1 NULL OBJECTS IN FRENCH AND ENGLISH * SARAH CUMMINS & YVES ROBERGE Université Laval University of Toronto 1. Introduction An important subject-object asymmetry in generative grammar has been the obligatory projection of a subject position (by the EPP or a feature of the inflectional layer of the clause) but not of an object position. Projection of an object position was considered to depend on lexical characteristics of the verb. However, languages seem to allow a wide range of possibilities for conventionally intransitive verbs to appear with a direct object (as illustrated for French and English in (1)), and for conventionally transitive verbs to appear without a phonologically realized direct object (2). 1 (1) a. Elle précisa qu'elle le mangerait «tout complètement», feula des baisers à blanc et raccrocha. (L:110) She added that she would eat him all up, growled air kisses, and hung up. b. Si Mike commence à bafouiller ses tirs, la sauce commence à prendre avec ses partenaires. (L:113) While Mike is beginning to splutter his shots, things are coming together for his teammates. c. Just how far the argument has come since Archie bellowed his brand of bigotry is evident in the first episode of 704 Hauser Street. (ECP) d. Two young German women wept tears of shame for their country as the car left. (ECP) (2) a. La lune, si t'y mets une porte et tu regardes la nuit, tu peux être fier de ton boulot.(gourio:153) If you put a door on the moon and you watch at night, you can be proud of your work. * We would like to thank Denis Bouchard, Diane Massam, Philippe Prévost, Michelle Troberg three anonymous reviewers and the members of the Asymmetry Project. This work is supported by a grant from the Social Sciences and Humanities Research Council of Canada (Di Sciullo ). 1 Abbreviations: L: Larjavaara (2000); GP: García Velasco & Portero-Muñoz (2002); ECP: English Canadian Press (newspaper corpus); PCF: Presse canadienne française (newspaper corpus); BNC: British National Corpus (general corpus).
2 2 b. C'est pas lui qui l'a écrit, son livre, le pape, c'est quelqu'un qui lui écrit... (Gourio:153) The Pope didn't write his book himself, someone writes for him. c. Why then do the psychic gifts often seem to tease, confuse and obstruct? (BNC) d. This is a rhetorical platitude that presents the posture of a freedom fighter, when really it's the same old argument: Don't bite the hand that feeds. (Toronto Globe & Mail) These possibilities cannot be attributed solely to lexical properties of the verb; if this were the case, certain verbs would always be able to appear without their objects regardless of the construction or discourse context, and others would never be able to appear without an object. As we will show, this is not the case. Rather, following Roberge (2003), we propose that null or implicit objects can be attributed to a Transitivity Requirement (TR) just as null subjects are ultimately due to the EPP. Recoverability for the EPP is morphologically based, as is evident in null subject languages, while recoverability involving the TR may also be semantically and pragmatically based; as we will show below, such recovery may be based on information derived from the verb's lexical semantics and Generalised Conversational Implicatures (formalised as in Levinson 2000) involved in the interpretation of reduced nominal forms. The factors that contribute to licensing superficial intransitivity the absence of an overt object may include lexical semantics, functional elements, discourse factors, and trans-clausal structural elements. This view is supported by a comparative study of null object possibilities in French and English. 2. On transitivity The concept of transitivity has been interpreted as a continuum in certain works, and a distinction has been proposed between syntactic transitivity and semantic transitivity; see, among many others, Blinkenberg (1960), Desclés (1998), Hopper et Thompson (1980), Lazard (1994). Surprisingly little is ever said about the object position itself. The hypothesis in Roberge (2003) is that there exists a Transitivity Requirement (TR), whereby an object position is always included in VP, independently of the lexical choice of V. The empirical motivation of this hypothesis is the well documented evidence (see in particular Blinkenberg (1960), Larjavaara (2000)) that any transitive verb has the potential to appear without a direct object and any unergative verb has the potential to appear with a direct object. To account for these facts, there must be a mechanism to generate the direct-object position, either optionally or obligatorily. The TR represents the second, more restrictive, possibility and conveys the concept of transitivity as a property of the predicate (the VP), rather than as a property of the lexical content of V. The TR is the internal-argument counterpart to the EPP. In other words, the configuration in (3) order irrelevant is given by UG:
3 3 (3) V 2 V Obj The TR is similar to the EPP in that (1) it targets a position and not necessarily the nature of the element occupying this position; (2) the end result varies depending on lexical choice and the merger and movement operations involved in the derivation. It differs from the EPP in that (1) it does not target a Spec position; (2) it is active in the thematic layer of the clause rather than the inflectional layer. For the purpose of our discussion, we define unexpressed objects interpretatively: there is an x such that x is (1) phonologically null, (2) involved in the event denoted by the VP, and (3) not an external argument Towards a typology of null objects Two recent studies Larjavaara (2000) on French and García Velasco & Portero Muñoz (2002) on English address the issue of null objects (NOs) comprehensively, while taking account of previous work on this topic. The findings of these two studies show clear similarities between the two languages. Both studies distinguish two types of objects: GP call the two types indefinite and definite null objects, while L refers to generic and latent null objects. Examples of the two types are illustrated in (4) and (5): (4) indefinite/generic a. Do you write? (GP) b. Wild Guns est un jeu qui défoule. (L:88) Wild Guns is a game that destresses. (5) definite/latent a. Do you like? I love! (GP) b. «Tu as lu les pages?» Il avait lu. (L:43) Did you read the pages? He had read. Both studies note characteristics of one or the other type. GP point out that definite objects are typically a non-first-order entity; L notes that the latent object often has propositional content. The two agree that indefinite or generic null objects do not have a contextually available referent. GP point out that generic null objects can give rise to an activity rather than an accomplishment reading of the verb; L notes that null objects can focus attention on the activity. Both point out that the lexical characteristics of the verb can help to identify the referent of the null object. GP note that null objects are often found in fixed phrases, while L describes a wider context of de-actualisation as being 2 Note that this definition correctly excludes empty object positions that are directly linked to an element in external argument position such as in passives, unaccusatives and perhaps middles. However, it leaves open to a null object interpretation an eventual unexpressed object position in unergative VPs. The definition also includes null oblique objects, although we will not discuss them here.
4 4 favourable to null objects. And both note several structural contexts that favour a nonovert object. These contexts are summarized and illustrated in (6) to (12). (6) sequences of verbs a. He will steal, rob, and murder. (GP) b. Elles ont caressé, pétri, étreint, pénétré... (L:97) They have caressed, kneaded, clasped, penetrated. (7) imperatives a. Push hard. (GP:) b. Fais voir. (L:50) Show. (8) contrastive uses a. He theorises about language, but I just describe. (GP:) b. Seulement moi, je n'assassine pas, je ressuscite. (L:91) Only I don't murder, I resuscitate. (9) infinitive a. This is a lovely guitar, with an uncanny ability to impress and delight. (BNC) b. Pour compenser, j'ai décidé d'adopter dorénavant cette graphie. (L:85) To compensate, I have decided to use that spelling from now on. (10) generic present tense a. There are those who annihilate with violence who devour. (BNC) b. Un peintre dérange bien moins qu'un écrivain. (L:83) A painter disturbs much less than a writer. (11) dative pronoun (French) J'étais où quand tu lui avais donné? (L:39) Where was I when you gave to him? (12) ça as subject (French) Ça flingue à tout va là-dedans. (L:91) They're shooting like crazy in there. In a third study, Goldberg (2001) investigates unexpressed objects of causative verbs (those that entail a change of state in the patient argument) in English. She concludes that the option of leaving these arguments unexpressed depends largely on factors relating to information structure: the unexpressed object is typically neither topical nor focal, and the verb is emphasized somehow, by being iterative or generic, by being contrasted with another verb, or by having a narrow focus.
5 5 All of these authors implicitly or explicitly adopt the position that the missing argument is not syntactically represented: syntactically the verb is intransitive. In a generative framework, this position finds a counterpart in Rizzi (1986: ), who proposes that both the arbitrary third-person human interpretation, meaning people in general or some people, and the prototypical-object interpretation, where the verb's lexical semantics identify the object, are available lexically to saturate the argument's theta role and block projection. Thus, the verbs are intransitive in syntax. The absence of a syntactic object explains why, in Rizzi s account, the type of sentence exemplified in (13) is impossible in English: there is no object that can bind the anaphor or be modified by the adjective. However, such sentences are grammatical in Romance; hence several accounts (Rizzi 1986; Authier 1989; Roberge 1991) posit a syntactically present null object. (13) a. Ce gouvernement rend malheureux. * This government makes unhappy. b. Une bonne bière reconcilie avec soi-même. * A good beer reconciles with oneself. Under the TR, the object position is projected and the verb remains transitive in syntax in both English and French. Although we do not find sentences like those in (13) in English (as shown by the ungrammaticality of the glosses), there is nonetheless evidence that a null object has an effect on syntax in both English and French. 3 For example, null objects can enter into a network of relationships with compatible pronouns, and sometimes require coreference, either with pronouns or with another null object. (14) a. Ce roman amuse quand on le prend avec humour. This novel amuses if one takes it with a sense of humour. b. Qui aime bien châtie bien. Who loves well punishes well. c. His attitude intimidates, until you figure out he's a phony. d. It's better to reuse than to recycle. Null objects can serve as the argument of a secondary predicate: (15) a. Les steaks, moi, je préfère manger saignant. Steaks, I like to eat rare. b. Vous avez acheté en solde? Did you buy on sale? c. Beat until thick and lemon-coloured. 3 The TR redefines the notion of NO and broadens the range of phenomenon it subsumes. It is beyond the scope of this paper to deal with all of them; we will not explore here the structure exemplified in (13) or attempt to explain the differences it highlights between Romance languages and English. Moreover, we do not investigate NOs with clausal or propositional characteristics.
6 6 A syntactically-represented null object is required to account for the availability of a parasitic gap interpretation for sentences such as (16). (16) Which document did the spy memorize before eating? This shows the necessity, even under a lexical account, of projecting an empty argument position. A lexical account, moreover, would require three mapping patterns for verbs such as eat: transitive with overt object, transitive with null object, and intransitive, for the prototypical-object or activity reading. Lastly, a null object can receive further specification. (17a) shows further specification of the NO of a transitive verb, while (17b) shows an attempt to further specify the argument of an unaccusative. The result is uninterpretable, presumably because the argument has moved. The impulse is to try to interpret (17b) as a transitive to supply a null object. (17) a. C'est une chose si douce que de louer, et surtout ses amis. (L:82) To praise is such a sweet thing, and especially one's friends. b. *C'est une chose si difficile que de partir, et surtout ses amis. To depart is such a difficult thing, and especially one's friends. These facts argue against both the lexical and the constructional accounts, which treat such sentences as objectless. Rizzi's (1986) general discussion also leaves unexplained instances of null objects that receive neither the arbitrary human nor the prototypical object interpretation, such as those in (18). Instead, elements of the linguistic and extralinguistic context come into play. It seems obvious that such information is not part of the lexical entry of the verb. (18) a. Lifting his arm to strike, he felt a grip of iron around his wrist, restraining him. (BNC) b. On voit que ce n'est pas lui qui lave. (L:86) You can tell he's not the one who washes. c. M. Jospin, maintenant, régularisez. (protesters' banner referring to the situation of immigrants without papers) (L: 55) Mr. Jospin, now, regularize. d. When you don't have money and you have to work hard to accomplish in life, it's not that easy to just throw it down... (Toronto Globe&Mail) Moreover, if the absence of an overt object could be explained entirely in semantic and pragmatic terms, we would expect English and French null objects to be substantially the same. But in fact, there is a subset of L's latent objects in French that have no counterpart in English. Examples are shown in (19). In these cases, there is a specific linguistic referent in the context, and the only interpretation is that this antecedent is the referent of the null object.
7 7 (19) a. On lui tendit une main...vexé, il négligea. (L: 48) A hand was extended to him. Annoyed, *he ignored. b. Si un mec t'offre un café balance lui à travers la gueule. (L:50) If a guy offers you a coffee, *throw in his face. c. Nikel m'a dit de prendre une boîte bleue dans le vestiaire. J_ 'ai prise. (L:59) Nikel told me to take a blue box from the locker. *I took. These absent objects, which are taken as definite and referential, resemble null arguments discussed by Huang (1984), Farrell (1990), Cardinaletti (1990), among others, and analysed as variables bound by a null topic or as null pronouns. In either case, the object is taken to be syntactically present. This is the position we adopt for the full range of null objects in French and English, by virtue of the TR, and we turn now to the issue of how these null objects are licensed and recovered. 4. Recoverability of NOs Null objects are diverse, and so are the means of their recovery. We propose that there are three means of recovering the identity or reference of NOs: (1) internally, through material in IP; (2) through discourse, involving referential NOs; and (3) by binding from the left periphery, i.e by a topic. We take up each of these in turn. 4.1 Internally-licensed NOs All of GP's understood objects, all of Goldberg's omitted arguments, all of L's generic absent objects and many of her latent absent objects can be considered to be internally-licensed, recovered through material in the IP. A primary means of recovery comes from lexical characteristics of the verb, as with the true prototypical object interpretation. Note that the prototypical object of psychological verbs, which are commonly found with NOs in both English and French, is in fact the arbitrary thirdperson affected human interpretation (see (13) above). (20) a. La magie des séries, c'est de surprendre, de dépayser. (L:98) The magic of the playoffs is in surprising, disorienting. b. Where Boulestin never falters or misleads is in the sureness of his taste and the sobriety of his ingredients. (BNC) c....the patter of the camp, grey-haired one between songs can irritate. (BNC) The identity of lexically-determined NOs can range from the vaguely predictable, as in (21a) ( the area around me ); to the narrowly determined, as in (21b) (a paper or envelope); to the entirely predictable, as in (21c) semantically, the only possible object of déciller is eyes. Examples (21a) and (21b) thus illustrate how the lexical-semantic contribution from the verb may be augmented by information from the linguistic and extralinguistic context, while (21c) shows an entirely lexical contribution. (21) a. «Ben, qu'est-ce que tu fais?» J'explore. (L:83) Hey, what are you doing? I'm exploring.
8 8 b. Dans ma hâte à décacheter, j'ai déchiré la feuille. (L:76) In my haste to unseal, I tore the page. c. Crystal claqua dans ses mains. On décilla. (L:54) Crystal clapped his hands. We opened. The internally-licensed NO is not formally linked to another linguistic element. It does not refer; it is not an anaphor and it is not in a relationship with a [+specific] nominal; in L s and GP s canonical cases, moreover, there is no contextually available referent. In English, when a referential interpretation is forced, a null object is impossible, as in (22), while in a similar context but without forcing reference, the null object is fine, as in (23). (22) a. What happened to that carrot? *I chopped. (Goldberg 2001:512) b. The door is open. *Didn't you lock? (23) a. What happened to all the vegetables? Well, Jacques has been chopping and dicing all afternoon. b. (pulling out of the driveway) Did you lock? Because the internally-licensed NO does not refer and is not anaphoric, pragmatics has a free hand in interpretation, and contextual factors can contribute to the inference of a specific reference. In fact, according to Levinson's I-principle (2000: 114), based on Grice's (1975) maxim of informativeness, hearers will seek out a maximally pertinent interpretation of such NOs, assuming rich connections with contextual information. This is illustrated by the sentences in (24). (24) a. We have to get rid of all the ugly dishes before your date arrives. Okay, you wash and I'll dry. (Goldberg 2001:515) b. Allez, envoie. (L :50) Come on, hand it over. c. I'll introduce. (one host to another before a talk) d. Même avec trois cuillerées de sucre en poudre, le breuvage reste amer. Leroy touille en comptant les miettes sur la toile cirée. (L:49) Even with three spoonfuls of sugar, the drink still tastes bitter. Leroy stirs, counting the crumbs on the oilcloth. Other factors that enhance recoverability are found within IP. These include the factors that contribute to de-actualisation, such as the generic present tense, the infinitive, ça as subject; see (6)-(12) above. Tenseless verb forms and non-referential tenses favour a nonreferential reading, while referential tenses, such as perfectives, favour a specific, referential reading. Although the correspondence is not perfect in either English or French (non-referential NOs are attested in sentences with, for example, perfective tenses), there is a clear tendency to associate specific, referential entities with specific events set at a
9 9 specific time; for this reason non-referential NOs can be less felicitous with referential tenses. The internally-licensed NO can be described as a null cognate object (NCO). Overt cognate objects, if unmodified, add no semantic information beyond that contained in the verb itself. The null cognate object is similar, and that is why constructions with NCOs are described as focussing on the action or on the verb. We liken the interpretation of a predicate containing an NCO to the thetic interpretation: an assertion is being made as to the existence of an object or of an event involving the object (Basilico 1998:542). But the object is not singled out from the event for a second judgment, such as assignment of a property as is the case in the categorical judgment We propose that the NCO is structurally a bare empty noun, similar to the empty NUMBER noun proposed by Kayne (2002). Kayne argues, on the basis of the adjectivelike properties of few (comparative and superlative forms, distributional facts), that few is in fact an adjective modifying a phonologically empty N with the semantic content NUMBER. NCOs can be conceived of as a similar N whose semantic content is derived from the verb, thus one which is semantically cognate to the verb: (25) V 2 V N cognate The NCO is available for all verbs. The difference between conventionally transitive verbs (such as manger 'eat') and 'unergatives' (such as dormir 'sleep') is that the NCO is the more marked object for the former class and the less marked object for the latter class (26a and b). Moreover, both classes can have objects that are semantically independent of the verb, as in (26c). (26) a. V V null cognate object 2 2 manger N edible dormir N sleepable eat sleep b. V V lexically-conditioned object 2 2 manger une pomme dormir un bon somme eat an apple sleep a good nap c. V V lexically-independent object 2 2 manger des claques dormir sa vie eat (i.e. receive) slaps sleep one's life away
10 Referential NOs Certain null objects, like those in (19) above and (27) below, have a referent that is identifiable from the linguistic or extralinguistic context. In this, they differ sharply from the internally-licensed NOs discussed in the previous section. (27) a. «Maîtrisez-vous vos interviews? C'est capital, les interviews.» Je maîtrise. (L:50) Do you control your interviews? Interviews are very important. I control. b. Ça provenait de deux planches à dessin dressées en guitoune. Crystal contourna. (L :88) It came from two drawing tables used as a tent. *Crystal bypassed. These NOs appear similar to NOs of colloquial German, discussed by Cardinaletti (1990) and exemplified in (28): (28) Habe es gestern gekauft. have it yesterday bought I bought it yesterday. Under Cardinaletti s analysis these null objects involve a base-generated empty operator locally binding a null pronominal variable; the content of pro is determined by the operator, which is compatible with third-person pronouns only. Kampen (1997) discusses similar examples in informal Dutch and presents evidence that the null topic is a null pronoun that is underspecified with respect to phi-features. Crucially, null pronominal variables may not appear if speccp is filled by lexical material: (29) a. *Gestern habe ich gesehen. yesterday have I seen b. *Wann hast du gesehen? when have you seen The null objects of French in (19) and (27) do not fall under the same constraints. Although third-person reference is most common, it is possible to construct acceptable examples with second- or first-person reference: (30) a. Crystal tient à toi. Mais les méchants veulent lui prendre. Crystal wants to keep you. But the bad guys want to take from him. b. Yan m'a vue pour la première fois à la bibliothèque, et tout de suite il a adoré! Yan saw me for the first time in the library, and right away he adored! Moreover, the presence of lexical material in SpecCP does not prevent the appearance of a null object.
11 11 (31) a. (in video store) Si on prenait Tigre et Dragon? Qui a vu? How about Crouching Tiger Hidden Dragon? Who has seen? b. Tu as lu les pages? Tu m'as dit que tu avais lu. Did you read the pages? You told me you had read. Therefore, there is no evidence for an operator in CP influencing the empty object in French. 4 Rather, this type of NO seems to correspond to a clitic, and its appearance can be seen as an instance of clitic-drop. The simplest approach to this construction would be to assume that it corresponds exactly to its counterpart with an accusative clitic; see Tuller (2000), Guasti & Cardinaletti (2003: note 17) and references cited there. Semantically, the sentences are equivalent to corresponding sentences with an object clitic, and the structural contexts for this type of NO are identical to those of clitics. (32) a. A: Pourqoui avoir choisi cette époque? B: Parce que j adore. (L:64) = Parce que je l adore. A: Why did you choose that period? B: Because *I love /I love it. b. Nikel m a dit de prendre une boîte bleue dans le vestiaire. J ai prise. = Je l ai prise. (L:77) Nikel told me to take a blue box from the locker. *I took /I took it. Moreover, the attested past participle agreement in (32b), while certainly unusual, is identical to the agreement that would be found if an accusative clitic were present. Thus, pro can appear without the clitic that is normally used to recover its feature contents. We do not address the issue of determining whether an empty or silent clitic linked to pro must be postulated or whether pro can appear on its own with default feature. The function of object clitics is to morphologically recover definite NOs; it seems that in these sentences, this recovery mechanism is manipulated for stylistic effect. English has no similar element able to recover NOs; therefore this option is not available and all counterparts to the clitic-drop examples are ungrammatical in English. 4.3 Topics in English and French French and English, as is well known, display contrasting behaviour of topics and the linked element in the matrix. Topics of the type in (33) are linked directly to an empty object in the matrix in English. In French, the link is mediated by a clitic as in (34), in the structure dubbed Clitic Left Dislocation by Cinque (1990). 4 A similar conclusion (i.e. that the null object in (27) is not a null variable) would be reached through an application of the tests proposed in Raposo (1986). For example, when talking about a safe, it would be acceptable in French to say: (i) J ai informé la police de la possibilité que la secrétaire ait ouvert _ à l insu de son patron. I informed the police of the possibility that the secretary might have opened without her boss s knowledge. Raposo uses the ungrammaticality in European Portuguese of a similar construction to argue that this type of null object is a variable in this language.
12 12 (33) a. Your book, I bought. *Ton livre, j'ai acheté. b. John, I can't stand. *Jean, je ne supporte pas. (34) a. Ton livre, je l'ai acheté. b. Jean, je ne le supporte pas. Rizzi (1997) posits Your book OP I bought ec for the English case in (33a), where the ec is a null constant licensed by the anaphoric operator. In Romance, according to Rizzi, clitics fill the same function of establishing the connection between the topic and the open position in the comment. We adopt this analysis, but note that the facts regarding topics, clitics, and ecs are somewhat more complex. Taking indefinite topics into account and incorporating the discourse-linked ecs discussed in 4.2 into the mix, the similarities and contrasts between English and French emerge as summarized by the data in (35) and (36). (35) Overt topic a. Token b. Type English That book, I hated. Wine, I bought. John, I can t stand. Bananas, I ll eat. French *Ce livre, j ai détesté. *Jean, je ne supporte pas. Vin, j'ai acheté. (cf. also Vin, j'en ai acheté.) Les bananes, je mange. (cf. also Les bananes, j'en mange.) (36) No topic or null topic a. definite (linguistic antecedent) b. indefinite (no linguistic antecedent) English Did you read the pages? A: You like? *He had read. B: I love! A: What do you think of my cake?? So, how would you rate? B: *I like. French Tu as lu les pages? Il avait lu. A: Que penses-tu de mon gâteau? B: J'aime! A: Tu aimes? B: J'adore! Alors, comment as-tu trouvé? In summary, non-definite null objects are acceptable in both languages, as in (35) and (36b); they are linked to an overt topic or an element in the nonlinguistic context. In English, a definite null object must be linked to an overt topic; compare (35a) and (36a) In French, a definite token null object is typically linked to an overt topic only via a clitic (35a). However, there are innovative instances of a definite token topic and a null
13 13 object linked without an overt clitic (see Fonágy 1985). In Fonágy s corpus, these cases most commonly involve the verbs connaître or aimer but are not restricted to these. (37) a. Jacques F., vous connaissez? (Fonágy 1985:5) Jacques F., do you know? b. Le yogourt X, il aime, il adore. (Fonágy 1985:8) X yogurt, he loves, he adores. c. La bleue, je prends. (Fonágy 1985:9) The blue one, I ll take. These cases show clear parallels with the clitic-drop cases discussed in 4.2 and represented in (36a). They are semantically equivalent to corresponding sentences with a clitic, and the sole structural difference is the absence of a clitic linking topic and NO. We hypothesize that clitic-drop is the strategy at work with the overt topics, as well as in the cases involving a null topic or no topic. Clitic drop can be seen as an extension of the general pragmatic strategy involved in the interpretation of nominal elements. We assume that a standard pattern in discourse is the sequence [lexical noun pronoun NO] in which all nominals are interpreted as coreferential (assuming no contradictory information). This is exemplified in (38): (38) J ai vu ton chien 1 dans le parc. Je l 1 ai caressé Ø 1. I saw your dog in the park. I petted it. The coreference between the pronoun and the NO is established by purely grammatical means, while that between the lexical noun and the pronoun is pragmatic and defeasible, involving further application of Levinson s I-principle, whereby a hearer infers from a lack of specification that there is no need for specification. The default here is for the hearer to assume coreference between the lexical noun and the pronoun. With clitic-drop, the same implicature comes into play, this time between the lexical noun and the NO. Unlike NCOs, the instances of clitic-drop are considered innovative and stylistically marked. Fonágy drew most of his examples from younger speakers and from advertising and considers that, in the latter case at least, the strategy is a deliberate attempt to appear hip and fresh. Larjavaara limited her corpus to recent works and chose literary texts that aim for a style that could be described in the same terms. It goes without saying that many such examples will be considered ungrammatical by speakers with other demographic profiles and different stylistic aims. However, as Lambrecht and Lemoine (1986:280) remark, data from spoken varieties must be included in a complete account of grammatical phenomena, especially in a language, like French, whose written variety is greatly influenced by prescriptive influences and differs significantly from spoken varieties. In addition, the label ungrammatical is of limited usefulness in assessing the stylistic effect of the strategy of clitic-drop. We hypothesize, as a matter for further research, that the stylistic effects noted by Fonágy and Larjavaara are due in part to the fact that the result of clitic drop is identical to a NO whose referent is physically salient in the
14 14 discourse. Noailly characterizes this use as deictic and describes its function as lending cohesion to the discourse. If NOs resulting from clitic-drop are similar, it is easier to begin to understand the immediacy they bring to the discourse and their effect of engaging the reader or hearer. Fonágy and others view the NOs resulting from clitic-drop as a fairly recent phenomenon in French. (We note also that most native speakers we have consulted agree that it is more typical of European French than of Canadian French.) However, Arteaga (1998) argues for a syntactically present NO in Old French, which she analyses as a null pronominal; certain examples she gives can be likened to current examples that we have treated as clitic-drop; compare, for example, (39) from Arteaga 1998, with (30a) above: (39) Tient une chartre, mais ne li puis tolir Ø. Il tient une lettre, mais je ne peux pas lui prendre. He holds a letter, but I cannot take from him. We note also the following example from a 14 th (Troberg 2004): century text of Middle French (40) Car les letres que li messages apportoit, c'estoit mes usages de regarder Ø avant toute oevre. Les lettres que le messager apportait, c'était mon habitude de regarder avant tout autre travail. The letters the messenger brought, it was my habit to look at before any other work. Further research is thus needed to establish clearly whether or not clitic-drop is an innovation in contemporary French. 5. Conclusion We have argued that the existence of null objects is largely determined by the Transitivity Requirement and that cross-linguistic variation is therefore predicted to occur mostly in the recoverability mechanisms particular grammars use. A systematic comparison of English and French null objects was used to support this claim. The data from these two languages lead to the conclusion that there are three types of null objects: 1. Bound: a bound variable or a null constant; 2. Discourse-linked: a null pronominal; 3. Internally-licensed: cognate null objects (predicted by the TR); a bare N. It was shown that French and English differ only in the availability of discourse-linked null objects and bound definite null objects (see 35 and 36 for a summary). We have characterized the French discourse-linked null object as a clitic-drop construction. This preliminary account is intended to sketch out the semantic and syntactic characteristics of the three types of NOs, and the role pragmatic principles play in their
15 15 recovery. 5 Under the TR, all NOs are syntactically represented; ensuring syntactic representation allows for an account of differences in referentiality and syntactic activity. REFERENCES Arteaga, Deborah On Null Objects in Old French. Romance Linguistics. Theoretical Perspectives ed. by Armin Schwegler, Bernard Tranel & Myriam Uribe-Etxebarria, Amsterdam: John Benjamins. Authier, J.-Marc Arbitrary Null Objects and Unselective Binding. The Null Subject Parameter ed. by Osvaldo Jaeggli & Ken Safir, Dordrecht: Reidel. Basilico, David Object Position and Predication Forms. Natural Language and Linguistic Theory Blinkenberg, Andreas Le problème de la transitivité en français moderne : essai syntacto-sémantique. Copenhagen: Coll. Historisk-filosofiske Meddelelser. Cardinaletti, Anna Subject/Object asymmetries in German null-topic constructions and the status of speccp. Grammar in Progress ed. by Joan Mascaró & Marina Nespor, Dordrecht: Foris. Cinque, Guglielmo Types of A' Dependencies. Cambridge: MIT Press. Desclés, Jean-Pierre Transitivité sémantique, transitivité syntaxique. La transitivité ed. by André Rousseau, Villeneuve-d Ascq : Presses universitaires du Septentrion. Farrell, Patrick Null Objects in Brazilian Portuguese. Natural Language and Linguistic Theory Fonágy, Ivan J aime. Je connais. Verbes transitifs à objet latent. Revue Romane García Velasco, Daniel & Carmen Portero Muñoz Understood Objects in Functional Grammar. Working Papers in Functional Grammar 76. University of Amsterdam. Goldberg, Adele Patient arguments of causative verbs can be omitted: the role of information structure in argument distribution. Language Science Gourio, Jean-Marie Brèves de comptoir. Paris: Laffont. Guasti, Maria Teresa & Ana Cardinaletti Relative Clause Formation in Romance Child s Production. Probus Hopper, P.J. & S.A Thomson Transitivity in Grammar and Discourse. Language Further work is needed on: 1) the distinction between non-referential yet not generic NCOs (often with inferred reference to a salient entity in the non-linguistic context) and referential NOs resulting from clitic drop; 2) the issue of NOs in the process of language change and grammaticalisation. Referential NOs seem to be gaining ground in spoken French; is this is a sign of a shift from the status of a syntactic language to that of a pragmatic language (Huang 2000: 261)? Use of null anaphora, like reliance on a topic-comment structure, also increasingly prominent in spoken French, is associated with the latter type.
16 16 Huang, James On the Distribution and Reference of Empty Categories. Linguistic Inquiry Huang, Yan Anaphora. A Cross-linguistic Approach. Oxford: Oxford University Press. Kampen, Jacqueline van First Steps in Wh-movement. Delft: Eburon. Kayne, Richard On the Syntax of Quantity in English. Ms., New York University. Lambrecht, Knud & Kevin Lemoine Vers une grammaire des compléments zéro en français parlé. Absence de marques et représentation de l'absence 1 ed. by Jean Chuquet & Marc Frid, Rennes: Presses universitaires de Rennes. Larjavaara, Meri Présence ou absence de l objet. Limites du possible en français contemporain. Helsinki: Academia Scientiarum Fennica. Lazard, Gilbert L actance. Paris: Presses universitaires de France. Levinson, Stephen C Presumptive Meaning: The Theory of Generalized Conversational Implicature. Cambridge: MIT Press. Noailly, Michèle Les mystères de la transitivité invisible. Langages Raposo, Eduardo On the Null Object in European Portuguese. Studies in Romance Linguistics ed. by Osvaldo Jaeggli & Carmen Silva-Corvalan, Dordrecht: Foris. Rizzi, Luigi Null Objects in Italian and the Theory of pro. Linguistic Inquiry The Fine Structure of the Left Periphery. Elements of Grammar ed. by Liliane Haegeman, Dordrecht: Kluwer. Roberge, Yves On the Recoverability of Null Objects. New Analyses in Romance Linguistics ed. by Dieter Wanner & Douglas A. Kibbee, Amsterdam: John Benjamins Transitivité verbale, grammaticalisation et sémantisation, colloquium Représentations du sens linguistique II, Montreal, May. Troberg, Michelle La proposition topique-commentaire en moyen français: la reprise facultative du clitique accusatif. Ms., University of Toronto. Tuller, Laurice Null Objects in Deaf French. Ms., Université de Tours.