On the interpretation of postverbal subject positions Frank
Transcription
On the interpretation of postverbal subject positions Frank
On the interpretation of postverbal subject positions Frank Drijkoningen (Utrecht) & Brigitte Kampers-Manhe (Groningen) This version is identical to the version published as Recherches de linguistique française et romane d'Utrecht XX, pp. 29-43. 0. Introduction In this paper we wish to make a short general overview of facts of French pertaining to issues with respect to (in)definiteness effects and semantic and pragmatic differences in postverbal positions (note 1). The (in)definiteness effects and restrictions are relatively well-studied crosslinguistically. However, most of the analyses are based on the opposition between impersonal constructions and ‘normal’ constructions with a preverbal subject NP. As an example, consider the effects obtaining in the pairs Jean est arrivé / Trois linguistes sont arrivés versus *Il est arrivé Jean / Il est arrivé trois linguistes. Recently, the so-called counter-indefiniteness effect (originating in De Cornulier 1974, see also Kupferman 1983) has been linked to these issues by Kayne & Pollock (1998, ms.) The counter-indefiniteness effect opposes questions like Quel gâteau a mangé Jean? To *Quel gâteau a mangé quelqu’un? In this paper we will concentrate on the (in)definiteness data without looking at the ‘normal’ (‘personal’) constructions. In other words, we oppose and compare the impersonal construction to the varieties of inversion occurring in French. In practice, we consider here: the impersonal construction, stylistic inversion, complex inversion, subject clitic inversion and locative inversion. Among these constructions, Stylistic Inversion will be the most central one throughout. Alongside this renewed interest in the definiteness effects in questions, there have been other developments in this area. With respect to French, Kampers-Manhe (1998) elaborates and analysis in which Focus plays a crucial part as far as Subject Inversion is concerned in subjunctive clauses. This entails that Stylistic Inversion is not homogeneous from a more semantic and pragmatic perspective. Proponents of HPSG also argue for a certain breakdown of the familiar classification, also at the non-syntactic level, see e.g. Marandin (2000b). And finally, the influence of Focus in Stylistic Inversion is made really dominant in the framework of Optimality Theory, as evidenced by Legendre (1998). On the theoretical level of generative grammar, since Rizzi (1997) there have been several proposals for the integration of notions like topic and focus in the syntactic structure. Inspired by these functional syntactic heads, Kayne & Pollock’s explanation of the counter-indefiniteness effect in Stylistic Inversion is based on the use of a Topic projection. Their analysis (making use of remnant movement) is a syntactic alternative to analyses of Stylistic Inversion that are based on multiple separate movements (e.g. Drijkoningen 1997). In the first section we will consider root clauses and put definiteness central. In the second section we will discuss embedded clauses, and where we will also consider focus. 1. Root clauses In this section we will consider the data as to three different properties. First, as our contribution bears on subjects, we will look at agreement. Second, we will discuss the contrast between definites and indefinites. Recall that we focus on a comparison between the impersonal construction and the inverted structures rather than on a comparison with normal non-inverted structures. Finally, we will take a closer look at indefinites, as those do not form an homogeneous class. Besides ‘ordinary’ indefinites we have integrated the data on ne .. aucun and on generic un. Both these elements are traditionally analyzed as indefinites, but do have special interpretative properties. The interesting point in this paper will be that the two elements differentiate between the constructions under discussion. 1.1. Agreement Agreement in the impersonal construction takes place with the preverbal subject (3rd person singular ([-feminine]), as shown in (1a/b). As shown in (1c/d) participle agreement is included in the same rule, i.e. absent. (1) (a) (b) (c) (d) Il est arrivé trois paquets. There is arrived three packages * Il sont arrivés trois paquets There have arrived three packages Il est arrivé une fille. There is arrived a girl * Il est arrivée une fille. There is arrived+FEM une fille. In Stylistic Inversion agreement takes place with the postverbal subject, as shown in (2a/b). Participle agreement coincides, as shown in (2c/d), i.e. present. (2) (a) (b) (c) (d) Où ont joué les dix enfants? Where have played the ten children * Où a joué les dix enfants? Where has played the ten children Où est allée cette fille? Where is gone+FEM that girl * Où est allé cette fille? Where is gone that girl The agreement pattern in Locative Inversion is identical to that in Stylistic Inversion, as shown in (3). (3) (a) (b) (c) Dans cette forêt ont vécu des singes In that forest have lived apes * Dans cette forêt a vécu des singes In that forest has lived apes Dans cette forêt ont été tués plusieurs soldats In that forest have been killed+PLUR several soldiers (d) * Dans cette forêt ont été tué plusieurs soldats In that forest have been killed several soldiers Although agreement clearly generalizes, locative inversion and stylistic inversion are most often kept different on the theoretical level. One of the causes is the tendency to use indefinites in locative inversion, while stylistic inversion prefers definites. This is the other aspect of the discussion we engage upon in this paper. In subject clitic inversion the verb agrees with the clitic (as shown in (4)). In complex inversion, in most cases the correct result is obtained with either one, the NP or the clitic (as shown in (5)), because the clitic is a copy of the NP (or co-indexed). However, there are facts that point to the relevance of the clitic. If the clitic and NP are copies (and/or coindexed) but do not coincide with respect to phi-features, the clitic determines the agreement. The fact are given in (6a/b) for spoken French and in (6c/d) for written French. (4) (5) (6) Viendra-t-il Comes+FUT+3rd person he Jean viendra-t-il? John comes+FUT+3rd person he (a) Jean et moi on devrait rester. John and me one(we) should+3rd person singular stay (b) * Jean et moi on devrions rester. John and me one(we) should+1st person plural stay (c) Pourquoi Jean et moi ne devrait-on pas rester? Why John and me should+3rd person singular – one(we) not stay (d) * Pourquoi Jean et moi ne devrions-on pas rester? Why John and me should+1st person plural – one(we) not stay In sum, there are two generalizations for agreement. The first one is a generalization across the impersonal construction and complex inversion. The second one is a generalization across stylistic inversion and locative inversion. Regular subject clitic inversion could fall under either one. (7) (A) (B) If there are two subjects, one ordinary and one clitic, the phi-features of the clitic determine the agreement on the verb: (a) Il (clitic) est arrivé trois paquets (subject NP) (b) Jean et moi (subject NP) on (clitic) devrait rester If there is only one subject, the phi-features of this NP determine the agreement on the verb. (a) Où ont joué les dix enfants (subject NP)? (b) Dans cette forêt ont vécu des singes (subject NP). In both the Government and Binding framework (henceforth GB) and the Head Driven Phrase Structure Grammar (henceforth HPSG) these generalizations can be properly explained. Generalization (7A) is explained as a more general property of clitic in the sense that clitics are more intimately associated with the verb than the NP (e.g. Hulk (1982) for GB and Miller (1991) for HPSG (note 2). Generalization (7B) is explained in GB by the fact that the postverbal subject in previous stages of the derivation has passed through the preverbal position in IP where agreement features are dealt with, e.g. Kayne & Pollock (ms.). An extension of such an analysis also accounts for participle agreement (e.g. Drijkoningen (1999)). In HPSG the notion of ‘first argument’ is used, related to the fact that agreement marking is dealt with lexically (note 3). 1.2. Definites and indefinites As has been mentioned briefly in the introduction, the standard definiteness effect is found in the impersonal construction. The contrast is repeated in (8). As has been mentioned above, stylistic inversion has a reverse effect, dubbed the counterindefiniteness effect. The contrast is repeated in (9). (8) (a) (b) (9) (a) (b) * Il est arrivé Jean There is arrived John Il est arrivé quelqu’un There is arrived someone Quel gâteau a mangé Jean? Which cookie has eaten John * Quel gâteau a mangé quelqu’un? Which cookie has eaten someone In (8) and (9) we inserted standard indefinites. If one considers the grammaticality of the impersonal construction as a test, an NP containing aucun (in which case ne is cliticized to the verb) qualifies as an indefinite, as shown in (10). (10) (a) (b) Il est arrvé trois paquets. There is arrived three packages Il n’est arrivé aucun paquet. There not-is arrived any package. It is independently known (Henriette de Swart, p.c.) that aucun has a special status in French. In practice, the translation of (10b) into English is subject to discussion, with respect to the alternative ‘there have arrived no packages’ and/or ‘there have arrived zero packages’ and/or ‘there hasn’t arrived any package’. We will leave this aside, and take along aucun in our discussion, be it as a special indefinite. Its special nature eventually has a function in this paper. One traditionally classifies un among indefinites. However, the semantic interpretation of un does not always lead to the standard indefinite reading. More specifically, the indefinite article can be interpreted generically, generic un. Surprisingly, this particular use of the indefinite article is impossible in the impersonal construction, as shown in (11). (11) (a) (b) Une femme va chez sa mère quand elle se querelle avec son mari. A woman goes to her mother when she has an argument with her husband * Il va une femme chez sa mère quand elle se querelle avec son mari. There goes a woman to her mother when she has an argument with her husband. Just as with aucun, there is something special about the generic reading of the indefinite article. The peculiarity of this reading blocks the use of the impersonal construction, and thus forms one of the testing possibilities for the properties of postverbal positions in inversions. Schematically, we thus have the following basis for further discussion. (12) Impersonal construction Stylistic Inversion Definite Indefinite * OK OK * Special aucun OK indef Special indef Generic un * The contrast between definites and indefinites for impersonal constructions and stylistic inversion suggest that the effects are essentially the “reverse”. The idea is “what is possible in impersonal constructions is impossible in stylistic inversion and what is possible in stylistic inversion is impossible in impersonal constructions”. The term counter-indefiniteness effect captures this intuition properly. But we will show below that it cannot be an overall “reverse” effect without further specification, due to the data on the special indefinites. 1.3. Indefinites and indefinites As we have just concluded, the qualification ‘indefinite’ needs some precision, and more precisely, we have added two rather special indefinites in order to proceed with further testing of what is possible in inverted structures. Let us first complete the schema in (12) with the data for Stylistic Inversion. Kayne & Pollock (ms.) cite (13a). (13) (a) (b) Quel livre ne comprend aucun linguiste? Which book doesn’t any linguist understand Il n’est arrivé aucun linguiste There hasn’t arrived any linguist. In this respect, the inverted NP of Stylistic Inversion has the same properties as the postverbal NP in the impersonal construction, (13b). In other words, the so-called counter-indefiniteness effect does not extend to all indefinites that are allowed in the impersonal construction. This seems to entail that the counter-indefiniteness effect is not simply a “reverse” effect. There is at least one set of facts that show that the impersonal construction and stylistic inversion are identical. Next, we add a new fact, concerning generic un. (14) (a) (b) Où va une femme quand elle se querelle avec son mari? Where goes a woman when she has an argument with her husband * Il va une femme chez sa mère quand elle se querelle avec son mari. There goes a woman to her mother when she has an argument with her husband. These data confirm the “reverse” effect. From this perspective, generic un occupies positions reserved for definites. At this point, the schema in (12) has become (15). Impersonal construction Stylistic Inversion Definite Indefinite * OK Special aucun OK OK * OK indef Special indef Generic un * OK Let us now take into consideration locative inversion, for which we showed above that it pattern with stylistic inversion as far as agreement is concerned. The picture for locative inversion is less clear, although the general tendency will be sufficiently solid. The two relatively standard sentences that show a relevant contrast between definites and indefinites are given in (16). (16) (a) (b) Dans la forêt vivait un vieil ermite In the forest lived an old hermite ?? Dans la forêt vivait le vieil ermite. In the forest lived the old hermite. However, additional data can cast doubt on the validity of this contrast. They are twofold. First, although quite normal indefinites like un vieil ermite are accepted, the grammaticality strongly drops when the NP is of the type we used in the examples above, quelqu’un. (17) (a) (b) ?? Dans la forêt vivait quelqu’un. In the forest lived someone. Dans la forêt VIVAIT quelqu’un. In the forest LIVED someone. When uttered with a neutral phonetic pattern, (16a) is fine, but with a similar neutral phonetic pattern (17a) is severely less acceptable. (17a) can be repaired by using a non-neutral phonetic pattern with an accent on the verb, as in (17b). Although this fact points at the relevance of Focus, it is unclear why this should improve the grammaticality of the examples with quelqu’un and crucially not those with un vieil ermite. Second, there are some examples that do allow for definites: (18) (a) (b) Dans cette forêt vivait le vieil ermite. In that forest lived the old hermite Dans cette île mourut Napoléon On that island died Napoleon. These examples involve a deictic determiner internal to the locative PP, which phonetically is relatively prominent. If the deictic element is omitted, ungrammaticality ensues: (19) * A Ste. Hélène mourut Napoléon Although this feasibly point at the relevance of prominent Deixis, it is unclear why this should improve the grammaticality of these examples with definite postverbal NPs. (note 4). Besides these qualifications to the basic pattern in (16), which cast doubt on the contrast, let us look at the special indefinites. They confirm the basic tendency in (16), as shown in (20). (20) (a) (b) Dans la forêt ne vivait aucun singe. In the forest didn’t live any ape. * Chez sa mère vit une femme après avoir quitté son mari. At her mother’s lives a woman after having left her husband (20a) is on a par with the impersonal construction, while the ungrammaticality of (20b) also shows that this construction is on a par with the impersonal construction. Because of these data we give the contrast of (16) the benefit of the doubt, knowing that there are additional issues to be settled in the field of the interaction of indefinites with Focus and Deixis. Finally, let us turn to complex inversion. Our judgments point at a generalization with Stylistic Inversion in this respect (despite other profound syntactic differences). They are given in (21). (21) (a) (b) (c) (d) (e) Quand Jean est-il arrivé? When John has-he arrived Quel livre aucun linguiste ne comprend-il? Which book no linguist understands-he not Pourquoi une femme n’aurait-elle pas les mêmes droits qu’un homme? Why a woman should she not have the same rights as a man ?? Quel gâteau quelqu’un a-t-il mangé? Which cookie someone has-he eaten ?? Où trois chats dorment-ils? Where three cats sleep-they At this point the schema in (15) can be completed to (22). (22) Impersonal construction Stylistic Inversion Locative Inversion Complex Inversion Definite Indefinite * OK Special aucun OK indef Special indef Generic un * OK * OK OK ?? OK OK * OK ?? OK OK 1.4. Conclusions From the overview in (22), one general observation and two generalizations emerge. The particularity of aucun revealed by studies on negation is confirmed in syntax. Among the NPs considered, those containing aucun are the only ones that are insensitive to any restriction in the area of definiteness. In all postverbal positions under consideration, NPs with aucun are straightforwardly accepted. The first generalization is a generalization across the impersonal construction and locative inversion: grammaticality of indefinites and impossibility of generic un, alongside questionable grammaticality (or ungrammaticality) for indefinites. The generalization has been elaborated elsewhere also, e.g. in Hoekstra & Mulder (1990), who analyze existential and locative predication with one formalism. The second generalization is a generalization across stylistic inversion and complex inversion: grammaticality of definites and possibility of generic un, alongside questionable grammaticality (or ungrammaticality) for indefinites. This generalization is new, and stems basically from our judgments in (21d) and (21e) which entail a counter-indefiniteness effect in complex inversion. We will return to this shortly below. It is impossible to generalize the impersonal construction and stylistic inversion. The agreement pattern already indicated this, the definiteness criteria confirm it. Hence, in GB syntax, the NP of the impersonal construction should occupy a different position than the NP in stylistic inversion. The proposal by Kayne & Pollock (ms.), generation of the NP in the standard preverbal position in structures with stylistic inversion, captures this. It is also impossible to generalize stylistic inversion and locative inversion from the dominant perspective of this paper. Despite the fact that the agreement patterns are identical, the definiteness criteria point at fundamental differences. In discussing the counter-indefiniteness effect, Kayne & Pollock insist on the fact that the effect obtains only with what they call ‘real indefinites’. The notion of ‘real indefinites’ can be fleshed out and made more specific in view of the data we discussed. Proper names in Stylistic Inversion are grammatical, which is not surprising as they are definite. NPs with aucun are also grammatical. From a pragmatic point of view sentences like Il n’est arrivé aucun paquet do not introduce a new entity in discourse. Rather, the sentence states that this potential new entity is not there. So a definition of ‘real indefinites’ as ‘new entities in discourse’ is feasible. This definition logically includes quelqu’un, introductory “par excellence”. Finally, under this definition generic un also fits in: A generic statement does not entail the introduction in discourse of the entity that the statement bears on. As a consequence, a pragmatic formulation of the counter-indefiniteness effect is: “do not pose questions about entities that you introduce in discourse while posing the question”. In terms of file-cards, the situation depicted amounts to the following for e.g. * Quel gâteau a mangé quelqu’un? (23) Make a new file card for unidentified person (quelqu’un). As a new file card this card does not have information on it. Now answer my question: What did this unidentified new person eat? This pragmatic interpretation of the notion of ‘real indefinites’ can be fully linked up a pragmatic way of reading the analysis of Kayne & Pollock. Their analysis of the counter-indefiniteness effect is that the NP in question has to be moved to a higher projection, Topic Phrase, and that ‘real indefinites’ cannot be topicalized. Hence, the subject NP in Stylistic Inversion is a ‘previously introduced existing entity’. Pragmatically speaking, the analysis states “do pose questions about entities that are known, i.e. topics”. We feel this analysis is on the right track. On the other hand, the pragmatic basis of the eventual explanation looks relatively construction-independent and, moreover, also language-independent. This intuition of the existence of a more general property of questions comes back in our judgments above for about complex inversion – which we repeat for convenience: (21) (d) (e) ?? Quel gâteau quelqu’un a-t-il mangé? Which cookie someone did-he eat Où trois chats dorment-ils? Where the cats sleep-they On the judgment level, we find these clearly infelicitous without any context. In starting a discourse fragment, the questions are ungrammatical. In running discourse, however, they can be more acceptable, but then there is a strong relation with other reading differences that have been noted in the literature, e.g. the strong reading of prenominal indefinites, as in (24). (24) (a) (b) Trois chats dorment par ici. Three cats are sleeping over here. Il dort trois chats par ici. There are sleeping three cats over here. We can imagine a discourse in which (24a) is not properly heard, of the type: (25) (a) (b) A. Trois chats dorment xxx Three cats sleep xxx B: Où trois chats dorment-ils? Where do three cats sleep The issue is close to the possibility of ‘second occurrence’, discussed in the literature on focus (Helen de Hoop, p.c.) In sum, our intuition is that (21d/e) have a counter-indefiniteness effect in the intended sense “no introduction of a new entity”, and that the grammatical instances have standard additional effects in terms of strong readings, specific readings or second occurrence nature – the additional effect being “the entity was already introduced before”. The details for this approach are left open for future research (note 5). 2. Embedded clauses In embedded clauses, complex inversion is impossible, and will therefore be left out of consideration. The impersonal construction is possible in embedded sentences, but there is a general consensus that there are no special properties when compared to root clauses; therefore the construction will not be further discussed either. Hence, locative inversion will be discussed, as well as Stylistic Inversion. Stylistic Inversion will comprise the bulk of the discussion. 2.1. Locative inversion The grammaticality of locative inversion in embedded clauses is a topic of hard judgment-oriented debate in a number of languages, e.g. for English and Dutch. The discussion is linked to the possibility of having a locative PP occupy the standard subject position, and is therefore theoretically relevant. In practice, French is not different from English or Dutch in this respect. The data are given in (26) for indefinites and (27) for definites. (26) (a) (b) (c) (27) (a) (b) (c) (d) ? Il a dit que dans la forêt vivait un vieil ermite He has said that in the forest lived an old hermite ? Il a dit que dans la forêt ne vivait aucun singe. He has that that in the forest didn’t live any ape. * Il a dit que chez sa mère vit une femme après avoir quitté son mari. He has said that with her mother lives a woman after having left her husband. ??? Il a dit que dans la forêt vivait le vieil ermite. He has said that in the forest lived the old hermite ? Il a dit que dans cette forêt vivait le vieil ermite. He has said that in that forest lived the old hermite ? Il a raconté que dans cette île mourut Napoléon. He has told that on that island died Napoleon. * Il a raconté qu’à St. Hélène mourut Napoléon. He has told that at St. Helens died Napoleon. Generally speaking, the sentences are less grammatical than their root counterparts, without therefore being fully ungrammatical. Crucially, however, the definiteness properties are identical to the ones described above. As the reader may verify, the deletion of one question mark gives the judgments in the first section. In other words, there are no special definiteness effects in embedded clauses with locative inversion. If they are considered grammatical, the generalization noted above holds – a generalization with the impersonal construction. 2.2. Stylistic Inversion The counter-indefiniteness effect is observable under the proper circumstances, as shown by (28). (28) (a) (b) (c) (d) * Il a demandé quel gâteau a mangé quelqu’un He has asked which cookie has eaten someone. Elle a demandé où habite un homme après son divorce. She has asked where lives a man after his divorce. Elle a demandé où ne vivait (plus) aucun singe. She has asked where didn’t live any ape (any more). ?? Elle a demandé à quel endroit dormaient trois chats (note 6) She has asked at which place slept three cats. Nevertheless, there is an issue in embedded clauses, which stems from the optionality of the inversion. Both (29a) and (29b) are grammatical: (29) (a) (b) Il a demandé où dormaient les trois chats. He has asked where slept the three cats Il a demandé où les trois chats dormaient. He has asked where the three cats slept. The question to be dealt with in this case is whether there are situations in which (a) is preferred over (b) or the reverse. These have been argued to involve Focus. In what follows we will argue for Focus as the intervening property in structures in which stylistic inversion is otherwise optional. First, consider ‘contrastive focus’. (30) (a) (b) (c) (d) (e) Il a demandé quel gâteau a mangé Jean. He has asked which cookie has eaten John Il a demandé quel gâteau Jean a mangé. He has asked which cookie John has eaten. Mais non, il veut savoir quel gâteau a mangé Paul / PAUL. No, he wants to know which cookie has eaten Paul ?? Mais non, il veut savoir quel gâteau Paul a mangé. No, he wants to know which cookie Paul has eaten. Mais non, il veut savoir quel gâteau PAUL a mangé. No, he wants to know which cookie PAUL has eaten. The postverbal position can be contrastive on its own, with the possibility of a strong accent. The preverbal position is unfelicitous in contrastive situations, but can be rescued by putting a strong accent. So, contrastive focus causes a strong accent and/or inversion, but excludes non-inversion without accent. The impressionistic (dynamic) discourse definition we have in mind for contrastive focus is: “delete an existing entity inside a specified predicate and insert another existing entity at the emptied position”. (31) Mais non, il veut savoir quel gâteau a mangé Paul. No, he wants to know which cookie has eaten Paul Manger (x, x=Jean, y, y=gâteau) Æ manger (x, x=Paul, y, y=gâteau) Before continuing, let us also define impressionistically ‘narrow focus’: insert an existing entity inside a specific predicate at the position that is open (by the question). (32) Qui a mangé ce gâteau? Jean Who has eaten this cookie? John Manger (x, x=?, y y=gâteau) Æ Manger (x, x=Jean, y, y=gâteau) Wide focus corresponds to the construction of a predicate and adding of its arguments: (33) Qu’est-ce qui s’est passé? Jean a causé un accident What happened? John causes an accident 00 Æ causer (x, x=Jean, y, y=un accident) Let us then consider the data which relate to the subjunctives in French. In situations of wide focus, inversion is optional without side effects, as shown in (34). As subjunctive is the licensing property of inversion, the same type of inversion is ungrammatical with indicatives, as illustrated in (35). (34) (a) (b) (c) (35) (a) (b) (c) Que veux-tu? What do you want Je veux que Paul vienne. I want that Paul comes. Je veux que vienne Paul. I want that comes Paul Que dis-tu? What do you say Je dis que Paul viendra I say that Paul will-come * Je dis que viendra Paul I say that will-come Paul In cases of narrow focus however, the pattern that we saw above with contrastive focus re-occurs. (36) A. B. Qui veux-tu qui vienne? Who do you want that comes? (a) J’aimerais bien que vienne Paul. I would like that comes Paul (b) ?? J’aimerais bien que Paul vienne.(note 7) I would like that Paul comes (c) J’aimerais bien que PAUL vienne. I would like that PAUL comes. Narrow focus causes either an accent or stylistic inversion. Like contrastive focus, it is incompatible with a preverbal un-accented NP subject. It is relevant to note that this pattern is also dependent on the subjunctive, as shown in (37). (37) A. B. Qui crois-tu qui a dormi? Who do you think that has slept * Je crois qu’a dormi Georges. I believe that has slept Georges. Thus, although narrow focus may entail inversion, it may only entail inversion in constructions in which it is otherwise allowed to invert. One cannot say that narrow focus triggers inversion in French embedded clauses. Narrow focus triggers inversion only if the other properties of the clause allow for inversion. Finally, consider the data for contrastive focus in an otherwise narrow focus context. The patterns observed above are confirmed. (38) (a) Qui croit-elle qui a dormi? (b) (c) (39) (a) (b) (c) Who does she think that has slept * Elle croit qu’a dormi Georges She believes that has slept Georges * Mais non, elle croit qu’a dormi Jacques / JACQUES No, she believes that has slept Jacques Qui veut-elle qui vienne? Who does she want that comes Elle veut que vienne Paul She want that comes Paul Mais non, elle veut que vienne Raoul. No, she wants that comes Raoul. Contrastive focus may entail inversion, but it is not possible to say that in French contrastive focus causes inversion. It only causes inversion if inversion is otherwise allowed. 2.3. Conclusions In embedded clauses the basic properties of the constructions in root clauses are identical. The observations and generalizations made in section 1.4. do not deserve further qualification and in fact are simply confirmed. This does not entail that there are no issues in embedded clauses. One of the standard syntactically problematic issues is that Stylistic Inversion is generally considered to be optional, not directly in root clauses, but generally in embedded clauses. In this domain we have paid particular attention to the pragmatic function of otherwise definite NPs. The conclusion is that French has a possibility of syntactically encoding focus in the postverbal position. But this possibility is severely restricted. The possibility may only be instantiated if the sentence itself allows for inversion for independent reasons; the contrast between subjunctives and indicates is the source of this observation. In indicatives focus has no influence on word order, because the indicative does not allow for the inversion in the first place. Subjunctives do supply a proper environment for inversion. When the precondition of allowing for inversion is satisfied, inversion is favored in situations of both narrow and contrastive focus. In wide focus inversion it is not favored, such that it is completely optional. An alternative for inversion is heavy accenting of the preverbal position, while the inverted position may additionally receive this heavy accent. In cases of narrow and contrastive focus, using the preverbal position with neutral accent is unfelicitous. There are a number of issues for future research, besides the linking between the issues of indefiniteness of section 1 with the focus issues for definites of section 2 (that is, the behavior of indefinites vs. definites under focus). We mention two here, relating to the data of focus. In (39) we gave examples with ergative verbs, which are uncontested. In the area of agentive verbs, there is a relatively strong judgment disagreement between Kampers-Manhe (1998) and Marandin (2000a) (note 8) (40) Je voudrais que travaille Marie (a: *, b: ok) I would like that works Mary Above we gave the data for narrow and wide focus, but there also exists ‘intermediate’ focus, defined as the situation in which the VP without the subject is new information: (41) (a) (b) (c) Que veux-tu que Pierre fasse? What do you want that Peter dos Je veux que Pierre travaille I want that Peter works * Je veux que travaille Pierre. I want that works Peter. Hence, these seem to point at the relevance of another property, in this case Topic. The pattern is neither identical to the pattern in narrow focus (where inversion is strongly preferred), nor to the pattern in wide focus (where inversion is fully optional). Topichood could block inversion also when it is otherwise licensed. 3. Conclusion Postverbal subject position in French illustrate an unexpected wealth of semantic and pragmatic differences. A statement that French is a language with a rigid word order that has no direct way of encoding semantic and pragmatic differences would appear to be too strong. References Abeillé, A. & D. Godard, 2000. “Construction impersonnelle”. Ms. Cited after Marandin (ms.) Cornulier, B. de, 1974. “Pourquoi et l’inversion du sujet non clitique”. In: C. Rohrer & N. Ruwet (eds), Actes du colloque franco-allemand de grammaire transformationnelle, Vol. 1, Niemeyer, Tubingen, pp. 139-163. Drijkoningen, F., 1997. “Morphological Strength: NP positions in French”. In: H. van Riemsdijk et al. (eds), Rightward Movement. Benjamins, Amsterdam. Drijkoningen, F., 1999. “Past Participle Agreement in French; AGRO? AGRA!”. In: R. van Bezooijen & R. Kager (eds), Linguistics in the Netherlands 1999. Benjamins, Amsterdam, pp. 41-52. Hoekstra, T., & R. Mulder, 1990. “Unergatives as copular verbs; location and existential predication”. The Linguistic Review 7, pp. 1-79. Hulk, A., 1982. Het clitisch pronomen “en”: een dwarsdoorsnede van de Franse syntaxis. Ph.D. Dissertation, Utrecht University. Kampers-Manhe, B., 1998. “Je veux que parte Paul: A neglected construction”. In: A. Schwegler, B. Tranel & M. Uribe-Etxebarria (eds) Romance Linguistics: Theoretical Perspectives. Benjamins, Amsterdam, pp. 130-141. Kayne, R., & J.-Y. Pollock, 1998, ms. (unpublished). “New thoughts on Stylistic Inversioin”. Both a handout at the Amsterdam conference on Inversion in Romance (1998), and a paper to appear in a volume edited by A. Hulk, to be published by Oxford U.P. Kupferman, L., 1984. “Syntaxe et conditions pragmatiques”. Linguisticae Investigationes VII/2, pp. 385-400. Legendre, G., 1998. “Focus in French Stylistic Inversion”. Handout of the paper presented at the Amsterdam conference on Inversion in Romance (1998). Marandin, J.-M, 2000a. “Unaccusative inversion in French”. Handout of a paper presented at the PICS meeting in Paris in May, 2000. [related to Marandin (to appear) “Unaccusative Inversion in French”, in Selected Papers of Going Romance 1999]. Marandin, J.-M, 2000b. “Unité et diversité de l’inversion en français”. Ms. CNRS/Paris 7. Miller, P., 1991. Clitics and constituents in phrase structure grammar. Ph.D. Dissertation, Utrecht University. Rizzi, L., 1997. “The fine structure of the left periphery”. In: L. Haegeman (ed), Elements of grammar. Handbook of Generative Syntax. Kluwer, Dordrecht, pp. 281-337. Notes 1. This paper has been presented at the workshop on “Information Structure in French” organized in the framework of PICS, in Amsterdam, December 2000. This work is still in progress: the text reproduced here is close to the oral version of December 2000, and does not always fully integrate the new data and discussions that have arisen since. The goal of PICS is to come to a rather theory independent overview of the relevant data. 2. For HPSG (a model for which we are not specialists) what is said here seems to favor an analysis of the Miller (1991) type over that of Abeillé & Godard (2000) – in which the impersonal subject looks as treated as the only subject, and the second (NP) subject as ‘demoted’ to object status. That is, as far as the agreement patterns are involved. 3. We are not aware of an analysis of participle agreement in HPSG. 4. An exception must be made for the use of this construction in a list. The sentence would be all right in an enumeration of different places where celebrities died: A Paris mourut Louis XVI; A New York John Lennon, …. 5. Nevertheless, the use of stylistic inversion in (25b) remains ungrammatical. This issue is left open for further research. 6. Our judgments on the ungrammaticality of (25d) indicate that the contrast that is central in the proposal by Kayne & Pollock (1998) may be less severe in embedded clauses. 7. In fact, we feel that these are ungrammatical. The speakers at the workshop did not fully agree with the judgment ‘plainly ungrammatical’. 8. The whole discussion group of the workshop remained consistently divided on that point during the discussion.