, achievement, state), but are not necessarily aspectually defined.

“Unaccusative” little v: this functional head may or may not have a phonological realization and may or may not be affixal. It can be glossed as BECOME or HAPPEN, depending on whether a Thing, Event, or State is brought about. This v is not compatible with an external argument. Unaccusative little v has no phonological realization in French.

“Existential” little v: Harley suggests that there may also be a little v that simply asserts the existence of a State, Event, or Thing which turns up as ‘be’ in many languages.

1.2.5 Roots and Conflation

Following Hale and Keyser (1993, 2002) and Marantz (1997), we assume that verbs are formed in the syntax by the combination of an acategorial root and a verbalizing head, little v. Roots are listemes, lexical entries having idiosyncratic meaning, which is to say that the meaning of a root is not linguistically derived from anything else. Roots can express properties (ouvr- ‘open’, large- ‘wide’, petit- ‘small’ ), events (parl- ‘talk’, saut- ‘jump’), things (chat- ‘cat’, eau- ‘water’) or relations (a- ‘at/to’, dans ‘in’, pré- ‘pre-’). Some roots, depending on their meaning, easily combine with verbalizing heads to become verbs (nag- ‘swim’, envoi- ‘send’), whereas others do not. Things often become nouns, properties easily become adjectives and relations are often realized as prepositions and prefixes. These associations are not, however, hard and fast rules. As the sentences in (6) demonstrate, roots can be inserted under almost any syntactic category. Furthermore, roots may or may not be affixal. Roots in French are often affixal, whereas the English glosses show phonologically independent roots.

Hale and Keyser (2002:47) propose that roots combine with a verbalizing head by a conflation mechanism concomitant of Merge.

We use the term to refer to the “fusion of syntactic nuclei” that accounts for derivations in which the phonological matrix of the head of a complement (say, N) is inserted into the head, empty or affixal, that governs it, giving rise to a single word.

Conflation therefore occurs upon the merger of two sister heads, one of which has a defective phonological signature (p-sig). The phonologically defective head copies the p-sig of its sister in order to meet full interpretation conditions at PF. A simple example would be the derivation of a denominal verb like baver ‘to drool’:

(10) vPbav- 3 v √Pbav- “bav-” The unergative verb baver is derived from the conflation of the p-sig of the root bav- (denoting a Thing, thus noun-like), with a verbalizing head v whose p-sig is defective (null). For economy reasons, the copied p-sig bav- is only pronounced once, in its uppermost inflectional position.

1.3 Thematic indirect objects in French: a typology

Although some of the following verb phrases share the same superficial syntax, they derive from distinct underlying argument structures.

(11) a. Marie donne le livre à sa mère.

‘Marie gives the book to her mother.’

b. Marie retourne à la bibliothèque.

‘Marie returns to the library.’

c. Marie ressemble à sa mère.

‘Marie resembles her mother.’

d. Marie ment à sa mère.

‘Marie lies to her mother.’

e. Marie plaît à sa mère.

‘Mary’s mother likes her.’

f. Marie aide sa mère.

‘Marie helps her mother.’

g. Marie casse le vase.

‘Marie breaks the vase.’

The argument structure attributed to a verb is not only based on how it appears in simple declarative sentences, but also on the various permutations that it demonstrates. For instance,

22 sentences (11c) and (11d) are superficially very similar, yet the object of (11c) cannot be null under any circumstances whereas that of (11d) can easily be realized as a null generic object.

(12) a. * Marie ressemble.

‘Marie resembles.’

b. Marie ment.

‘Marie lies.’

Likewise, although the verbs in (11f) and (11g) present the same superficial syntax, they behave very differently. Verbs like casser are often permitted in inchoative constructions while verbs like aider do not permit such an alternation, shown in (13a) and (13b), and, unlike casser, the object of aider can be null, in (13c) and (13d).

(13) a. Le vase casse facilement.

‘The vase is easily broken.’

b. * Jean aide facilement.

‘John helps easily.’ (in the sense of ‘Jean is easily helped’)

c. Marie aime aider (les gens).

‘Marie likes to help (people).’

d. Marie aime casser *(les choses).

‘Marie likes to break (things).’

One of the principle goals in a theory of argument structure is to capture verb classes in terms of their common syntactic properties. The work of Hale and Keyser propose an ontology of syntactic configurations that describe many characteristic behaviours of the major verb classes. Fine-grained distinctions within classes are attributed to the meaning of verb roots, often of an “encyclopaedic” character in the sense of Marantz (1997). Hale and Keyser discuss the contrast between splash and smear, for example, attributing the fact that smear does not appear in the middle construction to its particular meaning which, unlike splash, requires an agent to apply a given substance to a surface. An agent-manner feature, for example, necessitates an external argument thereby ruling out the middle construction. We follow their work, in conjunction with

23 that of Harley, in order to account for the distinct diachronic behaviour of two-place verbs like aider. The reason these verbs underwent a change is first attributed to their argument structure, distinct from other verb classes taking IOs in French. Differences between verbs within the class of interaction and reflection verbs are then accounted for by considering their particular meaning and selectional properties.

A well defined typology of indirect objects is necessary if we are to identify what sets the indirect object of aider-type verbs apart from all other IOs in French in order to explain why only they were replaced by an analogous direct object. Much of the following discussion is taken from Roberge and Troberg’s (2007a) account of indirect objects in French.

We understand the indirect object in French to be a subset of all DPs introduced by the preposition à. This set includes thematic, or verbal arguments (14), and non thematic, or event arguments (15), but excludes circumstantials (16).8

(14) a. Jean a offert des fleurs à Marie.

‘Jean gave flowers to Marie.’

b. Les invités vont bientôt arriver à la maison.

‘The guests are going to arrive at the house soon.’

c. Jean appartient à une équipe de foot.

‘Jean belongs to a soccer team.’

(15) a. Je lui ai lavé la/sa voiture. (Hoekstra 1995:127)

‘I washed the/his car for him.’

8 Leclère (1978) and Barnes (1985) propose the terms lexical and non lexical to describe the distinction between verbal and event IOs. We choose to use the terms thematic and non thematic in order to avoid confusion with other meanings of the word lexical. For a formal account of non thematic indirect objects in French, see Roberge and Troberg (2007b).

b. Un type comme ça, il te tue son père sans hésiter. (Jones 1996:300)

‘A guy like that, he’ll kill his dad on ya without a second thought.’

(16) a. J’ai chanté à Montréal la semaine passé.

‘I sang in Montreal last week.’

b. Paul prend toujours son goûter à 16h.

‘Paul always has a snack at 4 o’clock.’

Various tests distinguish IOs from circumstantials. The sentences in (17) show that while the circumstantial complement à Montréal can be separated from the verb chanter, verbal arguments such as à Marie and à la maison in (18) cannot.

(17) a. Chanter à Montréal, c’est ce que j’ai fait.

‘Sing in Montreal, that’s what I did.’

b. Chanter, c’est ce que j’ai fait à Montréal.

‘Sing is what I did in Montreal.’

c. J’ai chanté à Montréal la semaine passé et Jean en a fait autant à

Toronto.

‘I sang in Montreal last week and Jean did the same in Toronto.’

d. – Qu’est-ce que Jean a fait à Montréal? – Il a chanté.

‘– What did John do in Montreal ? – He sang.’

(18) a. * Offrir des fleurs, c’est ce que Jean a fait à Marie.

‘Give flowers, that’s what John did to Mary.’

b. * Arriver, c’est ce que les invités ont fait à la maison.

‘Arrive, that’s what the guests did at the house.’

c. * – Qu’est-ce que Jean a fait à Marie? – Il a offert des fleurs.

‘– What did Jean do to Marie? – He gave flowers.’

These diagnostics indicate selectional and structural relations that hold between the thematic IO and the verb phrase which are absent for circumstantials.

The valency change discussed in this study concerns thematic indirect objects. Thematic IOs can be classified in a number of ways; we adopt in part the classification of dative complements in Melis (1996, 2004) by grouping verbs by valence (trivalent vs. bivalent), by their tendency to express a dynamic or stative event, and further still according to commonly held meanings.9

IOs occurring with three-place verbs fall into one main group: dynamic verbs of transfer of possession, information or location (19a-c), and verbs of comparison (19d).

(19) a. Bill a remis les clefs au concierge.

‘Bill returned the keys to the concierge.’

b. Bill a annoncé son départ au conseil.

‘Bill announced his departure to the council.’

c. Bill a conduit son père à l’aéroport.

‘Bill drove his father to the airport.’

d. Jean préfère les pâtes au riz.

‘Jean prefers pasta to rice.’

IOs occurring with two-place verbs fall into two groups. The first consists of dynamic predicates: verbs of interaction, (20a,b), of reflection (20c), and change of location verbs (20d). The second consists of stative predicates: locative verbs (20e), relation verbs (20f), and psych verbs (20g).

(20) a. Marie a parlé à Bill.

‘Marie spoke to Bill.’

9 We do not discuss epistemic dative constructions of the type Je lui trouve mauvaise mine. See Roberge and Troberg (2007a) for discussion and analysis.

b. Marie a renoncé à sa famille.

‘Marie has abandoned her family.’

c. Marc pense à Sophie.

‘Marc thinks of Sophie.’

d. Jean est allé à Montréal.

‘Jean went to Montreal.’

e. Bill habite à Paris.

‘Bill lives in Paris.’

f. Ce cadeau ne correspond pas vraiment à mes attentes.

‘This gift doesn’t really meet my expectations.’

g. Bill plaît à Marie.

‘Marie likes Bill.’

1.3.1 Three-place verbs of transfer and comparison

Trivalent verbs select a relation. This relation can denote transfer of possession, transfer of location, comparison, or association. The preposition à permits all such meanings. A non exhaustive list is given in (21) and some examples are provided in (22). Note that some traditional unergatives are included among the trivalents (crier, chuchoter, hurler, etc.). This is meant to show that they may be coerced into occurring with a relation, as shown in (22c), in which case they adopt the same argument structure as traditional three-place verbs like donner.

(21) Verbs of transfer : abandonner, attribuer, adresser, apporter, allouer, confier, céder, conférer, donner, distribuer, infliger, lancer, léguer, envoyer, expédier, jeter, offrir, prescrire, renvoyer, retourner, remettre, rendre, arracher, dissimuler, enlever, épargner, interdire, ôter, retirer, susciter, voler, annoncer, apprendre, commander, communiquer, confesser, demander, dire, jurer, montrer, prouver, rappeler, crier, chuchoter, hurler, etc.; verbs of comparison : assimiler, comparer, confronter, opposer, préférer, substituer, adjoindre, associer, jumeler , juxtaposer, unir, etc.

(22) a. Marie lui a confié son secret.

‘Marie entrusted her secret with him.’

b. On a envoyé les troupes au front.

‘The troops were sent to the front.’

c. Paul a crié des injures à Régis.

‘Paul screamed insults at Régis.’

d. Le juge a retiré le permis à un chauffard.

‘The judge took away the license from a bad driver.’

e. Jean préfère les pâtes au riz.

‘Jean prefers pasta to rice.’

f. Le média a tendance à associer le mouvement d’extrême droite au terrorisme.

‘The media tends to associate the extreme right with terrorism.’

We propose the configuration in (23) as the argument structure associated with these verbs used trivalently. It describes a verbal root that selects as its complement a relation mediated by the preposition à. The event is introduced by an “agentive” little v, and brought about by an external subject. As such, the verb can be seen as a lexical causative; the subject brings about the relation between the direct and indirect object. A similar structure is proposed in Hale and Keyser (1993, 2002) for locative and locatum verbs in a number of different languages.

(23) vP 3 DPi 3 v √P 3 √ PP 3 DPi 3 P DP g à

Note that we follow Cuervo (2003) in merging the verbal root between little v and the PP. As sister to null little v, the root conflates with it at Merge, producing a verb with the phonological signature of the root.

The two internal DPs are licensed syntactically by virtue of the two positions projected by the preposition, and semantically by their relative position in the structure. The specifier of the general relation introduced by the preposition à is always interpreted as a Figure, a moving or conceptually moveable entity, while the complement of à is construed as the Ground, in this case a point of reference (Talmy 2000). The specific meaning of the two internal arguments comes from the particular meaning of the verbal root with which it combines and the denotation of their referents. With verbal roots denoting transfer, the Figure (the direct object) is generally interpreted as the theme while the Ground (the complement of à) can take on the meaning of source (22d), recipient (22a), (22c), or goal (22b). Some roots simply establish a relation between two entities, as in (22e) and (22f). In these cases, the IO is always interpreted as the general property to which the DO is compared or associated.

The lower DP receives its case from the prepositional head, while the higher DP must raise to check accusative case in the specifier of a functional projection. We will assume that this position is SpecvP. Finally, we contend that all indirect objects headed by à in French are exactly as they seem: prepositional phrases. It is sometimes assumed that dative IOs are in fact DPs, for which à would not be a preposition, but rather an inflectional case marker. We present a number of arguments in section 1.3.6 which demonstrate that such a claim is untenable.

1.3.2 Change of location verbs

Otherwise known as unaccusative verbs, change of location verbs denote punctual events in which a theme argument comes to be in a given location. These verbs express a relation between the grammatical subject and the indirect object. Some of the most common of these verbs are given in (24) and some examples follow.

(24) Change of location verbs : aller, arriver, descendre, entrer, monter, partir, parvenir, rester, retourner, sortir, tomber, venir, advenir, apparaître, etc.

(25) a. Marie est allée à Paris.

‘Marie went to Paris.’

b. Diane est partie à Boston.

‘Diane left to Boston.’

c. Sophie est revenue à la maison.

‘Sophie returned to the house.’

We assume the traditional unaccusative analysis of these verbs (Perlmutter 1978; Burzio 1986) according to which the grammatical subject is really a raised internal argument. Hale and Keyser propose a structure very similar to that of three place verbs of transfer and comparison since the relation is essentially the same.

(26) vP 3 vBECOME √P 3 √ PP 3 DP 3 P DP g à

The only difference between unaccusative verbs and three-place verbs of transfer and comparison is the absence of an external argument. According to the inventory of little v’s in section 1.2.4, the formal difference between these two types of verbs boils down to the kind of little v which introduces the event. In the case of change of location verbs, it is the little v with the meaning of BECOME that merges with the root to give an achievement reading. Since VoiceP is incompatible with predicates headed by “unaccusative” v, no external argument can be merged to this structure. Finally, following Burzio’s generalization, “unaccusative” little v does not have

30 case features. Consequently, the theme argument in the specifier of P raises to the inflectional layer, becoming the grammatical subject, acquiring nominative case and satisfying the EPP.10

1.3.3 Locative and relation verbs

Locative and relation verbs assert the existence of a relation between the grammatical subject and the indirect object. Locatives situate the subject relative to a particular place while relation verbs abstract away from concrete locations and establish a general association between the subject and the indirect object. Some common locative and relation verbs are given in (27) and (29) respectively. Some examples follow each.

(27) Locative verbs : confiner, demeurer, habiter, résider, rester, vivre, manquer, etc.

(28) a. Il demeure au 45 rue d’Ulm.

‘He resides at 45 rue d’Ulm.’

b. Ils vivent à Montréal.

‘They live in Montreal.’

c. Une touche de majesté manque à ce bâtiment.

‘This building lacks a touch of grandeur.’

(29) Relation verbs : appartenir, correspondre, préexister, ressembler, succéder, survivre

(30) a. Jean appartient à l’équipe.

‘Jean belongs to the team.’

b. Bill ressemble à son père.

‘Bill resembles his father.’

Locative and relation verbs are derived in essentially the same way as unaccusative verbs:

10 The theme argument remains in situ in impersonal constructions: Il est arrivé à Paris une caravane d’Espagne.

(31) vP 3 vBE √P 3 √ PP 3 DP 3 P DP g à

Unlike change of location verbs, however, locative and relation verbs denote states, not achievements and are thus headed by an existential little v, glossed here as vBE. The arguments of the verb are licensed syntactically and semantically by the configuration projected by the prepositional head in the same way as three-place verbs of transfer and change of location verbs. When the IO (the Ground) refers to an actual place, then the subject is located relative to that place. When the IO is not a place, the spatial relation is metaphorical; the Ground is a point of reference, usually a property, relative to which the Figure is defined. Since existential little v does not have case features, the DP in the specifier of P must raise to become the grammatical subject, impersonal constructions notwithstanding.

Although locative and relation verbs do not belong to the class traditionally defined as unaccusative, they pass a number of classic tests suggesting a derived subject analysis. Locatives allow en cliticization with respect to the subject, en relating to an element within the verb phrase, illustrated in (32). Locatives can also appear as non finite sentential complements with relative clauses. The latter constructions are difficult but acceptable with typical unaccusatives but definitely out with unergatives, as shown in (33a) and (33b) (examples adapted from Labelle, 1992:384). Crucially, locative verbs show the same contrast (33c), patterning more like unaccusatives than unergative verbs.

(32) a. Il en est déjà arrivé plusieurs à la fête.

‘There has already arrived several of them at the party.’

b. Il en habite/vit beaucoup à Montréal.

‘There are many of them living in Montreal.’

(33) a. ? L’homme que je croyais être arrivé/parti en retard est en fait arrivé/parti tôt.

‘The man that I thought to have arrived/left late in fact arrived/left early.’

b. * L’homme que je croyais avoir téléphoné/toussé n’était pas en fait, celui que

j’ai vu dans le train.

‘The man that I believed to have phoned/coughed was not in fact the one that

I saw in the train.’

c. ? L’homme que je croyais avoir habité à Montréal, a en fait habité à Denver.

‘The man that I believed to have lived in Montreal in fact lived in Denver.’

The facts also support a derived subject analysis for relation verbs. First, the verbs in (29) cannot appear with a DO independently of the subject. This suggests, in the context of stative relation verbs, that no such position is available. The internal argument position is presumably occupied by the grammatical subject.

(34) a. * Jean appartient ce contrat à l’équipe.

‘Jean belongs this contract to the team.’

b. * Une touche de majesté manque quelque chose à ce bâtiment.

‘This building lacks something a touch of grandeur.’

c. * Bill ressemble sa mine à son père.

‘Bill resembles his expression his father.’

Second, as with unaccusatives, many of these verbs can be used in impersonal constructions. The DO is then often heavy and extraposed.

(35) a. Il appartient à l’équipe de faire signer un contrat à ce joueur.

‘It is the club’s duty to have this player sign a contract.’

b. Il incombe au gouvernement de prendre les mesures nécessaires.

‘It is incumbent upon the government to take the necessary measures.’

c. Il manque à ce bâtiment une touche de majesté.

‘There lacks a touch of grandeur in this building.’

Finally, the grammatical subject of some relational verbs patterns like the derived subject of unaccusative verbs in constructions with non finite sentential complements with relative clauses. Again, the important point here is the contrast: like (36a), (36c) is more acceptable than the same construction in (36b) containing an unergative verb.

(36) a. ? L”homme que je croyais être arrivé en retard, est en fait arrivé

très en avance.

‘The man that I believed to have arrived late in fact arrived very early.’

b. * L’homme que je croyais avoir téléphoné/toussé n’était pas en fait, celui que

j’ai vu dans le train.

‘The man that I believed to have phoned/coughed was not in fact the one that

I saw in the train.’

c. ? L’homme que tu croyais avoir appartenu à l’équipe de foot est en fait

pianiste.

‘The man that you believed to have belonged to the soccer team is in fact a

pianist.’

1.3.4 Psychological verbs

Psychological predicates express the existence of an experiencer who has an emotional reaction to a theme. In French, experiencers can be subjects (37a), direct objects (37b), or datives (37c).

(37) a. Jean aime le film.

‘John likes the film.’

b. Le film a effrayé Jean.

‘The film frightened John.

c. Le film plaît à Jean.

‘John likes the film.’

This study is concerned with psychological verbs that take dative object experiencers. Some such verbs are given in (38), followed by a few examples.

(38) Psych verbs : agréer, convenir, démanger, (dé)plaire, importer, nuire, peser, profiter, répugner, suffire, tarder, etc.

(39) a. Le nez lui démange.

‘Her nose is itching her.’

b. La robe bleu plaît toujours à Marie.

‘Marie still likes the blue dress.’

c. Les sentiments des autres leur importent très peu.

‘The feelings of others matter very little to them.’

Our analysis of sentences that express psychological states follows in part Harley’s (2002a) extension of HAVE relations in double object constructions to psychological predicates. She claims that experiencers are construed as possessing a psychological state. Such a HAVE relation can be represented in two ways depending on if the language in question expresses possession with a verb meaning ‘have’ (have languages), or whether they express it as the combination of a verb meaning ‘be’ and a locative preposition (have-not languages). French commonly uses the verb avoir ‘have’ to express psychological states in which the subject is the experiencer, as illustrated in (40a) and (40b). Irish is a have-not language that expresses psychological states using ‘be’ and an experiencer introduced by a locative preposition (40c).

(40) a. Jean a faim.

‘Jean is hungry.’ Literally ‘Jean has hunger’

b. Jean a peur.

‘Jean is scared.’ Literally ‘Jean has fear’

c. Tá eagla roimh an bpúca ag Ailill (Noonan 1993; Harley 2002a)

BE fear before the Puca at Ailill

‘Ailill fears the Puca’

Note however, that while French generally expresses possession with the verb avoir ‘to have’, it can also be expressed using the verb être ‘be’ and the spatial preposition à, as shown in (41a). This relation can be similarly expressed in the absence of the verbal predicate être, in (41b).

(41) a. Le stylo est à Jean.

‘The pen is John’s.’ Literally ‘The pen is at John’

b. As-tu vu le stylo à Jean?

‘Have you seen John’s pen?’

We suggest that in French, psychological predicates taking dative experiencers are derived in much the same way as they are in Irish by employing the basic possessive construction underlying the sentences in (41). Compare, for example, the sentence in (41a) with a psych-verb construction:

(42) Le stylo plaît à Jean.

‘John likes the pen.’

The difference between (41a) and (42) is that the psych construction is composed of an additional layer of meaning: John has pleasure and that pleasure is associated with the pen. We propose the following derivation for psychological predicates taking dative experiencers:11

11 Possessor constructions like (41a) are derived from an argument structure very similar to that of locative and relation verbs; see the structure in (31).

(43) vP 3 vBE SCplais- “plaire” 3 DP aPplais- 5 3

le stylo aplais- nplais- 3

nplais- PP 3 P DP g 5 à Jean

The lower portion of the structure in (43) describes the adjunction of a non relational prepositional phrase, à Jean, to a bare noun denoting a property, plais- ‘pleasure’ in this case. The adjunct PP modifies the property by implicating it in a relation in which it is the theme and the animate complement of à is the possessor. This is exactly the same relation as that which holds between John and the pen in (41b). The experiencer DP in (43) thus receives its case and interpretation from the spatial preposition à.

The upper portion of the structure in (43) expresses the association of the property plais- , now adjectival, with le stylo in the form of a small clause. The existence of this association is asserted by an “existential” little v. ‘The pen is pleasing’ would be the interpretation of the upper portion of the structure. Since “existential” little v is neither associated with case features nor a Voice phrase, le stylo, raises to the inflectional layer to satisfy the EPP, becoming the grammatical subject and receiving nominative case. Finally, conflation of the adjective plais- with null vBE creates the verb plaire.

This structure differs from previous proposals for psych verbs in two important ways. First, the configuration in (43) derives a static event as opposed to an event introduced by an agentive little v (proposed in Arad (1998), for instance). The grammatical subject of a psychological state is thus derived from a vP internal position, supported by Burzio’s (1986) diagnostic test for derived subjects. The faire causative construction licenses the external subject me, in (44a), but not the derived subject, in (44b).

(44) a. La robe bleue m’a fait penser à Pierre.

‘The blue dress made me think about Pierre’

b. * Le fait que j’avais porté la robe bleue m’a fait plaire à Pierre.

‘The fact that I wore the blue dress made me appealing to Pierre.’

Second, the structural relation between the theme and the experiencer does not follow Belletti and Rizzi’s (1988) influential proposal for Italian psych verbs in which the experiencer of verbs like preoccupare ‘worry’ and piacere ‘like’ is claimed to be structurally higher than the theme. The experiencer-over-theme relation in Italian is justified based on binding evidence where the experiencer can bind an anaphor contained within the theme constituent regardless of linear precedence. In structural terms, (45) is possible because the anaphor propri is bound by Gianni at external Merge (Belletti and Rizzi 1988:321).

(45) I proprii sostenitori preoccupano Giannii.

‘His own supporters worry Gianni.’

This is not true, however, of dative experiencers in French. They do not bind local anaphors, as shown in (46a), just as an anaphor contained in the subject of a basic transitive sentence is not bound by the object, in (46b). The sentence in (46a) is presumably not possible because the dative experiencer does not asymmetrically c-command the theme in French in the same way that the direct object in (46b) does not asymmetrically c-command the external subject. Compare these sentences with (46c), however, in which an antecedent can indeed bind an anaphor that has linear precedence.12

(46) a. * Ses propresi films plaisent à Denys Arcandi.

‘His own films appeal to Denys Arcand.’

12 According to our informants, the sentences in (46a) and (46b) become somewhat acceptable in a context where the experiencer (Denys Arcand) has already been mentioned. (46a) is marginally acceptable in a very restricted, highly contrastive context such as the following: Ses propres films plaisent à Denys Arcand, pas les autres. Il déteste les autres. These contexts must be excluded, however, when assessing binding relations established at external Merge.

b. * Ses propresi films décrivent Denys Arcand i.

‘His own films describe Denys Arcand.’

c. Ses propresi partisans préoccupent Jeani.

‘His own supporters worry Jean.’

Contrasts such as these support our claim that at the level of the vP, the dative experiencer does not asymmetrically c-command the theme. 13

1.3.5 Verbs of interaction and reflection

Dynamic two-place indirect verbs denote a range of activities that are initiated by an agentive or causer subject. Some can be generally construed as denoting some form of social interaction between the subject and the object which involved in the action, but not directly affected by it. Animate objects are usually commutable with dative pronouns. Verbs of reflection refer to directed actions such as thinking, reflecting, dreaming, etc., the object of which is not implicated in the activity at all, but is rather a simple point of reference, much like a location. In these cases, animate objects are commutable with disjunctive pronouns. A partial list is given in (47) and some examples follow.

13 Our treatment of French dative experiencers differs, therefore, from the experiencer-over-theme analysis given for Italian and Spanish. Note however, that examples like (i) and (ii), which are used to support this hierarchy, are crucially different from French in that these Romance languages allow dative subjects which have linear precedence over the anaphor.

(i) Al directori le gustaron sus propiasi películas. (Cuervo 2003:6, ch.4) ‘The director liked his own films’

(ii) A Giannii piace chiunque accetti le propriei idee. (Belletti and Rizzi 1988:335) ‘Gianni likes whoever accepts his own ideas.’

If we follow Cuervo’s (2003) analysis of Spanish dative experiencers, the source of variation between French and Spanish psych verbs would reside in the category of the dative argument. Cuervo argues for a dative-marked DP introduced by an applicative head. We claim, on the other hand, that the dative experiencer in French is a PP which establishes an abstract spatial relation with the experience within the vP.

(47) Interaction verbs: acquiescer, attenter, céder, compatir, contrevenir, coopérer, déférer, désobéir, échapper, forfaire, mentir, obtempérer, obéir, obvier, parer, parler, pourvoir, présider, réagir, recourir, remédier, renoncer, répondre, résister, sourire, souscrire, subvenir, succomber, suppléer, vaquer, veiller, etc.; verbs of reflection: penser, réfléchir, rêver, songer, etc.

(48) a. Pierre ne lui ment jamais, à Marie.

‘Pierre never lies to Marie.’

b. Nous obéissons à nos parents.

‘We obey our parents.’

c. Elle refuse de renoncer à son frère.

‘She refuses to exclude her brother from her life.’

d. Il n’ose pas lui résister.

‘He doesn’t dare stand up to her.’

e. Henri lui a parlé hier, à son patron.

‘Henri spoke to his boss yesterday.’

f. Remédier au manque de main-d’oeuvre en faisant venir des ouvriers

étrangers.

‘To meet the lack of man-power by recruiting foreign workers.’

(Le Grand Larousse)

g. Veiller à la bonne tenue de ses enfants.

‘To attend to the good manners of one’s children.’

(Le Grand Larousse)

h. Il ne pense plus qu’à elle.

‘He no longer thinks of anything but her.’

i. J'ai rêvé à vous cette nuit. 14

‘I dreamed of you last night.’

We propose the configuration in (49) as the argument structure of these two-place verbs. It describes an Event denoting root which takes as its complement a non relational PP (the IO). The root conflates with an agentive little v at Merge. This configuration is like any unergative verb (see the derivation of baver in (10)), only here, of course, the root selects a complement.

(49) vP 3 v 3 √ PP 3 à DP

Since these verbs denote activities and select agentive or causer subjects, the subject is semantically and syntactically licensed by VoiceP. The merger of the root and a PP follows from the fact that there are no formal restrictions on external Merge. In principle, any constituent should be able to merge with the root assuming it meets that root’s selectional properties; the prepositional phrase must be compatible with the various semantic roles that these verbal roots assign to their complement. The non relational PP can thus be understood as a lexical case marker since its expression is not the result of a structural requirement, but rather depends on the general semantic role assigned by the verb.

1.3.6 Thematic indirect objects are prepositional phrases

It is often assumed that the dative indirect object in French is a determiner phrase and not a prepositional phrase. A careful consideration of the data, however, shows that the classic examples which have been used to demonstrate that à is an inflectional case marker (a bound morpheme that does not affect the category of the element with which it occurs) rather than a prepositional head do not in fact illustrate this at all. Since a correct description of the categorial

14 “Malgré la tradition puriste, rêver, au sens propre, se construit le plus souvent avec à dans la langue actuelle: J'ai rêvé à vous cette nuit. On peut admettre que j'ai rêvé de vous est plus élégant, mais il paraît difficile de considérer rêver à comme franchement incorrect” (Dupré, 1972:2286; TLFi).

41 status of dative IOs is crucial to a formal representation of thematic IOs in French, the main arguments of the debate are reviewed. We conclude that there is no piece of evidence showing that dative IOs are DPs. All thematic IOs in French are therefore taken to be PPs.15

Thematic indirect objects are traditionally classified paradigmatically according to the pronouns with which they commute. Following the terminology used in Herslund (1988), dative IOs, in example (50), are commutable with the clitic pronouns me, te, lui, nous, vous, leur and se. Neuter IOs, in example (51), commute with y, and the disjunctive pronouns à elle/lui/elles/eux/cela. Locatives, in (52), commute with y and là.16

(50) a. Jean lui a offert des fleurs, à Marie.

‘Jean gave her (Mary) flowers.’

b. Marc leur parle, aux invités.

‘Marc is speaking to them (the guests).’

(51) a. Jean pense à elle, à sa mère.

‘Jean is thinking about her (his mother).’

b. On ignore comment y remédier, à ce mal.

‘They don’t know how to put it (this trouble) right.’

(52) a. Les invités vont bientôt y arriver, à la maison.

‘The guests are going to arrive at the house soon.’

b. Marie habite là, au fond de la rue.

‘Mary lives there, at the bottom of the street.’

15 To be more precise, following a view of lexical items as outlined in section 1.2.5, à is acategorial but compatible with the structure associated with a preposition.

16 Neuter IOs are generally associated with verbs whose object is a point of reference (i.e., penser à quelqu’un), as opposed to dative IOs which are to some degree involved in the event (parler à quelqu’un).

Since verbs tend to occur with one particular set of pronouns, they are often referred to as being either dative, neuter, or locative, but these terms can be misleading in French. Verbs do not always belong exclusively to one single paradigm. A case in point discussed in Herslund (1988) is the fact that some so-called “dative” verbs can occur with both lui and y, depending on the nature of the complement of à:

(53) a. Ce bonhomme lui ressemble, à mon perroquet.

‘This fellow resembles my parrot.’

b. Ce bonhomme y ressemble, à un perroquet.

‘This fellow looks like a parrot.’

Herslund suggests that the use of y corresponds to the représentation intensionelle (the concept) of its antecedent whereas the dative pronoun corresponds to the représentation extensionnelle (the individual).

Similarly, some verbs can occur with both neuter and locative pronouns, depending on whether the object is animate or not:

(54) a. Paul court à elle, à sa mère.

‘Paul runs to his mother.’

b. Paul y court, au magasin.17

‘Paul runs to the store.’

The terms “dative”, “neuter”, and “locative” are therefore not always appropriate ways of referring to specific classes of verbs in French. As pointed out in Herslund (1988), given that the choice of pronominal form in French is often due to features internal to the complement of à, it is difficult to adopt the commonly accepted proposal that the dative indirect object is structurally distinct from neuters and locatives. If this were the case, then the categorial status of the object of ressembler in (53) would depend on the reference of the DP, which, needless to say, seems

17 This sentence can also have the interpretation “Paul is running while at the store”.

43 completely counter intuitive. A number of arguments for a categorial distinction between datives on the one hand and neuter and locatives on the other, have nevertheless had a surprising degree of success in the literature. They are discussed below.

Since Seelbach (1970) and Vergnaud (1974), there has been much debate about the differences between dative, neuter and locative IOs. Kayne (1975) argues that there are in fact two homophonous à’s in French which behave differently with respect to clitic transformations: the à heading neuter and locative IOs is a preposition just like sur, dans, avec, etc., while dative à, although considered to be a preposition structurally, behaves differently and has a separate entry in the lexicon. Namely, it undergoes deletion when followed by a pronominal complement (which eventually raises to incorporate with the inflected verbal head). Ruwet (1982) and Jaeggli (1982) take a stronger stand, arguing that there is not only a lexical difference, but also a structural difference between dative à and other prepositions. They argue that dative à should be treated as an inflectional Case feature, not the head of a prepositional phrase.

Coordination structures

Vergnaud’s (1974) coordination structures are often cited in order to show the apparent difference between dative IOs and other prepositions, including neuter and locative à. In examples (55a-c), the preposition sur and the à heading the neuter and locative IOs need not be repeated in the second conjunct of two coordinated DPS, while the dative à must be repeated, as shown in (55d).

(55) a. Ils se sont assis sur la table et les chaises. (Vergnaud 1974)

‘They sat on the table and the chairs.’

b. Tu penses à Paul et la directrice. (Jaeggli 1982:29)

‘You are thinking about Paul and the director.’

c. Tu vas aller à Rio de Janeiro et Buenos Aires. (Jaeggli 1982:29)

‘You are going to go to Rio de Janeiro and Buenos Aires.’

d. Ils ont parlé à Marie et *le/au directeur. (Vergnaud 1974)

‘They spoke to Marie and the/to the director.’

Herslund (1988:317) points out however, that the contexts in (55) illustrate less the DP status of the dative IO and more the general restriction on à in coordination structures. He proposes that it must be repeated when the two DPs are considered separately, but needn’t be if they form in some sense a natural group construed as a single goal or recipient as in (56). A similar view is adopted in Comorovski (1990) and in Zaring (1991).

(56) Tu as envoyé cette lettre à Paul et Marianne? (Herslund 1988:317) ‘Did you send that letter to Paul and Marianne?’

Clearly à can coordinate two DP complements just as any other preposition can. We suggest, however, that the restriction is better captured not from a semantic perspective, but rather by considering the phonological status of à. A second look at the examples in (55) reveals that à can evidently coordinate DPs as long as the determiner does not normally form a contraction with it (au, aux):

(57) a. [P à [[DP Paul] et [DP la directrice]]]

b. [P à [[DP Rio de Janeiro] et [DP Buenos Aires]]]

Expanding the coordination structures, no contracted forms would normally arise and all PPs are therefore acceptable: à Paul and à la directrice; à Rio de Janeiro and à Buenos Aires. On the other hand, when a contracted form would normally arise in one or more of the conjuncts, namely those headed by the determiners le and les, coordination is not possible:

(58) * [P à [[DP Marie] et [DP le directeur]]].

Expanding the coordination structure above, à Marie is acceptable, but à le directeur is not. À and le must form the contraction au. Examples (55b) and (55d) are therefore not minimal pairs. The former would be just as unacceptable as the latter if its second conjunct were headed by the determiner le:

(59) * Tu penses à Paul et le directeur.

In this case, coordination must occur between the two PPs in order for the sentence to be acceptable:

(60) [PP à Paul] et [PP au directeur]

Similarly, replacing le directeur in (55d) with la directrice renders the sentence acceptable for now no contraction is required in either conjunct; the preposition à can coordinate them both:

(61) Ils ont parlé à Marie et la directrice.

Coordination restrictions imposed by contracted forms are present not only for dative à but for neuter and locative à as well. Locative, neuter and dative IOs all behave in the same way in coordination structures, as illustrated in (62) to (64).

(62) a. * Tu vas aller à Rio de Janeiro et les îles Caïmans.

b. Tu vas aller à Rio de Janeiro et Buenos Aires.

(63) a. * Tu penses à Marie et le directeur.

b. Tu penses à Marie et la directrice.

(64) a. * Tu parles à Marie et le directeur.

b. Tu parles à Marie et la directrice.

The differences in (55) can therefore be attributed to the phonological character of à which restricts its occurrence in coordination structures. The examples in (55) show no evidence of a categorial difference between dative à and the à corresponding to neuter and locative IOs.

Vergnaud presents another set of examples that shows an apparent similarity between dative indirect objects and simple DPs. In (65a), a coordinated structure containing two PPs cannot serve as the antecedent of a relative pronoun, whereas two coordinated dative indirect objects (65b) and two simple DPs (65c) can.

(65) a. * Il a compté sur l’homme et sur la femme qui se sont rencontrés hier.

‘He counted on the man and on the woman who met yesterday.’

b. Il a parlé à l’homme et à la femme qui se sont rencontrés hier.

‘He spoke to the man and to the woman who met yesterday.’

c. Il a vu l’homme et la femme qui se sont rencontrés hier.

‘He saw the man and the woman who met yesterday.’

We suggest that the difference between (65a) on the one hand, and (65b) and (65c) on the other, can be better understood in light of the fact that the latter are both selected and coordinated under one VP while PPs headed by more lexical prepositions like sur are situated outside of the vP as circumstantials. This is the case for compter which is essentially an unergative verb that has, in its metaphorical sense, lexicalized a circumstantial complement. These kinds of PPs are thus adjuncts, positioned outside of the VP.

The coordinated internal arguments in (65b) and (65c) denote the unified object of one event of speaking or seeing. In contrast, two coordinated circumstantial complements are not restricted to denoting a single event; (65a) can refer to two separate events of counting on someone. Since a relative pronoun can clearly only refer to two or more participants when they are construed as a single object and coordinated under one event, (65a) is ungrammatical. Again, these examples do not provide any evidence for the DP status of dative IOs.

Chomsky (1957:35) observes that one of the governing principles of coordination is that only constituents of the same syntactic category can be coordinated. Ruwet (1982) uses this to argue that à is not a prepositional head, since it cannot be coordinated with a more lexical preposition:

(66) a. * Les flics grouillaient à et devant la Sorbonne. (Ruwet 1982:319)

‘The cops were swarming at and in front of the Sorbonne.’

b. Il y avait un tas de flics devant et derrière nous. (Ruwet 1982:318)

‘There were a ton of cops before and behind us.’

As pointed out in Miller (1992), the problem with examples like (66a) is most likely caused by the fact that à lacks semantic parallelism with other prepositions. A different choice of conjunction, however, permits coordination between à and its directional equivalent vers. The sentence in (67) is possible, for example, in the context of a hitchhiker asking a driver about where he or she is headed to.

(67) Est-ce que vous allez à ou vers Paris?

‘Are you headed to or towards Paris?’

Since coordination is possible between à and another preposition, à must also be a preposition.

Floating quantifiers

Jaeggli (1982:29) introduces a data set that shows similarities between dative IOs and DPs on the one hand and PPs and neuter IOs on the other. Examples taken from Perlmutter (1972) illustrate that quantifiers cannot be floated from PPs and neuter IOs in which the object of the preposition has been relativized, in (68a) and (68b). This is claimed to contrast with dative IOs and simple DP complements, in (68c) and (68d).

(68) a. * Ces femmes, avec qui j’ai parlé (avec) toutes, ...

b. * Ces femmes, à qui j’ai pensé (à) toutes, ...

c. Ces femmes, à qui j’ai parlé à toutes, ...

d. Ces femmes, que j’ai toutes vues, ...

What Jaeggli doesn’t mention, however, is that examples like (68c) are considered to be ungrammatical or rarely accepted in Kayne (1975:105) where in fact, they are used to argue for dative à as a preposition in the syntax. Our informants confirm Kayne’s analysis, finding (68c) to be just as ungrammatical as the (a) and (b) examples. But even if we take (68c) to be relatively acceptable, Herslund (1988) makes the more relevant observation that there is nevertheless a clear difference between (68c) and (68d): quantifier raising is blocked for the sentence in (68c) where toutes is preceded by à.

(69) * Ces femmes, à qui j’ai à toutes parlé, ... (Herslund 1988:318)

If à were not a head, but an inflectional case-marker, then one would expect the quantifier in (68c) to be able to undergo the raising operation available to these kinds of quantifier phrases illustrated in (68d). This is evidently not the case – presumably because à toutes is a PP.

A/A principle

Finally, Herslund points out that datives obey the A/A principle. If datives were really DPs, one would expect the extraction of a PP from a dative complement to yield a grammatical utterance, yet this is not so. While extraction of a PP is possible out of the DP in (70), it is not possible out of the neuter IO in (71), or the dative in (72) (examples from Herslund 1988:46).

(70) a. Claude a tué [DP la belle-mère de Simone]

‘Claude killed Simone’s mother-in-law.’

b. De qui Claude a-t-il tué [DP la belle-mère e] ?

‘Of whom did Claude kill the mother-in-law?’

(71) a. Claude a pensé [PP à la belle-mère de Simone]

‘Claude thought about Simone’s mother-in-law.’

b. * De qui Claude a-t-il pensé [PP à la belle-mère e] ?

‘Of whom did Claude think about the mother-in-law ?’

(72) a. Claude a plu [PP à la belle-mère de Simone]

‘Claude was pleasing to Simone’s mother-in-law’

b. * De qui Claude a-t-il plu [PP à la belle-mère e] ?

‘Of whom did Claude appeal to the mother-in-law?’

A natural explanation for the ungrammaticality of (71b) and (72b) can be attributed to the PP status of both the neuter and the dative arguments.

To conclude, there is not one piece of clear evidence that dative à behaves any differently, structurally speaking, than locative or neuter à or any other lexical preposition for that matter. Our analysis therefore treats all IOs in French as PPs.

1.4 Chapter summary

This chapter outlined the main goals of the dissertation, the object of study, and the basic assumptions that form the framework of the analysis.

Our main goal is to demonstrate the connection between regularity in form-meaning mapping and regularity in language change. When regularities within verb classes are characterized by a common formal representation, then systematic valency change can be linked to a change in a grammatical element associated with a particular argument structure. This way of considering valency change can do two things: on the one hand, it can distinguish structural changes from non structural changes. On the other hand, it can contribute to the theory of argument structure, motivating a well articulated typology of verbs and identifying dependencies that govern the morphosyntactic expression of objects in a given language.

The empirical basis of the analysis consists of an often-noted valency change in the history of French, exemplified by the verb aider. The change involves dynamic two-place verbs whose indirect object was replaced by a direct object without any apparent change in the meaning of the verb. In the following chapters, we will show that the indirect objects of these verbs changed in a unified way in response to a single common cause and are set apart from all other verbs taking indirect object complements. The cause is attributed to inability of the preposition à to encode directionality from the 16th century. Its exclusive effect on verbs like aider is attributed to the distinct argument structure of these verbs in conjunction with the specific licensing requirements of first and third order indirect objects in French.

We adopt a generative grammar approach to language change (Lightfoot 1979, 1991, 1998, 2006) and a neo-constructionist approach to the representation of argument structure (Hale and Keyser 1993, 2002; Harley 1995, 2002b, 2004, among others). A detailed classification and formal description of indirect objects in French was proposed. The derivation of the various verb classes in French forms the backbone of our explanation for why the valency change was limited to a specific group of two-place verbs. Finally, in light of commonly held assumptions about the category of thematic dative arguments in French, we went to some lengths to demonstrate that à is structurally a preposition as much as sur, dans, etc. All indirect objects in French which are headed by à, be they dative, neuter or locative, are therefore treated as prepositional phrases.

Chapter 2 The verbs

Historical grammars make mention of verbs which no longer have the same argument realization, or the same possibilities of argument realization, as they once had. It is common to find therein inventories of verbs that have undergone various kinds of syntactic and semantic changes.18 An analysis of the changes is not the aim of these grammars, but the presentation of the material and the few remarks that are offered clearly communicate the popular view at the turn of the century that most changes in argument realization are cases of random and isolated change, summed up, for example, in the first volume of Brunot’s Histoire de la langue française:

Comme aucune différence de nature ne sépare, dans la plupart des cas, transitifs et intransitifs, à toute époque un verbe passe d’un de ces états à l’autre, et des verbes intransitifs en latin se sont trouvés transitifs en ancien français. Je ne parle même pas de cas où des verbes qui comportaient le datif en latin se construisent en français avec l’accusatif sans préposition – c’est là simple accident grammatical... (Brunot, Histoire de la langue française, vol.1, 236)

Marchello-Nizia (2003:371) aptly states, however, that “the task of historical linguistics is indeed to seek out regularities in the multitude of diverse changes which occur in languages.” This thesis intends to contribute to the domain by showing that some changes in argument realization, long considered random, are indeed regular. The regularity of the change examined in the present study not only reveals a common time course, but also common structural and semantic properties shared by verbs like aider and applaudir for instance, traditionally considered to belong to different classes of verbs. Regularity such as this allows us to ask what it is that changed within the system underlying the derivation of a verb phrase to produce the valency change.

18 See for example, Brunot (1966), Brunot et Bruneau (1969, §272), Colin (1897), Clédat (1887, §432-434), Étienne (1895, §328-330), Fournier (1998: 96-101), Gougenheim (1974: 127-129), Haase (1914, §58-62), Nyrop (1930, vol.6, §163-190).

The intuition behind the present study was that the replacement of an IO by a DO appeared to be somewhat systematic in two-place verbs, despite the way valency change is framed in most French historical grammars. Drawing mainly on the data provided in these grammars (see note 18), we established a firm list of twenty two-place verbs whose indirect object was replaced by a direct object and which crucially underwent no shift in meaning.

(1) Two-place verbs having undergone a change in argument expression: IO → DO

The verbs in (1) are the focus of our study and were selected based on sufficient evidence of the change and the absence of a shift in the meaning of the verb. In section 2.1, the argument realization of each one is discussed and illustrated by examples. 19

19 Note that we do not consider the productive dative/accusative alternation in the possessor or goal argument of three place verbs of transfer and communication in Old French to be related to the dative/accusative alternation involving aider-type verbs, contra Herslund 1980; see chapter four, section 4.2.1.2. In other words, the loss of the double object construction is not considered to be related to the valency change in aider-type verbs, substantiated by the fact that the double object construction disappears from the texts some two hundred years before the valency change in aider-type verbs even begins.

Similarly, we do not consider the valency change in aider-type verbs to be related to the variation and change attested in the dative/accusative-marked subject of impersonal constructions involving psych verbs such as chaloir ‘to matter/to bother’, convenir ‘to be convenient’, doloir ‘to regret/to deplore’, etc., as described in Mathieu (2006). We apply Eythórsson’s (2002) “Dative Sickness” analysis of Icelandic to Old French according to which the dative/accusative variation is driven by thematic considerations: accusative experiencers are replaced by dative

Section 2.2 discusses a number of other two-place verbs in French that have also undergone a change of valence but which do not correspond to the kind of change that characterizes the verbs in (1). For some, sufficient evidence of change is not available. For others, the change of valence is associated with a shift in the meaning of the verb. Potentially more problematic, however, are verbs which appear to have undergone an opposite change. These verbs take both a direct and an indirect object at earlier stages in the language, but, unlike the aider-type verbs, have retained their indirect object and no longer occur with a direct object. If our hypothesis is correct, and a change in the preposition à removed the necessary licensing environment for certain indirect objects, then we expect to see a unidirectional change. Verbs which appear to have undergone an opposite valency change would present potential counter examples to our claim. We take these verbs up in some detail in section 2.2.3 in order to show that they do not in fact pose a problem for our analysis and are rather the effects of various other possible causes.

2.1 Verbs under consideration

This section gives a brief description of the evolution of each of the twenty verbs in (1) identified as having undergone the same change in argument realization. Verbs are often compatible with more than one construction, or argument structure, and these verbs are no different. This study only considers their evolution in the following basic two-place construction that may or may not allow an optional finite/non finite clause:

(2) Subject-Verb-Object (que/ de / à...)

Aider, empêcher, persuader, prier, and supplier can occur with an optional finite/non finite clause. In these cases, the change in the expression of the object holds whether or not it occurs with the optional clause. Note that Lüdi (1978) considers requérir ‘request/ require/ demand’ as having undergone the same valency change as prier and supplier. We do not include it in our

experiencers based on morphosyntactic leveling of an irregular paradigm (the accusative) in favour of a more productive and frequent pattern (the dative).

53 study since the verb never occurs in a clearly two-place construction as described in (2). Rather, the change seems to be limited to requérir (à) quelqu’un que/ de...

Some verbs can also occur in three-place constructions, reflexive and pronominal constructions, but the evolution of the expression of the object in these instances is not taken into account. Commander, épargner and contredire occur in three-place constructions for which the direct object or clausal complement and the IO are obligatory. In these constructions, the IO never alternates with a DO and does not change. As for reflexive and pronominal constructions, given that the presence or absence of past participle agreement is the only way to distinguish between direct and indirect object pronouns, they are not informative.

In what follows, we illustrate the meaning of each of the verbs in (1) and their use with both a direct and an indirect object. Examples are taken from various periods: Medieval French (14th and 15th centuries), Pre-classical French (16th century), Classical French (17th and 18th centuries), and Modern French (19th and 20th centuries).20 The rich resource of dictionaries, historical grammars, scientific articles, and the formidable electronic data bases that are now available have been enormously helpful in documenting how these verbs were used before and how they are employed today. The reader is referred to the bibliography for a complete list of the resources that were consulted.

2.1.1 Aider

In Medieval and Classical French, aider can be followed either by a direct object or an indirect object and these can be both animate and inanimate (3a-f). Aside from IOs denoting an event (those containing an action noun), in (3j), the IO construction is no longer used in Modern Standard French (see also chapter one, footnote 2). The verb has undergone no apparent change of meaning and has always had the sense of ‘to help, to contribute toward something’. Certain grammarians have argued that the difference between the indirect and direct constructions corresponds to a difference in meaning, but empirical evidence does not confirm this distinction, as summarized in the TLFi:21

20 See section 3.1.2 for a explanation outlining why Old French is not taken into account. 21 See also Grevisse (1961:287-9).

Le verbe aider hésite entre le régime dir. et le régime indir. Certains grammairiens, dont l’Acad., estiment qu’à cette différence de const. correspond une différence de signif. (aider qn jouirait d’une plus large ext. et pourrait notamment servir à designer une aide morale; aider à qn ne pourrait exprimer qu’une aide matérielle ou phys. de caractère momentané). L’usage ne confirme guère cette distinction. En revanche, il y a lieu de souligner que la constr. aider à qn, habituelle en a. fr. et fréq. dans la langue classique, est auj. très vieillie. (TLFi, aider, I.B.1, Rem.)

(3) a. qui ayda bien aux Flamans à avancer l’oeuvre

‘who helped the Flemish to advance the work’

(Commynes, Mémoires t.2, 1489: vi, viii, 301; BTMF)

b. les Romains ne leur aideroient point en leurs afferes

‘the Romans would not help them with their concerns’

(Simon de Phares, Recueil des... astrologues, c.1494-1498, f° 93 v°; DMF1)

c. et [le conte de Warvic] y aida bien le duc

‘and the count of Warwick indeed helped the duke there’

(Commynes, Mémoires t.1, 1489, iii, iv, 192; BTMF)

d. et aussi il aidoit sa fortune par courtoisie et bel lengage

‘and also he contributed to his fortune through courtesy and good language’

(Bersuire, Les Decades de Titus Livius, 1354, I, 9, 60; BTMF)

e. se elle meismes ne met les mains a la paste [...] pour les aidier.

‘if she herself doesn’t get her hands dirty ... in order to help them.’

(Pizan, Le Livre des trois vertus, 1405, I, 8, 31; BTMF)

f. la chrestienté enseigne que la foy et le salut est par l’ouye, et que la veuë

y nuict plus qu’elle n’y ayde.

‘Christianity teaches that faith and salvation are attained by listening, and that

seeing harms more than it helps.’

(Charron, De la Sagesse, Trois Livres, 1601, p.89; Frantext, ARTFL)

Modern French:

g. c’est à toi que je dis, père, aide tes enfants!

‘it’s you I’m telling, father, help your children!’

(Claudel, P., Les Choephores, 1920, p.928; Frantext, ARTFL)

h. je lui tendis la main [...] pour l’aider à se relever.

‘I held out my hand to him... to help him get up.’

(Giraudoux; Le Petit Robert)

i. si cet ordre […] a un but, c’est d’aider le libre progrès

‘if this order, … has a goal, it’s to facilitate free progress’

(Michelet, J., Le Peuple, 1846, p.359; TLFi)

j. Le débat aiderait à la compréhension du problème.

‘Discussion would contribute towards an understanding of the problem.’

(Collins Robert)

2.1.2 Applaudir

Before the 17th century, the complement of applaudir, either animate or inanimate, always appears as an IO (4a-c). Subsequent centuries however, show an alternation between the IO and an analogous DO (4d-f), noted, for example, in Féraud’s Dictionaire critique de la langue française:

Il [applaudir] régit la prép à des persones; aplaudir aux Comédiens, aux Acteurs. Il en est qui pensent qu'on dit également bien, tout le monde lui aplaudit, ou l'aplaudit.

M. de Vailly le dit ainsi dans le Rich. Port. — L'Acad. ne met d'exemple de ce régime direct qu'avec aplaudir, signifiant aprouver. "Chacun l'a aplaudi de cette bonne action. Il a fait une harangue que tout le monde a aplaudie. — Les Poètes le font indifféremment actif ou neutre.

The indirect object has all but disappeared from use in Modern French, save instances where it is inanimate and refers to an event or a state (4j). In these cases, the verb must appear with an IO; applaudir à quelque chose has the sense of strongly approving something, without any show of applause. The change in argument realization did not bring about a change in meaning; followed by an object, applaudir has always meant something along the lines of ‘to approve, to show one’s approval’

(4) a. ils escrivoient des comedies, ou applaudissoient à ceux qui en joüoient devant

eux.

‘they would write plays, or they would applaud those who were acting in front

of them.’

(Balzac, J.L. Guez, Dissertations politiques, 1654, p.430; Frantext, ARTFL)

b. Il ne luy applaudissoit pas seulement, mais il se rompoit les mains à luy

applaudir.

‘He didn’t just applaud him, he went wild applauding him.’

(Balzac J. L. Guez, Le Barbon, 1648, p.710; Frantext, ARTFL)

c. en aplaudent a la gloire persienne contre le nom rommain

‘while hailing the Persian glory against the Roman name’

(Bersuire, Les Decades de Titus Livius 1354, I, 9, 33; BTMF)

d. Je crois voir des farceurs qu’applaudissent des fous

‘I believe I see comics whom fools are applauding’

(LaMotte, A. de, Fables, 1719, p.171; Frantext, ARTFL)

e. s’il avoit observé qu’on ne l’applaudît pas.

‘if he had noticed that they weren’t applauding him.’

(Moliere, Le Misanthrope, 1667, p.486; Frantext, ARTFL)

f. Je ne dis pas un mot que chacun n’applaudisse

‘There isn’t a word I say that goes unapplauded’

(Boursault, E., Esope et la Cour, 1701, p.119; Frantext, ARTFL)

Modern French:

g. Applaudir un acteur, un orateur

‘To applaud an actor, an orator’

(Le Petit Robert)

h. si bien que les acteurs l’applaudirent et lui prophétisèrent un succès.

‘and so the actors applauded him and told him he would be a success’

(Gautier, T., Le Capitaine Fracasse, 1863, p.178; Frantext, ARTFL)

i. Et Lafayette, avec les idéologues, applaudit le changement.

‘And Lafayette, along with the ideologues, praised the change’

(Adam, L’Enfant d’Austerlitz, 1902, p.357; TLFi)

j. J’applaudis à votre initiative.

‘I applaud your initiative’

(Le Petit Robert)

2.1.3 Assister

In Medieval French, assister à quelqu’un/quelque chose not only had the meaning of ‘to lend help or to assist someone/something’, it could also mean ‘to be present at an event’, ‘to stand by someone’, or ‘to frequent someone’. This study has only taken the former meaning into consideration since it is only in this sense that we observe the replacement of the IO by an analogous DO. Assister still occurs with an IO when it is used in the locative sense.

Before the 17th century, the complement of assister ‘to help’ alternates between an IO and a DO without any change in meaning (5a-f), while only a DO is permitted in Modern French (5g,h).

(5) a. Les autres assistent soigneûsement à leurs Pères et Mères infirmes.

‘The others carefully help their ailing fathers and mothers.’

(St. Fr. de S., in Féraud, 1787-88, Dictionaire critique de la langue française)

b. ains de luy assister de ma personne et de mes moyens pour l’accomplissement

de son oeuvre.

‘and so to help him personally and financially to accomplish his work. ’

(Garasse le Pere F., Doctrine Curieuse Beaux Esp., 1623, p.152; Frantext,

ARTFL)

c. personne ne voulant aller à cette guerre. [...] il fit mourir quelques-uns de

ceux qui [...] ne vouloient pas assister à leur republique en cette grande

adversité.

‘No one wanting to go to this war... he had a few killed of those who

... didn’t want to come to the aid of their republic in this time of great

adversity.’

(Coeffeteau, Histoire Romaine, 1623, p.237; Frantext, ARTFL)

d. les admonitions que les evesques faisoyent au duc breton, de n’assister de ses moyens le comte rebelle ‘the reprimands that the Breton duke received from the bishops for not coming on his own to the aid of the rebel count’ (Fauchet, Declin Maison de Charlemagne, 1602, p.46 in Frantext, ARTFL)

e. Et pource que autant difficile seroit l’assister en ce chasteau

‘and because it would be so difficult to help her in that castle’

(Helisenne de Crenne, Les Angoysses douloureuses..., 1538, p.451; ARTFL)

f. tout puissant, de ta force assiste ma foiblesse

‘all mighty, with your strength help my weakness’

(Aubigné, Les Tragiques, 1616, p.45 in Frantext, ARTFL)

Modern French :

g. Nous étions trois, au début, pour assister notre patron

‘We were three, at first, to help our employer’

(G. Duhamel; Le Petit Robert)

h. les fonctionnaires locaux, [...] les assistaient et contrecarraient leurs

adversaires à tous moments.

‘the local public servants ... helped them and foiled their opponents at every

turn.’

(Lidderdale, Le Parlement fr., 1954, p.27; TLFi)

2.1.4 Commander

This study only considers the verb commander in two-place constructions in the sense of to command, that is, being in charge of a person or group of persons of lower social or military rank, or in the sense of dominating a location. This verb was usually followed by an IO (6a-c), but begins to occur more frequently with a DO from the 16th century (6d,e,f). IOs eventually disappeared from use, save in contexts where the object refers to a part of the subject’s body or some emotional or intellectual aspect of the subject’s life (6j).

Note that we do not consider commander when it used in three-place constructions, those, for example, that denote the transmission of an order or the recommendation of someone (commander à quelqu’un de faire quelque chose ‘to order someone to do something’; commander quelqu’un à quelqu’un d’autre ‘to recommend someone to someone else’). These latter uses of commander pattern with other three-place verbs (see chapter one, section 1.3.1) and did not undergo a valency change.

(6) a. Le Prince commande à ses sujets, le pere à ses enfants, le maistre à ses

valets, le Capitaine aux soldats.

‘The prince commands his subjects, the father his sons, the master his

servants, the captain the soldiers.’

(Dictionnaire de l’Académie Française, 1694)

b. mais non pas qu’il leur commande, ny qu’elles luy obeyssent.

‘but not that he commands them, or that they obey him.’

(Charron, De la Sagesse, Trois Livres 1601, p.67; Frantext, ARTFL)

c. Les seules bonnes histoires sont celles qui ont esté escrites par ceux mesmes

qui commandoient aux affaires, ou qui estoient participans à les conduire

‘the only good accounts are those which have been written by the very ones

who were in charge of business, or who participated in directing it’

(Montaigne, Les Essais - Book II, 1595, p.418; Frantext Encore, ARTFL)

d. le Général Gallas commanda le Baron De Fernamond, [...], et le Colonel

Bamberg

‘General Gallas had authority over the Baron of Fernamond... and Colonel

Bamberg’

(Arnauld d’Andilly, Memoires, t.1, 1667, p.352; Frantext, ARTFL)

e. et qui étoient ceux qui les commandoient par files et par demi-files

‘and who were the ones who were commanding them by file and half-file’

(Arnauld d’Andilly, Memoires, t.1, 1667, p.332; Frantext, ARTFL)

f. celuy qui commandoit la garnison en absance du gouverneur eust permis

ledict armement des quartiers.

‘the one who was commanding the garrison in the governor’s absence had

allowed the said arming of the neighbourhoods.’

(Peiresc, Lettres t. 3 1634-1637, p.30; Frantext, ARTFL)

Modern French:

g. Aujourd’hui, il n’y a que des gens médiocres qui ont à commander des gens à

peu près de leur force.

‘Today, you only find mediocre people who have under their orders people of the same calibre.’

(Merimée, Lettres à la comtesse de Montijo, t.1, 1870, p.262; TLFi)

h. Il n’aime pas qu’on le commande

‘He doesn’t like it that they have authority over him’

(Le Petit Robert)

i. Commander les opérations, un siège

‘to direct operations, a siege’

(TLFi)

j. Je ne commande plus à mes mains. Il [Pitteaux] commandait encore à son

visage, il fit sa moue, sa terrible moue.

‘I no longer have control over my hands. He (Pitteaux) could still control his

face; he put on his pout, his terrible pout.’

(Sartre, Le Sursis, 1945, p.116; TLFi)

2.1.5 Congratuler

Congratuler, in the sense of ‘to congratulate someone’ occurs relatively infrequently in the corpus. Most occurrences of the verb are with a DO (7c), but IO complements are nevertheless noted in the metalinguistic sources (7a,b). We assume therefore, an alternation between an IO and a DO complement without any appreciable difference in meaning. Only the direct construction is used in Modern French (7d).

(7) a. Donner occasion à quelqu'un de congratuler à son ami

‘Give someone occasion to congratulate his friend’

(Nicot, 1606, Thresor de la langue française)

b. En ce jour où l’Eglise est occupée à leur congratuler

‘On this day when the Church is busy congratulating them’

(Boss., Toussaint, 1649; Colin 1897:151)

c. M. De Voltaire a écrit à M. D’Alembert pour le congratuler sur le courage qu’il a eu ‘Mr. De V. wrote to Mr. D’A. to congratulate him on the courage he had’ (Bachaumont, Memoires Secrets Depuis 1762, 1763, p.180; Frantext, ARTFL)

Modern French:

d. Congratuler l’heureux père sur la naissance de sa fille

‘To congratulate the lucky father on the birth of his daughter’

(Le Petit Robert)

2.1.6 Contrarier

Contrarier is a two-place verb that means ‘to impede or oppose. It can signify the verbal expression of one’s opposition towards something, or an attempt to thwart or impede the progress of something. The verb can also have a psychological value, meaning to annoy or

63 bother someone. Until the classical period, contrarier occurs either with an IO or a DO (8a-g), but only the DO is used in Modern French (8h,i,j).

(8) a. ce sont les commandements de mon pere ausquels mon honneur ne permet

que je contrarie.

‘these are my father’s orders which my honour forbids me to oppose.’

(Urfé, L'Astrée t.1, 1607, p.404; Frantext, ARTFL)

b. de capter la bienvueillance du roy Alexandre mieulx que lui contrarier

‘better than to oppose King Alexander is to win his favour’

(Simon de Phares, Recueil des plus célèbres astrologues, c.1494-1498,

f° 63 r°; DMF1)

c. la petitesse de mon rude engin [...] contrariant à mon vouloir et au desir

ardent que j'’avoye

‘the pettiness of my primitive ruse... running counter to my will and the ardent

desire that I had’

(Robertet, Oeuvres, Epîtres, XIX, 1451, p.176; BTMF) d. En cela point n’y contrarie.

‘In this, nothing contradicts it.’

(Miracle de st-Nicolas et d’un juif, 1480, p.153; BTMF)

e. Polygame, qui jamais ne contrarioit personne et estoit de tous bons accords

‘Polygame, who never annoyed anyone and who was always agreeable’

(Fail, Les Baliverneries d’Eutrapel, 1548, p.689; Frantext Encore, ARTFL)

f. et sçavoir s’ils s’acordent avec les autheurs approuvez, ou s’ils les contrarient

‘and to know if they are in agreement with the approved authors or if they are

against them’

(Fauchet, Fleur Maison de Charlemagne, 1601, p.33; Frantext, ARTFL)

g. Vous contrariez tout ce qu’on dit.

‘You contradict everything they say.’

(Dictionnaire de l’Académie française, 1694)

Modern French:

h. Était-ce un homme qu’on avait habitude de contrarier dans ses desseins?

‘Was it a man with whose plans they were in the habit of interfering?’

(Claudel, Le Soulier de satin, 1944, 1re part., 1re journée, 5, 957; TLFi)

i. Paul cherche à contrarier sa soeur.

‘Paul seeks to annoy his sister.’

j. La tempête contrariait la marche du navire

‘The storm was impeding the ship’s progress’

(Le Petit Robert)

2.1.7 Contredire

In both Medieval and Modern French, contredire has the general sense of ‘to speak out against someone or something, to contradict, contest’ and occurs either with an IO or a DO (9a-f). Féraud notes that the IO was still used in the 17th century, but that only a DO is used by his contemporaries:

Autrefois on employait ce verbe neutralement avec le régime du datif. Les Dieux ont prononcé: loin de leur contredire, C’est à vous a passer du côte de l’Empire. Rac. ‘Elles ne contredisent point au témoignage extérieur des Écritures.’ Boss. — L'Acad. a dit aussi, dans ses Sentimens sur le Cid, ‘Ce discours nous paroit contredire à celui que le Poète lui fait tenir maintenant.’ Leibnitz dit toujours, contredire à... On ne le dit point aujourd’hui. On dirait, loin de les contredire, etc. Elles ne contredisent point le témoignage, etc. Ce discours parait contredire celui, etc. (Féraud, Dictionaire critique de la langue française)

Although the TLFi cites two occurrences of contredire followed by an inanimate IO, the Petit Robert considers it a strictly transitive verb (9g,h). This concurs with native speaker judgements that only the transitive construction is acceptable today. Until the 16th century, some uses of contredire with a DO tend to mark a slightly more affected object, having the sense of ‘to oppose, refuse, or impede someone or something’. These uses were also included in the quantitative study. Any trivalent uses of the verb (i.e., contredire quelquechose à quelqu’un ‘to forbid someone something’) were excluded since they show no variation in the expression of the indirect object and since they are not used today.

(9) a. Cestui, au moyen de sa science des estoilles et de sa prudence contredist au

roy Alixandre.

‘This man, through his knowledge of the stars and his wisdom, spoke out

against King Alexander.’

(Simon de Phares, Recueil des plus célèbres astrologues, c.1494-1498, f° 55

v°; DMF1)

b. Si Musigene eust osé luy contredire

‘If M. had dared to contradict him’

(Sorel, Polyandre I, 1648, p.368-369; Brunot, vol.3(2), 546)

c. Tiercement se le fait contredist a sa bonté come de dampner Saint Pierre et

saulver Judas le traytre

‘Thirdly, if the act contradicts his goodness such as by damning Saint Peter

and offering salvation to Judas the traitor’

(Somme abr., c.1477-1481, p.161; DMF1)

d. par Dieu je n’y contrediray ja

‘by God I will never oppose it’

(Gerson, Sermon pour la Noël, 1404, p.305; DMF1)

e. se le serf et la serve d’aucun seigneur se marient ensemble contredisant le

seigneur

‘if the male and female servant of a landowner get married, opposing the

landowner’

(Du Sacrement de mariage, c.1477-1481, p.50; DMF1)

f. Diane alors ordonna à Delphire de respondre à ce que Tomantes avoit dit, si

toutesfois elle y vouloit contredire quelque chose.

‘Diane then ordered Delphire to respond to what Tomantes had said, if, however, she wanted to say something against it.’

(Urfé, L’Astrée t. 4, 1619, p.336; Frantext, ARTFL)

Modern French :

g. Prendre plaisir à contredire tout le monde

‘To take pleasure in contradicting everyone’

(Le Petit Robert)

h. Les événements ont contredit ses prédictions, ses espérances

‘The events have contradicted his predictions, his hopes’

(Le Petit Robert)

2.1.8 Dominer

Much like commander, dominer ‘to dominate, command, have power over’ is widely used with an indirect object in Medieval and Pre-classical French (10a-c). Dominer sur occasionally appears in the same contexts and only one occurrence of a direct object (10d) was located in the references we consulted (DMF1, Huguet). In the first edition of the Dictionnaire de l’Académie Française (1694), however, there is no mention of dominer à quelqu’un/quelque chose. Dominer sur is the preferred form, with a small note concerning its use with a DO: “On s'en sert quelquefois activement. Il domine tout un pays, ou un grand pays.” A century later, Féraud

67 observes that dominer can freely alternate between dominer sur and the direct construction. He adds that some still use dominer followed by an IO, but that the IO is undesirable:

Figurément, il se dit assez indiféremment ou comme actif avec le régime absolu, ou comme neutre avec la prép. sur. “Dominer les passions ou dominer sur les passions. Cette montagne domine la ville, ou domine sur la ville.” La Touche préfère le 2d: j'aimerais mieux le 1er. L’Acad. les mets tous deux sans remarque. — * Le P. La Rûe ne met ni l’un ni l’autre de ces deux régimes: il done la préférence à la prép. à. "Votre salut, nécessité qui domine à toutes les aûtres. “Comme tous les membres du corps ont besoin les uns des aûtres, aussi ont-ils besoin d'un chef, qui domine à tous les aûtres.” On doit dire dans cette ocasion, domine sur. L'actif n'y serait pas bon: la prép. à n’y vaut rien. (Féraud, Dictionaire critique de la langue française)

In Modern French, dominer is only followed by a direct object.

(10) a. Et pour ce, cellui ouquel tel puissance dommine doit par droit estre appellé

sires et maistres des autres hommes pour cause des nobles euvres de vertu qui

lui font eschever toute villenie.

‘And because of this, the man who is guided by such power must rightly be called

lord and master of other men for the noble works of virtue which have him put an

end to all wickedness.’

(Pizan, Livre de la Paix, 1412-1413, p.174-175; DMF1)

b. Ceulx qui ont accoustumé de dominer aux aultres... ne sont point cruelz

envers ceulx quilz veinquent.

‘Those who are accustomed to commanding others... are not cruel towards

those whom they conquer.’

(Seyssel, trad. de Thucydide, 1527, I, 8 (23r); Huguet, Dictionnaire de la

langue française du seizième siècle)

c. La royne des Amazones, qui avoit nom Thalestris, et qui dominoit à tout le

pays qui estoit entre les rivieres de Phasis et de Thermodon, le vint trouver.

‘The queen of the Amazons, who was named Thalestris, and who reigned over the

entire region between the Phasis and the Thermodon rivers, came to find him.’

(Amyot, trad. de Diodore, 1554, XV, 8; Huguet, Dictionnaire de la

langue française du seizième siècle)

d. Car ainsin firent les bons Romains, et par ce dominerent en toute la

monarchie du monde et donnerent loiz desquelles encores nous usons.

‘For the good Romans did it this way, and so had control of the world’s

monarchy in every way and established laws that are still used.’

(La Sale, Jehan de Saintré, 1456, 77; DMF1)

Modern French:

e. Despote, tyran qui domine un peuple.

‘A despot, tyrant who rules a people.’

(Le Petit Robert)

f. Sa tête dominait son coeur

‘His head ruled his heart’

(Stendhal; Le Petit Robert)

2.1.9 Empêcher

Empêcher, in the general sense of ‘to prevent someone from doing something or to prevent something from happening’, could be followed by either a direct or an indirect object in our corpus (11a-f). In fact, the TLFi cites examples with IOs (followed by an infinitival phrase) up until the early 20th century. The Petit Robert reflects current use, however, showing empêcher used only with a DO (11g-i).

(11) a. ilz empeschent l’un a l’autre et ne puelent leurs cops frapper

‘they are checking one another and are unable to make a hit’

(La Sale, Salade, c.1442-1444, p.240; DMF1)

b. Cela n’empêcha pas à ces opiniastres D’adorer Belphegor

‘That didn’t prevent those tenaciously opinionated people from adoring

Belphegor’

(Racan, Psaumes, 1660, p.278; Brunot, vol. 3(2), 546)

c. la neccessité de gaigner leur vie (...) leur a par aventure empechié de savoir...

‘the necessity of earning a living ... by chance prevented them from knowing’

(Pizan, Trois vertus W.H., c.1405, p.207; DMF1)

d. Il croyoit que l’archiduc étoit de ses amis et qu’il ne lui empêcheroit pas

‘He believed that the archduke was a friend and that he wouldn’t hinder him’

(Malherbe, Lexique, III, 17th c., p.151; Goyens, 1998:482)

e. Les esliz d’Israel empedeçat

‘He stood in the way of the chosen ones of Israel’

(Psautier d’Oxford, 1120, 77, 35; Goyens, 1998:481)

f. Quant aux armes... ils estoient si accoustumez à les avoir sur le dos qu’elles

ne les empeschoient non plus que leurs membres

‘As for weapons... they were so accustomed to wearing them on their backs

that they didn’t bother them any more than their limbs’

(Montaigne, II, 16th c., 9, ii, 104; Huguet Dictionnaire de la langue française

du seizième siècle)

Modern French:

g. Pour empêcher les autres de travailler

‘To prevent the others from working’

(Mac Orlan; Le Petit Robert)

h. Ça ne l’empêche pas de dormir.

‘He doesn’t lose sleep over it.’

(Le Petit Robert)

i. Car que sert d’interdire ce qu’on ne peut pas empêcher ?

‘For what good is it to forbid that which one cannot prevent ?’

(Gide; Le Petit Robert)

2.1.10 Ennuyer

Until the 17th century, ennuyer could be followed either by a direct or indirect object, having the sense of ‘to bore, weary, bother, put out, annoy’ (12a-e). In Modern French however, it only takes a direct object (12f,g). Occurrences of ennuyer used in the reflexive construction were not considered since there is no morphological distinction between direct and indirect reflexive pronouns.

(13) a. que au chevalier ennoya le grant delay de son pardon.

‘that he knight was irritated by the long delay in his pardon.’

(La Sale, Salade, c.1442-1444, p.106; DMF1)

b. Car bien croy qu’il li enuia

‘For I believe that he upset her’

(Machaut, Le Jugement dou Roy de Navarre, 1349, p.264; DMF1)

c. ou que ce discours luy ennuyast

‘or that this speech bored him’

(Urfé, L’Astrée, I, 1607, p.17; Frantext, ARTFL)

d. Toutesfois pour n’ennuyer ce berger

‘However, so as not to annoy this shepherd’

(Urfé, L’Astrée, I, 1607, p.75; Frantext, ARTFL)

e. affin que les loux n’ayent nul vent de luy qui les ennuye

‘so that the wolves wouldn’t pick up his scent, which bothers them’

(Phebus, Livre de Chasse, 1387, p.239; BTMF)

Modern French :

f. Conférencier qui ennuie son auditoire

‘A speaker who bores his audience’

(Le Petit Robert)

g. Sans crainte d’ennuyer votre Excellence

‘Without fear of bothering Your Excellency’

(Dumas père, Monte-Cristo, t.1, 1846, p.633; TLFi)

2.1.11 Épargner

In Medieval French, the complement of épargner alternates between a direct and indirect object when the verb is used in two-place constructions with the meaning of ‘to spare, to save’ (13a-c). Only a direct object is used in these contexts in Modern French (13d,e). Note that trivalent uses of épargner, i.e., épargner quelquechose à quelqu’un ‘to spare/save somebody something, pattern with verbs of transfer like ôter (see chapter one, section 1.3.1) in not undergoing any valency change. These latter uses of the verb do not therefore figure in our quantitative study.

(13) a. et occioient [...] ce qu’ils trouvoient de sexe masculin, espargnans aux

femmes seulement.

‘and they killed ... whom they could find in the way of men, sparing only the

women.’

(Le Baud, Histoire de Bretagne, 1505, ch.3; Huguet, Dictionnaire de la

langue française du seizième siècle)

b. Et ainsi espargnier a pecune seroit mal

‘And so saving money would be bad’

(Oresme, Le Livre de Ethiques d'Aristote, c.1370, p.403; BTMF)

c. et a la verité, c’est chose naturele de esparnier la personne miserable.

‘and in truth, it’s a natural thing to spare the poor person.’

(Evrart de Conty, Le Livre des eschez amoureux moralisés, c.1400, p.565;

DMF1)

Modern French:

d. Épargner son adversaire

‘To spare one’s adversary’

(Le Petit Robert)

e. Épargner une somme d’argent

‘To save a sum of money’

(Le Petit Robert)

2.1.12 Étudier

The complement of étudier ‘to study’ was most commonly an indirect object in Medieval and Pre-classical French (14a-e), although it could also take a direct object in the same contexts (14f,g).22 This verb only takes a DO in Modern French (14h,i). The reflexive three-place construction, i.e., s’étudier à quelque chose ‘to apply one’s self to something’, was also used in similar contexts, but did not undergo a change in valency and is therefore not taken into account in the quantitative study.

(14) a. Et si n’est pas par raison vraysemblable que tous ceuls qui ont estudié as arts

et as sciences eüssent ignoré telle chose

‘And if it is not for some plausible reason that all those who studied the arts and sciences had been unaware of such a thing’

(Oresme, Le Livre de Ethiques d’Aristote, c.1370, p.116; DMF1)

b. Mais j’ai estudié au livre de Jason.

‘But I studied the book of Jason.’

(Guesclin, 8960; Littré, Dictionnaire de la langue française)

c. J’estudiay aux loix deux ans.

‘I studied the laws for two years.’

(B. de la Grise, trad. de Guevara, l’Orloge des princes, I, 3; Huguet,

Dictionnaire de la langue française du seizième siècle)

22 Note that some transitive uses of étudier, observed in Brunot (vol.3(2), 546), are also excluded from the quantitative study: “Bouhours commenta [...] étudier avec un régime sans préposition existe, mais souvent avec le sens particulier d’affecter : et je n’étudiai cette douleur menteuse Qu’à cause qu’en effet j’étois un peu honteuse (Corn., I, 228, Mél., v. 1423).” The object in this particular meaning, which has disappeared from use, does not appear to alternate with an IO.

d. La plupart n’étudient pas aux choses pour lesquelles il faut étudier

‘Most of them don’t study the things for which one should study’

(Malherbe, 17th c., II, 559; Brunot, vol.3(2), 546)

e. Ceux qui ont le mieux estudié à la connoissance de l’ame

‘Those who have best applied themselves to the knowledge of the soul’

(de la Pinelière, Le Parn. ou Critique des Poètes, 17th c., 19; Brunot,

vol.3(2), 546)

f. par gens qui avoient estudié livres de telle science

‘by people who had studied books on such branches of knowledge’

(Oresme, Le Livre de Ethiques d'Aristote, c.1370, p.99; DMF1)

g. Aucun pourroit dire que ceste science n’est pas si necessaire, car au temps

passé plusieurs roys et princesses ont très bien gouverné qui oncques n’estudierent

politicques

‘A person could say that this knowledge is not so necessary, for in the past, a

number of kings and princesses who never studied politics governed very well’

(Oresme, Prol. 14th c.; Littré, Dictionnaire de la langue française)

Modern French

h. Je voudrais que les jeunes filles étudiassent le latin comme les petits garçons

‘I would like young ladies to study Latin like little boys’

(Stendhal, Amour, 1822, p.217; TLFi)

i. Ces montagnes sont les lieux du monde les plus favorables pour étudier la

nature

‘These mountains are the most favorable places in the world for studying

nature’

(Bern. de St-P., Harm. nat., 1814, p.213; TLFi)

2.1.13 Favoriser

Favoriser ‘to support, to treat someone favourably’ alternates between taking an indirect and a direct complement in Medieval French (15a-f). By the 17th century, the DO is strongly preferred to the IO. In the Commentaires (vol.2, 799-800), Corneille observes that “favoriser gouverne toujours l’accusatif” and the Académie Française confirms it further on in the context of several verbs that are best used with a direct object: “On a approuvé cette Remarque tant pour le verbe servir, que pour prier et favoriser.”

Modern French only accepts a DO complement for this verb (15g-i).

(15) a. Mais en chose inique et mauvaise favoriser a celluy que on dit son amy, [...]

c’est chose dampnable

‘But in wrong and unjust things, to show favour to he whom one calls a

friend,... is a detestable thing’

(Fillastre, Le Traitie de conseil, 1472, p.251; BTMF)

b. quand fortune leur favorisoit tant que ensemble les faisoit deviser

‘when fortune showed them favour as much as it divided them as a group’

(Cent Nouvelles Nouvelles, c.1456-1467, p.546; DMF1)

c. pour favoriser aux sincères intentions d’un amant

‘in order to lend help to a lover’s sincere intentions’

(Yver, Le Printemps (extraits) 1557, p.1242; Frantext Encore, ARTFL)

d. certain nombre de gens d’armes qui avoient favorisié et favorisoient le conte

d’Armaignac

‘a certain number of soldiers who had supported and were supporting the

count of Armaignac’

(Fauquembergue, Journal t.1, 1421, p.166; BTMF)

e. Se je vouloye complaire aux hommes en leurs vices et les y favoriser

‘If I wanted to show my complicity with men and support them in their vices’

(Fillastre, Le Traitie de conseil, 1472, p.252; BTMF)

f. elle favoriseroit mon opinion

‘she favoured my opinion’

(Marguerite de Navarre, L’Heptaméron, 1559, p.718; Frantext Encore,

ARTFL)

Modern French :

g. L’examinateur a favorisé ce candidat

‘The invigilator favoured this candidate’

(Le Petit Robert)

h. Les événements l’ont favorisé.

‘The events were in his favour.’

(Le Petit Robert)

i. Le maire a favorisé ce plan d’urbanisme

‘The mayor favoured this city plan’

(Le Petit Robert)

2.1.14 Insulter

The complement of insulter ‘to insult, to affront someone or something’ alternates between an IO and a DO through the 17th and 18th century (16a-e). Corneille notes, for instance, “Insulter est un mot généralement receu. On dit, Insulter quelqu’un, insulter à quelqu’un, Insulter contre quelqu’un.” (Commentaires vol.2, 898). The IO is rarely used however in Modern French, and only ever in limited contexts such as that in (16i). Both the DO and the IO correspond to the same general meaning and are used in similar contexts, save the sense of physically attacking someone or something. This latter meaning, current until the 17th century, was only used with a DO (16d); we have found no occurrence of an IO complement in this sense.

(16) a. en prenoient lieu d’insulter aux patriciens

‘they took the opportunity to insult the aristocrats’

(Vertot, Histoire des Revolutions, 1719, p.462; Frantext, ATILF)

b. Un autre, pour lui insulter, disoit...

‘Another, in order to insult him, said...’

(Fénelon, Les Aventures de Télémaque, 1699, p.322; Frantext, ARTFL)

c. qu’il avoit insulté au malheur du poëte Theophile

‘that he had jeered at the misfortune of the poet Theophile’

(Sorel, La Bibliothèque françoise, 1664, p.107; Frantext, ARTFL)

d. Dans une grande colère, le corps se trouve plus prêt à insulter l’ennemi et à

l’abattre

‘In great anger, the body is more apt to attack the enemy and to bring him

down’

(Bossuet, De la Connaissance de Dieu, 1704, p.142; Frantext, ATILF)

e. Il l’insultoit dans tous ses discours

‘He insulted him in all that he said’

(Vertot, Histoire des Revolutions, 1719, p.447; Frantext, ATILF)

Modern French:

f. Insulter une femme par des propos offensants

‘To insult a woman by making indecent proposals’

(Dictionnaire de l’Académie Française, 1935; TLFi)

g. Mon père les insulta dans sa colère

‘My father insulted them in anger’

(Saint-Exupéry, Citadelle, 1944, p.857; TLFi)

h. j’insulte toute sa vie en déclarant que je veux retourner au métier comme nos

grands parents!

‘I’m insulting his whole life by declaring that I want to go back to the trade like

our grandparents!’

(Vallès, J. Vingtras, Enf., 1879, p.378; TLFi)

i. Le luxe des riches insulte à la misère des déshérités.

‘The abundance of the rich is an insult to the poverty of the disinherited.’

(Le Petit Robert)

2.1.15 Offenser

Offenser ‘to offend, to harm’ is followed by a direct object from its earliest attestations, yet we find occurrences of it in our corpus used with an indirect object as well (17a-d). While offenser used transitively could mean both to offend someone and to physically harm someone, when it occurs with an IO it only has the sense of offending. In Modern French, offenser only occurs with a direct object (17e-f) and is no longer used to mean bringing physical harm to someone.

(17) a. Se ung homme marié habite a la femme de son compere ou autre mariee, il

meismes se clost la porte de paradiz encontre au visage, ne jamais ne lui est

ouverte, fors par cellui a qui tant a offensé, quant il luy a pardonné.

‘If a married man has sexual relations with his friend’s or another’s wife, he is

the one who closes the door to paradise on himself; it is never opened to him,

save by the husband whom he has so offended, when the husband has

forgiven him.’

(Les Evangiles des Quenouilles 2, 1466, p.132; BTMF)

b. s’ilz ont point offencé ou dit mestier

‘if they have not harmed the said profession’

(Recueil des documents concernant le Poitou, t.7, 1420, p.358; DMF1)

c. ... d’autant que toute leur meschanceté et tous leurs desseins se dressent contre

la majesté de Dieu, pour le despiter, et offenser le genre humain par mille

moyens

‘...so much that all their wickedness and all their ruses rise up in opposition to

the majesty of God, in order to vex him, and to bring harm, in thousands of

ways, to the human race’

(Paré, Des Monstres et prodiges, 1585, p.100; BFPC).

d. il l’a offensé en son honneur, en sa personne.

‘he offended his honour, his person.’

(Dictionnaire de l’Académie Française, 1694)

Modern French :

e. Don Gormas fait tout ce qu’il peut pour offenser Don Diègue

‘Don Gormas does all that he can to offend Don Diègue’

(Péguy; Le Petit Robert)

f. Il a fallu que je déclarasse par écrit n’avoir pas eu l’intention d’offenser le

gouvernement

‘I had to declare in writing that it was not my intention to offend the

government’

(Flaubert, Corresp., 1851, p.139; TLFi)

2.1.16 Persuader

According to the lexicons and dictionaries that were consulted, persuader could be used with either a direct or an indirect object until the 18th century with no apparent meaning difference between the two forms: persuader (à) quelqu’un (que /de /à,,,). Féraud observes that some of his contemporaries use a direct object with persuader, but that a dative object is preferable:

[...] il régit le datif de la persone et la prép. de devant les verbes: Il lui a persuadé de venir. Quelques Auteurs avec ce régime des verbes mettent le nom de la persone à l'acusatif, contre l'usage le plus autorisé. “Il travaille à les persuader de les suivre.”

MOREAU. Je crois que, à leur persuader vaut mieux. L'Acad. ne met d'exemple que du datif. (Féraud, Dictionaire critique de la langue française)

Persuader used with an indirect object is considered to be old or literary in the TLFi and is no longer used in Modern French.

(18) a. Comment Pantagruel Persuade à Panurge prendre conseil de quelque fol.

‘How Pantagruel persuades Panurge to take council from some fool.’

(Rabelais, Le Tiers livre, 1546, p.272; Frantext, ARTFL)

b. Celuy qui persuade à un autre de faire un crime, n’est guere moins coupable

que celuy qui le commet. ‘He who persuades another to commit a crime is hardly less guilty than he who commits it.’ (Dictionnaire de l’Académie Française, 1694)

c. persuadez la de m’aymer

‘persuade her to love me’

(Amboise, Les Neapolitaines, 1584, p.168; BFPC)

Modern French :

d. J’ai fini par le persuader d’en prendre une [garde-malade]

‘In the end, I persuaded him to take on a sick-nurse’

(Ramuz, A. Pache, 1911, p.305; TLFi)

e. Je l’ai persuadé de terminer l’ouvrage qu’il avait commencé.

‘I persuaded him to finish the work that he had begun.’

(Le Grand Larousse)

2.1.17 Prier

Prier ‘to ask, to beseech someone to do something’ or ‘to pray (to a god)’ appears most often with an animate indirect object in Medieval French, but in Pre-classical French a direct object begins to appear in the same contexts (19a-f): prier (à) quelqu’un (que/ de...). By the 17th century, the DO is considered preferable by Vaugelas and his contemporaries: “Les anciens disaient aussi prier à Dieu, et même quelques-uns disent encore je prie à Dieu au lieu de dire je prie Dieu” (Vaugelas, Remarques, 270).23 The TLFi makes no mention of an IO, and in Modern Standard French, prier only takes a direct object complement (19g,h).24

(19) a. [Il] prioit as dieux et soushaidoit que il eüst la gorge plus longue que le col

d’une grue.

‘He prayed to the gods and hoped that he would have a throat longer than the neck of a crane.’

(Oresme, Le Livre de Ethiques d’Aristote, c.1370, p.222; DMF1)

b. et luy prierent de venir loger avec eulx

‘and they asked him to come and stay with them’

(Les Cent nouvelles nouvelles, c.1456-1467, p.396; DMF1)

23 See also the Commentaires (789-800, vol.2).

24 Prier is still attested with a dative clitic. The dative clitic is felt to be acceptable but not “good French” by many. The sentence in (i) is “okay” according to my informants, but when the clitic is replaced by a noun phrase, as in (ii), the sentence becomes clearly unacceptable.

(i) Aminata, en s’adressant à sa mère, lui prie de parler au jeune homme pour qu’il lui fiche la paix.

‘Adressing her mother, Aminata begs her to speak to the young man so that he leave her alone.’

(ii) * Aminata prie à sa mère de parler au jeune homme pour qu’il lui fiche la paix.

‘Aminata begs her mother to speak to the young man so that he leave her alone.’

c. et luy priay affectueusement de me dire le secret

‘and I asked him/her affectionately to tell me the secret’

(Paré, Des Monstres et prodiges, 1585, p.101; BFPC)

d. la bonne dame priera Dieu devotement

‘the good lady will pray devoutly to god’

(Pizan, Le Livre des Trois vertus, c.1405, p.56; DMF1)

e. et me dit que l’aviez prié de passer par cy en son chemin

‘and he told me that you had beseached him to pass by here on his way’

(Amboise, Les Neapolitaines, 1584, p.212; BFPC)

f. Prier quelqu’un de quelque chose, de faire quelque chose.

‘To beseech someone something, to do something.’

(Dictionnaire de L’Académie française, 1694)

Modern French:

g. Prier la Vierge, les saints d’intercéder pour nous

‘To pray to the Virgin Mary, the saints, to intervene for us’

(Le Petit Robert)

h. Le Père Nicolle reçut une lettre du curé Puyoo qui le priait, en termes

pressants, de passer chez lui

‘Father N. received a letter from the parish-priest Puyoo, who beseeched

him, in pressing terms, to come by his home’

(Toulet; Le Petit Robert)

2.1.18 Secourir

In the general sense of ‘to help, aid, remedy’, secourir could be followed by either a direct or an indirect object (20a-f). This is no longer the case in Modern French, in which only the direct object is acceptable (20g-i).

(20) a. le contract [...] avoit esté pour secourir au Roy en sa grant neccessité et

besoing

‘the contract... had been to assist the king in his time of great need’

(Fauquembergue, Journal, 1417-1420, p.123; DMF1)

b. Les bouviers et pasteurs auront leurs chambres prés de leur bestial, pour leur

secourir plus tost.

‘The cattlemen and shepherds will have their rooms near their live-stock, in

order to come more quickly to their aid.’

(Cotereau, tr. Columelle, I, 6; Huguet, Dictionnaire de la langue française du

seizième siècle)

c. pour secourir aux frontieres

‘in order to defend the borders’

(Baye, Journal, 1400-1417, p.212; DMF1)

d. il te peut assez souvenir comment je n’ay rein espargné : mais ainsi y ay

secouru comme si je n’eusse aultre thesor en ce monde que de te veoir...

absolu et parfaict.

‘you may remember well enough how I spared nothing: but so attended to

it as if I had no other treasure in the world but to see you... absolute and

perfect.’

(Rabelais, II, 8; Huguet, Dictionnaire de la langue française du seizième

siècle)

e. Lors Jhesucrist qui ses servans regarde Et qui sequeure en tous lieux ses amys

‘Now Jesus Christ, who watches over his servants and who everywhere help his friends’

(La Vigne, Le Mystère de saint Martin, 1496, p.356; DMF1)

f. a la voye Se sont mis moult hastivement, Pour la secourre

‘they quickly took to the road, in order to come to her aid’

(Pizan, Le Livre de la Mutacion de Fortune, III, 1400-1403, p.68; DMF1)

Modern French :

g. Secourir des skieurs en danger

‘to help skiers in danger’

(Le Petit Robert)

h. Ils ne peuvent ni ne savent les soigner en santé, ni les secourir dans leurs

maladies

‘They neither can nor know how to look after them in health, or to take care

of them when they are ill’

(Destutt de Tr., Comment. sur Espr. des lois, 1807, p.379; TLFi)

i. Il n’y avait nul espoir de secourir Audenarde. La ville ne pouvait manquer d’être prise, du moins par famine

‘There was no hope of defending Audenarde. The city would certainly be taken, at least by famine’

(Barante, Hist. ducs Bourg., t.1, 1821-24, p.162; TLFi)

2.1.19 Servir

In the sense of serving someone or serving a cause, servir is attested early on as taking either an indirect or direct object (21a-f). According to Vaugelas, the IO was already dated by the 17th century:

Servir, regit maintenant l’accusatif et non pas le datif comme il faisait autrefois, et comme s’en sert ordinairement Amyot et les anciens Escrivains; Par exemple, ils disoient, il faut servir a son Roy et a sa patrie, pour dire il faut servir son Roy et sa patrie, comme on parle aujourd’hui. (Vaugelas, Remarques, 270)

Note that we have not included in our quantitative study the stative uses of servir describing, for example, the usefulness of something: ce livre lui a beaucoup servi ‘this book was of great use to him’. These kinds of uses have undergone no change in argument realization, in line with most other verbs used in stative contexts.

(21) a. [Charles-Quint] Veult... pour servir à Dieu abandonner le monde

‘Charles-Quint wishes to withdraw from the world in order to serve God’

(Du Bellay, Regrets, 1558, p.111; Huguet, Dictionnaire de la langue

française du seizième siècle)

b. Pour, en servant à Dieu, servir aussi... à l’Estat et à Sa Majesté.

‘In order, by serving God, to also serve the State and his Majesty’

(Aubigné, Histoire universelle, IX, 3, 1616-18; Huguet, Dictionnaire de la

langue française du seizième siècle)

c. Car nous veon que le plus des gens sont trop enclins a delectacion et servent

as delectacions.

‘For we see that the more people are too inclined to pleasure, the more they

are slaves to pleasure.’

(Oresme, Le Livre de Ethiques d'Aristote, c.1370, p.497; DMF1)

d. Et semblablement monseigneur de Bourgongne fist serement pareil de servir loyaument le Roy et obeir à lui

‘And similarly, M. de Bourgogne made the same oath to loyally serve the king and to obey him’

(Fauquembergue, Journal, 1417-1420, p.157; DMF1)

e. Dont j’en cognois aucuns qui, pour estre vrays amoreux et de bien loialment servir leurs dames

‘of which I know some who, in order to be true suitors and serve their ladies

loyally and well’

(La Sale, Jehan de Saintré, 1456, p.9; DMF1)

f. Mais toutesvoies, celui qui les sert [les dieux et noz parens] et honeure selon

sa puissance, il est vertueus.

‘But however, he who serves and honours them [the gods and our family] according to his ability, is virtuous.’

(Oresme, Le Livre de Ethiques d’Aristote, c.1370, p.451; DMF1)

Modern French :

g. Immense difficulté de servir Dieu vraiment. Dieu veut tout

‘Immense difficulty to really serve God. God wants everything’

(Bloy, Journal, 1894, p.154; TLFi)

h. Un étranger qui a passé par une grande ville d’Italie est moins connu par son

nom que par celui de la dame qu’il servait

‘A stranger who has passed through a large Italian city is less known by his own name than by the name of the lady that he was courting’

(Stendhal, Rome, Naples et Flor., t.2, 1817, p.170; TLFi).

i. Bien servir son pays, l’État

‘To serve one’s country, the state, well’

(Le Petit Robert)

2.1.20 Supplier

In the general sense of ‘to implore, beseech, beg’, supplier appears to be followed exclusively by an indirect object in Medieval and Pre-classical French in constructions of the type supplier à quelqu’un (que/ de + infinitive) (22a-e). From the 17th century however, only a direct object is attested with the verb (22f-i).

(22) a. la estoit ung chevalier qui se disoit le plus dampné de tous les aultres [...] lui

suppliant de incontinent voulloir oïr la confesse de ses abominables pechiez.

‘there was a knight who considered himself the most damned of all the

others... begging him at once to hear the confession of his abominable sins.’

(La Sale, Salade, c.1442-1444, p.104; DMF1)

b. Einsois li doy humblement supplier D’avoir sa grace.

‘And so he must humbly ask his mercy.’

(Machaut, La Fonteinne amoureuse, c.1361, p.161; DMF1)

c. quant elle supplia a Jupiter, le soverain dieu, elle ne ly recorda pas les

benefices qu’elle ly avoit faiz.

‘when she implored Jupiter, the supreme god, she did not remind him of the

favours she had offered him.’

(Oresme, Le Livre de Ethiques d’Aristote, c.1370, p.253; DMF1)

d. Mabile luy supplie la luy envoyer pour la reconforter

‘Mabile begs him to send it to her in order to comfort her’

(Le Premier livre d’Amadis de Gaule, c.1540, p.255; Lüdi, 1978:161)

e. Ils se mettoient à genoux devant luy, en luy suppliant de leur vouloir

pardonner

‘They knelt before him, pleading to him for forgiveness’

(Amyot, Pomp. 16th c., 37; Littré, Dictionnaire de la langue française)

f. Supplier aucun

‘To beg someone’

(Nicot, 1606, Thresor de la langue françoyse, tant ancienne que moderne)

g. Supplier la Cour de retenir la cognoissance de la cause principale

‘To entreat the court to keep the main case in mind’

(Nicot, 1606, Thresor de la langue françoyse, tant ancienne que moderne)

Modern French:

h. je tombai à ses genoux, je joignis les mains, je le suppliai par Jésus-Christ de

m’épargner

‘I feel at his feet, I joined my hands, I begged him in the name of Jesus Christ

to spare me’

(Chateaubriand, Mémoires d’outre-tombe, t.1, 1848, p.82; TLFi)

i. Je la suppliais de m’aimer quand même

‘I begged her to love me all the same’

(Radiguet; Le Petit Robert)

2.2 Exclusions

The twenty verbs described in the preceding section are not the only two-place verbs which allow both a direct and indirect object earlier in the history of French and which only occur with one of these forms in Modern French. A number of other verbs are mentioned in the literature as having undergone a similar evolution where the indirect form of the internal argument disappears from use leaving the direct object as the only licit form. Still others have undergone a valency change in the opposite direction. The following sections show that whatever variation or change of valence that these verbs have undergone, it is not the same kind of change as we see with the aider-type verbs. We therefore do not assume the same cause of change for these verbs and they do not figure in the main analysis.

2.2.1 Insufficient evidence

The transitive verbs in (23) could evidently be used with an indirect object at earlier periods in the history of French. This could suggest that they follow the same type of valency change as the aider-type verbs in (1), but since a survey of the text-based corpus does not provide further examples, it is more likely that the occurrences with an IO are hapax, fixed expressions, or limited to very specific contexts. It therefore does not seem correct to speak about a change in the expression of the object. Rather, these verbs appear almost exclusively with a DO in Medieval French, and this is still the case in Modern French. With only two or three examples of these verbs used with an IO, we can arrive at no conclusion about them outside of the fact that there was some flexibility in the way the object was expressed. We nevertheless note the examples that are offered in the literature, followed by an example from Modern French.

(23) Small showing of an alternation: DO/IO → DO

conjurer ‘to plot’ obliger ‘to require’ endoctriner ‘to instruct’ occasionner ‘to bring about’ éviter ‘to avoid’ poursuivre ‘to pursue’ guigner ‘to ogle, eye’ prévenir ‘to prevent’ incommoder ‘to inconvenience, bother’ supporter ‘to support’ nécessiter ‘to require’ surmonter ‘to overcome’

(24) CONJURER

a. Assassiner sa patrie et conjurer à sa ruine sont les marques de grandeur et

d’autorité

‘To assassinate one’s fatherland and plot its ruin are the marks of greatness

and authority’

(Malherbe, II, 17th c., 153; Brunot, vol.3(2), 546)

b. Conjurer la perte, la mort d’un tyran

‘To plot the ruin, the death of a tyrant’

(Le Petit Robert)

(25) ENDOCTRINER

a. Cum Diex endoctriné li ha

‘As God has instructed him’

(Boron, Le Roman de l’estoire dou Graal, 1199, 2558; Herslund, 1980:53)

b. Si vous endoctrinez un enfant dans cette science, qui donne peu d’idées

‘If you instruct a child in this branch of knowledge, which gives few ideas’

(Chateaubriand; Le Grand Larousse)

(26) EVITER

a. Plusieurs luy font regir le datif, et disent: eviter aux inconveniens

‘A number of people have it govern the dative, and say “avoid to a

difficulty’

(Vaugelas, Remarques, 389)

b. pour éviter à querelle et à bruit

‘in order to avoid quarrelling and noise’

(Chapel., Guzm., D’Alf., 17th c., III, 12; Brunot vol.3(2), 547, n.1)

c. Je n’ai pas pu éviter cet inconvénient

‘I wasn’t able to avoid this difficulty’

(Le Petit Robert)

According to the Robert Historique, éviter is originally used with an indirect object in two-place constructions. If this is indeed the case, the IO fell out of use long before the 14th century. Our search of the main FRANTEXT database from 1300 to 1649 turned up 232 occurrences of éviter used with an inanimate DO and only 9 occurrences of it used with an inanimate IO, 8 of which are used by one author (Fauchet) (animate objects in general are rarely attested with this verb and none are used with an IO). By the 17th century, the IO is limited to fixed expressions as suggested in Vaugelas and later confirmed by Corneille in the Commentaires.

Plusieurs luy font régir le datif, et disent éviter aux inconveniens, mais très-mal, et ce qui a donné lieu a cette faute, c’est que l’on dit ordinairement, pour obvier aux inconveniens; mais éviter, régit l’accusatif, et obvier le datif. (Vaugelas, Remarques, 389)

On dit en parlant de procédures, pour éviter aux frais, c’est une phrase particulière autorisée par l’usage en matière de Palais, mais hors de là, la Remarque de M. de Vaugelas est très-bonne; éviter ne doit jamais regir le datif, et c’est une faute de dire on ne peut éviter à son malheur. (Commentaires, vol.1, 473-74)

(27) GUIGNER

a. guignant sous son capuchon a la pauvre femmelette

‘eyeing the poor young woman from under his hood’

(N. du Fail, Eutr., 17th c., vol.2, p.8; Brunot, t.2, 438)

b. Guigner une femme au passage

‘To eye a woman in passing’

(Le Petit Robert)

(28) INCOMMODER

a. failloit il que ce feust en incommodant à mon roy? ‘was it necessary that it entailed inconveniencing my king?’

(Rabelais, Gargantua, 1534, ch.31, t.1, 118; Huguet, Dictionnaire de la langue

française du seizième siècle)

b. La fumée l’incommode. ‘Smoke bothers him/her.’

(29) NECESSITER

a. l’Impôt du droit d’aide necessite [...] à une dépense énorme à prélever sur

son produit

‘The tax for assistance charges requires ... him to deduct an enormous amount

from his yield’

(Coulom., 17th c., 515; Brunot vol.10(1), 518)

b. Un train de vie fastueux nécessite de grandes dépenses.

‘An extravagant lifestyle requires large expenditures.’

(Le Grand Larousse)

(30) OBLIGER

a. obliger aux seigneurs à avoir sur les lieux des officiers de justice

‘to oblige land owners to have law-officers on site’

(Champi., 17th c., 405; Brunot vol.10(1), 518)

b. Ses parents l’ont obligé à travailler.

‘His parents required that he work.’

(Le Petit Robert)

(31) OCCASIONNER

a. que les dîmes ne soient plus à l’avenir perçues en nature... raison qui

occasionne à des grands procès entre les pasteurs et leurs ouailles

‘that in the future, taxes no longer be perceived as reason in itself... for

bringing about large lawsuits between the ministers and their flocks’

(Germ., 17th c., 622; Brunot vol.10(1), 518)

b. Pour savoir ce qui pouvait occasionner ce retard

‘In order to know what could have caused this delay’

(Laclos; Le Petit Robert)

(32) POURSUIVRE: two occurrences of poursuivre with an IO were found, both clearly

having an aspectual interpretation.

a. ...les Volces (...) le tuerent [Coriolan] tantost : pour ce qu’il n’avoit poursievy

a la destruction de Romme que tant ilz heoient.

‘The Volces... recently killed him because he hadn’t proceeded with

destroying Rome, which they hated so.’

(La Sale, Sale D., 1451, p.105; DMF1)

b. Et a sa mort je poursuivray

‘And I will continue in my efforts to kill him’

(Act. des Apost., vol.1, 140c (G., Compl.); Huguet, Dictionnaire de la langue

française du seizième siècle)

c. Et puis, il a poursuit son chemin.

‘And then, he continued on his way.’

(33) PREVENIR

a. Tu devois prevenir sagement au dommage

‘You had to wisely avoid harm’

(Ollenix du Mont Sacré, Second livre des bergeries, 1592, fo 50 vo;

Goyens 1998:479)

b. A celle fin de prevenir Au mal qui vous pourroit venir

‘To this end to prevent any harm that might come to you’

(Myst. du V. Testament, c.1480, 19495; TLFi, prévenir, Étymol. et Hist.)

c. Et nous en advertit, afin d’y prevenir

‘He warned us about it, in order to prevent it from happening’

(Desportes, Elégies, II, 48, 16th c.; Goyens 1998:480)

d. nouveaux produits infaillibles pour prévenir la peste

‘new, infallible products to prevent the plague’

(Camus, La Peste, 1947; Le Petit Robert)

Malherbe proclaims that “prévenir à quelque chose est ‘parlé allemend’” (Malherbe, IV, 354; Brunot vol.2(3), 547), suggesting that it occured often enough to warrent comment. Given the nature of our corpus, however, it is not surprising that very few solid examples with an IO were found.

(34) SOIGNER

a. Il dira que soignez seulement à vos estudes, et qu’il vous donnera femme

quand il en sera temps.

‘He will say apply yourself exclusively to your studies, and he will give you a

woman when the time comes.’

(Larivey, Laquais, 1579, I, 3; Huguet, Dictionnaire de la langue française du

seizième siècle)

b. Qui soignera à vostre mesnage durant vostre absence?

‘Who will take care of your domestic affairs during your absence?’

(Dictionnaire de L’Académie française, 1694)

e. mais il ne soignait pas les détails

‘but he didn’t tend to the details’

(Stendhal; Le Petit Robert)

By the 18th century, soigner is still used with an IO, but it is considered regional and of a lower register. Attestations with an IO are therefore rare in the corpus after this period.

Aûtrefois on le faisait neutre avec les choses au datif; soigner à ses afaires; et actif avec les persones: soigner un malade, soigner des enfans. "Les orangers veulent être bien soignés." Aujourd’hui il ne se dit que de la seconde manière: la première est du Peuple, en certaines Provinces, et à Paris même. = Au figuré, on le dit élégamment du style. "Compilateur laborieux, qui ne soignoit pas assez son style." Sabat. Trois Siècles, etc. (Féraud, Dictionaire critique de la langue française)

Féraud’s comment seems to imply that soigner may have often occurred with an IO, but this is not substantiated by our corpus. Indirect objects are rarely found.

(35) SUPPORTER

a. pour supporter à tous les besoins de l’Etat

‘in order to take responsibility for all the needs of the State’

(Ep., 17th c., 607; Brunot vol.10(1), 518)

b. Ne pas vouloir s’exposer à supporter les frais d’un procès

‘To not want to leave oneself open to assuming the costs of a lawsuit’

(Le Grand Larousse)

(36) SURMONTER

a. pour ... surmonter à toutes les difficultés, relativement aux finances

‘in order... to overcome all of the difficulties relative to finance’

(Beine, 17th c., 267; Brunot, vol.10(1), 518)

b. En elles deus les vertus intellectuelles et morales ont eu un doux combat à

qui surmontroit.

‘In the two of them, intellectual and moral virtue had a gentle battle over

which he would prevail.’

(Aubigné, Lettres de poincts de science, 8 (I, 447); Huguet, Dictionnaire de

la langue française du seizième siècle)

c. Surmonter les difficultés, la fatigue, un handicap, une maladie.

‘to overcome difficulties, fatigue, a handicap, an illness’

(TLFi)

Brunot (vol.10(1), 518) observes that surmonter à is quite common in Classical French, but this is not reflected in the corpus.

2.2.2 Loss of the IO but no replacement form

Another group of two-place verbs, given in (37), also present the possibility of taking an indirect object complement in Medieval French but only take a direct object in Modern French. At first blush, the change ressembles that described for the aider-type verbs.

(37) Valency change: IO1/DO2 → DO2

administrer ‘to serve’ éclairer ‘to bring light’ entendre ‘to listen, heed’

The valency change that these verbs undergo, however, is not at all the same. What has actually happened, is that the direct object has remained along with the particular meaning associated with it, while the indirect object has fallen into disuse: IO1/DO2 → DO2. Note the contrast with that of aider type verbs: IO1/DO1/2 → DO1. For verbs like aider, there is no obvious meaning difference between their use with an IO in Medieval French and that with a DO in Modern French. This fact alone allows us to assume a purely structural change: the DO replaces the IO because the latter is no longer licit in a given context. This is clearly not the case for the verbs in (37).

The IO of administrer, éclairer and entendre was not replaced by an analogous direct object. Without an extremely close study of how and when these IOs began to decline in frequency, it is not clear if the loss of the IO is structural or lexical. Further research may very well suggest that the same underlying cause that brought about the replacement of the IO with aider-type verbs also underlies the loss of the IO with the verbs in (37). In fact, the same reasoning applies to any dynamic two-place verb which could occur with an IO in Medieval French but which no longer can in Modern French. Habiter, for instance, could occur with an animate IO in Medieval French: habiter à quelqu’un ‘to have sexual relations with someone’; see example (17a), section 2.1.15. The fact that only the animate IO is infelicitous in Modern French while inanimate IOs are still perfectly acceptable, i.e., habiter au troisième étage, is in line with the general valency change that we are looking at.

In what follows, we describe the valency change for the verbs in (37), justifying our decision to exclude them, at least for now, from the inventory of verbs that underwent a systematic structural change.

ADMINSTRER

Administrer could be used bivalently or trivalently as a verb of giving. Concerning its bivalent uses, both the TLFi and the Dictionnaire Historique note that when used with an indirect object, administrer meant ‘to serve’, mainly in religious and civil contexts, as shown in example (38). The IO is present in our corpus until the 16th century (38b), after which point it appears to have fallen out of use.

(38) a. et vous a Dieu monstrer Dignes de li amenistrer A son autel.

‘and to show yourself worthy of God to serve him at his alter.’

(Miracle de saint Lorens, 1380, p.130; DMF1)

b. Mille miliers de serviteurs assistoient devant Dieu, et mille milions luy

administroient.

‘A thousand thousand servants knelt before God, and a thousand million

served him.’

(Calvin, Sermons sur le livre de Daniel, 1560, 13 (XLI, 459); Huguet,

Dictionnaire de la langue française du seizième siècle)

In contrast, the transitive use of administrer either had the sense of ‘to take care of someone’ in a military context, (39a), or ‘administer, govern or direct public affairs’, shown in (39b). The current sense of the word parallels the (b) example, shown in (40). The valency change therefore concerns the loss of the IO form, but does not involve its replacement by an analogous DO.

(39) a. Molt tost le poroit Dieus amenistrer

‘Very quicky could God care for him’

(Aiol et Mirabel, éd. Raynaud et Normand, 141; Etymol. et Hist., TLFi)

b. de bien gouverner et administrer ledit prioré

‘to govern and run well the said convent’

(Fauquembergue, Journal II, 1421-1430, p.229 ; DMF1)

(40) Le maire administre la commune.

‘The mayor runs the village council.’

(Le Petit Robert)

ECLAIRER

The verb éclairer followed by an indirect object meant ‘to bring light to someone or something’ in a more or less literal sense (41a). Brunot quotes Malherbe: “éclairer à quelqu’un, c’est lui fournir de la clarté” (Malh., IV, 352; Brunot vol.3(2), 549). In Modern French however, this verb

99 only occurs with a direct object and has a slightly different meaning, ‘to make someone or something light or bright’, and so allows the figurative sense ‘to enlighten someone’, in (41b).

(41) a. – Mabile... print un flambeau... et esclaira à Oriane.

‘Mabile took a torch and lit the way for Oriane.’

(Amadis, I, 15; Huguet, Dictionnaire de la langue française du seizième

siècle)

b. éclairer le peuple

‘enlighten the people’

(Hugo; Le Petit Robert)

Féraud’s commentary shows that meaning of éclairer when used with a direct object is distinct from that when used with an IO.

Au propre, répandre la clarté sur... "Le soleil éclaire la terre. "Le flambeau éclaire toute la chambre. [...] = 2°. Au figuré, doner de la clarté à l’esprit. "Les bonnes lectures éclairent l’esprit, sans corrompre le coeur. "Seigneur, éclairez mon entendement. = 3°. Instruire de ce qu’on ignôre; détromper. V. n. Aporter de la lumière à quelqu'un, pour lui faire voir clair. Il se dit au propre, et régit le datif (la prép. à.) Éclairez à Monsieur; et en sous-entendant le régime, éclairez; allez éclairer. "Euryclée éclairoit à ce jeune Prince. M. de. Dacier, Odyssée; c. à. d., portait le flambeau devant lui, comme l’Auteur l’a dit plus haut. (Féraud, Dictionaire critique de la langue française)

The two constructions therefore have distinct meanings. The indirect object and the meaning associated with it fell out of use while the causative, direct construction has remained.

ENTENDRE

Entendre also shows a DO/IO alternation where each form is associated with a more or less distinct meaning. Use with the indirect object is based on the original Latin meaning intendere ‘to stretch toward’ in the sense of ‘listen’ and ‘pay attention’, and is attested as early as the Chanson de Roland.

100

(42) a. Or, mon ami, entendez a moy

‘Now, my friend, listen to me’

(La Sale, Jehan de Saintré, 1456, p.59; DMF1)

Huget’s Dictionnaire de la langue française du seizième siècle contains many examples of entendre à quelque chose, but very few examples with an animate indirect object. Subsequent dictionaries only cite it in fixed expressions such as ne savoir à qui entendre. While entendre followed by a direct object could also mean ‘to listen’, its main meaning with a direct object was ‘to hear’. This verb no longer occurs with an indirect object and no longer denotes an event of listening, heeding, etc. Clearly, the direct object in Modern French, is a continuation of the same form found in Medieval French.

2.2.3 Stative verbs

We have excluded sentir and ressembler from the inventory of two-place verbs that have undergone a systematic valency change because they denote a state, not a dynamic event. They appear to be the only stative bivalent constructions to have undergone such a change. In chapters four and five, we argue that because two-place stative verbs have a different argument structure than two-place dynamic verbs like aider, they are not subject to the same cause of change.

A preliminary examination shows some fluctuation between these verbs used in a relational sense, and thus with an IO, and them used in an attributive sense, with a DO. Relational constructions establish a comparison (ressembler ‘to resemble’ and sentir ‘to smell like’), while attributive constructions establish equivalence (ressembler ‘to seem’ and sentir ‘to smell of’). For independent reasons, these forms would have evolved towards one construction and meaning as opposed to the other.

(43) a. Relational

Oeillet se dit giroflée, pour ce qu’il sent au clou de girofle

‘Oeillet calls himself clove, because he smells like cloves’

(R. Franc., Merv. de Nat., 17th c., 273; Brunot vol.3(2), 547)

101

b. Attributive

Tu sens la bière/ mauvais.

‘You smell of beer/ bad.’

(44) a. Attributive

Je le puis bien ressembler, car je suis luy mesme.

‘I may very well seem like him, for I am him.’

(Amboise, Neapolitaines, 1594, p.219; BLFP, ATILF)

b. Attributive

Ces deux... meisrent plus de peine à ressembler bons et vertueux que de

l’estre.

‘These two… went to more pains to appear good and virtuous than to actually be

that way.’

(Amadis, II, 20; Huguet, Dictionnaire de la langue française du seizième

siècle)

c. Relational

La jeune fille ressemble à sa mère.

‘The young girl resembles her mother.’

2.2.4 Opposite change

Some verbs show an opposite valency change to that of aider-type verbs. The two-place verbs in (45), for instance, can occur with both a direct and indirect object in Medieval and Pre-classical French, but unlike aider-type verbs, they only take an indirect object in Modern French.

102

(45) Opposite change: DO/IO → IO

aspirer ‘to aspire to’ pardonner ‘to pardon, absolve’ attenter ‘to make an attempt on’ prétendre ‘to aspire, lay claim to’ échapper ‘to escape’ renoncer ‘to give up, forgo’ mentir ‘to lie’ surseoir ‘to defer, postpone’ nuire ‘to harm, bother, be hurtful’ survivre ‘to survive’ obéir ‘to obey’

If our hypothesis is correct, and a semantic change in the preposition à removed the necessary licensing environment for certain indirect objects, then we expect a unidirectional change. A group of verbs which change in the opposite direction, whose DO is replaced by an analogous IO, are potential counter examples to our account. It can be demonstrated, however, that the valency change in these verbs resembles in no way that of aider-type verbs.

First, for many verbs, the DO – the form that is now obsolete – is used very infrequently. This is not the case with aider-type verbs for which the obsolete form is often the dominant form in Medieval French. In these cases, we take a similar position to that in section 2.2.1. The IO in Modern French is considered to be a continuation of the IO in Medieval French. The rare presence of a DO could be due to any number of factors: hapax, fixed expression, restricted context, etc.

Second, in a number of cases, there is a meaning difference associated with the two forms and it can be shown that the DO and the particular meaning associated with it, simply fell out of use. The IO attested in Medieval French is thus the IO that we use today in Modern French; the current expression of the object has undergone no change.

The general conclusion, therefore, is that the present form of the object of the verbs in (45) has undergone no change. The IO attested in Medieval French is the same IO that we use today:

IO1/DO2 → IO1. An explanation for why the IO is permitted to persist while that of the aider- type verbs is replaced by a DO is taken up in chapter four where it is demonstrated that these verbs are among many whose IO undergoes no change. Three properties distinguish the aider- type verbs from others that occur with an IO and it is the combination of these properties that determines if a verb undergoes the valency change in question.

103

We first take up the verbs that show little evidence of a DO.

NUIRE

Nuire is mentioned in some historical grammars as having once accepted a direct object as well as an indirect object. This is thinly confirmed in our corpus where we found only three reliable occurrences of a direct object. Interestingly, each occurrence is used in a clearly agentive sense, as illustrated below:

(46) Que gangnes tu a venir nuire Un povre hermitte?

‘What do you stand to gain from coming to do a poor hermit wrong?’

(Miracle de st Guillaume, c.1347, p.23; DMF1)

One explanation for the possibility of a direct object could be related to the fact that the meaning of nuire can fade from denoting a psychological state, compatible with a dative object experiencer, to an intentionally caused event, compatible with a patient-like direct object.

OBEIR

Only 2% (9/457) of all tokens in the text-based corpus between 1300 and 1600 correspond to obéir used with a direct object. To speak of a change of valence would therefore seem to be rather farfetched.

(47) a. obeïr son serment et promesse

‘respect one’s oath and promise’

(La Sale, Salade, c.1442-1444, p.103; DMF1)

b. L’infante lui dit que la plus grande beauté d’une femme étoit d’obéir son

mari

‘The child told him that the greatest beauty of a woman was to obey her

husband’

(Malherbe; Haase 1914:135; Herslund 1988:218)

104

One possible explanation for the reason that obéir shows a slight IO/DO alternation would be related to the kind of event it denotes. In section 4.2.2 we propose that obéir belongs to the class of resistance verbs, based on Talmy’s (2000) theory of force dynamics and thus licenses its IO much like résister, céder, etc. But since an act of obeying represents the least possible degree of resistance, it is possible that this verb is construed as an event that does not involve force, along the lines of écouter, respecter, etc., taking a direct object. One piece of evidence that supports this idea is the absence of variation we see in the object of its derivative désobéir. We propose that the two verbs do not pattern in the same way because désobéir clearly only denotes resistance to force. Without any relevant variability in its meaning, désobéir is not susceptible to variation in the realization of its object in the same way as obéir is.

The following discussion demonstrates that the DO/IO alternation associated with the rest of the verbs corresponds to a clear meaning difference between the two forms. The meaning implied by the DO fell out of use while the IO and it’s associated meaning has remained.25

ASPIRER

Aspirer used with a direct object meant to desire to have something, shown in (48a). With an indirect object, it means to aspire to something, generally to a situation (48b). Crucially, the example in (48a) cannot simply be “modernized” by replacing the DO with an IO, as illustrated in (48c). This is because the specific meaning aspirer had with a DO was and is not available with an IO.

25 Surseoir is the only verb for which we have not been able to identify any real meaning difference that corresponds to the two different forms of the object. This verb remains an exception to our analysis:

(i) Et, finablement, fu conclu de surseoir la pronunciacion d'iceulz arrestz

‘And it was finally decided to defer the sentence of those arrested’

(Fauquemberge, III, 1431-1435, 97; DMF1)

105

(48) a. En vous je vy, et en vous je respire; Autre richesse au monde je n’aspire.

‘In you I live, and in you I breathe; other wealth in the world I do not desire.’

(Ronsard, Elegies, var. 1572-1573, IV, 389; Huguet, Dictionnaire de la langue française du seizième siècle) b. Aspirer à un titre

‘To aspire to a title’

(Le Petit Robert)

c. #? En vous je vis, et en vous je respire; à autre richesse au monde je n'aspire.

ATTENTER

Attenter used with a direct object means to attempt something (49a), but with an indirect object, it implies that what is attempted is a criminal act against someone (49b). The DO has fallen out of use in French, along with its meaning. Crucially, it was never replaced by an analogous IO, confirmed by the oddity of (49c) in which the DO in (49a) is replaced by an equivalent IO.

(49) a. Le despit et la honte leur fit prindre une resolution pour attenter une chose

difficile

‘Spite and shame made them determined to attempt a difficult thing’

(La Noue, Disc. pol. et mil., XXVI, 1, p.703, Huguet, Dictionnaire de la langue française du seizième siècle)

b. Attenter à la vie de quelqu’un

‘To make an attempt on somebody’s life’

c. #? Le despit et la honte leur fit prindre une resolution pour attenter à une chose

difficile

ECHAPPER

Échapper ‘to escape’ was often used in place of its near synonym éviter ‘to avoid’. With an IO, échapper generally means ‘to escape’, maintaining the idea that the grammatical subject has

106 been, but is no longer located at the place or situation denoted by the IO. In contrast, échapper followed by a direct object appears to always means ‘to avoid’ in our corpus. The particular meaning entailed by the direct object is confirmed by Andry in the Commentaires (vol.2, 574).

Echapper un danger, echapper d’un danger. Monsieur de Vaugelas les confond (ce me semble) dans ses Remarques, mais ils sont neanmoins fort differens; échaper un danger c’est l’éviter, & échaper d’un danger c’est s’en tirer aprés y être tombé. Je retranche les exemples, la chose est assez claire d’elle-même.

MENTIR

With a direct object, mentir meant ‘to betray, deceive’ (50a). Both its meaning and its use with a direct object have disappeared in Modern French. Today, mentir occurs with an indirect object and has the meaning ‘to tell someone a lie’ (50b). Again, the indirect object did not simply replace the direct object. The meaning of the verb has shifted to denote a manner of communicating; the current expression of its object corresponds to that meaning change.

(50) a. tu as menti l’amour, je m’en doutais

‘you have been unfaithful to love, I suspected as much’

(Restif de la Bret., M. Nicolas, 1796, p. 171; TLFi)

b. Paul lui a menti, à Jean.

‘Paul told Jean a lie’

PRETENDRE

Prétendre, much like aspirer, had a slightly different meaning when used with a direct object. Used transitively, it could mean ‘want’, for example, shown in (51a). In Modern French, prétendre takes an IO and means ‘to aspire’ (51b). The direct object in (51a) cannot simply be replaced by an indirect object, confirming the meaning difference between the two forms, as shown in (51c).

107

(51) a. Ce que je prétends de vous, c’est de m’aider à ...

‘What I want from you is to help me...’

(Gobineau, Pléiades, 1874, p.178; TLFi)

b. Prétendre à un titre

‘To aspire to a title’

(Le Petit Robert)

c. #? Ce a quoi je prétends de vous, c’est de m’aider à ...

PARDONNER

We propose that the IO of pardonner is always interpreted as the recipient in a three-place construction where the DO is implicit, i.e., lui pardonner (quelque chose/de...). In contrast, the DO is construed as an affected patient-like object, moving from guilty to absolved. It is significant that all examples we have found with a DO are never followed by an infinitive clause: * le pardonner de..., suggesting further that one object form was not simply replaced by another. Recall, for example, that with the aider-type verbs, both the direct and the indirect object could occur with an infinitival clause.

RENONCER

The modern meaning of renoncer is to ‘give up’. This meaning is associated with an IO which generally denotes or implies a state or situation, illustrated in (52a-c). In contrast, when used with a DO, renoncer was often synonymous with renier ‘to publicly deny, disown’, or ‘to go back on’. Crucially, the example in (52d) must be glossed by renier in Modern French, not by renoncer.

(52) a. Renunçant aux dieux interditz, Le serviray devostement.

‘Renouncing forbidden gods, I will serve him devotedly.’

(La Vigne, Saint Martin, 1496, 499 ; DMF1)

108

b. Paul a renoncé au vin, paraît-il.

‘Paul has given up wine, it seems.’

c. Paul a renoncé au mariage.

‘Paul has given up on the idea of mariage.’

d. Je la déteste... Et la renonce pour ma fille.

‘I hate her... and disown her as my daughter.’

(Molière; Le Petit Robert)

SURVIVRE

Survivre shows a rather robust DO/IO alternation. Vaugelas observes the following :

Ce verbe regit le datif, et l’accusatif tout ensemble, comme il a survescu tous ses enfants, et il a survescu a tous ses enfans. Il depend apres cela de l’oreille, de mettre tantost l’une, tantos l’autre, selon qu’elle le juge plus a propos. (Vaugelas, Remarques, 267)

In the Commentaires, however, the Académie specifies that while animate objects may occasionally appear as direct objects, inanimates never do:

[...] il est vray qu’on peut dire survivre à quelqu’un et survire quelqu’un, mais ce verbe a plus souvent le regime du datif, surquoy il faut remarquer que s’il gouverne quelquefois l’accusatif pour les personnes, comme en cet exemple, il a survescu son pere, il ne le gouverne jamais pour les choses. Ainsi, il n’est point permis de dire survivre sa gloire, survivre sa reputation. Il faut dire tousjours survivre a sa gloire, a son honneur, a sa reputation. (Commentaires, vol.1, 313)

These comments suggest that when the verb is used with an inanimate indirect object, it always has the sense of continuing to exist despite potentially adverse forces. It implies resistance to metaphorical death, in a way. Taking an IO, it patterns, as expected, with events denoting resistance to force. On the other hand, while surviving someone can also denote an implicit

109 resistance to death on the part of the subject, it can also denote a simple spatio-temporal relation between two individuals, mediated, for example, by the preposition/preverb sur. A relation analysis could also apply to surseoir used with a DO.

In sum, for reasons concerning the semantic evolution of lexical items, the indirect object associated with the verbs in (43) and their corresponding meaning have prevailed over the direct object and the meaning it entailed. Most important to our study is the fact that these verbs do not undergo a purely structural valency change and thus do not pose as counter examples to our claim about the change in aider-type verbs.

2.3 Chapter summary

Twenty verbs, exemplified by aider, were identified as having undergone the same valency change in the history of French. These verbs permitted an indirect object complement at an earlier period, but only occur with a direct object complement in Modern French. For each one, it was demonstrated that there was no change in the meaning of the verb. In other words, what the verbs denoted with an indirect object is what they denote with a direct object in Modern French. This is a crucial point to make, since our claim is that the valency change is purely formal.

An array of other verbs that appear to present a valency change were also discussed in some detail in order to justify their exclusion or irrelevance to the present study. Some are excluded on the grounds that there is not enough evidence of a change. Others involve the loss of a form, but no change in the expression of the object. The evolution of these verbs is therefore best characterized as semantic, not structural. The present study is only concerned with structural valency change, in which one form is replaced by another without entailing any change of meaning.

Finally, this chapter illustrates just how common the variation between a direct and indirect object complement really was in Medieval and Pre-classical French. The range of verbs permitting both a DO or an IO appears far broader than what we see in Modern French. This would correlate in part, as we develop further on, with the fact that dative objects are licensed in a wider range of contexts in Medieval French. It would also follow from a greater degree of flexibility in the expression of certain kinds of objects in the absence of a strong written standard. For verbs like obéir, renoncer, survivre, etc., given the overlap in meaning between the

110

IO and the DO, the form of the object that a speaker chooses to employ is likely based more on frequency than meaning.

111

Chapter 3 Quantitative study

Recall that in chapter two, twenty verbs were identified as having undergone the same change in the expression of their object:

(1) Two-place verbs having undergone a change in argument expression: IO → DO

All of the verbs in (1) used to allow an indirect object in two-place constructions but generally only permit a direct object in Modern Standard French. Crucially, no obvious semantic change accompanied the change of valence. The structural change is characterized as follows:

(2) V-IO1/DO1/2 → V-DO1

This chapter builds on the general observation that by the 18th century, the IO of many of the aider-type verbs in (1) had been all but replaced by a DO (see Brunot 1966:439, vol.2; Goyens 1998). Our quantitative study of nine of the verbs shows that the IO clearly began to decrease in frequency between the 16th and 17th centuries and had disappeared or was rarely used by the 18th century. A survey of the remaining verbs illustrates the same tendency.

Concomitant changes are of particular interest in diachronic linguistics because unity of change often suggests unity of the analysis, namely, the same underlying cause. The particular case of aider-type verbs presents at least two hypotheses. The first would attribute a common cause of change to an external force acting on the language. It has been proposed for instance, that the

112 great writers, grammarians, and Remarqueurs of the 17th century were the driving force behind the loss of certain IO/DO alternations and the evolution towards a single complement form (i.e., Malherbe, Vaugelas, Patru, Corneille, and other members of the Académie Française). This assumption is implicit in many works which address variation and change of valence in French, and explicitly stated, for example, in Kilroe (1989:187-194). However, there is simply no evidence of any reasoned choice, on the part of anyone, for one complement form over another.26 No preference is shown based on analogy with other French verbs or on the valence of Latin cognates. In fact, if parallels with Latin morphological case had any influence at all in the matter, many of the verbs in (1) would have maintained a dative IO (see chapter four, table 1).

These points aside, there nevertheless seems to be a correlation between the use of a single complement form and the undeniable increase in attention to uniformity of expression from the 17th century onward. Our quantitative study however, shows that the grammarians and Remarqueurs were not in fact the cause of the valency change. The shift from an indirect to a direct object was well underway before the generation of influential writers, grammarians and Remarqueurs was born. Indeed, if the elite of the 17th and 18th centuries had any influence at all, it was simply to accelerate a change that was already in the works.

Given that the replacement of the IO by a DO occurs during the same limited time period for all of the aider-type verbs in (1), another hypothesis would be that the change in argument realization is the effect of an abstract underlying change in the grammar. A common assumption in diachronic linguistics is that different superficial changes which occur during the same period of time may be linked to the reanalysis of a single grammatical item, the repercussions of which are the concomitant “reflexes”, or superficial changes. When the cause of change is grammatical, it can reveal structural commonalities, otherwise not apparent, among various syntactic derivations. The quantitative study that follows strongly supports the hypothesis of a common underlying cause affecting the valence of two-place aider-type verbs. The grammatical analysis that defines the cause of the change is discussed in chapter five.

26 See Schøsler (To appear) for evidence that the change was not top-down.

113

Section 3.1 outlines the corpus and methodology employed in the quantitative study. A detailed evolution of the IO of nine of the verbs in (1) is presented in section 3.2. Section 3.3 provides a synthesis and interpretation of the results. The data show that the actuation of the change is limited to a 150-year period between the 16th and 17th centuries, while the total duration of the change varies from verb to verb. A survey of the remaining verbs in 3.4 corroborates the results of the quantitative study. Section 3.5 concludes.

3.1 Corpus and methodology

This section gives an overview of the various corpora employed in the quantitative study and the methodology we used to treat the data.

3.1.1 Corpus

Our quantitative study examines the evolution of the IO and the DO used with nine verbs from the list of verbs given in (1): aider, applaudir, assister, commander, contrarier, contredire, ennuyer, favoriser, and insulter. These verbs were selected over the others based on their frequency in the texts, as some verbs are better represented in the corpus than others. They are also representative both of verbs having evolved directly from a Latin cognate and of verbs that were borrowed from Latin during the Middle Ages.

A variety of electronic data bases were searched, containing a total of almost 4000 texts in poetry, prose and technical prose from the 14th to the 19th century. The ARTFL project was our main source of data: the Frantext data base, the Encore collection, its collection of texts by French Women writers, and in some cases, the Textes de Français Ancien. All of the data from the Middle French period was obtained from the Base Textuelle du Moyen Français (BTMF), made available for public consultation by the Laboratoire ATILF. In order to ensure adequate coverage of the 16th century, we consulted other data bases as well: Cornucopia d’Epistemon, the Corpus Lemieux, Rentexte (Textes de la Renaissance), and several on-line texts available through the Université de Rennes 2. Although we make use of a number of other metalinguistic sources elsewhere in this dissertation, none were employed in the quantitative study of the nine verbs listed above.

114

3.1.2 Methodology

In the graphs that follow, the frequency of the IO was tracked over time and was controlled for animacy; animate objects are represented by the solid line while the inanimates are represented by the broken line. The vertical axis indicates the frequency of the IO in percent, comparing the number of IO tokens with the total number of occurrences of DOs and IOs for each verb. The horizontal axis tracks the change in frequency over time. It is divided into fifty year intervals save the first two centuries which are collapsed into 100 year intervals, as there appears to be little change in frequency during this period.

On the horizontal scale, we chose to plot the data against the birthdate of the writer and not against the time of composition or publication. For example, we examined four works by François Rabelais appearing between 1532 and 1552 in which the verb aider was used with a DO twice and with an IO eight times. This data corresponds to the interval 1450-1499 since Rabelais was born in 1490. Plotting the data this way allows us to evaluate the kind of object Rabelais uses with aider along side other authors born during the same period (Marguerite de Navarre, Bonaventure de Périers, etc.), rather than with Michel de Montaigne, for instance, born some forty years later, in 1533, and whose Essais were published in 1595. The assumption here is that the grammars of speakers who are of approximately the same age will have more points in common than the grammars of speakers separated by several decades, dialectal variation not withstanding.

This choice is not uncontroversial and follows from our belief that the grammar established at childhood determines the form of the object that will appear with aider-type verbs in the adult speaker. In other words, we do not believe that the choice of a DO or an IO with these particular verbs develops during the adulthood of speakers. We propose that the child develops one of three possible grammars which generates predictable output in adult speech. The child speaker learns that IOs are either licit with aider-type verbs or they are not, or the speaker develops a grammar in which the two forms are in competition (the existence of two competing grammars). In the first case, we expect to see the author using IOs robustly with the verbs in (1), but this does not

115 rule out the use of DOs with the verbs in the same contexts.27 Alternatively, if the child speaker develops a grammar in which IOs are illicit with two-place aider-type verbs, then we expect these verbs to occur (almost) exclusively with DOs in the adult learner. We do not expect to see an IO/DO alternation in their writing. Finally, if the child learner acquires a grammar where the IO and the DO are in competition, then like the first case, we expect to see both forms in the adult writing, except that the IO may not be used as robustly. Organizing the data in terms of the authors’ birthdates allows us, in principle, to identify the kind of grammar each author has acquired as a child. This is in line with the Minimalist idea that during the process of acquisition, a child associates specific features with grammatical elements which in turn determine possible derivations. Once the properties of a grammatical element are acquired, there is no reason to believe that they are ever lost in the adult.

Since the data is broken down by author, regional differences should also be identifiable. Yet, while our analysis reveals some variation among authors as to what form of the object is preferred, there are no obvious dialect patterns suggesting how the valency change may have spread through the language. For instance, both Jean Froissart (b.1337) and Jean Gerson (b.1363) use a direct object after the verb aider far more than their contemporaries. Froissart is from Picardie (but traveled extensively) and Gerson is a Champenois. These two authors may be representative of a more widespread tendency, hinting at the Northeastern dialects as a possible origin of the change, but our data is far from conclusive. From another point of view, Béroalde de Verville (b.1556), Honoré d’Urfé (b.1567) and Charles Sorel (c.1602) tend to use an indirect object after aider significantly more often than their contemporaries. Both Sorel and Béroalde were born and lived in Paris (along with many other contemporary writers in our corpus) whereas Urfé is from the south and spent much of his time in the Savoy region. If these authors give an indication as to where the change was late to take hold, the study offers no further insights. The diffusion of the valency change and its possible relation to regional or literary dialects is certainly an area that merits further study. Our analysis of the data simply defines the overall trend of the valency change. Given that the general tendency shows a more or less

27 Recall that a number of examples in chapter two show that DOs can express the same semantic role as IOs with aider-type verbs, while the reverse is not true.

116 uniform diffusion of the change, we follow the commonly held view that the syntax of Medieval French is sufficiently homogeneous to justify a unified treatment.

Texts dating from the Middle French period (14th and 15th centuries) to the 19th century were searched for each of the nine verbs, taking into account alternate spellings, the range of possible inflections, etc. Note that the entire Old French period (9th to 13th century) was excluded from the quantitative study since, during this period, there is significant overlap between the form used for dative pronouns and that used for tonic accusative pronouns. For instance, the distinction between li (dative) and lui (tonic form of accusative le) is not always reliable, not to mention the fact that li was also the tonic form of accusative la. The pronominal form li in a sentence like Guillaume li aide is thus ambiguous; it could either be a dative pronoun or a tonic accusative pronoun. Similarly, all forms of lor are also unreliable, as it was at once the form of the third person plural dative pronoun and the tonic form of the third person plural accusative pronoun. In addition, tonic pronouns are employed as a rule in Old French before nominal forms of the verb (infinitives and forms in –ant). This again makes any formal distinction between the accusative and dative case impossible in such contexts: Pour lui querre s’achemina ‘He took to the road in order to go and look for him’ (BibleM, Rois, 15584; Buridant 2000, §356). Data from the 9th to 13th century was therefore excluded on the grounds that pronominal forms are unreliable during this period.

Each occurrence of a direct or indirect object was controlled for the construction in which it was found, its meaning and its form. As explained in section 2.3, only constructions of the S-V-O (que/ à/ de...) type were considered. In addition, only meanings for which the IO was replaced by an equivalent DO were counted. For example, only contexts in which assister à meant ‘to help’ were noted, meanings such as ‘to attend, to be present, to frequent’ were not.

Also excluded from the analysis was the object of two or more coordinated verbs occurring before the 17th century. Until the 17th century, the object, be it a pronoun or full nominal form, could be governed by two or more coordinated verbs even if it bears a different case for each verb: si l’appelay et demanday dont il venoit ‘and I called him and asked him from whence he was coming’ (A. de la Sale, Jehan de Saintré, 103, 4; Zink 1997:197). While the case agreement of the clitic is often assigned by the closest verb, it is not systematic, and for full noun phrases, case agreement is not predictable at all (see Fournier 1998:102, 104).

117

All nominal forms of the object were counted as were most occurrences of third person direct object pronouns (l’ le, la, les) and indirect object pronouns (lui, leur, y). For authors born during the 14th century, the following two pronominal forms were excluded: l’ due to potential ambiguity between an elided accusative (le, la) and an elided dative pronoun (lui) (see Martin- Wilmet 1980, §262), and lui/leur directly followed by a nominal form of the verb (eg., pour lui aider) since in these contexts, such pronouns were still occasionally used for both the dative and the tonic form of the accusative (Zink 1997:207-8).

3.2 Evolution of individual verbs

This section summarizes the evolution of the form of the object of nine verbs from the list in (1): aider, applaudir, assister, commander, contrarier, contredire, ennuyer, favoriser and insulter.

3.2.1 Aider

Table 1: aider: 14th to 19th c., frequency of the IO in % (animate, inanimate and clitic)

IO ANIMATE IO INANIMATE LUI, LEUR IO/ total [+an] IO/ total [-an] cl/total IO [+an] 1300-1399 78 (190/245) 91 (21/23) 67 (128/190)

1400-1499 85 (102/120) 92 (11/12) 76 (78/102)

1500-1549 67 (75/112) 66 (19/29) 71 (53/75)

1550-1599 61 (107/175) 56 (29/52) 73 (78/107)

1600-1649 47 (70/149) 49 (41/83) 76 (53/70)

1650-1699 10 (22/227) 58 (28/48) 68 (15/22)

1700-1749 16 (14/90) 58 (37/64) 93 (13/14)

1750-1799 3 (9/290) 60 (50/84) 56 (5/9)

1800-1849 0 (0/180) 53 (26/49) 0 (0/0)

118

Graph 1: aider: 14th to 19th c., evolution of the frequency of animate and inanimate IOs

General observations:

During the 14th and 15th centuries, the animate IO is clearly the dominant form of the complement and appears to be in stable alternation with the animate DO, remaining between 79% and 85% of all tokens. From the beginning of the 16th century, however, there is a clear and constant decrease in frequency of the IO; it undergoes a sharp drop in the last half of the 17th century and is almost obsolete by 1750. The inanimates follow a similar evolution until the second half of the 17th century. At this point, their numbers do not fall off sharply; rather, the IO continues to occur between 50% and 60% of the time. Interestingly, the fourth column of the table suggests that as animate IOs occur less frequently in the texts, it is the clitic form that is preferred over the nominal: 13 out of 14 occurrences of animate IOs are clitics between 1750 and 1799.28

The data permit us to witness a near complete change in the expression of the object. For animate IOs, replacement of the IO by a DO begins with the generation of authors born at the beginning of the 16th century and ends with those born in the 19th century.

28 Schøsler (To appear) claims that the rise of the accusative form begins in the pronominal system and then spreads to full nominal forms. Our data does not track the possible differences between the pronominal and nominal forms of the accusative.

119

Particulars:

Inanimates which can be interpreted as autonomous individuals in the texts are classified as animate objects in the study. For example, (3a,b) are classified with the animate DOs, whereas (3c,d) are treated as inanimates. This said however, the distinction is not always clear.

(3) a. le geste aide le mot, l'oeil explique le coeur

‘hands speak when words cannot, as eyes do for the heart’

(Lamartine, Jocelyn, 1836, 623; Frantext, ARTFL)

b. Aidier nature a evacuer [la matière digérée], se elle encommence a faire par

lieu convenable.

‘To help nature to evacuate (the digested matter), if it begins to do it in a

convenient place.’

(Saint-Gille, Les Amphorismes Ypocras, 1362, 57; BTMF)

c. Ils [les muscles] doivent recourber le corps sur les côtés lorsqu'ils agissent

séparément, [...], raccourcir le corps, et aider ainsi la progression.

‘They (the muscles) must bend the body sideways when they act separately ...

shorten the body, and help it thus to move forward.’

(Cuvier, Leçons d'anatomie comparée, 1805, 435; Frantext, ARTFL)

d. Ta femme a bien aidé à notre refroidissement

‘Your wife has indeed contributed to our falling out’

(Balzac, Le Contrat de mariage 1842, 273; Frantext, ARTFL)

What appears to be happening with the inanimate IOs in graph 1, is that there is an initial and sudden replacement of all inanimate non eventive IOs by an analogous DO: namely personified inanimates, like those in (4).

120

(4) ... aide aux fractures des oz

(Saint-Gille, Les Amphorismes Ypocras, 1362, p.82, section v, 22 ; BTMF)

These kinds of objects are subject to the same forces of change as the animates. This micro change is completed by the beginning of the 17th century at which point the inanimate IOs level off and there is no apparent change in the frequency, diverging dramatically from the animates whose frequency continues to decrease. Interestingly, the IOs that remain are generally all eventive. Recall that aider followed by an eventive (inanimate) IO is still felicitous in Modern French; see chapter one, example (2c) and chapter two, example (3j). In chapters four and five, we discuss the reoccurrence of the correlation between eventive IOs and resistance to the valency change.

3.2.2 Applaudir

Table 2: applaudir: 14th to 19th c., frequency of the IO in % (animate, inanimate and clitic)

IO ANIMATE IO INANIMATE LUI, LEUR IO/ total [+an] IO/ total [-an] cl/total IO [+an] 1300-1399 (1/1) --- (0/1)

1400-1499 ------

1500-1549 100 (9/9) (2/2) 89 (8/9)

1550-1599 100 (5/5) 100 (8/8) 80 (4/5)

1600-1649 90 (18/20) 90 (38/42) 50 (9/18)

1650-1699 45 (15/33) 98 (142/145) 47 (7/15)

1700-1749 39 (11/28) 90 (112/124) 27 (3/11)

1750-1799 21 (6/28) 75 (85/114) 33 (2/6)

1800-1849 5 (3/60) 65 (62/96) 0 (0/3)

121

Graph 2: applaudir: 14th to 19th c., evolution of the frequency of animate and inanimate IOs

General observations:

Applaudir is a late 14th century borrowing from Latin and only occurs intransitively in the corpus until the 16th century at which point it is used exclusively with an indirect object. The first occurrences of a direct object are attested in the first half of the 17th century, and are followed by a dramatic decrease in the frequency of IOs in the second half of the same century. By the 19th century, the IO has all but been replaced by a DO in the texts. Like aider, the inanimate IOs follow a similar evolution to the animates until the 17th century, at which point they maintain their numbers while the animates undergo a rapid decline in frequency. Inanimate IOs persist in the texts, slowly diminishing from the 18th century. Contemporary speakers find applaudir followed by an indirect object to be odd, or archaic. Finally, there are not enough tokens to say whether the dative clitic is the preferred form or not as the frequency of the animate IO decreases; as it stands, dative clitics do not occur as frequently as the full nominal forms.

The data allow us to observe a near complete change in the expression of the object. For animate IOs, the shift to a direct object complement begins with the generation of authors born in the 17th century and ends with those born in the 19th century.

122

3.2.3 Assister

Table 3: assister: 14th to 17th c., frequency of the IO in % (animate, inanimate and clitic)

IO ANIMATE IO INANIMATE LUI, LEUR IO/ total [+an] IO/ total [-an] cl/total IO [+an] 1300-1399 (1/1) (2/2) (0/1)

1400-1499 (1/1) (2/2) (0/1)

1500-1549 69 (18/26) 60 (3/5) 67 (8/9)

1550-1599 5 (3/63) 60 (6/10) 100 (3/3)

1600-1649 3 (1/31) (0/3) 100 (1/1)

Graph 3: assister: 14th to 17th c., evolution of the frequency of animate and inanimate IOs

General observations:

The verb assister is a Middle French borrowing from the Latin assistere. Its early uses with an IO have the meaning of ‘to attend, to be present at an event’, ‘to frequent or to stand by someone’. With a DO, it often means ‘to elect or nominate someone’. None of these senses of the verb underwent a valency change and are thus not represented here. Rather, only in the sense of ‘to help’ does assister show a change in argument realization. This particular meaning of the

123 verb begins to appear robustly in the corpus at the beginning of the 16th century at which point the IO is clearly the dominant form of animate objects at almost 70% of all tokens. By the end of the same century however, the frequency falls dramatically to 5%. From 1650 onward, the IO is rarely attested with assister in the sense of ‘to help’; the DO has effectively replaced the IO in this context. Inanimate IOs follow the same evolution as the animates, differing only in that they persist in the texts at a constant rate for 50 years longer than the animates. The few animate IOs that appear after 1650 are all clitic forms.

Finally, it is unclear if the IO of assister only began to change frequency in the last half of the 16th century or if, in fact, it began earlier. It is quite possible that our data have not captured the beginning of the shift from an indirect object complement to a direct object complement.

3.2.4 Commander

Table 4: commander: 14th to 19th c., frequency of the IO in % (animate, inanimate and clitic)

IO ANIMATE IO INANIMATE LUI, LEUR IO/ total [+an] IO/ total [-an] cl/total IO [+an] 1300-1399 (4/4) (1/1) (1/4)

1400-1499 (2/2) (1/1) (1/2)

1500-1549 98 (87/89) 100 (35/35) 20 (17/87)

1550-1599 50 (28/56) 87 (13/15) 4 (1/28)

1600-1649 25 (36/143) 42 (25/59) 19 (7/36)

1650-1699 27 (28/105) 38 (8/21) 25 (7/28)

1700-1749 28 (24/86) 31 (19/62) 13 (3/24)

1750-1799 17 (21/123) 39 (12/31) 0 (0/21)

1800-1849 30 (29/96) 29 (17/59) 3 (1/6)

1850-1899 15 (6/39) 25 (8/32) 17 (1/6)

124

Graph 4: commander: 14th to 19th c., evolution of the frequency of animate and inanimate IOs

General observations:

Note that we only considered commander in the sense of ‘to command, be in charge of a person or group of persons of lower social or military rank’, or ‘to dominate a location’. Taking into account the few occurrences of animate IOs during the 14th and 15th century, the indirect object appears to be by far the preferred form at almost 100% until the first half of the 16th century. At this point, the relative frequency of animate IOs drops suddenly to 50% of all tokens gradually declining to 15% by the end of the 19th century. The IO appears to have been completely replaced by an analogous direct object in contemporary French, as speakers no longer accept commander used with an IO, save in the sense of having control or influence over one’s feelings or a body part (see chapter two, example (7j)). The inanimate IOs follow the same basic evolution as the animates, differing only in that there is a fifty year lag before they take too a sharp drop in numbers at the beginning of the 17th century. Once the IO begins to decline in frequency, the dominant form appears to be the full nominal form, not the clitic.

Like aider and applaudir, commander presents a near complete change in argument realization if we take into account the data from the 14th and 15th centuries. Since the IO constitutes 98% of all tokens at the beginning of the 16th century, it is reasonable to assume that the valency change begins with authors born during the second half of the 16th century. The IO is completely replaced by the DO in the 20th century.

125

Particulars:

Inanimate complements are classified as follows: any word that denotes a group of people, such armée, garnison, troupe, etc. is considered to be animate. Note, however, that in cases such as (5a), the pronominal form y indicates that the naval army is construed as unindividuated, a situation rather than a group of individuals. All other inanimate objects are classified as inanimates: emotions, body parts, elements of nature, a ship or flotilla, a nation, laws, opinions, business, and places. We maintain this general rule, regardless of whether some of the latter notions could arguably be classified as animates either by way of metonymy or personification. The way in which inanimate objects are conceived is highly subjective and varies according to author and context, as shown by the variation in the type of pronoun employed in examples (5b) and (5c). Such variation inevitably introduces a certain amount of error into the results, but we do not feel that it affects the general trend in graph 4.

(5) a. On dit que le roy est d'accord avec les hollandois pour dresser une armée

navale à laquelle chacun contribue force vaisseaux, et que l'archevesque de

Bordeaux sera celui qui y commandera pour le roy.

‘They say that the king agrees with the Dutch about raising a navy to which

each will contribute a good number of ships and that the archbishop of

Bordeaux will be the one who will command it for the king.’

(Patin, Lettres de Gui Patin, 1630-1672, 1649, 136; Frantext, ARTFL)

b. Constantinople fut toûjours regardée comme la capitale de tant de régions. Sa

situation semble faite pour leur commander.

‘Constantinople was always considered as the capital of a great many regions.

Its location seems perfect for governing them.’

(Voltaire, Essay sur l'histoire générale, 1756, 281; Frantext, ARTFL)

126

c. Si ton coeur me conserve une eternelle foy, si j'y commande seul, s'il n'écoute

que moy...

‘If your heart will forever believe in me, if I alone command it, if it only

listens to me...’

(Godeau, Poésies chrestiennes, 1633, 27; Frantext, ARTFL)

3.2.5 Contrarier

Table 5: contrarier: 14th to 18th c., frequency of the IO in % (animate, inanimate and clitic)

IO ANIMATE IO INANIMATE LUI, LEUR IO/ total [+an] IO/ total [-an] cl/total IO [+an] 1300-1399 38 (3/8) (3/3) 100 (3/3)

1400-1499 (3/3) 100 (5/5) 100 (3/3)

1500-1549 56 (5/9) 80 (4/5) 80 (4/5)

1550-1599 59 (10/17) 74 (23/31) 70 (7/10)

1600-1649 0 (0/6) 6 (1/16) 0 (0/0)

1650-1699 (0/1) (0/4) 0 (0/0)

1700-1749 0 (0/5) 0 (0/10) 0 (0/0)

127

Graph 5: contrarier: 14th to 18th c., evolution of the frequency of animate and inanimate IOs

General observations:

The corpus does not provide a robust number of occurrences of the verb contrarier. Consequently, graph 5 only offers a general impression of how the IO evolved from the 14th century. Regardless of the uncertainty as to how this verb was used during the 14th and 15th centuries, it is nevertheless worthwhile noting that animate IOs undergo a dramatic drop in frequency at the beginning of the 17th century. They fall from over half of all animate tokens in the 16th century to none at all a century later. The inanimate IOs disappear from the corpus at the same time, but here the data shows that they begin decreasing from the beginning of the 16th century. The sudden loss of the IO is supported by meta-linguistic texts. Nicot’s (1606) Thresor de la langue françoyse, tant ancienne que moderne records contrarier exclusively with an indirect object complement while less than a century later, the 1694 edition of the Dictionnaire de l’Académie Française shows contrarier used only with a direct object.

In sum, the inanimate IOs present a complete change in argument realization. The change begins with the generation of authors born at the beginning of the 16th century and ends 100 years later. As for the animates, it is difficult to tell whether the data capture them in mid change from the 16th century onward, or whether the valency change only really begins in the 17th century.

128

Particulars:

If an inanimate object was clearly personified, it was classified among the animates. For example, mon cuer in (6a) was counted among the animates, whereas vostre contentement in (6b) was not.

(6) a. Mais vo douce maistrie Maistrie mon cuer si durement Qu' elle le contralie

‘But your gentle authority rules my heart so hard, it tortures it’

(Machaut, Les Chansons baladees,1377, 584; BTMF)

b. quel malheureux demon a voulu contrarier vostre contentement

‘what ill-seeking devil wanted to disturb your happiness’

(Urfé, L'Astrée t.4, 1624, 398; Frantext, ARTFL)

3.2.6 Contredire

Table 6: contredire: 14th to 18th c., frequency of the IO in % (animate, inanimate and clitic)

IO ANIMATE IO INANIMATE LUI, LEUR IO/ total [+an] IO/ total [-an] cl/total IO [+an] 1300-1399 38 (15/39) 51 (28/55) 33 (5/15)

1400-1499 45 (5/11) 90 (36/40) 40 (2/5)

1500-1549 64 (14/22) 80 (30/38) 50 (7/14)

1550-1599 30 (3/10) 63 (15/24) 100 (3/3)

1600-1649 10 (6/59) 24 (14/58) 33 (2/6)

1650-1699 5 (2/37) 11 (5/45) 100 (2/2)

1700-1749 0 (0/22) 2 (1/57) 0 (0/0)

129

Graph 6: contredire: 14th to 18th c., evolution of the frequency of animate and inanimate IOs

General observations:

The frequency of the animate IO, at 64% of all animate tokens at the beginning of the 16th century, begins to decline during that same century and is no longer present in the corpus by the beginning of the 18th century. Inanimate IOs follow a very similar evolution. The dative clitic appears to be the dominant form once the animate IO begins to decrease in use, but the numbers are not very substantial. It is worth noting that some uses of contredire followed by a DO tended to mark a slightly more affected object, having the sense of ‘to oppose, refuse, or impede someone or something’. According to the Dictionnaire Historique and an examination of the data, this sense of the verb fell out of use in the 16th century, which would explain why we see a gradual and relative rise in frequency of the IO during the 15th and the beginning of the 16th century. Contredire presents a complete change in argument realization. It clearly begins with the generation of authors born in the second half of the 16th century and is completed by the 18th century.

130

3.2.7 Ennuyer

Table 7: ennuyer: 14th to 17th c., frequency of the IO in % (animate, inanimate and clitic)

IO ANIMATE IO INANIMATE LUI, LEUR IO/ total [+an] IO/ total [-an] cl/total IO [+an] 1300-1399 95 (53/56) --- 57 (30/53)

1400-1499 93 (13/14) --- (1/1) 77 (10/13)

1500-1549 38 (3/8) --- 100 (3/3)

1550-1599 19 (3/16) --- 100 (3/3)

1600-1649 0 (0/12) --- 0 (0/0)

1650-1699 0 (0/15) --- 0 (0/0)

Graph 7: ennuyer: 14th to 17th c., evolution of the frequency of animate IOs

General observations:

A quintessential object experiencer verb, ennuyer essentially only occurs with animate objects in our corpus. The IO is by far the preferred complement form during the 14th and 15th centuries, but drops suddenly to less than 40% of all tokens for authors born in the first half of the 16th century. The frequency continues to decline and by the 17th century, the IO has completely

131 disappeared from the texts. Ennuyer presents a complete change in argument realization, starting at the beginning of the 16th century and ending at the beginning of the next.

3.2.8 Favoriser

Table 8: favoriser: 14th to 17th c., frequency of the IO in % (animate, inanimate and clitic)

IO ANIMATE IO INANIMATE LUI, LEUR IO/ total [+an] IO/ total [-an] cl/total IO [+an] 1300-1399 50 (6/12) (1/3) 50 (3/6)

1400-1499 45 (10/22) 80 (4/5) 40 (4/10)

1500-1549 7 (2/28) 18 (5/28) 50 (1/2)

1550-1599 1 (1/85) 4 (3/85) 100 (1/1)

1600-1649 0 (0/33) 0 (0/32) 0 (0/0)

Graph 8: favoriser: 14th to 17th c., evolution of the frequency of animate and inanimate IOs

General observations:

Although favoriser is not robustly attested in our corpus, the animate IO nevertheless shows a clear evolution from comprising almost half of all tokens in the 14th and 15th centuries, to disappearing from use by the end of the 16th century. The inanimates, whose frequency is initially much higher than the animates, follow a similar evolution. They are completely replaced

132 by an analogous DO by the beginning of the 17th century. At no point do pronominal clitics appear to be favoured over nominal forms. For the animate IO, favoriser presents a complete change in argument realization. The change begins at the beginning of the 16th century and is almost complete by the end of the same century. Favoriser used with an indirect object is definitively absent from the corpus by the beginning of the 17th century.

3.2.9 Insulter

Table 9: insulter: 14th to 19th c., frequency of the IO in % (animate, inanimate and clitic)

IO ANIMATE IO INANIMATE LUI, LEUR IO/ total [+an] IO/ total [-an] cl/total IO [+an] 1300-1399 ------

1400-1499 ------

1500-1549 ------

1550-1599 (1/1) (3/3) 0 (0/1)

1600-1649 68 (13/19) 88 (37/42) 15 (2/13)

1650-1699 38 (31/82) 71 (42/59) 35 (11/31)

1700-1749 5 (1/19) 72 (13/18) 0 (0/1)

1750-1799 11 (6/55) 63 (35/56) 0 (0/6)

1800-1849 0 (0/70) 3 (1/33) 0 (0/0)

133

Graph 9: insulter: 17th to 19th c., evolution of the frequency of animate and inanimate IOs

General observations:

Insulter is a Latin borrowing, and although the Dictionnaire Historique and the TLF claim the earliest attestation to be in Bersuire (1352-56), it does not appear with a complement in our corpus until the 16th century. Nevertheless, animate IOs are frequently found in the texts at the beginning of the 17th century, but rapidly decline in frequency from that point until they are all but replaced by a DO in the 18th century. The IO is by far the preferred form for inanimate objects, persisting a century longer than the animates. Interestingly, it too undergoes a rapid decline in frequency, virtually disappearing from use with authors born at the beginning of the 19th century. For example, while Chateaubriand (b. 1768) uses an inanimate indirect object 18 times out of 30, Hugo (b.1802) only uses an inanimate indirect object one time out of 26.

Particulars:

The DO/IO alternation does not appear to correspond to any significant meaning difference, save the particular sense of physically attacking someone or something, which was only used with a DO and which is considered archaic today. Although the DO with this meaning did not replace an equivalent IO, it was not excluded from the data since it was often hard to judge whether it was used concretely or metaphorically. Moreover, insulter in the sense of physically attacking someone is not prevalent enough in the corpus to appreciably affect the results. If it had been successfully excluded, the relative frequency of the IOs would simply be slightly higher.

134

3.3 Summary and analysis

This section provides a summary and discussion of the results of our quantitative study. An interpretation of the results is offered with attention to some of the more important assumptions concerning the graphic representation of the diffusion of the change. The initial period of stability, when there is no valency change, is discussed at some length, as is the inception of the valency change. Given that the evolution of all of the animate IOs follows the typical S-shaped curve while the inanimates do not necessarily do so, much of the discussion relates only to the animates. The significance of the two patterns of change (animate vs. inanimate) is taken up in chapters four and five.

Seven of the nine verbs studied in detail present a complete evolution of the IO allowing a comparison of the point at which the change in argument realization began for each verb and the diffusion of the change that followed. For all of the seven verbs, the actuation of the change clusters between 1500 and 1649, while the subsequent rate of replacement of the IO by a DO varies widely from verb to verb. The limited period within which the IOs began to change strongly supports the hypothesis of a common cause of change and rules out random causes such as analogy or the notion that each verb underwent a subtle semantic change which disfavoured the use of an IO. Furthermore, since the change can be shown to begin in speakers born before the influential grammarians and Remarqueurs of the 17th century, the latter are necessarily ruled out as the cause of the valency change. Our hypothesis is therefore supported. The common cause of change must be located in the abstract representation of a grammatical element common to all of the aider-type verbs listed in (1).

3.3.1 The S-shaped curve

Much recent work treating language change notes that the time course of the diffusion of a linguistic replacement typically follows an S-shaped curve (Croft 2000; Denison 2003; Kroch 1989, etc.). Change tends to proceed slowly in the beginning, then dramatically picks up momentum to finally slow down once again as the last vestiges of the older form linger on. In Denison’s words, “What actually happens much of the time is more like ‘slow slow, quick, quick, slow’.” (p.55). An idealized graph of language change would look like that in Figure 1.

135

100 -

80 -

60 -

40 -

20 -

0 time

Figure 1: Idealized graph of the diffusion of a linguistic replacement (adapted from Denison 2003:56)

Kroch (1989) observes that the S-shaped curve is a characteristic property of changes that have been studied quantitatively, and notes that it is not unique to language change, but is equally characteristic of domains such as population biology which also study changing percentages of alternating forms over time. It is not surprising then, that the S-shaped curve reappears in the graphs of the seven verbs whose animate IO presents a near complete evolution: aider, applaudir, commander, contrarier, contredire, ennuyer, and favoriser. Note, however, that because our study measures the frequency of the form that is being replaced (IO), rather than the replacement form (DO), our graphs are actually upside down S-shaped curves. If the data had been robust enough to provide a quantitative description of earlier uses of assister and insulter, we would expect that the valency change in these verbs would also follow the same trajectory. Note however, that the decrease in frequency of the inanimate IOs does not always follow an S- shaped curve. This is important, suggesting that the inanimates may not have been subject to the same mechanism of change as the animates.

3.3.2 Initial period of stability

Verbs showing a complete change suggest an initial period of stability, a period where there is relatively no change in the frequency of the DO and the IO. We have allowed up to a 5% drop in frequency to constitute a stable interval. Graph 10 illustrates the period during which the indirect object of applaudir and commander appears to be the only one used in the texts.

136

Graph 10: Initial period of stability for applaudir and commander

Similarly, the graphs of aider, contrarier, contredire, ennuyer, and favoriser also present an initial period of stability, but in these cases, it is a period of stable alternation between an IO and a DO complement: 29

Graph 11: Initial period of stability for aider, contrarier, contredire, ennuyer and favoriser

29 Recall that the rise in frequency of the IO of contredire at the beginning of the 16th century is a result of the loss of one of its meanings, ‘to oppose or impede’, which was exclusively transitive. The period from the 14th to the 16th century can thus otherwise be considered a period of stability.

137

The fact that these periods of stability are relatively long suggests that during this time the expression of the object was not subject to change. We do not assume, for instance, that at some earlier period in the history of the language which was not examined, i.e., Old French, the verbs only took an IO as a complement. We feel that it is unnecessary and likely incorrect to make the assumption that at some point in the history of each of these verbs, only one form of the object was used. This would imply that all alternations represent a change in progress, which would be wrong. Languages abound with productive and stable alternations. The existence of this initial period when there was no change of valence allows us to accurately locate in time the actuation of change, taken up in the next section.

Before considering the point at which the valency change began for each verb, the notion of two forms in stable alternation must be considered. Within a generative grammar framework, stable alternations must be justified since free variation, or optionality, runs contrary to the way in which the theory accounts for the logical problem of language acquisition. Given that children rapidly acquire a productive grammatical system on the basis of very little experience, a minimalist approach to language acquisition assumes that a child acquires her grammar based on the interaction between her linguistic experience (PLD) and universal grammar (UG). Based on positive evidence only, a child’s developing grammar must be able to rule out certain utterances according to abstract knowledge about the developing grammar, and not according to what she does not perceive. Otherwise, we would not expect a child to acquire language as quickly and systematically as it does and we would not expect adult speakers to make definitive grammaticality judgments about the language they speak.

If case marking were optional, for instance, without any real semantic or syntactic motivation, then one would expect children to generate utterances like that in (7a) from time to time along side similar sentences like that in (7b).

(7) a. * Paul a écrasé à la mouche et lui a tué.

‘Paul crushed to the bug and killed to it.’

138

b. Paul a écrasé la mouche et l’a tuée.

‘Paul crushed the bug and killed it.’

This kind of optionality is neither attested in French nor cross-linguistically (see Levin (1999) and Levin and Rappaport Hovav (2005:21) and references cited therein), and it does not seem to arise as a common grammatical error in children, either. French children do not produce sentences like (7a) because very early on, they learn an abstract rule that directly or indirectly excludes IOs with verbs that denote events in which an animate agent acts on and causes a change of state in a patient.

If optionality were not excluded from the learning algorithm, then the child learner would make errors like that in (7a) and would require negative evidence or corrective measures in order to conclude that écraser à quelque chose is incorrect. It has been shown, however, that negative data is generally not available to children (see Pinker 1989, and references cited therein), so the knowledge that (7a) is incorrect must come for free, that is, it must follow from a related parameter learned early during a child’s language development.

When grammars allow for alternating forms, the alternations must therefore be motivated, either morphosyntactically, semantically, or phonologically. Verbs provide particularly good examples of how stable and productive semantic alternations are common to human language and extend linguistic expression. When a verb is used in two different constructions, which at first blush seem to have the same meaning, differences between the two can usually be identified. Locative alternations illustrate this point:

(8) a. Jean a chargé le foin dans le camion.

‘Jean loaded hay into the truck.’

b. Jean a chargé le camion de foin.

‘Jean loaded the truck with hay.’

While the pair of sentences in (8) appear to present two options for describing the same event, they are not truth-conditionally equivalent. (8a) describes a causative event in which Jean causes hay to go into the truck whereas (8b) describes a change of state in which the truck goes from a

139 state of not being loaded to one of being loaded. Only (8b) entails that the truck is affected in some way, with a natural interpretation in this case that it has been completely loaded.30 In contrast, (8a) may describe a scenario in which a significant part of the wagon is still empty.

Returning to the issue at hand, we assume that aider, contrarier, contredire, and favoriser once had a productive IO/DO alternation. Since there is no reason to believe that this alternation is driven by morphosyntactic or phonological factors, there must be semantic differences underlying the IO/DO alternation. This is particularly difficult to argue in historical linguistics as subtle differences in meaning are most often invisible to the linguist who must rely on textual evidence alone. Despite this handicap, there is nevertheless some indication suggesting that a stable, semantically motivated alternation is a reasonable assumption to make.

As shown in chapter two, semantically motivated IO/DO alternations have indeed been identified for several of the verbs listed in (1). The DO tends to have a more affected interpretation in these instances, while the IO is less affected by the actions or intentions of the subject. When contrarier has the meaning of tormenting or bothering someone, its complement always appears as a DO whereas the IO designates an entity toward whom the subject simply demonstrates some kind of opposition. Similarly, until the 16th century, contredire could have the particular meaning of ‘to oppose, refuse, or impede someone or something’ and in this sense, its object is always a DO. The IO complement, however, refers to the entity against whom the subject speaks out. Furthermore, it has been noted that when insulter had the particular meaning of physically attacking someone or something, it was only used with a DO, but the object could be realized as an IO when its referent is simply insulted or affronted.

It is important to note the asymmetry that exists between the way a DO can be used and the way an IO can be used. While the DO is construed as an affected object, it can also be used in the same context and can convey the same meaning as the IO. The reverse is not true, however, for

30 Following Tenny (1994), the direct internal argument of the verb is unique with respect to affectedness. An affected argument is specifically one that undergoes a change of state and this change effected in the object measures out the event. In (8b), the truck can be construed as the entity which measures out the hay-loading. In (8a) on the other hand, the truck is not the direct internal argument of the verb and is therefore not interpreted as undergoing any particular change of state.

140 the IO. An IO never takes on the role of an affected object. The use of the verb insulter in Medieval French illustrates this asymmetry very well. Example (9a) shows that the DO can have an unaffected meaning, i.e., the object of an insult, while (9b) shows that it can also be construed as affected, i.e., the object of a physical attack. In contrast, the IO only ever has the unaffected interpretation, as shown in (10a). We have found no occurrence of it in the corpus as the object of a physical attack, demonstrated in (10b).

(9) a. Il l’insultoit dans tous ses discours

‘He insulted him in all that he said’

(Vertot, Histoire des Revolutions, 1719, p.447; Frantext, ATILF)

b. Dans une grande colère, le corps se trouve plus prêt à insulter l’ennemi et à

l’abattre

‘In great anger, the body is more apt to attack the enemy and to bring him

down’

(Bossuet, De la Connaissance de Dieu, 1704, p.142; Frantext, ATILF)

(10) a. en prenoient lieu d’insulter aux patriciens

‘they took the opportunity to insult the aristocrats’

(Vertot, Histoire des Revolutions, 1719, p.462; Frantext, ATILF)

b. * Dans une grande colère, le corps se trouve plus prêt à insulter à l’ennemi et à

l’abattre

What is at stake therefore, is a DO which is not limited to a particular interpretation (it can denote a more or less affected object), and which, at some point, replaces certain IOs which are limited in their interpretation.

Another piece of evidence in support of our claim that the stable IO/DO alternation corresponds to a meaning difference is that semantically motivated IO/DO alternations exist in other Romance languages for the translational equivalent of verbs such as aider, favoriser, prier, etc. Strozer (1976) argues that certain dative IOs in Spanish have a greater degree of participation in the event than DOs; examples from Strozer (1976:453).

141

(11) a. Tienes que ayudarla. Ha caído del tejado y no puede moverse

‘You have to help her. She’s fallen off the roof and can’t move’

b. Tienes que ayudarle con las preparaciones

‘You have to help her with the preparations’

Strozer explains that in (11a), in which the object (la) of ayudar ‘to help’ is direct, “the person needing help cannot move. She is helpless, not in control, almost as if she were a physical object”. In contrast, when the object of ayudar is realized as a dative (le), as in (11b), it receives a different interpretation: “she participates in the preparations, and the indirect object indicates this participation” (Strozer 1976:453).

García (1975) proposes a similar analysis for the IO/DO alternation in Spanish, and concludes that “...literal or potential activity – that which distinguishes live being, and particularly people, from inanimate – also motivates the choice of le vs. lo” (Garcia 1975:326). A number of examples illustrating the meaning difference between the two forms are offered, but she stresses that the difference entailed by the use of the direct or indirect object may be very subtle and can often only be determined from an understanding of the entire context. Assuming this is the case for 14th and 15th century uses of aider-type verbs, it is not surprising that such differences have escaped analysis.

Delbecque (1998) further demonstrates that the IO/DO alternation in Spanish two-place verbs is quite productive, citing semantically similar verbs to those in (1) such as entorpecer ‘hinder’, estorbar ‘perturb’, favorecer ‘favour’, dominar ‘dominate’, regir ‘govern’, among others. She argues that in active events the direct construction corresponds to a purely affected participant whereas the indirect construction involves an autonomous, reactive participant.

This semantically motivated alternation also exists in Latin, illustrated in Blake (2001:145). When the verbs temperare and moderor ‘to temper’ or ‘to restrain’ refer to restraining or controlling an entity external to the agent, they are used transitively (12a), but where they refer to exercising self-restraint, they take a dative complement (12b).

142

(12) a. Ego moderor equum meum

I moderate horse.ACC my.ACC

‘I control my horse’

b. Ego moderor orationi meae

I moderate speech.DAT my.DAT

‘I moderate my speech’

Blake goes on to explain that “[w]ith self-restraint the restraint is directed towards oneself and the entity specified in the dative is only indirectly involved” (Blake 2001:145).

In sum, it was shown that there is a period that precedes the valency change during which time the IO and the DO are in a stable alternation. We argued that this alternation was semantically motivated. Although there is little direct evidence in favour of such an analysis, there is ample indirect evidence to support it. Establishing a period of no change is important because it allows us to determine the point at which the frequency of the IO relative to the DO begins to decline. Establishing the actuation of the change is crucial to our quantitative study for it allows us to determine if the valency of each verb began to shift at the same time. Clustering of this sort, as is shown in the next section, strongly suggests a unified change, a sound argument in favour of a common cause underlying the valency change.

3.3.3 Actuation

In the previous section, an initial period of stability was identified for seven of the nine verbs. During this time, there was no valency change.31 The period that directly follows, when the IO begins to decline in frequency, signals the actuation of the change in argument realization, summarized in graph 12. Each verb is represented by two points. The first indicates the end of the period during which there is no change in the IO. The second point describes the first indication of a significant decline in the frequency of the IO: the actuation of the change.

31 Recall that for assister and insulter, the data was not robust enough to establish a period of stability.

143

Graph 12: Actuation of the valency change for aider, applaudir, commander, contrarier, contredire, ennuyer, favoriser

Considering the second point only, graph 12 shows that the actuation of the change clusters between the 16th and the beginning of the 17th century. The frequency of the IO of aider, ennuyer and favoriser begins to decline in the first half of the 16th century, followed by commander and contredire at the end of the same century, and then by applaudir and contrarier during the first half of the next. The actuation of the change is therefore clearly located within a 150-year period; authors born between 1500 and 1649 are the first to show an increasing preference for a DO over an IO for these seven verbs.

Recall that in section 1.1.1 the distinction was made between actuation, corresponding to a change in grammar, and diffusion of the change, associated with the spread of the new grammar in the language. This study is concerned with grammatical change and therefore focuses on the points at which the valency change first began – not on the evolution of the change. When changes occur during the same period of history, there is a strong possibility that they are each the effect of a common cause. The clustering of the actuation of change between 1500 and 1649 strongly suggest therefore that verbs change as a group.

As was outlined at the beginning of the chapter, there are at least two possibilities for a common cause of change: either the valency change is the effect of 17th century writers, grammarians and Remarqueurs, or it is the effect of an abstract, underlying change in speakers’ grammars. Graph 12 effectively rules out the first possibility since the valency change was already underway

144 before the influential writers were born. Vaugelas, for instance, was born in 1585, but the IO for five of the verbs in graph 12 had clearly begun to be replaced by an analogous DO before his birth – and before that of many other influential figures in the history of the French language. The change did not therefore come from the top; the elite did not therefore decide once and for all which form would be used and which would be discarded. They simply prescribed the form or forms that they accepted as correct, which often corresponded to those most commonly used at the time. In other words, they didn’t initiate change, they merely made it official and consequently, may have accelerated it. This is corroborated by Schøsler’s (To appear) study of the evolution of aider in informal vs. formal texts. She shows that the rise of an accusative object with aider began in récits and private letters well before it is seen in the formal writings of the 18th century. This rules out the normative tendencies of Classical and Post-classical French as a possible cause. In chapters four and five, we therefore develop the idea that the single cause of change is located in the abstract grammar of event structure.

3.3.4 Diffusion

This section discusses the propagation of the change, the period during which the animate IO declines in frequency and is eventually replaced by an analogous DO. Contrary to the period of stability, we do not expect the IO/DO alternation during this period to correspond to an appreciable meaning difference. Following Kroch (1989 and subsequent publications), we assume the alternation is the effect of coexisting grammars.

As sociolinguists have insisted (Labov, 1982; Weinreich, Labov, & Herzog, 1968), people live in linguistically heterogeneous environments and both learn and use their language under these conditions. Furthermore, the widespread occurrence of bilingualism and diglossia shows that people often know more than one grammatical system; [...] variation often reflects choices that are not categorically determined by linguistic principles at any level but instead are only probabilistically influenced by features of context and situation. In the case of replacement of one form by another, this is the expected circumstance, at least so long as the change is moving forward and does not turn into a stable alternation. (Kroch 1989:202, underlining is ours)

Theoretically, this assumption offers a way of accounting for what appear to be optional derivations for a verb phrase, and it accounts for the gradual spread of the new form through a

145 speech community. For the first point, recall, as discussed in 3.3.2, that within a Minimalist framework, a grammar should not permit optional operations. Yet, if a speaker may operate with more than one grammar, then each grammar can conceivably have a distinct derivation for the same basic verb phrase.32 Empirically, as long as the change is progressing, we shouldn’t expect to find the same kind of meaning differences between a verb used with an IO and the same verb used with a DO as we do during the period of stability.

In so far as the frequency of the animate IO begins to decline during the same time period for all of the verbs, it does not appear to be the case that they share the same rate of change. The verbs vary in terms of how long the IO persists in the texts. For example, it takes between 300 and 350 years for the frequency of the animate IO of both aider and commander to drop by 80%. On the other hand, it takes only 150 years for the IO of ennuyer to drop from 87% to 0%, and 50 years for the IO of favoriser to go from 59% to 0%. We note that these facts do not reflect Kroch’s (1989) Constant Rate Hypothesis for which it is proposed that different contexts of the same underlying grammatical change will proceed at the same rate. The hypothesis is substantiated by cases of syntactic reanalysis, and since it is not the case that the IO in French was reanalyzed as a DO, it is not surprising that the rate of change is not the same for each verb.33

The variable rates at which the frequency of the IOs decrease must reflect functional or stylistic factors related to each verb. One might speculate, for instance, that the reason the IO of ennuyer is replaced so quickly and definitively is related to the fact that this verb only occurs with animate IOs. In addition, it may not be a coincidence that the IO of aider and commander persists longer in the texts than that of other verbs. A possible reason that we see this lag may be related to the kind of constructions in which these verbs are often used. Aider is often followed by an infinitival complement (aider à faire quelque chose) as is commander (commander à quelqu’un de faire quelque chose). The next section presents some evidence that these contexts

32 Grammars are distinct as long as they differ by at least one parameter, and by parameter, this can mean a single formal feature of a listeme.

33 Note also that Kroch’s Constant Rate Hypothesis requires a sophisticated statistical treatment of the data, not available here, in order to demonstrate the uniform diffusion of an underlying grammatical change.

146 favour a dative object clitic, but any conclusion would require a close examination of the data, which surpasses the present study.

3.3.5 The dative clitic

The quantitative study controlled for the persistence of the dative clitic when the IO was declining in frequency based on the observation that pronominal clitics characteristic of earlier valency patterns seem to persist longer than their full nominal counterparts. In Canadian French, for instance, aider can be used with an indirect object or direct object clitic without any meaning difference. The difference between the two forms is purely one of register. Contrasting with the dative clitic, many of my informants hesitate with a full indirect object:

(13) a. Va lui/ l’aider à ranger sa chambre.34

‘Go and help her clean up her room.’

b. Va aider ? à Marie/ Marie à ranger sa chambre.

‘Go and help Marie clean up her room.’

Similarly, prier is still accepted in both Canadian French and French spoken in France. Again, the difference is a question of register. Note also that while the clitic is acceptable to some degree, the full nominal form is out:

(14) a. Marie, en s’adressant à sa mère, lui/ la prie de parler au jeune homme pour

qu’il lui fiche la paix.

‘Turning to her mother, Marie begs her to speak to the young man so that

he will leave her alone.’

b. Marie prie *à sa mère/ sa mère de parler au jeune homme pour qu’il lui fiche

la paix.

‘Marie begs her mother to speak to the young man so that he will leave her alone.’

34 The dative clitic is pronounced more like “y”: “Va y aider à ranger sa chambre”.

147

One might expect pronominal clitics to persist longer than full noun phrases since nominal forms are subject to licensing requirements in overt syntax. When a licensing environment is no longer available, neither is the noun phrase. A clitic on the other hand, is simply the spell-out of a functional head and is realized independently of the licensing requirements on its nominal counterpart. Acceptability depends on the frequency with which it is perceived to occur in the language and with the canonical meaning that the clitic has come to acquire based on its use with other verbs.

Our data illustrate that as the IO declines in frequency, the dative clitic tends to persist with aider, assister, contredire and ennuyer. This does not appear to be the case, however, with the other verbs. The precise factors that determine the retention of the clitic form have not been pursued here, but it seems safe to say that they are not syntactic (they do not depend on the presence of an infinitival clause, for example) or phonological. They likely depend on a complex interplay of verbal semantics, analogy and frequency.

3.3.6 Animacy

The effects of the valency change are different depending on the animacy of the object. Of all the nine verbs studied in detail, only the animate and inanimate indirect objects of three (contrarier, contredire and favoriser) present the same evolution.35 In all the other cases, the time course of the change with inanimates either mirrors that of the animates but lags behind 50 to 100 years (assister, commander, insulter), or it doesn’t follow the typical S-curve at all (aider, applaudir). Another difference between the animates and the inanimates is that some inanimate IOs are still acceptable, albeit in limited contexts, in Modern French; see the examples given in chapter two for aider, applaudir, commander, and insulter.

The animate IOs, or datives, in our study change in a predictable and uniform manner: they have been completely replaced by an analogous DO and the time course of their replacement follows a typical S-curve. In contrast, the inanimate IOs do not change in a uniform or predicable way. Some are still used in Modern French, and the time course of their replacement does not always follow an S-curve. It appears, therefore, that the underlying cause of the valency change did not

35 Recall that ennuyer does not occur with inanimate objects in our study.

148 have the same effect on the inanimates as it did on the animates. Rather, it appears as though the datives lead the change and that for the most part, the inanimates follow, perhaps by analogy. In chapter four, we take a closer look at the animate/inanimate distinction to find that in fact, it is better described in terms of the level of abstraction of the object. It turns out that first and third order entities (mainly animate) are replaced by a direct object, while second order entities (always inanimate) tend to resist the change.

3.4 General evolution of remaining verbs

A general survey was undertaken of the remaining 17 verbs in (1) in order to see if their IOs follow the same basic evolution as observed for the nine verbs examined in detail. For the survey, we did not search the textual data bases used in the quantitative study. Rather, we consulted a number of meta-linguistic texts: historical and contemporary dictionaries, lexicons, and grammars, Vaugelas’s Remarques, and the Commentaires of members of the Académie Française. Table 1 summarizes the results of the survey where a check indicates at least one reliable attestation of an IO during that period. We also note when the IO is qualified as either undesirable (bad), rare, or old. Note that the survey does not classify the occurrences by the author’s date of birth but rather by the date it was written.

Table 10: Survey of the use of an IO with 17 aider-type verbs

14th – 15th 16th 17th 18th 19th congratuler [+an] --- ✓ ✓ dominer ✓ ✓ --- (bad) empêcher [+an] ✓ --- ✓ (rare) épargner [±an] ✓ ✓ étudier [-an] ✓ ✓ ✓ offenser [±an] ✓ persuader [+an] --- ✓ ✓ ✓ (old) prier [+an] ✓ ✓ (bad) secourir [-an] ✓ ✓ servir [±an] ✓ ✓ (bad) supplier [+an] ✓ ✓

149

The results in table 1 mirror to a large extent those from the quantitative study. Although they only sketch an evolution of the IO, two things are clear: first, the IO was used productively in the 14th and 15th centuries and second, the 16th century is likely a turning point for the IO. The DO would have begun to replace the IO in the 16th century, since, by the 17th century the change is completed for half of the verbs and by the 18th century, the DO is the dominant form of the object in the corpus – save for persuader. The survey therefore offers a rough idea of when the IO was replaced by an equivalent DO for the rest of the aider-type verbs. It does not, of course, describe the process of change. We know, for instance, that prier is still marginally acceptable when used with a dative clitic in Modern French, but this is not reflected in our results.

Finally, the prolonged occurrence of the IO of persuader and prier is reminiscent of what we observed for aider and commander in the quantitative study. Interestingly, persuader and prier are also commonly used with an infinitival complement which, as suggested earlier, may be a favorable context in which the IO (perhaps only the pronominal form) persists.

3.5 Chapter Summary

The nine verbs examined in detail and the survey of the remaining verbs in (1) clearly show the effect of a change in the grammars of French speakers born after 1500. Speakers born in the early 16th century must have been exposed to PLD that differed significantly from the PLD to which previous generations of speakers were exposed. It caused these speakers to develop two grammars: one permitted aider-type verbs to occur with indirect objects, and one did not. Recall from chapter one, section 1.1.1, that actuation was stated as grammatical change, the reanalysis of at least one lexical or functional feature in an individual speaker. This is maintained, but in light of the idea of competing grammars, it must be restated as the adoption of an alternative inventory of features for a given listeme.

Speakers with these two grammars use both IOs and DOs in the same contexts with aider-type verbs and the use of DOs occurs more often than they do in the speech of their predecessors. The 16th century marks the inception of the change, but does not prevent newly introduced verbs like applaudir from being acquired as “dative” verbs. The actuation of the change may have taken place in a small group of acquirers, and as more were exposed to evidence of the new grammar (the one disallowing IOs with aider-type verbs), the valency change spread through the

150 population. This is language change, reflected in the recurrent S-curves which are, as Kroch observes, a general property of population dynamics. The loss of datives during the 16th and 17th centuries reflects a situation where more and more speakers propagate more and more evidence of an alternate grammar. Eventually, no later than the end of the 17th century, learners of French were no longer exposed to PLD that would trigger the old system. Dative IOs are consequently rarely used with aider-type verbs, save some lingering forms in specific contexts.

151

Chapter 4 Qualitative Study

This chapter provides a description of aider-type verbs with the goal of showing two things: on the one hand, what aligns aider-type verbs with transitive verbs and on the other, what distinguishes them from other verbs taking indirect objects in French. Our description should then shed light on why the IO was replaced by a DO and not some other object form, and why the IO of only these verbs changed and no others.

Section 4.1 defines the aider-type verbs in terms of Levin’ (1999) broadly defined natural class of non core transitive verbs. These verbs show a strong tendency to alternate between a direct and an oblique object. The analysis is based on two ideas. The first distinguishes between structurally and semantically licensed objects. Semantically licensed IOs are susceptible to changes in their morphosyntactic expression while structurally licensed IOs are not. The second idea is that the expression of semantically licensed objects depends on the particular system of oblique markers in a given language. Oblique markers, be they bound morphemes or prepositions, identify specific semantic roles whereas the direct (or accusative) object is the default case in most languages and is able to accommodate a range of semantic roles. In the absence of an appropriate oblique marker, semantically licensed objects tend to be expressed in the default case. We also demonstrate how Levin’s lexical approach to argument structure can be reinterpreted in structural terms, giving us a clear sense of the structural representation of a verb like aider. We propose that its sole internal argument is merged as sister to the verbal root and licensed exclusively by the meaning of the verb.

Section 4.2 demonstrates that the valency change in aider-type verbs is constrained in three principle ways. First, only dynamic two-place verbs underwent the valency change; the IO of trivalent verbs, unaccusatives and verbs denoting states does not undergo such a change. Second, only verbs unspecified for directionality undergo the valency change. Third, the change is mainly affects first and third order objects, following Lyons’ (1977) classification of nominal entities. These restrictions show that the valency change is principally constrained by argument structure. Further distinctions within the class of dynamic two-place verbs are of an “encyclopaedic” character involving the presence of directionality (or Path) as a feature of the verb’s meaning and the nature of the object referent. By clearly defining the parameters within which the valency

152 change takes place, we not only provide further evidence of a unified change, but we identify the properties that likely underlie the cause of the change. Section 4.3 concludes.

4.1 Aider-type verbs are NCTVs

Our description of aider-type verbs begins by defining them in terms of Levin’s 1999 “non core transitive verbs” (NCTVs). As opposed to “core transitive verbs”, NCTVs encompass a broad group of verbs of which the object does not conform to any uniform or general semantic role. Formally speaking, the object of these verbs is semantically as opposed to structurally licensed, and as such, presents much intra and crosslinguistic variation in its morphosyntactic expression.

4.1.1 Non core transitive verbs

This section follows Levin’s distinction between “core transitive verbs” (CTVs) and “non core transitive verbs” (NCTVs) and shows that the aider-type verbs conform in every way to the NCTV class.

In Modern Standard French, the verbs in (1a) are transitive. They take a direct object complement just as other transitive verbs do, such as those in (1b).

(1) a. aider/contredire/applaudir Jean

‘to help/contradict/applaud John’

b. couper/détruire/casser/ouvrir la porte

‘to cut/destroy/break/open the door’

In Levin’s (1999) examination of objecthood, verbs like those in (1a) fall within a broad group of crosslinguistically recognizable verbs described as “non core transitive verbs” (NCTVs). These are dynamic two-place transitive verbs whose participants are often expressed in the same way as “core transitive verbs” (CTVs), in (1b), but which clearly differ from them in some important semantic, syntactic and typological ways. On the basis of these facts, Levin proposes that CTVs and NCTVs have fundamentally different meanings and thus two distinct event structures. The object of CTVs is licensed by a complex causative event structure whereas the object of NCTVs is licensed only by the lexical content of the verb. Levin’s analysis is of particular interest to us for two reasons. First, the source of variation between these two types of transitive verbs can be

153 defined structurally, a desirable analysis from a neo-constructionist perspective. Second, the idea that the expression of structurally licensed arguments should be stable while semantically licensed arguments should not characterizes the valency change within the system of thematic IOs in French.

4.1.1.1 Semantic distinction between CTVs and NCTVs

Levin (1999) points out that while core transitive verbs denote events of the type “agent act on and affect patient”, non core transitive verbs do not easily fit this semantic profile. The internal arguments of NCTVs are difficult to subsume under the notion “patient”, understood to be an entity that undergoes a (physical) change of state. As Levin observes, the object of many NCTVs “cannot be readily assigned roles from the most common semantic role inventories”, demonstrated by her sentences in (2) and (3).

(2) The engineer praised the bridge

The engineer touched the bridge

The engineer avoided the bridge

The engineer owned the bridge

The engineer imagined the bridge

The engineer studied the bridge

(3) The engineer ignored the architect

The engineer praised the architect

The engineer greeted the architect

The engineer selected the architect

The engineer supervised the architect

The engineer fought the architect

The engineer met the architect

The engineer visited the architect

The engineer followed the architect

154

Note that a number of the verbs in (2) and (3) correspond to the French verbs in (4) which are transitive in Modern French, but which take an indirect object at an earlier stage of the language. Some of the examples in (2) can also find their French equivalents among the verbs discussed in chapter two, section 2.2.

(4) Two-place verbs having undergone a change in argument expression: IO → DO

The verbs in (4) belong to the broad class of NCTVs. Their internal arguments are not patients and many cannot be attributed a characterizable yet general semantic role. This can be illustrated through the use of deverbal process nouns associated with these verbs. Deverbal nouns inherit the argument structure of the verb from which they are formed. Since the morphology associated with noun phrases can be richer than verb phrases, differences in the way a deverbal noun marks its object indicate differences in selectional properties. Nominalizations in French distinguish between goal-like objects, broadly construed, and other types of objects by the choice of preposition used to introduce the object complement. Goal-like objects are introduced by the preposition à while patients and other roles are introduced by the preposition de. The examples in (5) and (6) show that while the objects of a good number of the verbs in (4) are goal-like, some escape any meaningful classification.36

36 The nominalizations in (5) and (6) show that there is no one semantic role that corresponds to all thematic indirect objects in Medieval French, a fact that also applies to the subclass of thematic dative objects. Alsina’s Case Assignment Convention (1996:175) in Romance languages predicts that for two-place verbs, dative case is assigned

155

(5) Goal-like objects

a. On ne nous a pas reconnu notre aide aux/*des naufragés.

‘Our help to/of the victims of the shipwreck went unrecognized.’

b. Mes congratulations au/*du père.

‘My congratulations to/of the father.’

c. On a condamné notre faveur aux/*des étudiants.

‘Our favour to/of the students was condemned.’

d. On ne nous a pas reconnu notre secours aux/*des naufragés.

‘Our help to/of the victims of the shipwreck went unrecognized.’

e. On ne nous a pas reconnu notre service aux/*des étudiants.

‘Our service to/of the students went unrecognized.’

f. On n'a pas remarqué nos insultes aux/*des comédiens.

‘Our insults to/of the actors went unnoticed.’

g. On n'a pas remarqué notre offense aux/*des comédiens.

‘Our offence to/of the actors went unnoticed.’

h. On n'a pas entendu ma prière à/*de Dieu.

‘They didn’t hear my prayer to/of God.’

i. On n'a pas entendu ma supplication aux/*des autorités.

‘They didn’t hear my plea to/of the authorities’

only to internal arguments that are thematic goals. If, in fact, this holds for Modern French, it clearly does not apply to Medieval French, or to languages like German and Icelandic as illustrated in Maling (2001). So, while Modern Romance may possess a [±dative] distinction for the mapping of goals to the syntax, ruling out recourse to structural vs inherent case distinctions, Germanic languages – including Medieval French – do not. Any mapping theory for these kinds of languages must allow for inherent case.

156

(6) Other types of objects

a. On a apprécié nos applaudissements *aux/des comédiens.

‘Our applause to/of the actors was appreciated.’

b. On lui a reconnu sa commande *à/de l'armée.

‘His command to/of the army was recognized.’

c. Il régnait sans aucune contradiction *à/de son autorité.

‘He ruled without any challenge to/of his authority.’

d. On a condamné sa domination *au/du peuple.

‘His domination to/of the people was condemned.’

The deverbal nouns in (5) and (6), originating from the list of verbs in (4), cannot therefore be attributed one basic selectional property. Furthermore, it is possible to show that although the nouns in (6) mark their objects in the same way as patient objects are marked, they are not in fact patients. Patient arguments undergo a change of state and thus always denote punctual events (achievements) which are incompatible with durative pendant ‘for’ adverbials, illustrated in (7).37 Rather, the nouns in (6) are derived from verbs denoting activities, not achievements, and most are perfectly acceptable with durative adverbials, as shown in (8). This confirms that the nouns in (8) are not derived from CTVs but are in fact are NCTVs.

(7) a. Elle a coupé le morceau de pain (#pendant cinq minutes).

‘She cut the piece of bread (for five minutes).’

b. Elle a détruit le document compromettant (#pendant cinq minutes).

‘She destroyed the compromising document (for five minutes).’

c. Il a tué l’oiseau (#pendant cinq minutes).

‘He killed the bird (for five minutes).’

37 All readings are non iterative.

157

d. Il a cassé le vase (#pendant cinq minutes).

‘He broke the vase (for five minutes).’

e. Elle a ouvert la porte (#pendant cinq minutes).

‘She opened the door (for five minutes).’

(8) a. Nous avons applaudi les acteurs pendant cinq minutes.

‘We applauded the actors for five minutes.’

b. Il a commandé ce régiment pendant cinq ans.

‘He commanded this regiment for five years.’

c. Ce despote a dominé le Chile pendant dix ans.

‘This despot controlled Chile for ten years.’

In sum, non core transitive verbs differ semantically from core transitive verbs in that they lack a unified and independent semantic characterization. While CTVs denote events in which the patient object undergoes a change of state, the only generalization that can be made about NCTVs is that the object does not undergo a change of state. The aider-type verbs listed in (4) conform to this description.

4.1.1.2 Syntactic distinction between NCTVs and CTVs

Rappaport Hovav & Levin (1998) and Levin (1999) note that NCTVs show more argument expression options than CTVs. Some NCTVs, for example, allow unspecified objects without recourse to generic or repetitive contexts, whereas change of state verbs do not, illustrated below.

(9) a. Marie (nous) a aidé à faire la vaisselle ce soir.

‘Marie helped (us) do the dishes this evening.’

b. Elle a commandé (notre regiment) pendant trois ans.

‘She commanded (our regiment) for three years.’

c. Jean a servi (l’armée) pendant vingt ans.

‘Jean served (the army) for twenty years.’

158

d. Il a prié (Dieu) ce matin.

‘He prayed (to God) this morning.’

(10) a. Marie a détruit *(le document compromettant) hier soir.

‘Marie destroyed (the compromising document) last night.’

b. Jean a tué *(le souris) dans la cuisine.

‘John killed (the mouse) in the kitchen.’

c. Marie a encore cassé *(un verre) ce soir lorsqu’elle lavait la vaiselle.

‘Marie broke (a glass) again tonight when she was doing the dishes.’

d. Il a ouvert *(le cadeau) avant de lire la carte.

‘He opened (the gift) before reading the card.’

Levin (1999) demonstrates other syntactic differences between NCTVs and CTVs involving surface contact verbs and resultative constructions in English. While these latter analyses are not applicable to French, the contrast illustrated in (9) and (10) nevertheless suggests that the internal object of NCTVs is licensed differently that that of CTVs.

4.1.1.3 Typological distinction between NCTVs and CTVs

Levin further observes that from a typological perspective, there is a great deal of inconsistency in the way the internal argument of NCTVs is expressed. On the one hand, if a verb in some language meets the CTV semantic profile, then its translation equivalents in other languages are invariably transitive verbs.38 On the other hand, the translation equivalents of NCTVs need not be transitive, so that one language’s direct object can be another’s oblique. A number of NCTVs in English, for example, are realized with an indirect object complement in French:

38 As Levin points out, her CTVs are roughly equivalent to what Andrews (1995) calls “primary transitive verbs”, and to Hopper & Thompson’s (1980) “highly” transitive verbs. See also Tsunoda’s (1985) cross-linguistic study of “prototypical transitive verbs”.

159

(11) a. (dis)obey one’s parents / (dés)obéir à ses parents

b. resist temptation / résister à la tentation

c. renounce one’s brother / renoncer à son frère

d. remedy a situation / remédier à une situation

e. answer someone, a question / répondre à quelqu’un, à une question

Verbs of authority, ruling and disposition, such as rule, direct, manage, govern, command, master, etc. are another group of NCTVs that show a good deal of crosslinguistic variation in the expression of their object. They are strictly transitive, for example, in Modern English and French, but take instrumental complements in Russian and dative complements in Lithuanian (see Nichols 1975), and they vary between dative and accusative in Medieval French and English. Again, this kind of variation is never attested for CTVs like kill, destroy, break, bend, and open.

Within the class of NCTVs there is not, however, an entirely random distribution of case marking. A number of researchers, for example, have remarked on the crosslinguistic tendency of helping and hindering verbs to take a dative complement. It turns out that the verbs in (4) more often than not take dative objects in other languages – just as they do in Medieval French. A sample of these verbs is given in table 1, with corresponding dative verbs in three other Indo European languages.39

39 See Blume (1998) for examples of how some of these verbs are used in a variety of other languages.

160

Table 1: Interaction verbs in French compared with dative equivalents in three other languages

FRENCH GERMAN CZECH CLASSICAL LATIN aider helfen pomáhat auxiliarī servir dienen sloužit servire applaudir applaudieren tleskat (intrans.) commander befehlen vládnout, vévodit imperare congratuler gratulieren gratulovat (acc.) contredire widersprechen odmlouvat contradicere contrarier (acc.) protikladnout incommodare empêcher (acc.) zabránit (acc.)

The French verbs in table 1 have a strong tendency to occur with dative objects in languages with relatively rich case systems. Medieval French is therefore no different than any other language with a rich case system in formally aligning the object of these verbs with that of verbs of giving and distinguishing them from patients and other highly affected objects. This tendency has led some researchers to propose a mapping algorithm for dative case in order to capture the conceptual regularity among such verbs. In Blume’s (1998) treatment of two-place interaction verbs in a number of Indo European and non Indo European languages, she claims that dative selection is systematic, contrary to claims that an inherent (idiosyncratic) lexical index determines dative assignment in these cases. Within a proto-role approach to thematic relations, she proposes a linking rule that restricts nominative/dative case frames to social interaction verbs denoting complex events in which one participant (the dative) strives consciously for a certain aim while the other (the nominative) performs an unspecified act that contributes to the achievement of the aim of the first one. Crucially, the arguments associated with each subevent express no proto-patient properties.40

The present study is not concerned with adopting any specific mapping algorithm; it is more interested in showing that within the broad class of non core transitive verbs, two-place interaction verbs form a recognizable crosslinguistic subclass which describes at once the

40 See Svenonius (2002) for a similar description of dative objects in Icelandic .

161 variable expression of the object and yet its clear tendency to be realized as a dative. In the following sections we propose an argument structure for these verbs that captures both the variability and the possibility of subregularities in the expression of the object, avoiding, as Blume does, recourse to lexical indexing for inherent case.

Finally, Levin observes that language internally, near synonyms of CTVs are always transitive whereas near synonyms of NCTVs are not. Given the shared meaning, it is not surprising that near synonyms of a CTV would themselves be CTVs, illustrated in (12) for French.

(12) a. couper, trancher, fendre

b. détruire, anéantir, écraser

c. tuer, assassiner, exécuter

d. casser, briser, rompre

e. ouvrir, déboucher, déplier

In contrast, the object of near synonyms of NCTVs does not show the same morphosyntactic expression. In Medieval French, many near synonyms of transitive verbs of helping (13a), hindering (13b) and authority and ruling (13c) take indirect objects:

(13) a. appuyer, seconder, soutenir (DO) vs. aider, assister (IO)

b. contrecarrer, agacer (DO) vs. contrarier (IO)

c. diriger, contrôler, gouverner (DO) vs. commander, dominer (IO)

Maling (2002:3) makes the same observation in Icelandic. Her near-minimal pairs in (14) show that the near synonym of many NCTVs takes a dative object.

(14) a. ‘drive’ keyra (acc.) vs. aka (dat.)

b. ‘help’ aðstoða (acc.) vs. hjálpa (dat.)

c. ‘finish’ klára (acc.) vs. ljúka (dat.)

d. ‘meet’ hitta (acc.) vs. mœta (dat.)

e. ‘protect’ vernda (acc.) vs. hlífa (dat.)

162

The French diachronic facts reflect a typological phenomenon; the aider-type verbs in (4) are non core transitive verbs. Transitive in Modern French, their translational equivalents are not always transitive crosslinguistically. In fact, many are often expressed with a dative or oblique object in languages with rich case-marking systems.

4.1.2 The argument structure of NCTVs

The previous sections demonstrate that the group of transitive French verbs in (4), whose object was expressed as an indirect object in Medieval French, pattern according to Levin’s (1999) description of non core transitive verbs. They contrast with core transitive verbs both intra and crosslinguistically with respect to transitivity and the semantic role of their object. These facts suggest a significant meaning difference, which Levin captures structurally within a lexicalist approach to argument structure.

For Levin, the difference between core and non core transitive verbs lies in what component of a verb’s lexical semantics licenses the object. Licensing is a way of defining what it is that permits a given constituent to appear as a legitimate argument of a verb phrase. According to her approach, the meaning of a verb is composed of two distinct parts, its idiosyncratic meaning and an event structure template. The latter are finite in number and are shared by verbs that belong to the same class. As simple activities, NCTVs are associated with the event structure in (15a), while CTVs are associated with complex causative events, in (15b) .

(15) a. NCTV: [x ACT ]

b. CTV: [[x ACT ] CAUSE [BECOME[y ]]]

The event structure of simple activities introduces an external argument x and allows for the insertion of a verb’s idiosyncratic meaning as the component of the activity. Note that there is no mention of an object in this event structure. In contrast, a complex causative event introduces an external argument x that performs some action which causes another argument y to take on a particular state of being. The variable y corresponds to the object of core transitive verbs. Note that it is an essential component of the event structure – it cannot be omitted and always has the same interpretation (the entity which undergoes a change of state, or patient). On the other hand, the object of activity verbs is not a component of the event structure. Its

163 interpretation and presence remains unspecified, capturing the semantic, syntactic, and typological facts discussed earlier.

Levin proposes that when activity verbs occur with an object, it is introduced, or licensed, by the idiosyncratic meaning of the verb – not by the structure. Objects of activities are underlined in order to distinguish them from objects that are an essential component of event structure. Transitive verbs like help would be represented as follows:

(16) [x ACT y]

Applying Levin’s lexicalist approach to the neo-constructionist idea that verbs are built in the syntax (not in the lexicon), we propose the basic structure in (17a) for simple activity verbs. The configuration is intransitive, just like Levin’s event structure in (15a). In cases where activities involve an additional participant, the root simply merges with a sister constituent, as in (17b).41 Since Merge is “free”, this operation is always sanctioned provided that the sister constituent is compatible with the meaning of the root.

(17) a. v b. v 3 3 v √ v √P 3 √ XP

The aider-type two-place verbs in (4) are associated with the structure in (17b). In Medieval French, the object of these verbs could be a PP, whereas in Modern French, it is restricted to DPs. 42

Following the work of Hale and Keyser (1993, 2002), Roberge (2002), and Cummins and Roberge (2005), an object position is always assumed, which would rule out the structure in (17a) and leave (17b) as the only legitimate configuration. We do not take a stand one way or the

41 See Nash (2000) for a similar proposal.

42 Infinitival phrases such as aider à quelqu’un à faire quelque chose and prier à quelqu’un de faire quelque chose are not considered to be complements of the verb, but rather circumstantials merged above the vP layer.

164 other on this matter, but merely point out that such a predetermined position for all verbs is entirely compatible with our analysis. The contrast Levin makes between structurally and semantically licensed objects would have to be restated, however, since both CTVs and NCTVs would, in this case, be syntactically licensed. The crucial difference would reside in the fact that the grammatical category, or case, of the internal argument of CTVs is a consequence of the structure associated with them whereas it is not for NCTVs.

4.2 Structural and semantic distinctions

This section describes the three properties that distinguish aider-type verbs from all other verbs that take thematic indirect objects.

4.2.1 Structural distinction

A central idea of Levin’s (1999) study is that typologically stable objects, such as those of CTVs, are licensed by the event structure associated with the verb. Typologically unstable objects, like those of NCTVs, are not. This dichotomy also holds true for the system of indirect objects in French. Structurally licensed IOs do not undergo a valency change, while IOs that are not structurally licensed are susceptible to change. Recall that only the thematic IOs of certain interaction and reflection verbs were replaced by an analogous DO. All other IOs occurring with dynamic and stative verbs were not subject to change, illustrated in table 2. A sample of the verbs that underwent the valency change is highlighted in grey, contrasting with other verbs and verb classes that took – and still take – an indirect object.

165

Table 2: Verbs taking thematic indirect objects in French

VERBS OF CHANGE OF LOCATIVE & PSYCH VERBS VERBS OF TRANSFER & LOCATION RELATION INTERACTION COMPARISON VERBS VERBS & REFLECTION attribuer aller confiner agréer céder ‘attribute’ ‘go’ ‘adjoin’ ‘suit’ ‘yield’ apporter arriver demeurer convenir coopérer ‘bring’ ‘arrive’ ‘remain’ ‘suit’ ‘cooperate’ donner descendre habiter démanger obéir ‘give’ ‘go down’ ‘live in’ ‘itch’ ‘obey’ lancer entrer manquer importer parler ‘throw’ ‘enter’ ‘lack’ ‘matter’ ‘speak/talk’ rendre monter résider manquer remédier ‘give back’ ‘go up’ ‘reside’ ‘miss’ ‘remedy’ arracher partir vivre nuire penser ‘pull off’ ‘leave’ ‘live’ ‘harm/bother’ ‘think’ retirer parvenir appartenir plaire réfléchir ‘take away’ ‘reach’ ‘belong’ ‘like’ ‘reflect’ voler rester correspondre peser aider ‘steal’ ‘stay’ ‘correspond’ ‘weigh’ ‘help’ annoncer retourner ressembler profiter applaudir ‘announce’ ‘return’ ‘resemble’ ‘benefit’ ‘applaud’ dire sortir succéder répugner dominer ‘say’ ‘go out’ ‘succeed’ ‘disgust’ ‘dominate’ comparer tomber préexister suffire empêcher ‘compare’ ‘fall’ ‘pre-exist’ ‘suffice’ ‘hinder’ préférer venir survivre tarder servir ‘prefer’ ‘come’ ‘survive’ ‘be impatient’ ‘serve’

In section 1.3, structures were proposed for each verb class taking thematic indirect objects in French. They are summarized in (18). Note that all of these verb classes, save interaction and reflection verbs, denote events involving a relation between two internal arguments. The portion of the structure that corresponds to this relation is emphasized in bold.

166

(18) a. VERBS OF TRANSFER AND COMPARISON b. CHANGE OF LOCATION VERBS offrir un cadeau à mon ami Sa soeur va au mariage

3 vP

vDO √Poffr- 3 “offrir” 3 vBECOME √Pall- √offr- PP “aller” 3

3 √all- PP DP 3 3 5 P DP DP 3 un cadeau g 5 5 P DP à mon ami sa soeur g 5 à le mariage

c. LOCATIVE AND RELATION VERBS d. PSYCHOLOGICAL VERBS Le garçon ressemble à son père Le film plaît à ma soeur

vP vP 3 3 vBE √Pressembl- vBE SCplais- “ressembler” 3 “plaire” 3 √ressembl- PP DP aPplais- 3 5 3 DP 3 le film aplais- nplais- 3 5 P DP n PP le garçon plais- g 5 3 à son père P DP

g 5 à ma soeur

e. INTERACTION AND REFLECTION VERBS remédier à la situation aider à mon ami (Medieval French)

vDO √Paid- “aider” 3 √aid- PP 3 P DP g 5 à mon ami

167

As described in chapter one, three-place verbs of transfer and comparison (18a) select a Figure- Ground relation. This relation can denote material or abstract transfer, comparison, or association between the DO and the IO. Change of location verbs (18b), otherwise known as unaccusatives, also involve a relation mediated by à. They denote punctual events in which a theme argument moves to a given location. They have the same basic structure as three place verbs of transfer; the difference between the two simply lies in the type of little v that introduces the event.

Unaccusative roots associate with vBECOME as opposed to vDO. Two-place locative and relation verbs (18c) also assert a relation. Locative verbs situate the theme relative to a particular place while relation verbs situate it relative to a property of the indirect object. The only difference between these verbs and those described in (18a) and (18b) is, again, in the type of little v that introduces the event. Locative and relation verbs are defined by a stative, or existential little v, as opposed to vBECOME or vDO. Psychological verbs (18d) express a relation between two entities as well. Experiencers of psychological predicates are construed as possessing a psychological state. The relation is thus between a possessor, the dative IO, and a possessed, a nominal root denoting a property or a state. It is crucial to observe that as simple activities, verbs of interaction and reflection (18e) do not express a relation between two entities.

A generalization of the structure common to the verb classes represented in (18a-d) is given in (19a). Note the contrast with (19b), the non relational structure associated with the object of interaction and reflection verbs. Many of these latter objects may or may not appear as an IO, as evidenced by the valency change in aider-type verbs and the crosslinguistic variation exemplified in (11), (13), (14), and table 1.

(19) a. RELATION b. NON RELATION

3 3 D/nP 3 √ DP/PP à DP 123 IO

All indirect objects in French, save those of interaction and reflection verbs, are therefore licensed as elements of the basic structure in (19a). In lexical semantic terms, the core predicate

168 involving the IO, [BE [x AT y]], is always specified for relational verbs.43 The IO is defined in terms of the predicate and the predicate in terms of the IO. The indirect object is a necessary element of the event structure and as such, it can be considered as “structurally” licensed.

The object of interaction and reflection verbs is not licensed this way. Interaction and reflection verbs require a Thing as a complement, but unlike the complement of relation verbs, this is not a grammatical category. The category of the complement of these verbs, be it a DP or a PP, is underspecified. Its morphosyntactic expression depends on the matching of the semantic role that the verb assigns to its internal argument and the general semantic role of the object, derived from its position in the complement of a particular preposition or from its status as a simple DP in the accusative case. Since there is no grammatical information associated with the complement of these verbs, it is not structurally licensed.

The difference between IOs that are structurally licensed and those that are not corresponds to where we find stability as opposed to variation and change in the expression of the indirect object in French. Structurally licensed IOs only occur in a Figure-Ground relation. Because this relation is expressed as a PP in French, the Ground will always be the complement of a preposition. As long as the preposition à has predication properties and expresses a basic spatial relation between two entities, the indirect object will be properly licensed as a necessary component of that predication. If, for example, à were no longer able to denote a relation between two entities, then we would expect the verbs in (18a-d) to have recourse to another preposition that could perform the same function.44

Non structurally licensed IOs are vulnerable to variation and change because a change in the meaning of à can change the semantic role it attributes to its complement. This will necessarily affect the kinds of verbs with which the IO may occur given that there must be mutual compatibility between the semantic role assigned by the preposition and the semantic role

43 What we mean here, is any verb that is used in a relational sense. Not all verbs that can occur in three-place constructions are exclusively relational (i.e., crier, chuchoter, hurler, sourire, etc.) 44 See Cummins et al. (2008) for further discussion of relation verbs in French.

169 assigned by the verb. Our task is thus to identify what changed in the meaning of à so that IOs no longer appear as the complement of certain two-place interaction verbs.

Finally, the formal description of NCTVs captures the empirical fact that there may be no obvious semantic difference between an NCTV that occurs with an oblique object and one that occurs with a direct object. The English and French verbs in (11) for instance, differ in the expression of their object, but do not appear to differ in their meaning. That the direct object and the oblique object of NCTVs are generated in the same syntactic position would account for this possibility, whereas an analysis that posits different positions for each object form would have some difficulty explaining the semantic equivalence.

Levin points out that it is the nature of the case or prepositional system that ultimately determines how the object of an NCTV is marked. Oblique markers identify specific semantic roles; the more fine-grained the case system, the more semantic roles are identified via the morphology. Conversely, when a language has few oblique markers, semantic roles are redistributed and collapsed into larger groups. Ultimately, semantic roles are expressed by the marker that best captures that role. Semantic roles that are not compatible with any of the available oblique markers are merged without case and obtain structurally assigned accusative, in line with Kiparsky’s Elsewhere Condition.

In Modern French, for example, an IO is the “best fit” for the semantic role of the object of obéir, and since it is licensed with this type of verb, it may be employed. On the other hand, dative case is not available in Modern English, so the object of obey takes the default accusative case.

In Medieval French, the “best fit” for the semantic roles of the aider-type verbs is an IO. When the IO is no longer available to these verbs, their object takes the default case, presumably for lack of a preposition that better captures these roles. Accusative case, which encompasses a broad range of semantic roles (see Levin’s examples in (2) and (3)), can ostensibly convey that which is conceptually necessary for the object of this group of verbs.

4.2.1.1 Structural vs. inherent indirect objects

One issue that our analysis raises is the difference between structural and inherent case. Structural case, commonly associated with accusative and nominative case, is assigned in a

170 particular structural position more or less independently of the verb. This seems to correspond to IOs which belong to a Figure-Ground relation; see Van Peteghem (2006) for a similar conclusion. Inherent case, on the other hand, is idiosyncratic and is said to be assigned directly by the verb in its complement position. This corresponds to the non structurally licensed IOs of interaction and reflection verbs.

As pointed out in Blake (2001) and Blume (1998), there is a strong cross-linguistic tendency for the oblique marker of aider-type verbs to be the same as that used for the objects of trivalent verbs like donner (i.e., dative). This does not necessarily mean, however, that the two types of datives are generated in the same structural position. Rather, what it means is that inherent case is not completely synonymous with idiosyncrasy. In most linking models, the lexical entry of a verb that assigns lexical case contains an inherent lexical index that prevents the object from receiving structural case and determines its specific oblique case. Within such models, the crosslinguistic tendency is not captured and, as Blume (1998:254) points out, “Such an analysis seems to suggest that the dative complement of helfen [to help] is purely idiosyncratic, attributable to an accidental development in the history of the language.” The same comment would apply to the dative complement of aider in Medieval French.

The challenge is to capture the structural distinction in French between the IO of relational verbs and that of interaction and reflection verbs, while maintaining the regularity with which IOs are assigned to the latter. Assuming that the lexical entry of the aider-type verbs has not changed over the course of the history of French (there has been no significant sound change and no obvious meaning change), it should not contain any specific information about the syntactic category of its internal argument. Synchronically, such information would, at the very least, make it difficult to account for productive case alternations; see chapter five, examples (12) and (13). The only information about the object of a given verbal root should be a notion of the semantic role it assigns to its internal argument. This role is completely idiosyncratic and is not associated with a discrete set of primitive thematic roles, exemplified by the broad range of meanings that can be attributed to objects of NCTVs in examples (2) and (3). This semantic role is an entailment of aspects of the encyclopaedic meaning of the verbal root. Crucially, the category or case of the argument is not specified. The lexical entry for the eventive root aid- would therefore look something like that in (20).

171

(20) Lexical entry for the verb aider

PHONOLOGY: /εd/

SYNTAX: √

SEMANTICS: ∈ of {Events}

ENCYCLOPAEDIC: some notion of what an event of helping implies which will entail “agentive” little v and the existence of a relatively autonomous participant whose own course of events benefits from what the other participant does (that associated with agentive little v).

Instead of associating a dative case index with the internal argument, the implicit notion of “beneficiary” should be sufficient to determine its syntactic form (this would correspond to a more general linking rule along the lines of what Blume (1998) proposes). The expression of the object depends on the compatibility of beneficiary with the meanings of the various case markers in French. Such matching of the semantic role attributed by the verb and those belonging to the accusative and the oblique cases is a separate algorithm, determined compositionally. There is no feature matching involved here. Indeed it is precisely this aspect of “inherent case” that is language specific and idiosyncratic. That most languages mark the beneficiary of a helping event the same way as the goal in an event of giving reflects strong conceptual similarities between the two roles.

4.2.1.2 Other proposals for the structure of verbs like aider

Our analysis of the argument structure of aider-type verbs closes with a discussion of two different structures that have been proposed in the literature. A number of studies have converged on the idea that two-place dative verbs are really covert three-place verbs of transfer with an implicit or incorporated direct object; see for example, Herslund (1980) for Old French, Melis (1996) for Modern French, Abraham (1983), Wegener (1991), McFadden (2003) for German, and Kress (1982) for Icelandic. Alternatively, in a revised version of his 2003 paper, McFadden (2006) proposes that the sole dative argument of German verbs like helfen ‘help’ is generated in a high applicative phrase. We argue that neither of these solutions correctly describes the dative argument of aider-type verbs in Medieval French. First, there is little correlation between the meaning conveyed by the two proposed structures and that of verbs like

172 aider. Second, these approaches do not reflect the similarities between the sole indirect object of dynamic events and direct objects more generally, evidenced by the synchronic, diachronic, and crosslinguistic alternation between the two forms.

According to proposals that aider-type verbs are underlyingly three-place verbs of transfer, Marie aide à Jean in Medieval French would be derived from something like Marie donne de l’aide à Jean ‘Marie gives Jean help’. The theme aide would be incorporated into a (null) light verb, resulting in the realization of only one internal argument, the dative beneficiary à Jean. The structure proposed in Herslund (1980) for aider in Old French is given below. (21) VP 3 VP NP/PP45 3 V NP ! ! FAIRE AIDE 123 aider Herslund (1980:51)

Proposals such as this are either advanced within the context of a unified analysis of thematic dative arguments, or they support theories like UTAH (Baker 1988, Larson 1988, etc.) which match specific theta roles to specific structural positions. In both cases, the dative object of two- place verbs is aligned with the dative object of a double object construction (Mary gives John help). Since the object of aider has goal-like properties, illustrated earlier in (5a), the logic is that it should merge in the same position as recipients. Following Larson, this would be the specifier of lower V.

45 Herslund’s structure is also meant to account for non prepositional dative arguments in Old French like that in the following double object constuction: (i) Lors a conté son oncle toute sa destinee (Herslund 1980:24) ‘He now told his uncle all that had happened to him’ He aligns verbs like aider with this construction so that objects which are not introduced by the preposition à are nonetheless taken to be datives. The same syntactic position accomodates both prepositional and non prepositional datives (Herslund 1980:31-33).

173

Derivations based on a double object construction, no matter what form they may take, should be ruled out in general, however, based on the well-documented empirical fact that double object constructions denote possession of the theme (Guéron 1985, Pesetsky 1995, Harley 2002a, Pylkkänen 2002, among others). If two-place dative constructions are derived from a double object construction, one would expect them to convey the basic meaning of the construction, yet this is not true. The object of commander, contredire, dominer, épargner, persuader, for instance, is clearly not the intended recipient of an implicit theme.

Finally, the parallel drawn between the argument structure of verbs like aider and donner does not account for the fact that a good number of two-place verbs tend to alternate synchronically and diachronically between taking a dative or an accusative object. In other words, if a structure like (21) represents the meaning and the syntactic instantiation of aider in Medieval French, how would this same verb look in Modern French? To assume that it is still derived from a double object construction would be difficult to argue from an acquisition point of view, to say the least, since this construction is no longer available in Modern French. Furthermore, positing a simple verb-object structure for Modern French would imply that aider was a syntactically derived verb in Medieval French and that it came to be construed as a lexical verb from the 16th century onward. This might be plausible as an isolated case, but in light of the systematic valency change that we see in verbs of this type, it would be highly unlikely that all of the verbs in (4) were reanalyzed as lexical verbs during the same period of time. Lexical changes are, as a rule, unsystematic.

A structure like that in (21) is clearly not compatible with two-place dative verbs. First, there is no clear semantic correlation between two-place and three-place verbs. Second, by encoding the dative object in a position distinct from the accusative, their common grammatical function is not captured. The synchronic and diachronic dative-accusative alternation suggests that the sole object of such verbs, be it dative or accusative, is generated as the sister of the verbal root. We have proposed, therefore, that verbs like aider are derived from a configuration involving one internal argument, not two, as suggested in (19b). Aid- is not therefore the direct object of a light verb, but rather a manner root from which the verb is derived. Its object, dative or accusative, simply merges as its sister.

174

Next consider the structure proposed in McFadden (2006:54) for German verbs like helfen ‘help’ whose sole object is dative, much like aider in Medieval French.

(22) Boris hilft den Wissenschaftlern

vP 3 DP v’ ! 3 Boris v ApplP 3 DP ApplP’ ! 3 den Wiss’lern Appl VP ‘the scientists’ ! hilft ‘helps’

The motivation for proposing the structure in (22) appears to be mainly theoretical. McFadden observes that an inherent case analysis for the object of German verbs like helfen is out of step with current research on dative objects. Within a neo-constructionist approach to argument structure and case assignment, the dative object’s properties should be derived from the way in which it is introduced into the structure, not from the morphology of its lexically assigned case. Following the pioneering work on applicative phrases in Marantz (1993), Anagnostopoulou (2002), Pylkkänen (2002), Cuervo (2003), etc., McFadden proposes that the object of verbs like helfen is introduced in the specifier of a high applicative head. Traditionally, the function of the high applicative is to introduce non core (or non thematic) dative arguments and apply them to the eventuality denoted by the vP complement. In this sense, helfen is essentially treated as an unergative verb to which a dative argument may or may not be applied.

Lüdi (1978) also equates the dative argument of helping verbs with that of non core dative benefactors. Indeed, both kinds of datives seem to convey a similar meaning: the action has beneficial consequences for the dative referent. This does not, however, imply that they must occupy the same position. The basic fact that dative benefactors are not inextricably linked to the core meaning of the verb suggests that they must be generated in a position outside of the vP. Since the dative argument of helping verbs is a core argument of the verb, the object is merged as its complement, as we propose in (19b). The dative in this vP internal position can acquire an

175 interpretation similar to that of a benefactor, since, just as the benefactor is in a relation with the event as a whole, the dative object of aider-type verbs in a relation with the eventive root. Neither are part of a Figure – Ground relation, for instance.

Even if one accepts that verbs like helfen are unergative, McFadden’s configuration in (22) does not capture the tight connection between meaning and syntactic structure maintained by the very theory of dative arguments after which it is modeled. The dative object of helfen has little in common with non core dative benefactors, the arguments traditionally generated in the high applicative phrase. For instance, in German, dative benefactors can be paraphrased by für ihn ‘for him’, shown in sentences (23a) and (23b). This contrasts with the dative argument of helfen which cannot be paraphrased by für ihn. (23c) and (23d) do not mean the same thing.

(23) a. Ich öffnete ihm die Tür.

I opened himDAT the door

‘I opened the door for him.’

b. Ich öffnete die Tür für ihn.

‘I opened the door for him.’

c. Sie half ihm.

She helped himDAT

‘She helped him’

d. Sie half für ihn.

‘She helped for him.’

If the dative argument of helfen were generated in the same position as benefactors, one would expect them to have the same basic interpretation. This is clearly not the case for German, or for French for that matter. Non core datives in French, including benefactors, are also analyzed as event arguments taking the vP as a complement (see Roberge and Troberg 2007b), similar in many ways to datives generated in high applicative structures. Some examples are given below:

(24) a. Jean lui a ouvert la porte.

‘Jean opened the door for her/him.’

176

b. Elle m’a attrapé un rhume.

‘She caught a cold on me.’

c. Les accidents lui arrivent souvent.

‘He often has accidents.’

Yet, while the indirect objects of the aider-type verbs were replaced by direct objects, dative benefactor arguments were not. If the dative argument of aider in Medieval French were generated in the same position as non core dative benefactors, one would expect both to be subject to the same change. Furthermore, dative benefactors in French are subject to a strict syntactic constraint: the verb with which they occur must be followed by an overt direct object. The dative object of verbs like aider in Middle French would violate this requirement.

In sum, the structures in (21) and (22) do not reflect the fact that a DO can be functionally equivalent to the dative argument of dynamic two-place verbs. In response to the wealth of synchronic, diachronic and typological data to this effect, we propose that the internal argument of two-place interaction verbs, be it dative or accusative, is generated in only one position: the complement of the verbal root.

4.2.2 Directionality

Section 4.2.1 showed that aider-type verbs have a particular argument structure that distinguishes them from most other verbs that took and still take an indirect object. Only the IO of non relational, dynamic two-place verbs was susceptible to replacement by an analogous DO. On its own, however, argument structure does not account for the specific group of non relational verbs that underwent the change. We now turn to the more fine-grained lexical properties of the verbs that make up this class in order to identify what determines whether a verb underwent a valency change or not. This section discusses the crucial role of lexical directionality.

Table 3 divides the class of dynamic two-place indirect verbs into those whose lexical semantics orient the action in a particular direction and those which do not have this meaning. Directional verbs are further grouped in terms of common meanings. It turns out that adirectionality

177 characterizes all aider-type verbs and distinguishes them from many other non relational verbs whose IO was not replaced by an analogous DO.

Table 3: Non relational verbs classified in terms of lexical directionality

NON RELATIONAL VERBS

ADIRECTIONAL DIRECTIONAL

RESISTANCE COMMUNICATION REFLECTION MOVEMENT attenter remédier accéder46 mentir penser aboutir compatir réussir acquiescer parler réfléchir accéder consentir satisfaire céder répondre rêver accourir contribuer souscrire contrevenir sourire songer aspirer coopérer subvenir déférer téléphoner atteindre croire suppléer désobéir courir faillir vaquer déroger échapper forfaire veiller obéir prétendre obvier aider obtempérer procéder parer applaudir réagir recourir participer commander renoncer pourvoir insulter résister présider étudier survivre prétendre etc. succomber

Verbs that express directionality specify how a Figure object (a moving or conceptually moveable entity) is oriented or moving with respect to a Ground (a reference landmark). Typical verbs of this type would be entrer ‘enter’, donner ‘give’ and voler ‘steal’ for example. Entrer is lexically specified to orient the subject towards the object just as offrir orients the direct object toward the indirect object, and as voler orients the direct object away from the indirect object.

Two-place verbs of verbal and non verbal communication like parler, mentir, sourire, etc., also have directional properties. They resemble in many ways three-place verbs of transfer such as annoncer, communiquer, montrer, etc., the only difference being that the information which is

46 In the sense of ‘concede’: accéder aux voeux de la population

178 transferred remains implicit. Some examples are given in (25). The use of sourire in example (25c) illustrates this point rather well, since the intended meaning of the verb sourire is paraphrased as the three-place verb of transfer montrer.

(25) Verbs of communication

a. Pierre ment souvent à ses amis.

‘Pierre often lies to his friends.’

b. Le ministre répondra aux journalistes au cours de la conférence de presse.

‘The minister will respond to journalists during the press conference.’

c. Elle lui sourit en lui montrant qu’elle le comprenait bien.

‘She smiled at him, showing him that she understood him.’

(Balzac; Le Petit Robert)

Similarly, verbs of reflection may also take on directional properties to become verbs of directed thought.

(26) Verbs of reflection

a. Il ne pense plus qu’à elle.

‘He no longer thinks of anything but of her.’

b. J'ai rêvé à un monde meilleur.

‘I dreamed of a better world.’

All verbs denoting a change of location in French are unaccusative and their indirect object is thus structurally licensed within a relation (aller, partir, retourner, etc.). The (abstract) movement verbs listed in table 3 are not unaccusative, but rather have the same argument structure as interaction and reflection verbs. They denote eventualities that are oriented in a particular direction, specified by way of their prefixes, as shown in the sentences in (27).

179

(27) Verbs of abstract movement

a. Nous procéderons à l’analyse du texte.

‘We will proceed with the analysis of the text.’

b. Elle devait y recourir, à son avocat.

‘She had to turn to her lawyer.’

c. Paul accourt au secours de son ami en danger.

‘Paul runs to the help of his friend in danger.’

Verbs of resistance also have directional properties. Following Talmy’s (2000) force dynamic principles, these verbs are lexical representations of the semantic category which encompasses how entities interact with respect to force, namely, resistance to force. They denote events in which the object (the agonist) is the force exerting entity while the subject (the antagonist) is considered for the way in which it reacts to the agonist. Varying degrees of resistance are expressed, from none at all (obéir) to very little (céder), to total resistance (désobéir). Some examples are given in (28). These verbs are all directional in that the antagonist exerts an opposing force to that implied by the agonist: Antagonist → g ←Agonist. The directionality inherent in these verbs is thus a function of the general concept of resistance to force.

(28) Verbs of resistance

a. Ici, il faut obéir aux autorités.

‘One has to obey the authorities here.’

b. Son père a succombé à un cancer.

‘Her father succumbed to cancer.’

c. Il n’ose pas lui résister.

‘He doesn’t dare stand up to her.’

Turning now to interaction verbs that do not express directionality, a number are given in table 3 and some examples follow.

180

(29) Adirectional dynamic two-place verbs

a. Il compatit à notre douleur

‘He sympathizes with our pain’

(Le Petit Robert)

b. Pour obvier à un accident, à une maladie

‘To take precautions against an accident, an illness’

c. Remédier au manque de main-d’oeuvre en faisant venir des ouvriers

étrangers

‘To meet the lack of man-power by recruiting foreign workers’

(Le Grand Larousse)

d. Veiller à la bonne tenue de ses enfants

‘To attend to the good manners of one’s children’

(Le Grand Larousse)

Note that the verbs that underwent a valency change are also unspecified for directionality. They do not denote events in which the object receives a message, in which it is the goal of a directed thought or movement, or in which it is met by an opposing force. Rather, they share many properties with the adirectional verbs listed in table 3, exemplified by the recurrent meanings of helping, hindering, and avoiding. Interestingly, however, some resemble the directional verbs discussed earlier. Certain verbs like applaudir, insulter, persuader, prier and supplier seem no different than (directional) communication verbs like parler, téléphoner, etc. Yet, while these verbs may be construed as manners of communication, the transfer of a message is ostensibly less important than the idea of exerting influence over what the object does or feels. Téléphoner and parler, for instance, identify their object as a recipient of a phone call or a verbal message. The fact that the call or message may change the course of action of the recipient is conceptually irrelevant. In contrast, the salient property of the object of adirectional verbs like insulter and prier is one of an actor of an implicit subevent whose course of action or state is (potentially)

181 influenced by what the subject says. This subevent can often be realized as an object-controlled infinitive clause or prepositional phrase, in (30). This is not available for recipients, in (31).47

(30) a. Je vous prie de vous taire.

‘I beg you to keep quiet.’

b. Elle lesi insulte dans leursi efforts.

‘She insults them in their efforts.’

(31) a. * Je vous ai téléphoné de venir plus tard.

‘I phoned you to come later.’

b. * Je vous ai parlé de venir plus tard.

‘I spoke to you to come later.’

The verbs that underwent the valency change, exemplified by aider, are adirectional. Non relational verbs can therefore be divided into directional and adirectional verbs. Aider-type verbs, along with others like obvier, remédier, veiller, etc. belong to the group of verbs that do not express a directed action. On the other hand, verbs of resistance, communication, reflection and movement all denote an action that is oriented with respect to the object complement.48

4.2.3 First, second and third order entities

Section 4.2.2 shows how the class of non relational dynamic two-place verbs can be divided into two subclasses: those that possess lexical directionality and those that do not. Aider-type verbs

47 Many of the verbs that underwent the valency change conform to Talmy’s (2000) approach to the organization of events in terms of entities interacting with respect to force. Applied to social interactions, these verbs describe a dynamic whereby the object has a tendency toward performing some action and the subject influences that action positively, negatively, or otherwise.

48 Regretably, we have not been able to identify a test that permits an unequivocal distinction between a verb possessing lexical directionality and one that does not. Our classification may thus appear rather impressionistic. Nevertheless, we maintain the distinction, since it fits squarely within the larger analysis that draws on other well substantiated criteria.

182 belong to the subclass of adirectional verbs. This section further narrows the focus to the class of adirectional verbs in order to identify the property that distinguishes aider-type verbs from the other verbs of this type.

Consider again the two classes of verbs proposed in table 3. The division of the non relational verbs into adirectional and directional verbs also reveals another striking distinction: the type of IO that occurs with the two classes of verbs. Following Lyons’ (1977) typology of nominal entities, in Modern French, adirectional verbs only occur with second order indirect objects, illustrated in (32), whereas directional verbs appear to have no restrictions on the type of IO that they occur with, shown in (33).

Lyons distinguishes nouns in terms of their level of abstraction. First order entities are material objects; they exist in space and time and are publicly observable. Any animate object is a first order entity, as is any individuated inanimate object like a book, a painting or a tree. Second order entities are located in time and are said to occur rather than exist. These include events, states, situations and places. Third order entities are abstract and exist outside of time and space. They correspond to concepts, facts, ideas, attitudes, possibilities and propositions.

(32) Adirectional verbs

a. remédier à cette situation/ * à cet enfant qui s’est cassé la jambe

‘to remedy this situation/ this child who has broken his leg’

b. attenter à la liberté de la presse/ * au président

‘to conspire against the freedom of the press/ the president’

c. parer au scandale/ * à un ennemi

‘to deal with a scandal/ an enemy’

d. satisfaire aux revendications du syndicat/ * à son employeur

‘to meet the demands of the union/ please one’s employer’

e. aider à la guérison de l’enfant/ * à Jean

‘to contribute to the recovery of the child/ to Jean’

183

f. applaudir à votre initiative/ * aux acteurs

‘to applaud your initiative/ the actors’

g. insulter au malheur de l’écrivain/ * à l’écrivain

‘to be an insult to the misfortune of the writer/ insult the writer’

h. commander à ses sentiments/ * à l’armée

‘to have command over one’s feelings/ the army’

(33) Directional verbs

a. céder à la tentation/ à sa belle-mère

‘to yield to temptation/ one’s mother-in-law’

b. obéir aux lois/ à son maître

‘to obey the laws/ one’s master’

c. survivre à une épidémie/ à son frère

‘to survive an epidemic/ one’s brother’

d. mentir aux clients

‘to lie to the clients’

e. parler aux clients

‘to speak to the clients’

f. sourire aux enfants

‘to smile to the children’

g. penser à une solution/ à son amie

‘to think of a solution/ of one’s friend’

h. recourir à son charme/ à son avocat

‘to resort to one’s charm/ turn to one’s lawyer’

184

From a synchronic point of view, the correlation between lexical directionality and the type of object a verb selects for could be seen as purely coincidental. The diachronic facts, however, show us that this is not the case. Adirectional verbs like aider, applaudir, commander and insulter used to permit first order indirect objects as well, but these, as we have seen, were systematically replaced by analogous direct objects from the 16th century. Recall, for example, that the quantitative study in chapter three revealed a striking difference between the evolution of animate and inanimate objects. As shown for aider, reproduced in graph 1 below, the frequency of the animate IOs (first order entities) shows a steady decline from the 16th century while the inanimates (generally second order entities) maintain their numbers.

Graph 1: aider: 14th to 19th c., evolution of the frequency of animate and inanimate IOs

Verbs like aider illustrate the dramatic loss of animate first order entities. Étudier provides an example of a verb that exclusively selects third order indirect objects in Middle French, as shown in (34). These kinds of objects are also replaced by a direct object.

(34) a. ... que tous ceuls qui ont estudié as arts et as sciences eüssent ignoré telle chose

‘... that all those who studied the arts and sciences had been unaware of such a thing’

(Oresme, Le Livre de Ethiques d'Aristote, c.1370, p.116; DMF1)

b. Mais j'ai estudié au livre de Jason.

‘But I studied the book of Jason.’

(Guesclin, 8960, Littré, Dictionnaire de la langue française)

185

c. J’estudiay aux loix deux ans.

‘I studied the laws for two years.’

(B. de la Grise, trad. de Guevara, l’Orloge des princes, 16th c., I, 3;

Huguet, Dictionnaire de la langue française du seizième siècle)

d. Ceux qui ont le mieux estudié à la connoissance de l’ame

‘Those who have best applied themselves to the knowledge of the soul’

(de la Pinelière, Le Parn. ou Critique des Poètes, 17th c., 19; Brunot,

vol.3(2), 546)

Adirectional verbs can and do therefore select for first and third order entities, but these may not be expressed as IOs in Modern French. Only second order objects are permitted. Not surprisingly, it turns out that within the group of adirectional verbs, only those verbs whose semantic selection is restricted to second order entities did not undergo a valency change. All other adirectional verbs that select first and third order objects underwent a change, illustrated in table 4.

Table 4 : Object selection of adirectional verbs

ADIRECTIONAL VERBS

SECOND ORDER OBJECTS FIRST, SECOND AND THIRD ORDER OBJECTS attenter aider compatir applaudir consentir assister obvier commander parer empêcher participer étudier pourvoir prier remédier servir satisfaire soigner etc. etc.

186

4.2.4 Generalisation

We have identified three properties that define the group of verbs that underwent the valency change and that distinguish them from all other verbs taking indirect objects. These properties have to do with argument structure, lexical directionality and the level of abstraction of the referent of the internal argument.

Verbs like aider have a non relational argument structure, setting them apart from many other types of verbs that take IOs, such as donner, préférer, arriver, ressembler, plaire, etc. Non relational verbs can further be divided into directional and adirectional verbs. Aider-type verbs are adirectional, contrasting with a large group of directional verbs like obéir, parler, penser and aboutir which have not undergone a valency change. Finally, within the small group of non relational adirectional verbs, those that underwent a valency change differ from those that did not in the kind of object they select. Verbs that did not undergo a change, like remédier, obvier and parer, select strictly second order entities (events, situations, places) whereas those that did, like aider, servir and prier, can select for first, second or third order entities.

(35) VALENCY CHANGE NO VALENCY CHANGE

 non relational constructions  relational constructions

 adirectional verbs  directional verbs

 first and third order entities  second order entities

Identifying the distinctive properties of the group of verbs that underwent the valency change not only supports a unified change, but it also allow us to narrow down the possible cause of the change. Since the change only affects a small, well-defined group of verbs to the exclusion of all others taking IOs, we presume that the cause is related to the distinctive properties of this group.

One way of looking at the cause of change would be to propose that it only targets first and third order IOs and that relational constructions and lexical directionality “protect” these IOs from replacement by a DO. In other words, first and third order IOs are only licensed in relational and directional contexts. The difference between Medieval and Modern French resides in the range of these contexts. This is not to say that aider-type verbs were relational or directional in

187

Medieval French, but rather, that directionality was available elsewhere in the derivation of a verb phrase, namely via the preposition à. This would allow first and third order IOs to occur in otherwise illicit contexts. When the preposition à could no longer encode directionality, the range of contexts in which these IOs could occur was consequently reduced.

One prediction that our analysis makes is that only first and third order IOs should have undergone the change. All second order entities should resist it. This is partially true for verbs like aider, applaudir, commander and insulter, but it does not hold for the other verbs whose inanimate (second order) IO is replaced by a DO along with the animates. Some examples are given in (36).

(36) a. La tempête contrariait (*à) la marche du navire.

‘The storm was impeding the ship’s progress.’

b. Les événements ont contredit (*à) ses prédictions, ses espérances.

‘The events have contradicted his predictions, his hopes.’

c. Épargner (*à) une somme d’argent.

‘To save a sum of money.’

d. Favoriser (*à) la privatisation.’

‘To favour privatization.’

Recall, however, that inanimates do not generally follow the same time course of change as the animates. For the nine verbs that were examined in detail, there is no case where the decline of the inanimate IO precedes that of the animate. Datives were clearly most directly affected by the valency change, and indeed, they appear to lead it. The second order entities follow the change, that is, when they occur with a verb that also selects first and third order entities, they change by analogy with those objects. Moreover, inanimate IOs encompass both second and third order entities. The decline in frequency of the inanimates also reflects the predicted replacement of the third order IOs by a DO.

188

4.3 Chapter Summary

In this chapter we defined aider-type verbs in terms of Levin’s (1999) non core transitive verbs. We then identified three properties that set aider-type verbs apart from all other verbs taking a thematic indirect object: their argument structure, the presence or absence of lexical directionality and the level of abstraction of the nominal complement.

The general picture that emerges is that inherent datives are most susceptible to change in French. Only directional verbs like parler, sourire, obéir, etc. occur with inherent datives in Modern French. Our proposal that lexical directionality licenses inherent dative objects follows from this observation. What we see in Medieval French is a broader range of dynamic two-place verbs that are able to occur with dative objects, a result, we claim, of the possible directional interpretation of à during that period. Directionality licenses inherent datives in French, and since directionality was available in the dative marker itself in Medieval French, there are more occurrences of inherent dative objects. In the next chapter we develop this idea, showing that when Path was no longer available as a derived feature of prepositions, dative objects would no longer be licensed with adirectional verbs like aider.

189

Chapter 5 Cause and effect

In this chapter, we explore a possible cause of the valency change that affected verbs like aider. In light of our quantitative and qualitative studies, there should be no question that the shift from an IO to a DO is the result of an underlying change in an abstract grammatical element associated with the derivation of these verbs and with their distinctive properties. Chapter three showed that they changed as a group within the same limited time period and chapter four identified clear structural and semantic constraints on the change.

The conclusion of our analysis thus far is that within the class of non relational dynamic two- place verbs, directionality licenses first and third order IOs. Directionality is only available via the lexical semantics of individual lexical items in Modern French, so we only find these kinds of IOs occurring with directional verbs like parler, sourire and obéir. This chapter develops the idea that directionality was also available via the preposition à in Medieval French, accounting for the broader range of dynamic two-place verbs that could occur with first and third order IOs. To answer the question of why directionality should license dative objects in these contexts, we propose an examination of the meaning of the preposition à and the selectional restrictions on its object.

Within a cue-based theory of language change (Lightfoot 1999, 2006), we must also account for how it happened that native speakers of 15th century Medieval French acquired a grammar in which the preposition à could be construed as directional, but speakers a century later acquired à as it is today: a purely locative preposition. We propose that along with the loss of the verb particle system came the loss of the single most robust source of unambiguous evidence that directionality was a derived property of prepositions.

The chapter is organized as follows: in section 5.1, we review previous studies that treat the valency change of aider-type verbs. Since none of them base their conclusions on a detailed quantitative and qualitative analysis, they do not account for the verbs as a group and therefore miss the structural and semantic constraints on the change. Section 5.2 focuses on the local cause of change. We considers the meaning of the preposition à and its selectional properties in light of

190 the apparent restrictions on first and third order IOs to relational or directional contexts. In section 5.3, we demonstrate that all prepositions, including à, could encode directionality via conflation with a directional functional head. The difference between Medieval and Modern French in this regard is thus the presence vs. the absence of such a head. The preposition à is purely locative in Modern French and can no longer be construed as directional. In section 5.4, we approach the valency change from a cue-based approach to language change. The loss of the verb particle system is argued to obscure the cue for a functional category encoding directionality. Speakers of French in the 16th century acquired a grammar in which directionality was not a derived property of prepositions, but a lexical property belonging to individual items. The preposition à was from that point on purely locative and, without the possibility of bringing directionality to the verb phrase, incapable of licensing first and third order IOs with aider-type verbs. Correlations with other structural changes are discussed. Section 5.5 briefly addresses the continued use of aider and prier with dative objects and 5.6 concludes.

5.1 Previous accounts

Changes in argument realization in French, particularly the replacement of the dative object of aider by an accusative, have inspired a number of accounts of the cause of the change. A variety of approaches, from rule-based analogical change to functional reanalysis, have proposed why the expression of these objects changes, but none can account for the group of verbs that underwent the valency change to the exclusion of all others, and no proposal attempts to account for why the change happened when it did. Our analysis hopes to contribute to a better understanding of the phenomenon by underlining the value of vast, searchable electronic data bases to diachronic linguistics. Patterns of change are now apparent where they were not before. It is these patterns of change that beg a re-examination of the cause underlying the systematic valency change.

5.1.1. Analogical change

Variation and change in argument realization easily lend themselves to explanations based on the force of analogy. Within a Neo-grammarian framework, analogy is considered to underlie most language change and in this sense, it is a unifying cause. Yet, since each verb or small group of verbs is susceptible to a different analogical force, the local causes of change differ. A change of valence in any given verb would depend on the influence of another verb with a similar meaning

191 but different argument realization, summed up in Brunot and Bruneau’s Précis de grammaire historique (1969, §272): Les changements de construction, amenés par l’analogie, sont extrêmement fréquents [...]. Le mécanisme est le suivant: par analogie, ‘il se rappelle de quelque chose’ apparaît et se développe à côté de ‘je me souviens de quelque chose’. C’est que les verbes sont groupés dans l’esprit suivant leur sens: les verbes peu communs tendent à adopter la construction des verbes plus usités.

[Changes in construction, brought about by analogy, are extremely frequent ... The mechanism is the following : by analogy, ‘il se rappelle de quelque chose’ appears and develops along side ‘je me souviens de quelque chose’. This is because verbs are grouped in the mind according to their meaning. Less common verbs tend to adopt the construction of more commonly used verbs.]

Nyrop, in his Grammaire historique, draws upon analogy to explain why some French verbs no longer have the same argument realization as their Latin cognates. The citation below illustrates just how commonly assumed analogy is in these cases. Its effect is so obvious that no explanation is needed. When a verb’s valence ceases to pattern after its Latin ancestor, then some sort of analogical rule particular to French must have applied to it: La construction française reproduit parfois celle du latin; on dit nuire, obéir, plaire à qn, parce qu’on disait primitivement nocere, obedire, placere alicui. Dans d’autres cas, il s’agit sans doute d’un développement français dû à une analogie quelconque.

(Nyrop 1930: 185, vol.6)

[French verb constructions sometimes follow from the Latin; one says nuire, obéir, plaire à qn, because one originally said nocere, obedire, placere alicui. In other cases, it is undoubtedly a question of a development particular to French due to analogy of some sort.]

Analogy can be seen as either random or systematic. In the former case, the force of analogy operates on individual verbs, as Brunot and Bruneau suggest for rappeler, and depends on the relative frequency of near synonyms. It can be considered systematic when, for instance, an analogical rule applies to a number of verbs at once. Lüdi’s examination (1978, 1983) of the verbs prier, supplier, requérir and aider proposes just this. He argues that the dative

192 complements of these four verbs used to pattern by analogy of meaning with dative benefactors. Benefactors are non core participants who benefit from the event described by the verb: Jean lui a ouvert la porte ‘Jean opened the door for him’.49 Lüdi reasons that the object of aider for example, is analyzed as a dative benefactor in Medieval French in that it accomplishes an activity by virtue of the help or instigation of the subject. He argues that in Modern French on the other hand, aider is subject to structural analogy with predicates like amener ‘bring, cause’ and conduire ‘lead’, which take a direct object followed by an infinitive clause, pointing out the structural similarities between amener/conduire quelqu’un à faire quelque chose and aider quelqu’un à faire quelque chose. What has changed in the course of the history of French is an underlying tendency or analogical rule; semantic analogy, which had precedence for certain verbs in Medieval French, has been replaced by structural analogy.

There are two drawbacks to Lüdi’s proposal. Theoretically, it is difficult if not impossible to define the nature of these underlying tendencies and analogical rules. Where do they come from and what causes them to change? This is a longstanding problem for the Neo-grammarian approach to language change, and the reader is referred to Lightfoot (1999, chapter 2) for an in depth discussion of the issue. Empirically, Lüdi’s analysis can only account for a subset of the verbs we have identified as undergoing the same valency change. It does not account for verbs which do not occur with infinitive clauses such as applaudir, assister, contrarier, etc.

5.1.2 IO/DO ambiguity and random evolution

Goyens (1998) examines the evolution of the valence of six verbs from Latin to Modern French, each showing a different pattern of change: obéir ‘to obey’, ressembler ‘to resemble’, mentir ‘to lie’, prévenir ‘to prevent, avert’, empêcher ‘to trouble, hinder’ and contredire ‘to oppose, contradict’. Drawing mainly upon historical grammars and dictionaries, Goyens makes two main observations: first, the form that each object takes today (either an IO or a DO) does not necessarily correspond to the case it took in Latin, and second, in Old and Middle French, the

49 Drawing a parallel between the dative argument of aider and dative benefactors is apparently quite tempting: recall McFadden’s (2006) analysis discussed in section 4.1.4.

193 complement of all six verbs alternates between a prepositional and non prepositional object before evolving towards its present form.

Goyens proposes that the availability of the double object construction in Old French lies at the heart of the valency change. Following Herslund (1980), two different derivations are available to certain two-place verbs such as aider, ennuyer, mentir, etc.: a derivation involving an IO with an overt preposition and one without. The two derivations are assumed to parallel those available to three-place verbs of transfer. Compare the expression of the object of aider with that of the recipient argument of donner, first as a prepositional dative, then as a non prepositional dative:

(1) a. donner quelque chose à quelqu’un/ aider à quelqu’un

‘give something to someone/ help à someone’

b. donner quelqu’un quelque chose/ aider quelqu’un

‘give someone something/ help someone’

Goyens suggests that the prepositional/non prepositional alternation in two-place verbs created a certain amount of ambiguity for speakers making it unclear whether the non prepositional object of a given two-place verb was a DO or whether it was a dative IO that could be realized without the preposition à. Speaker uncertainty would explain the alternation for the six verbs during the Middle French period and would eventually entail a valency change. Some two-place verbs that originally governed a dative complement would have been mistaken as governing a DO (the case for aider and contredire, for example). Conversely, verbs governing a DO might have been mistaken as governing an underlying dative complement (perhaps the case for ressembler and mentir). Various other factors may have also determined the current case marking of the object complement: the influence of Latin case marking, verbal prefixation, the increasing grammaticalization of the preposition à, and verbal semantics are suggested as possibilities. The result is an essentially random evolution toward the form these complements take today.

According to her account, one would expect to see DO/IO alternations in all two place verbs taking a DO or an IO, but this, of course, is not at all the case. Our examination of the facts shows that only a relatively small number of two-place verbs show such an alternation in Medieval French, whence the hypothesis of selective change with a single underlying cause.

194

Goyens examines a small number of verbs each presenting a different kind of valency change rather than focusing on verbs showing the same kind of change and so it is no wonder the changes are considered random and unrelated. She herself notes that to draw any firm conclusions about the changes, an examination of more verbs and the use of a more extensive data base would be required. This is precisely what we have proposed for the present study. We show through both qualitative and quantitative analysis of over 30 verbs that the replacement of certain IOs by an analogous DO is indeed a systematic change. Cases that appear to move in the opposite direction are not counter examples, but can be explained independently.50

5.1.3 Grammaticalization of the S-V-IO construction

Schøsler (2003, To appear) observes that the dative object of two-place verbs in Modern French often denotes the experiencer of an event, as in plaire à quelqu’un ‘to please/appeal to someone’. She proposes, in fact, that the S-V-IO construction has become increasingly specialized, based on analogy with the dative argument of psych verbs and the dative benefactor of impersonal constructions, so that in Modern French the construction itself expresses a psychological relation between the animate object on the one hand and the subject on the other. As a consequence, this pattern is claimed to favour psych verbs like plaire and nuire and to no longer be compatible with verbs that do not express a psychological relation. One piece of evidence Schøsler uses to support her case is the valency change in helping verbs such as aider, assister, secourir and servir. She reasons that because helping verbs do not express the psychological state of the object, they are no longer mapped to the meaningful S-V-IO valency pattern. This would explain their shift towards an S-V-DO structure, a default structure in Modern French, according to Schøsler, which expresses no particular meaning.

Schøsler’s account of the valency change in the helping verbs does not, however, provide a fruitful avenue of exploration with respect to the larger valency change that concerns us here. We do not find that the hypothesis captures a structural generalization in French. Psych verbs represent only a small number of two-place dative verbs in French; there are a number of verbs

50 See chapter two, section 2.2.3 and 2.2.4 for a discussion of why mentir and ressembler are not considered to have undergone a purely structural change. The object alternation for obéir is elaborated in 2.2.4.

195 with the same superficial structure which are clearly not object experiencer verbs (ressembler, survivre, succéder, appartenir, etc.), a point that Schøsler herself raises. She attributes these cases to fossilization (note 13). Furthermore, the kind of analogical change that she describes would be difficult to maintain in light of our quantitative study. The uniform and relatively sudden change that we observe would not be expected. Rather, one would predict that the dative is lost to helping verbs on a verb by verb basis, through a gradual process of morphosyntactic levelling.

5.1.4 Functional reanalysis

In his influential book Le Problème de la transitivité en français moderne (1960), Blinkenberg proposes a unified functional definition of transitivity. He claims that in French, transitivity – the mutual determination between the verb and its object – is expressed by both direct objects and indirect objects introduced by the preposition à. Central to his argument is the well-known variation and change in argument realization in Latin and French verbs. The author claims that the apparent free variation between DOs and IOs illustrates that the two forms have the same function. Verbs such as aider (p.180) and insulter (p.185) which show both synchronic and diachronic variation between an IO and a DO without any appreciable meaning difference, indicate, he argues, an absence of any functional distinction between the two forms.

The vacillation between an IO and a DO in the history of French is reduced to one cause: the general evolution towards a syntax centred around the notion of transitivity for which the phrase is the basic syntactic unit of meaning as opposed to the word. One consequence of this is that the functional meaning once expressed by case marking is now specified in the lexical content of the verb. This implies a mutual dependency between the verb and its object without any need for special marking on the object. The preposition à that heads the IO is thus essentially redundant, making the IO interchangeable with a DO.51

51 See Croft’s theory of hyperanalysis (2000:121-124) for a similar take on the loss of governed oblique case in Russian and Early Middle English. What Blinkenberg describes seems to correspond to the typological shift in French from a satellite-framed to a verb-framed language (Talmy 2000).

196

In order to account for the fact that certain verbs still take IOs in French, Blinkenberg suggests that since à no longer has any lexical content, many IOs are vestiges of an earlier time when case marking expressed distinct functions (early Latin, for example):

Un groupe important d’objets indirects introduits par à provient évidemment de la substitution commencée dans le latin tardif de à + acc. à l’ancien datif et continue ainsi dans un moule nouveau une très vieille tradition répartissant les objets en plusieurs groupes principaux. Cette tradition avait connu longtemps un affaiblissement sensible, qui est facile à constater en latin et encore davantage en ancien français. (Blinkenberg 1960:48)

[A large group of indirect objects introduced by the preposition à evidently comes from the substitution, beginning in Late Latin, of the former dative by à + acc., and so continues in a new mould the very old tradition separating objects into several main groups. This tradition has undergone for a some time now a significant weakening which is easy to see in Latin and still easier in Old French.]

Other IOs that do not correspond to the Latin dative such as that of applaudir (p.184), would have initially been introduced as a circumstantial complement of a sentence in which the verb is used intransitively. He speculates for instance, that speakers adjoined a circumstantial complement headed by the preposition à to the verb applaudir in order to specify – after the fact – what they were applauding. The evolution from a circumstantial complement to a verbal complement would have followed once the PP constituent was interpreted to be part of the verb’s meaning and once à no longer had the semantic content required of a circumstantial. Reanalysis of the PP as the object of the verb would have gone as follows: applaudir → applaudir, à quelque chose → applaudir à quelque chose.

Blinkenberg’s explanation of the directionality of the change – the fact that the IO of verbs like aider, insulter and applaudir has evolved into a DO in Modern French – follows from his central claim that the direct object is the ultimate expression of the mutual determination between the verb and its object. Exceptions to this evolution are two-place verbs whose IO has not become a DO, or IOs which do not correspond to a dative Latin cognate. These aberrant cases would have been subject to analogy:

197

Nous pensons qu’il faut attacher une plus grande importance à un autre facteur de variations dans la forme de la transitivité, à savoir l’affinité de sens entre verbes à transitivité différente. Dans de tels cas en effet, le sens du verbe peut amorcer une analogie qui change la forme de la transitivité; une synonymie complète ou partielle amène une contamination syntactique. (Blinkenberg 1960:49)

[We think that a greater importance must be attributed to another factor of variations in the transitive form, namely affinities of meaning between verbs with different transitivities. In such cases, the meaning of the verb can in effect initiate an analogy that changes the transitive form. Complete or partial synonymy leads to syntactic contamination.]

Indeed, functional equivalence between the IO and the DO coupled with the forces of analogy predict massive variation in valence and random distribution of IOs and DOs among verbs:

S’il est naturel de supposer à l’origine de l’opposition ancienne entre un objet à l’accusatif et un objet au datif une différence de sens, reposant sur les valeurs spécifiques des deux cas, dans la mesure où de telles valeurs soient arrivées jamais à une complète consistance, il faut avouer que pour la partie de l’évolution que l’histoire éclaire il est difficile de rendre compte de la répartition entre les deux groupes de cas par des critères sémantiques évidents. On trouve aussi bien en latin qu’en français des verbes qui sont apparentés pour le sens et dont cependant les régimes adoptent des formes différentes.

Une conséquence directe de cette répartition en apparence capricieuse des régimes est le nombre de changements de régime constaté par la grammaire latine. Apparentés par le sens, sans distinction autre que la distinction lexicale dans les emplois absolus, ces verbes sont facilement victimes d’une “erreur” de construction... (Blinkenberg 1960:49-50)

[If it is natural to assume that a difference in meaning is at the origin of the former opposition between an object in the accusative case and an object in the dative case based on the specific values of the two cases, in so far as such values have ever arrived at complete consistency, one must admit that for the part of evolution on

198

which history sheds some light, it is difficult to account for the division between the two cases by obvious semantic criteria. Verbs which have related meanings but for which their objects have different forms are found both in Latin and in French.

A direct consequence of this apparently random distribution of object case marking is the number of changes that Latin has undergone. Related by meaning, without any distinction other than the lexical distinction they have in intransitive uses, these verbs are easily victims of a construction “error”...]

Considering the facts however, Blinkenberg’s explanation of variation and change in argument realization in French is difficult to maintain. First, it is claimed that since Latin, verbal objects have been susceptible to free variation between dative and accusative case without any appreciable meaning difference. This is, however, not the case. Many of the verbs we have examined show semantically motivated variation where one form is not interchangeable with the other. Indirect objects in French are encoded as a point of reference in space, a Place, and, depending on the verb they occur with, generally do not imply any physical contact and have a certain degree of autonomy. In this respect, they are distinct from direct objects which can and often do denote objects that undergo some physical change and which are not construed as having any degree of autonomy in the event; see chapter two for examples from Middle French and section 5.3 where we discuss similar semantically motivated DO/IO alternations in Modern French. Of course, in order to maintain his thesis, Blinkenberg must emphasize the amount of verbs that allow DO/IO alternations, while at the same time downplaying the semantic differences between a verb used with a DO and the same one used with an IO. By contrast, we believe the meaning differences are too salient and too numerous to ignore. As explained in chapter three, we do not expect to see free variation of any kind unless a change is in progress involving two competing grammars.

Another fact that is not mentioned in Blinkenberg’s work is that DO/IO variation is unidirectional. The sole DO of a verb may alternate with an IO, but the opposite is not true. For example, the transitive use of toucher, in toucher la voiture, can alternate with toucher à la voiture, but a verb like parler used with an indirect object in parler à Jean cannot alternate with parler Jean. This asymmetry should not exist within a theory that posits functional equivalence between the IO and the DO. Furthermore, according to Blinkenberg’s thesis, any IO should, in

199 principle, be susceptible to variation and change, yet this is clearly not so. Some verbs never vary between a DO and an IO complement. The IO of directional verbs such as répondre, and résister, for example, and of verbs of comparison such as correspondre and préexister are never in free variation with a DO. Finally, the hypothesis of variation with a slow evolution towards a direct object complement would not predict the sudden and uniform change from an IO to a DO observed for the aider-type verbs.

While we do not agree that DOs and IOs are functionally equivalent in French, we concede that some may appear so for two reasons: first, DOs and IOs may occupy the same syntactic position relative to the verb (i.e., the complement position); second, the general meaning of the direct object is underspecified and overlaps with some of the semantic roles that correspond to the IO. An excellent example of this is the seamless replacement of the IO of aider-type verbs by a DO during the course of the history of French. Ostensibly, the direct object can sufficiently convey that which is conceptually necessary for the object of this group of verbs. One might ask then: if the DO can do what the IO does, why would the latter even appear with these verbs? Our answer would be that the IO, being more specific, better expresses the role of the object of interaction verbs than a DO, and that the grammar makes use of the IO in these contexts when it is available (this question is raised in Levin 1999 where she invokes Kiparsky’s Elsewhere Condition). Meaning differences, of course, notwithstanding.

Finally, we agree with Blinkenberg – and with Kilroe (1989) – that the loss of lexical content in the preposition à plays a central role in valency change. Where we differ from these two authors is just how the loss of directionality in à affected the status of the IO in French. Blinkenberg proposes that the inherent semantic value of the preposition à is reanalyzed as belonging to the verb with the consequence that à becomes completely devoid of meaning, semantically unnecessary, and is eventually omitted. Similarly, Kilroe (1989:187) suggests that à has become “superfluous” in Modern French, making it functionally equivalent to the DO. According to this approach, we expect to see massive variation with gradual change toward a DO, probably by lexical diffusion, exceptions due to analogy not withstanding. This of course, does not correspond to the empirical facts. The change from an IO to a DO is abrupt and highly selective.

200

5.1.5 Loss of inherent dative case

Van Peteghem (2006) argues that dative case is purely structural in Modern French based on the observation that most datives appear to depend on the presence of a direct object. The basic idea is that dative arguments are introduced into a syntactic configuration that already includes a direct object. She notes however, that two-place verbs like causer, mentir, sourire, parler, and téléphoner pose a problem for her hypothesis since the presence of a direct object is not immediately obvious. She admits that these could be exceptional cases of a non configurationally assigned dative case, or inherent dative case, licensed purely by the semantics of the governing verb. Taking the case of aider, she concludes that during the course of the history of French, there was some alternation between structural and inherent case, but that in Modern French, structural case is preferred.

Van Peteghem’s assessment is actually not far from what we propose. In chapter one section 1.3, and chapter four, section 4.2.1, we argue that while the majority of IOs in French are licensed structurally within a small clause relation (a PP), there exist a number of IOs that are not structurally licensed: namely, those occurring with the group of interaction and reflection verbs exemplified by remédier, résister, parler, penser and échapper, meaning that inherent dative case does in fact exist in Modern French. Among this group of verbs, dative case was shown to be limited to directional verbs. What has changed in the course of the history of French is the range of verbs with which inherent datives can occur; crucially, non directional verbs like aider are out.

Although Van Peteghem’s analysis is not concerned with language change, it is invoked to justify the apparent exceptions to her analysis. Implicit in what Van Peteghem writes is that, much like a parameter, inherent dative case either exists in a language or it doesn’t. Yet, there is no suggestion as to what it may depend on and thus why it may exist at one point in the history of a language and not in the next. One of the things our study shows is that a unified treatment of dative objects is not fruitful. Datives are just one of several kinds of indirect objects, and they are not all licensed in the same way. Their occurrence with a given verb is a function of a number of different interdependent properties, namely argument structure, directionality and the nature of the object.

201

5.2 The meaning of à

The preposition à, its meaning and its role in the valency change, is central to the second part of our claim. We propose that when the preposition à was no longer able to encode directionality, it was likewise no longer able to license first and third order entities as its object. The details of how à encoded directionality in Medieval French is taken up in the next section. This section begins by considering the meaning of the preposition and its selectional properties in light of the apparent restrictions on first and third order IOs to relational or directional contexts.

There is a general consensus in the literature that the main function of the preposition à is to localize a Figure (or target) X at a Ground (or landmark) Y. Following the work of Ruwet (1982), Vandeloise (1987, 1991) and Goyens et al. (2002), they show that in purely locative contexts, the prepositional object must already have an inherent locative value in order for the IO to actually be interpreted as such. In Ruwet’s words:

Il semble que le contenu sémantique de à (pour autant qu’à ait un contenu sémantique) soit insuffisant pour marquer à lui seul la valeur locative du complément introduit par cette préposition. En général on peut dire que, si on veut avoir un complément de lieu de forme à NP, il est nécessaire que le NP qui suit à ait lui- même, d’une manière ou d’une autre, une valeur locative intrinsèque, ce qui n’est pas nécessairement le cas quand il s’agit de compléments de la forme dans NP, sur NP, etc. (Ruwet, 1982:320)

[It seems that the semantic content of à (in so far as à has semantic content) is not sufficient enough to mark on its own the locative value of a complement introduced by this preposition. In general it can be said that, if one wants to have a complement of place of the form à NP, the NP following à must itself have an intrinsic localizing value in some sense; this is not necessarily the case when we consider complements of the form dans NP, sur NP, etc.]

(2) a. * Pierre était couché à son tigre apprivoisé. (Ruwet 1982:320)

‘Pierre was lying down at his tame tiger.’

202

b. Ce film passe au cinéma de mon quartier. (Ruwet 1982:320)

‘That film is playing at the movie theatre in my neighbourhood.’

Vandeloise (1991) specifies that the landmark is conceptualized as a single point, regardless of its actual physical dimensions. As a point of reference, the complement of à contrasts with that of prepositions such as dans, sur, etc. which require specific physical dimensions. His detailed examination of the restrictions on the object of the preposition à demonstrates that if it does not have an intrinsic locative value, like le cinéma in (2b), then it must call to mind a social routine. A concrete entity such as son tigre apprivoisé in (2a) is inappropriate since it is not intrinsically locative and it does not evoke some kind of social routine. Other concrete entities, however, like la fenêtre ‘the window’, evoke social routines and are thus acceptable objects of à:

(3) a. Jean est à la fenêtre.

‘Jean is at the window.’

b. La plante est à la fenêtre.

‘The plant is by the window.’

In example (3a), la fenêtre evokes a well-known social routine of standing in front of a window and gazing through it. Interestingly, example (3b) evokes an entirely different routine, one of placing a house plant on a window sill so that its leaves will be exposed to sunlight.52 What Vandeloise intends to show in these cases is that à la fenêtre does not simply mean to be at the window. If it did, we would not expect two completely different implications.

Goyens et al. (2002) further observe that when à is used in a locative sense, the complement of à is in fact restricted to second order entities (any DP that denotes a static or dynamic situation). As is now clear, first order entities (discrete physical objects) are very difficult, if not completely excluded in these contexts, as shown in (4) and (5); examples from Goyens et al. (2002:295-96).

(4) a. Il va à Lille/ Léa va au bureau chercher le livre.

‘He is going to Lille/ Lea is going to the office to get a book.’

52 Vandeloise discusses other restrictions on the object of the preposition à that are not taken up here.

203

b. Léa va au mariage/ au colloque/ à la réunion.

‘Lea is going to the wedding/ to the conference/ to the meeting.’

(5) a. * Va à Jean et demande-lui la vérité.

‘Go to Jean and ask him for the truth.’

b. ?* Va à la table et apporte-moi le livre.

‘Go to the table and bring me the book.’

If we accept that the principle function of à is to establish the co-occurrence of two entities X and Y, then these examples illustrate that the minimal condition that à imposes on its object is that it be a second order entity. In other words, second order entities should be licensed across the board as the object of à. On the other hand, in order for first and third order entities to be acceptable forms of the object of à, they must be licensed by other means.

One way that first and third order entities can actually occur as the complement of the preposition à is within a Figure-Ground relation in which there is an inference of possession. In this way, when a Figure object is “at” someone, it is construed as being in his or her possession (6a).53 If, on the other hand, the relation expresses co-location, then when a Figure object is “at” something, it can be interpreted as a relation of possession, in (6b), or of destination, in (6c). Abstract relations that denote association and comparison also permit first and third order entities as the object of à, as shown in (6d) and (6e).

(6) a. le stylo à Marie

Marie’s pen

b. l’homme au fusil

‘the man with the gun’

53 Co-location is a common way to express possession cross-linguistically; see chapter one, section 1.3.4 for some discussion in so far as this concept relates to psychological predicates.

204

c. un verre à vin

‘a wine glass’

d. Paul ressemble à son père.

‘Paul resembles his father.’

e. Paul préfère les pâtes au riz.

‘Paul prefers pasta to rice.’

While all the sentences in (6) denote static events, the basic inference of possession or co- location is also available in dynamic contexts. In these cases, the animate IO is interpreted as a Recipient and the inanimate IO as a Source or bounded Goal as shown in (7).

(7) a. Paul a donné des fleurs à Marie.

‘Paul gave flowers to Marie.’

b. On a enlevé une couche de peinture au banc.

‘They removed a coat of paint from the bench.’

c. Il faudrait ajouter quelques touches plus claires au portrait.

‘Some lighter touches would need to be added to the portrait.’

First and third order IOs are therefore licensed as the object of à in relational contexts which imply possession, co-location, or comparison with another object. Given the right inference, relational contexts in general can license first, second, and third order entities. (8) 3 DP 3 à DP [1, 2, 3 order]

Note that because à locates X at Y, Y is always the place at which X begins or ends up (or is anticipated to end up); in this sense, when Y is interpreted as a Goal it must be a bounded Goal. But what happens when there is no X, as is the case with two-place verbs such as parler, répondre, penser, etc.?

205

(9) a. Tu parles à Jean.

‘You speak to Jean.’

b. Tu réponds à la question.

‘You answer the question.’

c. Tu penses à Marie

‘You think of Marie.’

The examples in (9) show that there is no Figure object X located metaphorically or otherwise at the Ground, Y. Instead, the semantic contribution of the directional verb permits Y to be interpreted as un unbounded Goal; this means that the action is directed toward the participant along an imaginary path, but it does not necessarily terminate there. This is crucial because as an unbounded Goal, the referent of the object of à is not constrained to second order entities or to specific contexts in which it can be construed as a place at which some X exists. By allowing an unbounded Goal interpretation, directionality lifts the semantic constraints on the object of the preposition à and therefore provides another way in which first and third order entities can occur in this position.54

(10) 3 V[+dir] 3 à DP [1, 2, 3 order]

Interestingly, when the non relational construction in (10) is maintained, but the verb does not contribute any directional meaning, the IO is restricted to verbs that select second order entities in Modern French:

(11) a. On va remédier à cette situation/ * à cet enfant qui s’est cassé la jambe.

‘They are going to remedy this situation/ this child who have broken his leg.’

54 Inherent datives in Modern French therefore have the unique interpretation of unbounded Goal.

206

b. On va attenter à la liberté de la presse/ * au président.

‘They are going to conspire against the freedom of the press/ make an attempt on

the president.’

c. Il a paré au scandale/ * à l’ennemi.

‘He dealt with the scandal/ the enemy’

This restriction becomes even more apparent in light of the distribution of second order vs. first and third order IOs with verbs that allow DO/IO alternations. Verbs that denote events which imply some kind of directed motion can also occur with first and third order IOs, much like the verbs in (9). Verbs that do not, only occur with second order IOs, like those in (11).

Direction can be implied with verbs of contact and of pursuit, playing on the sense of contact on the one hand, and on the trajectory necessary to establish contact on the other. As the sentences in (12) illustrate, a directional reading of the verb emphasizes the process rather than the object with the net effect that the IO is construed as less affected than the DO. Blinkenberg (1960:176) describes the alternation as follows: “la présence de la préposition souligne l’élément de localisation qui est impliqué dans l’action et que la construction à objet direct “affecté” laisse inexprimé.” [the presence of the preposition underlines the element of localisation which is implicated in the action and that the “affected” direct object construction leaves unexpressed.].55

(12) a. Cogner/frapper/heurter/taper la porte/ à la porte56

‘To hit the door/ to knock at the door’

55 Alternations such as a) Je l’ai pardonné and b) Je lui ai pardonné in spoken/colloquial French involve the option of construing the person forgiven as either an affected theme in (a) or the recipient of an implicit theme in (b). This alternation is much like the locative alternation J’ai chargé le camion / J’ai chargé le foin dans le camion; the difference being that the direct object is implicit in the ditrasitive use of the verb pardonner.

56 Although the verb toquer ‘to knock’ is only used with an IO in Modern French, the expression of its object used to alternate, as other contact verbs, between a DO and an IO (see the TLFi).

207

b. Gratter/pousser (à) quelque chose (Blinkenberg 1960:176)

‘To scratch/push (at) something’

c. Ne touche pas ce vase/ à ce vase

‘Do not touch the vase/ keep your hands off of the vase’

d. Il a visé le coeur/ au coeur

‘He had the heart in sight/ he took aim at the heart’

e. Chasser le tigre/ au tigre (Blinkenberg 1960:47)

‘To hunt the tiger/ after the tiger’

f. Tirer un lapin/ au pigeon (Blinkenberg 1960:178-9)

‘To shoot a rabbit/ at a pigeon’

g. Goûter les canapés/ aux canapés

‘To taste the hors d’oeuvres/ to try the hors d’oeuvres’

h. Grimper un arbre/ à un arbre

‘To climb a tree/ up a tree’

Conversely, verbs that do not denote an event of directed action do not occur with first or third order IOs. In these cases, the alternation does not play on the level of affectedness of the object, but significantly modifies the meaning of the event so that the same object can generally not be used in both constructions.

(13) a. Atteindre la ligne d’arrivée/ à la perfection

‘To reach the finish line/ to reach a state of perfection’

b. Travailler la terre/ à un article

‘To work the land/ to work towards the publication of an article’

208

c. Croire son professeur/ à un principe57

‘To believe one’s professor/ to believe in a principle’

d. Regarder la foule/ à la dépense

‘To look at the crowd/ to think about the expense’

e. Réussir un repas/ à un examen

‘To succeed in making a good meal/ to obtain a good mark on an exam’

f. Satisfaire son employeur/ à toutes leur exigences

‘To satisfy one’s employer/ to meet all of their demands’

g. Consentir un prêt/ à un divorce

‘To authorize a loan/ consent to a divorce’

h. Parer un coup/ au plus pressé

‘To deflect a blow/ to attend to the most urgent things’

i. Suppléer un enseignant/à une carence alimentaire

‘To replace a teacher/ make up for a lack of food’

j. Présider un comité/ aux préparatifs

‘To preside over a committee/ to be in charge of the preparations’

k. Veiller une malade/ au maintient de l’ordre

‘To watch over an ill person/ to attend to the maintenance of law and order’

There are therefore two types of DO/IO alternations in Modern French. The first involves verbs that can have a possible directional interpretation. The DO and IO can be of the same type, most notably they can both be first order entities, and the meaning difference concerns the level of affectedness of the object. The second does not involve verbs that have possible directional

57 Note that croire à ‘to believe in’ cannot be followed by a person (i.e a first order entity), the preposition must change to en to accomodate these kinds of objects: croire en ses amis.

209 interpretations. The DO and IO may not always be of the same type since the IO is strictly second order, and the alternation entails a significant difference in the meaning of the event.

The DO/IO alternation we observe in Medieval French appears to correspond to the first type of alternation. Recall, for instance, that the direct and indirect objects of verbs like aider are essentially interchangeable; IOs are not, for instance, limited to second order entities. In addition, the evidence that we have of meaning differences between the IO and the DO indicates that it is a question of a more or less affected object. This would suggest that the alternation is based on the presence or absence of directionality. Unlike Modern French, however, directionality is not introduced via the lexical semantics of the verb, as it is with those in (12), but rather, we claim that it is introduced via the preposition à.58 If the preposition à itself once possessed directional properties, then we would indeed expect to see more verbs occurring with dative objects since directionality licenses first and third order IOs.

5.3 Directionality and à

In this section, we demonstrate that the preposition à is exclusively locative in Modern French, contrasting with Medieval French, where all prepositions, including à, could encode directionality. The difference between Medieval and Modern French in this regard is the presence vs. the absence of a functional category encoding direction into which a prepositional head can conflate.

Recall that in section 5.2, it was shown that directionality licenses first and third degree IOs in non relational constructions. In Modern French, this property enters the derivation via the lexical semantics of the verb. Now, assuming that in Medieval French, directionality was also available via the preposition à, then we would expect to see more first and third degree IOs in non relational constructions. This is precisely how we account for adirectional verbs like aider which occur with dative objects in Medieval French but do not in Modern French.

58 Recall that the aider-type verbs have not undergone any obvious change in meaning, so we do not assume that they once possessed lexical directionality. If they did, they would presumably stil permit a DO/IO alternation.

210

Directionality is a semantic feature associated with the way in which space-time is conceived. It is a geometric notion that specifies how a Figure object (a moving or conceptually moveable entity) is oriented or moving with respect to a Ground (a reference landmark).59 Every language has terms that are used to express direction and it appears that cross-linguistically, in addition to case marking and affixes, etc., all major syntactic categories can encode direction (see van der Zee and Slack 2003). Below are some examples of elements that express direction in French:

(14) a. Marie est entrée dans la gare.

‘Marie entered the station.’

b. Jean a mis le cap sur le nord.

‘Jean headed north.’

c. Votre chambre regarde vers la colline.

‘Your bedroom faces/looks towards the hills.’

d. L’oiseau s’est envolé du nid.

‘The bird flew away from the nest.’ (Kopecka 2006:88)

In Modern French, à is a spatial preposition that localizes a Figure object with respect to a Ground (Ruwet 1982; Vandeloise 1987, 1991, among others). It does not however, orient the Figure in any particular way with respect to the Ground and therefore does not encode the semantic property direction. The preposition à cannot, for instance, introduce the object of the perception verb regarder in (14c), since it cannot mark la colline as the place towards which the window is facing. Rather, à merely identifies la colline as a Place. In the adirectional context of regarder, the only interpretation à la colline can have is one of a circumstantial complement: an event of looking takes place at a hill.

(15) * Votre chambre regarde à la colline.

‘Your room looks while at the hill.’

59 We use the terms direction and Path interchangeably here, but they are often considered to be two separate notions.

211

The preposition à has a purely locative function in Modern French. It identifies its complement as a Place. Note however, that as a locative preposition, à is often used with other elements that imply a sense of direction. It should not however, be mistaken as itself encoding this property.

(16) a. Paul a donné des fleurs à Marie.

‘Paul gave flowers to Marie’

b. Paul vient à Paris la semaine prochaine.

‘Paul is coming to Paris next week.’

c. Paul a volé les fleurs à Marie.

‘Paul stole the flowers from Marie.’

d. De Kingston à Montréal, ça prend trois heures (Tremblay 1999:175)

‘From Kingston to Montreal, it takes three hours’

The lexical properties of the verbs in (16a-c) orient the Figure object with respect to the Ground: donner and venir orient the Figure (les fleurs, Paul) towards the Ground (Marie, Paris) while voler orients the Figure (les fleurs) away from the Ground (Marie). Similarly, the construction in (16d) provides the necessary sense of direction or intention.60 In all cases, the preposition à simply identifies its complement as a point of reference. The specific interpretation of the complement as a possessor or recipient (16a), a goal (16b) and (16d), or a source (16c), is derived compositionally.

In contrast, the preposition à can encode directionality in Old French (see Goyens et al. 2002; Kilroy 1989). À can be used, for instance, in the sense of vers ‘towards’ without identifying its complement as an endpoint. In the following examples, the subject is heading for the sea (but not necessarily reaching it) or turning in the direction of someone. These interpretations are not

60 Tremblay (1999) uses this example to show that the preposition à has semantic content since it can identify its complement as a goal. By contrast, we claim the goal interpretation of Montréal is derived from the construction. À Montréal by itself means ‘at/in Montreal’; the preposition is purely locative. Only when this phrase is placed within a context like that of (5d) is Montreal interpreted as the end point of a path.

212 possible in Modern French since à no longer encodes this directional meaning. A different preposition is necessary in these cases, like directional vers.

(17) a. Dunc vint errant dreitement a la mer.

[Alors il partit en allant tout droit vers la mer]

‘Then he left heading straight for the sea’

(La Vie de Saint Alexis, v.76; Goyens et al. 2002:284)

b. Il se tourne donc a son père

‘He thus turns towards his father’

(Bossuet, Car. de St.-Germ.; Bourciez, 1946:688-89; Kilroe 1989:109)

c. Et quant messire Enguerrant vist ce tresbel gros diamant (...) se tourna aux autres

seigneurs prouchains et en son langaige castellam leur dist...

‘And when Master Enguerrant saw this very beautiful and large diamond... he

turned to the other lords next to him and in Castilian said to them...

(La Sale, Jehan de Saintré, 1456, p.120; DMF1)

À also functions as a productive directional preverb in Medieval French.61 This element enhanced the meaning of the verbal base, giving it an additional sense of direction, goal, or passage from one state to another: baisser/abaisser ‘lower/lower down’; porter/apporter ‘carry/bring along’, river/arriver ‘skirt the shore/arrive at shore’, etc. (taken from Dufresne et al. 2000).

The preposition à was also used in a strictly localizing sense as well, much as it is today:

61 We assume that preverbs and prepositions are the same lexical item. See Dufresne et al. (2000) for arguments in favour of treating Romance prefixes as prepositional elements.

213

(18) a. Or fu Amis a la cort son parrain

[Maintenant, Ami s’en alla à la cour de son parrain]

‘Now, Ami went to his godfather’s court’

(Ami et Amile, v.2499; Goyens et al. 2002:284)

b. ...et disons comment il devint prestre a Romme...

[disons comment il devint prêtre à Rome]

‘let’s tell how he became a priest in Rome’

(Cent Nouvelles Nouvelles, p.286; Goyens et al. 2002:284)

In Medieval French, the preposition à encoded both location and direction, whereas it only encodes location in Modern French.

5.3.1 Directionality as a derived property

Consider now how direction was most likely encoded in Old French. Following Rooryck (1996) and later Svenonious (2007), direction can be encoded in two ways: semantically or grammatically. Direction is semantically encoded when it can be considered part of the inherent meaning of a lexical element. It is grammatically encoded when a lexical element acquires directionality through the course of a syntactic derivation, namely by conflating with a functional head possessing a directional value. It has been proposed that in languages like English and Dutch for which prepositions like under, in and behind can be either directional or adirectional, the property is derived. In languages like Modern French on the other hand, where there is no variability, the property is inherent to particular prepositions like vers. But what of Old French for which the preposition à could have both directional and non directional properties? If directionality was indeed derived in Old French, then the loss of directionality in à (our principle concern) would correspond to its relation with a functional head, and not to a semantic change particular to à.

One way to establish if directionality was derived in Old French would be to check for relevant similarities between Old French and English or Dutch. We might expect, for example, that other prepositions in Old French vary between a directional and locative reading. We saw that this was

214 true for à, and it also holds for other prepositions. The preposition en, for instance, could be used in both directional and purely locative contexts.

(19) a. en: directional (directional verb particle)

quant il vindrent a la porte, si la trouverent ouverte; si entrerent enz

‘when they came to the door, they found it open and they went in.’

(Artu, p.57; Tremblay et al. 2003)

b. en: directional (with manner of motion verb)

... des tresgrosses roches qui, a force de gens, se puent esbranler en bas. Sy

vous certiffie que pour les faire roller en bas...

‘... very big rocks which, with the strength of some people, could be sent below.

You will see that to make them roll to the bottom...’

(La Sale, Salade, c.1442-44, p.77; DMF1)

c. en: locative

Li cuens Amiles jut la nuit en la sale En un grant lit a cristal et a saffres.

‘Count Amiles spent the night in the room in a great bed inlaid with crystal and

sapphires.

(Ami et Amile, 1200, p.21; TFA, Frantext)

Since directionality can be shown to be present, but not an inherent feature of prepositions, we propose that it was derived by conflation with a functional head possessing that meaning. A simplified representation of how this would look is given in (20).

(20) FdirP 3 Fdir PP Ø 3 P DP à, en...

215

The functional head would be affixal with no phonological form, at least in this context, and therefore adopts, through conflation, the phonological signature of the prepositional head. Rooryck describes the function of this category as similar to an aspect phrase in the verbal functional domain. Its lexical entry would look something like that in (21).

(21) Lexical entry for Fdir

PHONOLOGY: Ø, +affixal

SYNTAX : F

SEMANTICS: direction

Although the details surrounding this head must be left for future research, what we have outlined so far is sufficient for our current purposes. With respect to the valency change, the most important idea that this proposal makes is that directional à, crucial to the licensing of IOs with verbs like aider, is entirely dependent on the visibility of the functional category in (20). We claim that the local cause of the valency change is the loss of this functional item near the end or at the beginning of the 16th century.

5.4 A cue-based approach to change

In this final section of our analysis, we adopt a cue-based approach to language change in order to examine why the indirect object of aider-type verbs was replaced by a direct object and why it happened when it did.

Recall that we adopt a generative grammar approach to language change of the kind argued for in Lightfoot (1979, 1991, 1999, 2006). It is assumed that structural changes in grammars occur when primary linguistic data shifts in some significant way from one generation to the next with the consequence that children draw a different conclusion about some aspect of their grammar than their parent’s generation did. We have proposed that the loss of a functional item encoding directionality underlies the change in argument realization whereby the indirect object of aider- type verbs was replaced by an analogous direct object. Within a cue-based approach to language change, it is nevertheless necessary to ask what specifically changed in the PLD of speakers born during the 16th and 17th centuries so that directionality was no longer perceived to be a derived property available to prepositions like à.

216

According to Lightfoot, children scan their linguistic environment for designated structures or “cues” which determine the architectural features of the developing grammar. When a child hears and understands an utterance, he or she has a mental representation of it; it is the acquisition of such abstract structures which determines the child’s grammar. Language change takes place when the expression of a particular unambiguous cue falls below some threshold of robustness so that it is no longer perceived and interpreted by the child as it was by the adult speaker. As a result, the mental representation of that particular cue and the set of utterances that it generates will not constitute a part of the learner’s grammar. With this mechanism of change in mind, we propose that learners of French assumed a directional functional head as long as the verb particle system was productive. When verb particles were no longer productively used by speakers born in the late 15th and 16th century, this head was no longer “visible” to subsequent generations.

5.4.1 Directional verb particles

Verb particles were used to orient and give additional direction to an event denoting change of location (22a), manner of motion (22b), ballistic motion (22c), and other activities (22d).

(22) a. Et cil vont arriere a la nef

[Et les autres regagnent le navire]

‘And the others go back to the ship’

(Partonopeu, 5901; Buridant 2000:540)

b. Il resaut sus encontremont

[Il fait un bond en l’air]

‘He jumps up in the air’

(Gormont, 279; Buridant 2000:542)

c. Gitez en donc ma part sus

[Jetez donc ma part en bas]

‘Throw my part down then’

(RenartR, IIIb, 4998; Buridant 2000:542)

217

d. Muntent en lors chevals, ariere se sunt mis

[Ils enfourchent leurs chevaux et ils prennent le chemin du retour]

‘They mount their horses and start back.’

(Becket, 2323; Buridant 2000:542)

In Buridant’s words:

[...] une particule séparée est un adverbe fondamentalement locatif jouant par rapport au verbe le rôle de vecteur sémantique en complétant son sémantisme, comme le font les particules séparables de l’allemand contemporain (hinausgehen « aller dehors » → « sortir ») ou les particules des phrasal verbs de l’anglais, appelés aussi verbes discontinus (to go out « aller dehors » → « sortir ») (Buridant 2000:540)

[... a separated verb particle is fundamentally a locative adverb having, with respect to the verb, the role of semantic vector, completing its meaning as do Contemporary German particles (hinausgehen “aller dehors” → “sortir”) or the particles of English phrasal verbs (to go out “aller dehors” → “sortir”)]

In order to appreciate the correlation between verb particles, directionality, and valency change, we begin with the assumption that the presence of a productive verb particle system in a language is an unambiguous cue indicating that a directional functional head is available to prepositions. The verb particle system depends on the possibility that locative prepositions can take on directional meaning, therefore generating a robust and salient cue that directionality is a derived property of prepositions.

According to Buridant, the particle system began to decline in the 15th century and quickly disappeared from texts written during the 16th century. According to our analysis, this salient structural change should have preceded and therefore created the necessary linguistic environment that triggered the decline in frequency of IOs first observed in the generation of authors born at the beginning of the 16th century. The correlation between the presence of a productive particle system and the robust use of IOs with aider-type verbs is substantiated by comparing the grammars of Christine de Pizan (b.1364) and Antoine de la Sale (b.1385), for example, with those of François Rabelais (b.1494) and Claude Fauchet (b.1530). The former use

218 verb particles very productively; particles such as arrière, avant, ens, sus, etc. are well attested with motion and activity verbs in our corpus. Some examples are given in (23). As expected, the frequency of IOs used with aider-type verbs is very high for these authors.

(23) a. Romulus arriere Regarde et .zzzxii. en a veüs

‘Romulus looks back and saw zzzxii of them’

(Pizan, Le Livre de la mutacion de fortune, 1400, t.3, p.180; BTMF)

b. et chascun arriere se retrait

‘and each one retreated back’

(De la Sale, Jehan de Saintré,1456, p.283; BTMF)

c. Quelquonques personne aille avant Est receüe sanz dongier

‘And whoever goes forward is received without threat’

(Pizan, Le Livre de la mutacion de fortune, 1400, t.1, p.136, BTMF)

d. quant ilz furent entrez ens, et il les vist porter ces choses

‘when they had all come in, and he saw them carrying these things’

(De la Sale, Jehan de Saintré, 1456, p.53, BTMF)

e. belle cousine, levez sus et venez veoir tant de si belles choses

‘lovely cousin, get up and come and see the many very beautiful things’

(De la Sale, Jehan de Saintré, 1456, p.158, BTMF)

A little more than a century later however, verb particles are rarely attested in works by Rabelais and Fauchet. For example, sus and ens are no longer used as such, while arrière and avant still occur as particles, but much less frequently and in limited contexts. Arrière survives in small measure with the verbs tourner and retourner, (24a), and avant is almost exclusively used with the verb entrer in the aspectual sense of ‘to continue’, (24b). Note that Fauchet still occasionally uses avant with motion verbs like marcher, shown in (24c).

219

(24) a. puis retournerent arriere, apres y avoir faict grand dommage

‘and they went back after having caused much damage there’

(Fauchet, Fleur de la maison de Charlemagne, 1601, p.198; Frantext,

ARTFL)

b. J’entre plus avant en ceste matiere

‘I still continue with this matter’

(Rabelais, Gargantua. 1534, p.112; Frantext, ARTFL)

c. À ceste cause il fit marcher son armée plus avant

‘and for this reason he had his army walk further on’

(Fauchet, Fleur de la maison de Charlemagne, 1601, p.162; Frantext,

ARTFL)

As anticipated, the regression of the proportion of IOs used with aider-type verbs is first apparent in the works of Rabelais, Fauchet and their contemporaries. This can be explained if one adopts the proposed analysis whereby the loss of a productive verb particle system created a linguistic environment in which new speakers of French acquired a grammar in which the functional category FdirP was not assumed. Directionality is then acquired as a lexical property of individual items. The preposition à would consequently be acquired as a purely locative element and indirect objects would no longer be licensed with aider-type verbs.

We have thus proposed a direct link between the loss of the verb particle system and the valency change in aider-type verbs:

220

(25) Medieval French: Pre-classical to Modern French:

particle system no evidence of particle system ↓ ↓ P can be directional/locative P has one interpretation ↓ ↓ directionality is derived directionality is lexical ↓ ↓ FdirP à is locative ↓ ↓ à can encode direction * aider à Jean ↓ ✓ aider à Jean

5.4.2 Preverbs

A number of phenomena conspire in Medieval French so that this stage of the language corresponds in many ways to Talmy’s (1985, 2000) description of a “satellite-framed” language. Talmy’s two-category typology concerns the way languages express events involving such things as motion and change of state. Modern French, like other Modern Romance languages, shows a preference for the lexicalization of path of motion in the verb and falls within the class of “verb-framed” languages. Example (26a) illustrates how the verb entrer expresses the core schema of the motion event, or the Path, while the gerund expresses the co-event, here, the manner.62 In contrast, “satellite-framed” Germanic languages express Path in a sister constituent to the verb, such as a prefix or a particle. The verb particle in in example (26b) from English expresses the Path, while the verb denotes the manner. Similarly, Early Medieval French expresses Path via a productive system of preverbs (27) and verb particles; see the examples in (22) and (23). Furthermore, examples (19a) and (19b) illustrate how Medieval French need not conform to the Modern French expression of manner of motion events in (26a), but can express Path via the preposition.

62 Kopecka (2006) argues that French is a hybrid system which has characteristics of both verb and satellite-framed languages. She proposes that directionality, or Path, is a lexical feature not only of verbs, but also of certain prefixes such as dé-, é-, re-, among others.

221

(26) a. La bouteille est entrée en flottant.

b. The bottle floated in.

(27) a. baisser ‘lower’ a-baisser ‘lower down’

b. porter ‘carry’ a-pporter ‘bring’

c. river ‘skirt the shore’ a-rriver ‘arrive at shore’

The presence of a functional category encoding direction is crucial for the verb particle system and for the grammaticality of sentences expressing manner of motion like that in (19b). It is also necessary for the preverb system. In Medieval French, the preposition à, for instance, functioned as a directional preverb in a productive process of verb prefixation. Directionality associated with à enhanced the meaning of the verbal base, giving it an additional sense of direction, goal, or passage from one state to another, as shown in (27). The most common effect of à-prefixation in Old French was to change the aspectual class of the verb from a durative activity, for instance, to an achievement. Without the sense of direction, à could not bring these kinds of meanings to the verbal base.

One might expect a correlation then, between the loss of the item associated with Fdir and the productive preverbs, but in fact, this is not at all what we see. In fact, what happened was quite the opposite: according to Tremblay et al. (2003), the loss of productive preverbs brought about the loss of the verb particle system which entailed the “invisibility” of the directional functional head.

The evolution of the preverb à- has been systematically studied in Dufresne and Dupuis (1998) and in Dufresne et al. (2000, 2001). Table 1, taken from Dufresne et al. (2001:37), shows the decrease over the centuries in new verbs formed by the addition of à-.

Table 1: Création des Vpréfixés dans l’évolution du français

e e e e e e e e e PÉRIODES 10 -13 s. 14 s. 15 s. 16 s. 17 s. 18 s. 19 s. 20 s.

# OCC. 312 24 18 12 1 3 2 2

Beginning with a baseline of 312 verbs used with the preverb à- in the 13th century, 24 new forms were created during the 14th century, 18 in the 15th century and 12 in the 16th century.

222

From the 17th century however, the process of à-prefixation is clearly no longer productive given that only eight new forms have since been created.

The loss of productivity is attributed to the semantic erosion of à (although other factors, such as the rise of the passé composé may have played a role). Dufresne et al. (2000, 2001) claim that to be a productive element in word formation, preverbs must have independent lexical content. As à lost its intensifying and inchoative value as a preverb, retaining only its directional value, it ceased to be an element used in the creation of new verb forms. It is important to note, however, as is pointed out in Dufresne et al. (2000, 2001) and in Tremblay et al. (2003) that the particular case of à- is part of a broader phenomenon involving a general loss of productivity in French verb prefixation from at least the 14th century onward.

Tremblay et al. (2003) build on this research to show that once the preverb à- is no longer used productively, its directional value was lexicalized by the verb itself. This entailed a sweeping change that not only affected morphologically complex verbs like apporter and aboutir, but motion verbs in general (entrer, partir, etc.). They argue that one of the consequences of this shift in the directional (or Path) feature was the rapid loss of verb particles during the 16th century.

Tremblay et al. (2003) argue that once directionality was lexicalized as a property of verb roots, the verb particle system became redundant and disappeared from use in the space of a century. This concurs with Buridant’s observation that “[le micro-system des particules séparées] voit son déclin s’amorcer dès le XVe siècle et s’amplifier au XVIe siècle pour ne plus subsister que sous forme de traces résiduelles dans le français littéraire du XVIIe siècle, non sans subsister plus largement dans les dialectes” (Buridant 2000:546) [The decline of the system of verb particles begins in the 15th century and accelerates during the 16th century to no longer exist but in the form of residual traces in the literary French of the 17th century, not without remaining to a larger extent in the dialects]. Directionality therefore remains an integral part of the verb phrase, but rather than being a property of preverbs and particles, it shifts to become a property of verb roots. The net result is a shift from a satellite-framed language to a principally verb-framed language.

According to the analysis presented in Tremblay et al. (2003), the following correlation can be made: the decline in the productivity of à-prefixation from the 15th century lead to the

223 lexicalization of directionality by many verbs, both complex and simple, which were previously unspecified for this property. Once direction became a general property of the verb, the particle system became redundant and quickly disappeared during the 16th century. A compelling correlation therefore exists among the indirect object of aider-type verbs, à-prefixation, and the verb particle system, which all show dramatic change during 16th and 17th centuries.

Finally, the analysis offered in Dufresne et al. (2001) suggests that the semantic erosion of à is at the source of this gradual decline in productivity. We propose, however, that the loss of productive preverbs may not be the result of gradual semantic changes in the preverbs/prepositions. Rather, it is likely one of the effects of the prosodic shift from words carrying their own individual accents (Old French) to a fixed oxytonic stress system (end of the Middle French period) in which accentuation of the phrasal group replaces that of the word. Words in Old French had initial as well as final stress: generally, the final syllable received primary stress and the initial syllable secondary stress. In this system, preverbs have a certain degree of prosodic salience, distinguishing them from the verb root. The strengthening of phrase- final stress and the cliticization of subject pronouns would gradually obscure the distinction between preverbs and their base. Eventually, the preverb’s semantic contribution of Path to the verb phrase would be reanalyzed as belonging to the verbal root itself. In so far as the preverb à was concerned, acquirers of French would have begun to perceive it less as a phonologically and semantically salient independent constituent and more as an element of individual verb roots.63

5.4.3 Verb restructuring

An important aspect of diachronic linguistics, especially within a minimalist framework, is to establish, when possible, correlations between a given change and other changes that are likely effects of the same underlying cause. Related and co-occurring effects strongly support our analysis, since a change in the abstract grammar is likely to produce more than one reflex in the utterances that are actually generated.

63 It would be interesting to map the diffusion of the change in order to identify the region at the forefront of the valency change. One could then check to see if the varieties of French spoken there also show early signs of the shift to an oxytonic stress system.

224

We would like to suggest that the loss of verb restructuring in French is a reflex of the loss of the functional item that encoded direction in Medieval French. A preliminary analysis is outlined here, but we leave the development of the analysis for further research.

Verb restructuring involves the merger of a verb having an incomplete thematic specification, such a modal or an aspectual, with a full-fledged lexical verb in the infinitive. This creates a verb complex which projects a common functional domain in which thematic arguments of the infinitive verb appear:

(28) car moult les amoit a veoir pres de lui

‘because he very much liked to see them by his side’

(Artu, 36, 89; Pearce, 1985:135; Martineau 2002:79)

Martineau (1990, 2002) shows that infinitive complements presenting restructuring effects in Medieval French are characterized by the subordinator à or Ø, but not by de. Moreover, the subordinator is always strictly adjacent to the infinitive. These properties distinguish them from other infinitive clause constructions. Furthermore, Martineau and Motapanyane (2000) propose that the infinitive clause in restructuring constructions is not a CP, but a reduced projection of VP that includes an aspectual phrase of which the morpheme à/Ø is the head. They argue that à/Ø is an affix, permitting the merger of the infinitive to its right and thus accounting for strict adjacency.

Restructuring disappeared during the 17th century (Martineau 1990). Since restructuring requires the infinitive to raise to the head of an aspect phrase, Martineau and Motapanyane (2000) and Martineau (2002) propose that the loss of this construction is linked to the decrease and eventual disappearance of general verb raising in infinitive clauses (CPs), given that both à and Ø also head CPs, along with de.

However, as pointed out in Martineau (2002), there is a potential problem with this analysis considering that the loss of restructuring actually precedes the decline of infinitive raising by a

225 century. As a possible explanation, she suggests that the general change involving infinitive raising would have begun more rapidly in structures involving an aspectual phrase.64

We suggest that a clearer correlation exists between the loss of restructuring constructions and the loss of the functional item associated with Fdir. Just as this category merges as part of the functional projection of a preposition, giving it an aspectual value, as Rooryck suggests, we propose that it may also merge as part of the functional projection of a verb. In this respect, it would be the very same FP proposed in Martineau and Motapanyane (2000). Only the null form of the head is realized with prepositions, but both the null and the à forms may occur with infinitives.

With the loss of the verb particle system, the functional head encoding direction was no longer assumed and the subordinators à/Ø would have simply been reanalyzed as a variant of the head of CP (as proposed in Martineau 2002). This alternative account does not therefore make a correlation between the sudden loss of restructuring and the gradual decline of infinitive raising, but rather between restructuring and the loss of the verb particle system. The latter correlation provides, at the very least, a better account of the chronology of the various changes.

Italian may appear to present a potential counter-example to the correlation that we have proposed between a directional FP, datives with aider-type verbs and verb restructuring. The Italian equivalent of aider, aiutare, has an accusative argument while Italian retains clitic- climbing and infinitive movement. French restructuring, however, differs in some very significant ways from the Italian, suggesting two completely different analyses.65 Crucially, Italian does not involve raising of the infinitive to a directional FP; see also Roberts (1997).

64 Hirschbühler and Labelle (1994) explain this change in terms of constraints on the establishment of Tense chains in the syntax.

65 Martineau and Motapanyane (2000) point out that in Italian, clitics and adverbs can intervene between the subordinator and the infinitive while French restructuring requires strict adjacency between these two elements. Moreover, Ian Roberts (p.c.) has remarked that there seems to be a wider set of verbs involved in the French restructuring than that found in Italian. Aimer, cited in example (28), for instance, is not a restructuring verb in Italian.

226

While we posit a connection between datives with helping verbs and restructuring in French, this does not appear to be the case for Italian.

5.5 The persistence of lui aider and lui prier

Finally, how do we account for uses of aider with a dative object, attested, for example, in Canadian French, Swiss French and Franco-Provençal?

(29) a. - Elle m’a dit qu’elle ne savait pas ce qu’il fallait dire.

- Je vais lui aider.

(Radio Suisse romande, 1.4. 1976; Lüdi 1983:87)

b. Votre aide financière pour un rider peut lui aider à changer sa vie...

‘Your financial aide for a rider can help him change his life...’

(“Ride à l’os”, cited June 4, 2008. Available at :

http://www.ridealos.com/ecole_rodeo_canada.html)

Similarly, prier can be used with a dative object in informal speech. Here, the dative tends to be limited to clear contexts of verbal communication in which prier means ‘to ask’.

(30) a. Marie, en s’adressant à sa mère, lui prie de parler au jeune homme pour

qu’il lui fiche la paix.

‘Turning to her mother, Marie begs her to speak to the young man so that

he leave her alone.’

b. Je leur ai prié de bien vouloir m'excuser.

‘I begged them to please excuse me.’

In line with Lüdi’s (1983) analysis of aider in Swiss French, we consider these forms to be vestiges of an earlier grammar. Clearly, uses of the dative with aider and prier do not point to microgrammars in which the preposition à can derive a directional interpretation. If this were the case, we would expect to see structures related to the presence of a functional category encoding direction. At the very least, we would expect datives used with other verbs like applaudir,

227 commander, servir, persuader, etc. We would also predict manner of motion events to be expressed like that in (19) and sentences similar to (17b,c) to be acceptable. Even verb restructuring should, in principle, be possible. We have found no evidence of any of this in our formal and informal examination of the facts. Finally, we would expect the dative to be used in both its nominal and clitic form, yet the clitic is by far the favoured form of the dative with aider and the only acceptable form with prier.

That the dative objects of aider and prier persist in some varieties of French seems to suggest that the semantic role that these particular verbs attribute to their complement is highly compatible with the most common goal-like interpretations of the dative clitic. It would not be correct to say, for instance, that the dative has survived by virtue of analogy with current dative constructions. If this were the case, then we would expect full nominal forms to be acceptable in the dative as well. What appears to be happening, is that archaic uses of the verb persist because conceptually, the dative is the best fit for the object and because dative clitics can continue to appear where dative nominals cannot. As mentioned in section 3.3.5, dative clitics are simply the spell-out of a functional head and are realized independently of the licensing requirements on their nominal counterparts. In this way, they escape the restrictions imposed on dative noun phrases and can continue to occur with adirectional verbs.

5.6 Chapter summary

In this chapter, it was established that directionality is necessary for the licensing of first and third order IOs in non relational contexts. Directionality is available via the lexical semantics of the verb in Modern French, but was also available via the preposition à in Medieval French, explaining the broader range of verbs with which these types of IOs could occur.

Based on variable locative/directional interpretations of prepositions in Medieval French, we proposed that directionality was derived via a functional head dominating the preposition. The loss of this functional item underlies the replacement of the IO of aider-type verbs by an analogous DO. By proposing an underlying structural change, our analysis accounts for both the selective nature of the valency change and the strong correlation between it and other structural changes in French, the effects of which are also observed during the 16th and 17th century.

228

A cue-based approach to language change allowed us to identify the loss of the verb particle system as the crucial change in the PLD which triggered the subsequent change in argument realization of aider-type verbs. Our analysis accounts for what caused the change, and how and why the change happened when it did.

229

Concluding Remarks

Diachronic generative syntax has concerned itself relatively little with argument structure, tending to focus instead on word-order change, grammaticalization and category changes. This dissertation shows that changes in argument structure can raise new questions for diachronic syntax by demonstrating that variation and change in the realization of grammatical functions do not have to be accompanied by changes in meaning and that they can be structurally motivated. Contrary to traditional views, changes in argument realization are not necessarily random and paths of change can be shown to be structurally determined. French alone presents other systematic changes in argument structure realization that are yet to be examined quantitatively and in light of recent theories. Among them, the loss of the double object construction and its possible correlates. Systematic changes in argument structure realization can offer a fruitful exploration of grammatical dependencies and can reveal the underlying property or properties upon which typologies are based.

One of the most exciting aspects of this study was to observe the striking regularity of the valency change, and to what extent this particular change serves to inform our understanding of the system of indirect objects in French. Its contingency on argument structure reveals a sharp distinction between structural and inherent indirect objects, but no structural difference between datives on the one hand, and neuter/locative indirect objects on the other. The change therefore supports a unified treatment of IOs as prepositional phrases, but does not support a unified analysis of the way datives or IOs more generally are licensed. Finer grained differences between dynamic two-place indirect verbs that underwent the change and those that did not allow us to identify licensing constraints on indirect objects involving the degree of abstraction of the object and lexical directionality.

The study as it stands points to a number of avenues for further research. The question of a functional item that encodes Path and its effects elsewhere in the grammatical system, not only in

French, but in other languages as well, demands closer examination. If FdirP turns out to be the parameter that underlies the satellite vs. verb-framed typological distinction exemplified by Germanic vs. Romance languages, then the correlation between satellite-framed languages and the availability of resultatives like John broke the vase open would have to be accounted for. This connection does not seem immediately obvious.

230

The link that was suggested between a directional functional item and verb restructuring in Medieval French also deserves more attention. French verb restructuring clearly differs from what is observed in Italian, for example, but to understand it better, similar systems need to be found elsewhere.

Finally, we have not compared the valency change in French to similar phenomena in other Romance languages. Such a comparison would offer a richer understanding of both the cue that triggered the change and the structural consequences of the cross-linguistic semantic and functional differences in the preposition à. The diffusion of the valency change, both temporally and geographically, was likewise not considered in the study. A closer look at the process of the change may permit a better understanding of the possible connection between it and the evolution of French toward a fixed oxytonic stress system.

231

References

Scientific works

Adams, Marianne Patalino. 1987. Old French, null subjects, and verb second phenomena. Doctoral dissertation, University of California, Los Angeles.

Alsina, Alex. 1996. The Role of argument structure in grammar: Evidence from Romance. Stanford California: CSLI.

Anagnostopoulou, Elena. 2005. Cross-linguistic and cross-categorial variation of datives. In Advances in Greek Generative Syntax, ed. M. Stavrou, and A. Terzi, 61–126. Amsterdam: John Benjamins.

Andrews, Avery D. 1995. The major functions of the noun phrase. In Lanuage typology and syntactic description I: Clause structure, ed. Timothy Shopen, 62-154. Cambridge: Cambridge University Press.

Arad, Maya. 1998. VP structure and the syntax-lexicon interface. Doctoral dissertation, UCL.

Baker, Mark. 1988. Incorporation. Chicago: The University of Chicago Press.

Barnes, Betsy. 1985. A Functional explanation of French nonlexical datives. Studies in Language 9(2):159-195.

Belletti, Adriana, and Luigi Rizzi. 1988. Psych-verbs and theta-theory. Natural Language and Linguistic Theory 6:291-352.

Blake, Barry J. 2001. Case. 2nd ed. Cambridge: Cambridge University Press.

Blinkenberg, Andreas. 1960. Le Problème de la transitivité en français moderne: essai syntactico-sématique. Copenhagen: Munksgaard.

Blume, Kerstin. 1998. A contrastive analysis of interaction verbs with dative complements. Linguistics 36(2):253-280.

Borer, Hagit. 2005. In Name only. Structuring sense. Vol.1. Oxford: Oxford University Press.

Burzio, Luigi. 1986. Italian syntax: A Government-Binding approach. D. Dordrecht: Reidel.

Chomsky, Noam. 1957. Syntactic structures. The Hague: Mouton & Co.

232

Chomsky, Noam. 1986. Knowledge of language: Its nature, origins, and use. New York: Praeger.

Chomsky, Noam. 1995. The Minimalist program. Cambridge, MA: MIT Press.

Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Step by step: Essays in honor of Howard Lasnik, ed. R. Martin, D. Michael, and J. Uriagereka, 89-155. Cambridge, MA: MIT Press.

Chomsky, Noam. 2001. Derivation by phase. In Ken Hale: A life in language, ed. M. Kenstowicz, 1-52. Cambridge, MA: MIT Press.

Clark, H. H. 1973. Space, time, semantics, and the child. In Cognitive development and the acquisition of language, ed. T.E. Moore. New York: Academic Press.

Colin, T. F. 1897. Notes on the syntax of the French verb in writers of the seventeenth century. Modern Language Notes 12(3):140-156.

Comorovski, Ileana. 1990. Verb movement and object extraction. In Proceedings of NELS XX, ed. J. Carter, R.-M. Dechaine, B. Philip, and T. Sherer, 91-105. Amherst: University of Massachusetts Graduate Linguistics Students Association.

Croft, William. 2000. Explaining language change: An evolutionary approach. London: Longman.

Cuervo, Maria Cristina. 2003. Datives at large. Doctoral dissertation, MIT, Cambridge, MA.

Cummins, Sarah, and Yves Roberge. 2005. A Modular account of null objects in French. Syntax 8(1):44–64.

Cummins, Sarah, Yves Roberge and Michelle Troberg. 2008. Les objets indirects en français: sens, représentations et évolution. Paper delivered at Les Français d’ici, May 22-25, University of Ottawa.

Delbecque, Nicole. 1998. Why Spanish has two construction frames. Leuvense Bijdragen 87: 387-415.

Denison, David. 2003. Log(ist)ic and simplistic S-curves. In Motives for language change, ed. R. Hickey, 54-70. Cambridge: Cambridge University Press.

233

Dufresne, Monique, Fernande Dupuis, and Mireille Tremblay. 2000. The role of features in historical change. In New approaches to old problems, ed. S. N. Dworkin and D. Wanner, 129-148. Amsterdam: John Benjamins.

Dufresne, Monique, Fernande Dupuis, and Catherine-Marie Longtin. 2001. Un changement dans la diachronie du français: la perte de la préfixation aspectuelle en –à. Revue québécoise de linguistique 29(2): 33-54.

Eythórsson, Thórhallur. 2002. Changes in subject case-marking in Icelandic. In Syntactic effects of morphological change, ed. David Lightfoot, 196-212. Oxford: Oxford University Press.

García, Erica C. 1975. The role of theory in linguistic analysis: The Spanish pronoun system.

Guéron, Jacqueline. 1985. Inalienable possesion. PRO-inclusion and lexical chains. In Grammatical representations, ed. H.-G. Obehauer, Jean-Yves Pollock, and Jacqueline Guéron, 43-86. Dordrecht, MA: MIT Press.

Goyens, Michèle. 1998. L’alternance entre construction accusative et dative dans l’histoire des verbes français. Leuvense Bijdragen 87:465-490.

Goyens, Michèle, Béatrice Lamiroy, and Ludo Melis. 2002. Déplacement et repositionnement de la préposition à en français. Lingvisticae Investigationes 25(2):275-310.

Hale, Ken, and Samuel J. Keyser. 1993. On argument structure and the lexical expression of syntactic relations. In The View from building 20, ed. K. Hale, and S. J. Keyser, 53-109. Cambridge, MA: MIT Press.

Hale, Ken, and Samuel J. Keyser. 2002. Prolegomenon to a theory of argument structure. Cambridge, MA: MIT Press.

Hale, Mark. 1998. Diachronic syntax. Syntax 1(1):1-18.

Harley, Heidi. 1995. Subjects, events and licensing. Doctoral dissertation. MIT, Cambridge, MA.

Harley, Heidi. 2002a. Possession and the double object construction. In Yearbook of Linguistic Variation, Vol.2, ed. Pierre Pica, and Johan Rooryck, 31-70. John Benjamins.

Harley, Heidi. 2002b. A minimal(ish) linking theory. Paper delivered at the Maryland Minimalist Mayfest, Maryland.

234

Harley, Heidi. 2004. Merge, conflation, and head movement: The First Sister Principle revisited. In Proceedings of the 34th Annual Meeting of the North East Linguistics Society, Vol.1, ed. Keir Moulton, and Matthew Wolf. University of Massachusetts, Amherst : GSLA Publications.

Harley, Heidi, and Raffaella Folli. 2005. Consuming results in Italian and English: flavors of v. In Aspectual Enquiries, ed. Paul Kempchinsky, and Roumyana Slabakova, pp. 95-120. Dordrecht: Springer.

Herslund, Michael. 1980. Problèmes de syntaxe de l’ancien français: compléments datifs et génitifs. Etudes Romanes 21. Copenhagen: Akademisk Forlag.

Herslund, Michael. 1988. Le Datif en français. Louvain: Peeters.

Hirschbühler, Paul and Marie Labelle. 1994. Changes in verb position in French negative infinitival clauses. Language Variation and Change 6:149-178.

Hoekstra, Teun. 1995. To have to be dative. In Studies in comparative Germanic syntax, ed. H. Haider, S. Olsen, and S. Vikner, 119-137. Dordrecht: Kluwer Academic Publishers.

Hopper, Paul J. and Sandra A. Thompson. 1980. Transitivity in grammar and discourse. Language 56:251-295.

Jaeggli, Osvaldo. 1982. Topics in Romance syntax. Dordrecht: Foris.

Jones, M. A. 1996. Foundations of French syntax. Cambridge: CUP.

Kayne, Richard. 1975. French syntax: The transformation cycle. Cambridge, MA: MIT Press.

Kayne, Richard. 1984. Connectedness and binary branching. Dordrecht: Foris.

Kilroe, Patricia. 1989. The grammaticalization of French à. Doctoral dissertation, University of Texas, Austin.

Kopecka, Anetta. 2006. The semantic structure of motion verbs in French : Typological perspectives. In Space in languages : Linguistic systems and cognitive categories, ed. M. Hickmann, and Stéphane Robert, 83-101. Amsterdam/Philadelphia: John Benjamins.

Kratzer, Angelika. 1996. Severing the external argument from its verb. In Phrase Structure and the Lexicon, ed. J. Rooryck, and L. Zaring, 109-138. Dordrecht: Kluwer.

Kress, Bruno. 1982. Isländische Grammatik. Leipzig: Verlag Enzyklopädie.

235

Kroch, Anthony. 1989. Reflexes of grammar in patterns of language change. Language Variation and Change 1:199-244.

Labelle, Marie. 1992. Change of state and valency. Journal of Linguistics 28(2):375-414.

Larson, Richard. 1988. On the double object construction. Linguistic Inquiry 19(3):335-391.

Leclère, C. 1978. Datifs syntaxiques et datif éthique. In Méthodes en grammaire française, ed. J.-C. Chevalier, and M. Gross, 66-75. Paris: Klincksieck.

Levin, Beth. 1999. Objecthood: An event structure perspective. In Chicago Linguistic Society 35, Part 1: Papers from the Main Session, 223-47.

Levin, Beth, and Malka Rappaport Hovav. 2005. Argument realization. Cambridge: Cambridge University Press.

Lightfoot, David. 1979. Principles of diachronic syntax. Cambridge: Cambridge University Press.

Lightfoot, David. 1991. How to set parameters: Arguments from language change. Cambridge, MA.: MIT Press.

Lightfoot, David. 1999. The Development of language: Acquisition, change, and evolution. Oxford: Blackwell.

Lightfoot, David. 2006. How new languages emerge. Cambridge: Cambridge University Press.

Longobardi, Guiseppe. 2001. Formal syntax, diachronic minimalism, and etymology: the history of French chez. Linguistic Inquiry 32:275-302.

Ludi, Georges. 1978. Die Alternanz zwischen Dativ und Akkusativ bei prier, supplier, requérir im 15. und 16. Jahrhundert. Vox Romanica 97:161-192.

Ludi, Georges. 1983. Remarques sur la construction aider à qn en français romand. In Dialectologie, histoire, et folklore: mélanges offertes à Ernest Shüle pour son 70e anniversaire, 85-97. Berne: Editions A. Francke.

Lyons, John. 1977. Semantics. London: Cambridge University Press.

Maling, Joan. 2001. Dative : The heterogeneity of the mapping among morphological case, grammatical functions, and thematic roles. Lingua 111:419-464.

236

Maling, Joan. 2002. Icelandic verbs with dative objects. Working Papers in Scandinavian Syntax

70:1-60.

Marantz, Alec. 1984. On the Nature of Grammatical Relations. Cambridge, MA: MIT Press.

Marantz, Alec. 1993. Implications of asymmetries in double object constructions. In Theoretical Aspects of Bantu Grammar, ed. S. Mchombo, 113-150. Stanford, CA: CSLI Publications.

Marantz, Alec. 1997. No escape from syntax: don’t try morphological analysis in the privacy of your own lexicon. In Penn Working Papers in Linguistics 4.2, Proceedings of the 21st Annual Penn Linguistics Colloquium, ed. A. Dimitriadis, L. Siegel, C. Surek-Clark, and A. Williams, 210-225. Pennsylvania: University of Pennsylvania Working Papers in Linguistics.

Marchello-Nizia, C. 2003. Changes in the structure of grammatical systems : The evolution of French. In Aspects of Linguistic Change, ed. A. Lodge, 371-385. Oxford : Oxford University Press.

Martineau, France. 1990. La montée du clitique en moyen français: une étude de la syntaxe des constructions infinitives. Doctoral dissertation, University of Ottawa.

Martineau, France. 2002. Grammaticalisation et changement de paramètre: la préposition à, élément introducteur d'un complément infinitif. Scolia 16:70-90.

Martineau, France, and Virginia Motapanyane. 2000. Infinitive subordinators and verb restructuring in French. In New approaches to old problems, ed. Dworkin, Steven N., and Dieter Wanner, 217-232. Amsterdam: John Benjamins.

Mathieu, Eric. 2006. Impersonal constructions in Old French and the agreement puzzle. In Proceedings of the 2006 Canadian Linguistic Association Annual Conference, eds. Claire Gurski and Milica Radisic. Available at: http://ling.uwo.ca/publications/CLA2006/Mathieu.pdf.

McFadden, Thomas. 2003. Underlying structures of German inherent Case. Write-up of talk delivered at the joint meeting of the FGLS and the SGL, London. Available at : http://ifla.uni-stuttgart.de/institut/mitarbeiter/tom/downloads/fglspap.pdf

237

McFadden, 2006. German inherent datives and argument structure. In Datives and other cases: Between argument structure and event structure, ed. Hole, Daniel, André Meinunger, and Werner Abraham, 49-77. Amsterdam: John Benjamins.

Melis, Ludo. 1994. Paradigmes de la valence verbale et réalisations nominales et pronominales. International Journal of Lexicography 7(2):142-157.

Melis, Ludo. 1996. The dative in Modern French. In The Dative. Vol.1. Ed. W. Van Belle, and W. Van Langendonck, 39-73. Amsterdam: John Benjamins.

Melis, Ludo. 2004. Cas, fonctions et rôles – des composantes nécessaires pour décrire les constructions verbales? Une enquête préliminaire des constructions verbales en à en français. In Semantische Rollen, ed. R. Kailuweit, and M. Humme, 168-183. Tübingen: Niemeyer.

Miller, Philip. 1992. Clitics and constituents in Phrase Structure Grammar. London: Garland.

Nash, Lea. 2002. Entre la flexion et le verbe: syntaxe, morphologie, acquisition. Document de synthèse pour l’habilitation, Université de Paris 7.

Perlmutter, David. 1972. Deep and surface structure constraints in syntax. New York: Holt, Rinehart and Winston.

Perlmutter, David M. 1978. Impersonal passives and the Unaccusative Hypothesis. Proceedings of the 4th Annual Meeting of the Berkeley Linguistics Society, 157–189. UC Berkeley.

Pesetsky, David. 1995. Zero syntax. Cambridge, MA: MIT Press.

Pinker, Steven. 1989. Learnability and cognition: The Acquisition of argument structure. Cambridge, MA: MIT Press.

Pylkkänen, Liina. 2002. Introducing arguments. Doctoral dissertation, MIT, Cambridge, MA.

Rappaport-Hovav, Malka, and Beth Levin. 1998. Building verb meanings. In The Projection of arguments: Lexical and compositional factors, ed. M. Butt, and W. Geuder, 97-134. Stanford, CA: CSLI Publications.

Roberge, Yves. 2002. Transitivity requirement effects and the EPP. Paper read at WECOL 2002, University of British Columbia, Vancouver.

238

Roberge, Yves, and Michelle Troberg. 2007a. Thematic indirect objects in French. Journal of French Language Studies 17(3):297-321.

Roberge, Yves, and Michelle Troberg. 2007b. Les objets indirects non thématiques en français. In Proceedings of the 2007 Canadian Linguistics Association Annual Conference, ed. Milica Rasidic, 14pp. Available at: http://www.chass.utoronto.ca/~cla- acl/actes2007/Roberge_Troberg.pdf

Roberts, Ian. 2007. Restructuring, head movement and locality. Linguistic Inquiry 28(3):423- 460.

Roberts, Ian, and Anna Roussou. 2003. Syntactic change: A Minimalist approach to grammaticalization. Cambridge: Cambridge University Press.

Rooryck, Johan. 1996. Prepositions and minimalist Case-marking. In Studies in Comparative Germanic Syntax, Vol.2, ed. Höskuldur Thráinsson, Samuel Epstein, and Steve Peter, 226-256. Dordrecht: Kluwer.

Ruwet, Nicolas. 1982. Grammaire des insultes et autres études. Paris: Seuil.

Seelbach, Dieter. 1970. Präpositionalphrasen mit de und à als Quellen der verbundenen Pronomen en und y. Inaugural Dissertation, University of Frankfurt.

Schøsler, Lene. 2003. Le rôle de la valence pour une classification sémantique des verbes. In La cognition dans le temps: études cognitives dans le champ historique des langues et des textes, ed. P. Blumenthal, and J.-E. Tyvaert, 145-159. Tübingen: Niemeyer.

Schøsler, Lene. To appear. Les verbes exprimant la notion d’aide. In Le français, d’un continent à l’autre, ed. Luc Baronian et France Martineau. PUL.

Strozer, Judith R. 1976. Clitics in Spanish. Doctoral Dissertation, UCLA.

Svenonius, Peter. (June 2007). Spatial P in English. Ms. Accessible at http://ling.auf.net/lingBuzz/000001.

Talmy, Leonard. 1985. Lexicalization patterns: semantic structure in lexical forms. In Language typology and syntactic description: Grammatical categories and the lexicon. Vol.3. Ed. T. Shopen, pp. 57–149. Cambridge: Cambridge University Press.

239

Talmy, Leonard. 2000. Toward a cognitive semantics: Concept structuring systems. Vol.1. Cambridge, MA: MIT Press.

Tenny, Carol. 1994. Aspectual roles and the syntax-semantics interface. Dordrecht: Kluwer.

Tremblay, Mireille. 1999. Du Statut des prépositions dans la grammaire. Revue Québécoise de Linguistique 27(2):167-183.

Tremblay, Mireille, Fernande Dupuis, and Monique Dufresne. 2003. Les prépositions dans l’histoire du français: transitivité, grammaticalisation et lexicalisation. Verbum 25(4):549- 562.

Tsunoda, Tasaku. 1985. Remarks on transitivity. Journal of Linguistics 21:385-396.

Vandeloise, Claude. 1987. La préposition à et le principe d’anticipation. Langue française 76:77- 111.

Vandeloise, Claude. 1991. Spatial prepositions: A case study from French. Chicago: The University of Chicago Press. van der Zee, E., and J. Slack. 2003. Representing direction in language and space. Oxford: Oxford University Press.

Van Peteghem, M. 2006. Le Datif en français: un cas structural. Journal of French Language Studies 16:93-110.

Vergnaud, J.R. 1974. French relative clauses. Doctoral dissertation, MIT, Cambridge, MA.

Wegener, H. 1991. Der Dativ – ein strucktureller Kasus? In Merkmale und Strukturen syntactisher Kategorien, ed. G. Fanselow and S. Felix, 70-103. Tûbingen: GNV.

Zaring, Laurie. 1991. On prepositions and case-marking in French. Canadian Journal of Linguistics 36(4):363-377.

Zink, Gaston. 1997. Morphosyntaxe du pronom personnel (non réfléchi) en moyen français (XIVe-XVe siècles). Genève: Droz.

240

Corpus Electronic text-based corpora The ARTFL Project, The University of Chicago. “Frantext”, “Frantext Encore”, “French Women Writers”. Available at: http://www.lib.uchicago.edu/efts/ARTFL/newhome/texts/

ATILF/ Équipe "Moyen français et français préclassique", 2003-2005, Base Textuelle du Moyen Français (BTMF). Available at: http://www.atilf.fr/dmf/index_textes.htm

Cornucopia d’Epistemon, ed. Marie-Luce Demone, Tours. Available at: http://www.etudes- francaises.net/nefbase/epistemon_tact.htm

Corpus Lemieux (UQAM, created by M. Lemieux, CRSH 410-89-1409 ; 410-93-1409) 16e siècle.

Rentexte. Textes de la Renaissance, ed. R. Woolridge, 1998. http://translatio.ens.fr/rw/rentexte/ (no longer available).

Université Rennes 2, CETM. On-line texts, Théâtre médiéval, Available at: http://www.uhb.fr/alc/medieval/#recherche

Metalinguistic references The ARTFL Project, The University of Chicago “Dictionnaires d'autrefois” Available at: http://www.lib.uchicago.edu/efts/ARTFL/projects/dicos/

ATILF/Équipe "Moyen français et français préclassique", 2004, Base de Lexiques de Français Préclassique (BFPC). Available at: http://atilf.atilf.fr/gsouvay/scripts/g2/showps.exe?host=interface_preclassique.txt

ATILF/Équipe "Moyen français et français préclassique", 2003-2005, Base de Lexiques de Moyen Français (DMF1). Available at: http://www.atilf.fr/dmf/

ATILF/Équipe. Trésor de la Langue Française Informatisé (TLFi). Available at: http://atilf.atilf.fr/tlf.htm

Brunot, Ferdinand. 1966. Histoire de la langue française: des origines à 1900. Paris: Armand Colin. [1905-1953]

Brunot, Ferdinand, and Charles Bruneau. 1969. Précis de grammaire historique de la langue française. 3rd ed. Paris: Masson et Cie.

241

Buridant, Claude. 2000. Grammaire nouvelle de l’ancien français. Paris: Sedes.

Clédat, Léon. 1887. Grammaire élémentaire de la vieille langue française. Paris : Garnier Frères.

Collins Robert French-English English-French Dictionary. 2002. Paris: Dictionnaires Le Robert.

Commentaires sur les Remarques de Vaugelas. 1936. Paris: Libraire E. Droz.

Dictionnaire de l’Académie Française. 1694; 1762; 1798; 1835. In “Dictionnaires d'autrefois”, The ARTFL Project.

Dictionnaire historique de la langue française. 1998. Paris, Dictionnaires Le Robert.

Druide informatique, Antidote-Dictionnaire v.5., Montréal, PQ, Canada.

Étienne, Eugène. 1980. Essai de grammaire de l’ancien français (IXe-XIVe siècles). Geneva: Slatkine Reprints. [1895]

Féraud, Jean-François. 1787-1788. Dictionaire critique de la langue française. Marseille: Mossy. In “Dictionnaires d'autrefois”, The ARTFL Project.

Fournier, Natalie. 1998. Grammaire du Français classique. Paris: Belin.

Furetière, Antoine (1694). Le Dictionnaire Universel. Paris, Dictionnaires le Robert, 1978.

Godefroy, Frédéric.1969. Dictionnaire de l'ancienne langue française et de tous ses dialectes du IXe au XVe siècle. Kraus Reprint. [1880-1902]

Gougenheim, Georges. 1974. Grammaire de la langue française du seizième siècle. Nouv. éd. entièrement refondue. Paris: A. & J. Picard.

Le Grand Larousse de la langue française, 1976, Paris: Librairie Larousse.

Greimas, Algirdas, and Teresa May Keane. 1992. Dictionnaire du Moyen Français. La Renaissance. Paris: Larousse.

Grevisse, Maurice. 1961. Problèmes de Langage, vol.4. Duclot: Gembloux.

Haase, Albert. 1914. Syntaxe Française du XVIIe siècle, 2nd ed. Paris: Librairie Delagrave.

Huguet, E. 1925-1967. Dictionnaire de la langue française du seizième siècle. Paris: Didier.

Littré, Émile. Dictionnaire de la langue française. In “Dictionnaires d'autrefois”, The ARTFL Project.

242

Martin, Robert, and Marc Wilmet. 1980. Manuel du Français du Moyen Âge. vol.2. Syntaxe du Moyen Français. Bordeaux: Sobodi.

Ménard, Philippe. 1976. Manuel du Français du Moyen Âge, vol.1. Bordeau: Sobodi.

Nicot, Jean. 1606. Thresor de la langue françoyse, tant ancienne que moderne. In “Dictionnaires d'autrefois”, The ARTLF Project.

Nyrop, Kristopher. 1899-1930. Grammaire Historique de la Langue Française, vol.4. Copenhagen: GBNF.

Le Petit Robert, 2006, Paris: Dictionnaires Le Robert.

Raynaud de Lage, Guy. 1993. Introduction à l’ancien français 2nd ed., G. Hasenohr. Paris: Sedes.

Vaugelas, C. F. de. 1981. Remarques sur la langue française. Paris: Éditions Champ Libre. [1647]