Object Clitics in Spanish and Catalan

OBJECT CLITICS IN SPANISH AND CATALAN Francesc Roca Universitat Autbnoma de Barcelona This paper studies the properties of object pronominal clitics in Spanish and Catalan and shows that they behave very differently in several ways. Contrary to what is generally asumed, we propose that direct object clitics and indirect object clitics are two different kinds of functional categories: the direct object is considered a DP generated in an A-position, and the indirect object clitic is analyzed as a functional category with Person features that acts as a sort of dative marker rather than as an element in au A-position. This analysis covers the main properties of each clitic and it is free from the main problems that usually affect previous hypotheses. 1. Introduction Within the Government and Binding framework, several proposals have been made to explain the nature and behaviour of pronominal clitics in Romance languages. From the syntactic point of view, this discussion can be reduced to the following three hypotheses, very briefly reviewed: the 'Movement Hypothesis', that assumes that the clitic is generated in an A position and that it moves towards the functional head occupied by the verb (see Kayne (1975, 1991)); the 'Affix Hypothesis', that suggests that the clitic is base-generated next to the verbal form (see Jaeggli (1982, 1986) and Borer (1983)); and the 'AGR Hypothesis', that puts forward that pronominal clitics and subject agreement morphemes are the same kind of element and that both head an AGRP (see Suñer (1988) and FernAndez Soriano (1989)). No matter what hypothesis on clitics we adopt, I0 and DO clitics are usually analyzed as if they were entirely equivalent. Although there is no doubt that these elements have a lot of things in common, it is also true that in some languages they do not behave exactly alike. When noted, these differences have been usually attributed to a lexical feature or to a very idiosyncratic Catalan Working Papers in Linguistics (CWPL) 1992: 245-280 Universitat Autbnoma de Barcelona property of one of the clitics (see Suñer (1988)) or to the properties of the argument that it represents (see Jaeggli (1986)), but never to the possibility of having a different status for each clitic. The only exception is Torrego (1990), who tries to acwunt for some differences between Castilian and American Spanish in tems of a distinction between the two object clitics.' This paper is organized as follows. Section 2 is devoted to the study of object clitics in Spanish and Catalan. We will see that I0 clitics differ from DO clitics in several ways and that, although they share some properties, it is possible to anribute a different status to each clitic. In section 3 we suggest an analysis that can acwunt for these differences. Mat of the data and constructions examined here belong to Catalan and Spanish and refer only to direct and indirect object clitics, but some of the observations can be extended to other related Romance languages too. 2. The Properties of Object Pronominal Clitics in Spanish and Catalan We think that in Spanish and Catalan there are some differences between the two object clitics that clearly call for a distinction in their status. Now, let us consider these differences, that any theory on clitics should be able to account for. At the moment, we limit ourselves to presenting them; we will not argue for or against any hypothesis. 2.1. Overt +features It is usually assumed that a pronominal, even if phonologically null, has some +-features. These features include the person, gender and number specification, as we can see in the socalled personal or strong pronouns: 1 This reference corresponds to a conme in a summer school, but in fact I think that this idea is being developed in recent work in progress. Unagereka (1992) also seems to accept some of the ideas of Torrego. (1) YO UP sgl nosotros [IP P~I ni [2~ sgl vosotros PP P~I 4 [3~ sg ml ellos 13~ ~1 ml ella [3~ sg fl ellas [3~ PI fl Pronominal clitics also reflect these features, but as (2) shows, while DO clitics have the same features as strong pronouns, 3rd person I0 clitics do not show gender distinction: (2) DO clitics: I0 clitics: me [lp sg] nos [lp pl] me [lp sg] nos [lp pl] te [2p sg] vos [2p pl] te [2p sg] vos [2p pl] 10 [3p sg m] la [3p pl m] le [3p sg] les [3p pl] la [3p sg fl las PP pl fl Given that as 1st and 2nd person object clitics are formally identical, from now on we will pay attention only to 3rd person clitics, the ones which show more differences. There are also other elements with +-features such as the different determiners and the possessive pronouns. Among these determiners, we call your attention on the definite article, that coincides with some object clitics in the specification of the $-features and even in its morphophonological form: (3) el [m sgl 10s [m sg] la [fsg] las [fsg] This parallelism can be seen as a first piece of evidence for analyzing the definite article and some object clitics as the same kind of element, that is as determiners that head a DP. In fact, this is what we will suggest and develop in the following sections. 2.2. Clitic-NP Doubling As is well-known, the clitic-NP doubling constructions are sentences in which a pronominal clitic and an NP in the argument position that the clitic refers to co-appear. The two object clitics differ clearly in this respect: whereas clitic-NP doubling is always possible with indirect objects, it is not always allowed with direct objects. The following examples show that the I0 clitic can double any kind of NP in dative position: (4) a. LuislediounlibroaMaria L. 3pDat gave a book to M. 'Luis gave Maria a book.' b. Les expliquk 10 sucedido a unos policías. 3pDat (I) told the happened to some policemen 'I told what happened to some policemen'. (5) a. Le duele la cabem a Juan. 3pDat hurts the head to J. 'Juan has a headache.' b. Le hice un tmje a Luis. 3pDat (I) made a suit to L. 'I made a suit for Luis.' c. Luis siempre le ha sido fiel a su esposa L. always 3pDat has been faithful to his wife 'Luis has always been faithful to his wife.' In (4) there are NPs2 with different inherent features and they all allow clitic-doubling. The direct object counterparts of these cases are clearly ungrammatical in peninsular Spanish: - 2 The categorial status of dative arguments is a controversial point. Here we assume that they are hTs, not PPs (see Branchadell (1992) and references cited there). (6) a. *LuislavioaManía. Luis her saw to Maria b. *Los expliquc? unos cuentos. them (I) told some tales (Obviously, these sentences are grammatical without the clitic.) The examples of (5) involve datives that are not subcategorized by the verb (see Branchadell (1992) for an analysis of the similarities and differences between subcategorized and nonsubcategorized datives). Here, there is no possible comparison with direct objects because these are aiways arguments of the verb. In addition to this, the I0 clitic can double any instance of Wh-phrase, while this is absolutely out in the case of direct objects in Catalan or peninsular Spanish and even in Amencan dialects that allow some instances of clitic-NP doubled direct objects: (7) a. ¿A quiCn le diste el libro? to whom 3pDat (you) gave the book To whom did you give the book?' b. hta es la chica a la que le regal6 rosas. this is the girl to the that 3pDat (I) gave roses This one is the girl to whom I gave roses.' c. A MARIA le di un beso. to M. 3pDat (I) gave a kiss 'It is Maria that I kissed.' (8) a. *¿A qui611 10 viste ayer? to whom him (you) saw yesterday b. *&tas son las rosas que se las regaic? a Maria. these are the roses that 3pDat them (I) gave to M. (8) c. *A MARIA la besé. to M. her kissed This contrast clearly shows that DO doubling is more constrained than I0 doubling. 2.2.1. Clitic-NP Doubling of Strong Pronouns. In the preceding paragraphs we have seen that the DO clitic NP-doubling is impossible in a large number of cases. This does not mean, however, that we cannot find any occurrence of this kind of doubling. As a matter of fact, in Catalan and peninsular Spanish a DO clitic can double a direct object when it is a strong pronoun. In this case the clitic is not only allowed but also required since its absence causes the ungrammaticality of the sentence. Here, the indirect object behaves exactly alike: (9) a. Lo vi a 61. (10) a. Le devolví el libro a ella. him(acc) (I) saw to he 3pD (I) returned the book to she 'I saw him.' 'I returned her the book.' b. *Vi aCI. b. *Devolvi el libro a ella This behaviour can be explained by attending to the nature of strong pronouns rather than to the properties of the two object clitics. Rigau (1988) observes that, among pronominal elements, clitics and empty pro act in a parallel way and are very different from strong pronouns. She points out the ability to act as a resumptive pronoun (ll), and the impossibility of strong pronouns to refer to a left dislocated constituent (12) or to count as variables bound by a quantified phrase (13): (1 1) a. Aquest Cs el nen que diuen que li van regalar un cavall. this is the boy that say that 3pD (they) PAST give a horse This is the boy that they say that they gave a horse to him.' b. *Aquest és el nen que diuen que li van regalar un cavall a ell. (12) a. A en Pere, li van regalar un cavall. to the Pere, 3pD (they) PAST give a horse 'Pere, they gave him a horse.' b. *A en Pere, li van regalar un cavall a ell. (13) a. Tothom; sap que proj es divertila everybody knows that pro have-fun-FUT 'Everyone knows that he will have fun.' b. *Tothom; sap que elli es divertih. In (lla) the clitic can be involved in the resumptive pronoun strategy and can act at LF as a variable bound by the operator of the relative clause, but in (1 lb), when it is related to a strong pronoun, it cannot. This would show that these strong pronouns cannot serve as logical variables at LF. Similarly, in (13b) the presence of the strong pronoun blocks the bound reading that pro, and aiso a clitic, allows. Moreover, she also notes that strong pronouns do not occupy the same syntactic position as other NP arguments. This can be seen in the following contrast: (14) a. Vam acostumar el nen a aixb. (we) PAST get-used the boy to this 'We got the boy used to this.' b. *El vam acostumar a ell a aixb. 3pAcc (we) PAST get-used to he to this c. El vam acostumar a aixb a ell. (15) a. Consideren en Pere molt intel.ligent. (they) wnsider the P. very intelligent They consider Pere very intelligent.' (15) b. Us consideren molt intel.ligents a vosaltres. 2pDat (they) consider very intelligent-pl to you-pl They consider you very intelligent.' c. *Us consideren a vosaltres molt intel.ligents. In (14) the verb acostumar 'get used' selects an NP and a PP, but if the direct object is a strong pronoun the order DO-PP is ruled out and the only possibility for the pronoun is to appear at the right of the PP. The same happens in (15): in (15a) the NP en Pere is placed in the Aposition, the subject position of the small clause, but the strong pronoun cannot stay there as (1%) shows. Consequently, Rigau claims that strong pronouns do not occupy an A-position and that they are in a peripheral position, an A'-position external to the VP, that can free them from becoming bound elements at LF and can explain their S-structure position. The presence of these pronouns would be licensed through coindexing with a pro or a clitic chain. So, under this view, the A-position of these constructions would be occupied by thispro and the strong pronoun would remain in a peripheral A'-position. Picallo (1991) also observes the same differences between possessive pronouns and strong pronouns inside nominal phrases. She assimilates the possessive pronoun in nominals to the empty pro in sentences and distinguishes these two elements from strong pronouns. She gives the following contrasts concerning proximate interpretation (16), quantifier binding (17) and denotative properties (18): (16) a. La Mariai diu que en Perej truca sovint a la sevai/j mare. the M. says that the P. phones often to the herlhis mother 'Maria says that Pere often phones to herlhis mother.' b. *La Mariai diu que en Perej truca sovint a la mare d'ellilj. (17) a. El temor de tot acusatj al seui fiscal. the fear of every accused to-the his Public Prosecutor The fear of every accused to his Public Prosecutor.' b. *El temor de tot acusati al fiscal d'elli. (18) a. La desaparici6 de les llibretesi. the disappearance of the notebooks b. La seva¡ desaparicib. the its disappearance C. *La desaparici6 d'ellesi. the disappearance of them The facts in (16)-(17) are identical to those of (1 1)-(13), and what (18) shows is that strong pronouns are more restrictive than pro or possessives: whilepro or possessives can denote any kind of object or set, strong pronouns can denote only denumerable or [+humanlanimate] entities. This is the reason for the ungrammaticality of (I&), where the pronoun elles 'them' should refer to the [-mim] NP les llibretes 'the notebooks'. In (18b) a possessive is used and there is no problem. If we accept this analysis, and we really accept it, then we have a plain explanation for the need of the clitic in (9) and (10). In these constructions the obligatoriness of the clitic would follow straightfonvardly from the Projection Principle: given that the strong pronoun occupies an A'- position, the A-position must be filled by some element in order to keep the 8grid of the verb up; this element will be the clitic, or the pro licensed by it, that in its turn can serve as licenser of the peripheral strong pronoun. 2.2.2. Bare NPs in I0 Position. Turning to I0 clitic doubling, it has been noted that the only NPs that do not allow it are bare NPs, which have a non-specific reading. In this respect, Fernández Soriano (1989) gives the following e~amples:~ This has also been observed by Jaeggli (1982: 59. fn. 39). who gives these examples: that, as a consequence, they can be linked only to NPs that are compatible with these features. We suggest that, according to its similarity with the definite determiner, the DO clitic in Spanish and Catalan has the same features as this determiner, and it is interpreted as [+defJ. The same criterion will be applied to the Catalan clitic en, but with the slight difference that this element seems to receive a partitive interpretation. Thus, the ungrammaticality of (22) and (27) in Spanish and Catalan follows from the fact that there is no DO clitic form compatible with an indefinite NP: both the [+deu pronominal clitic, the only one that Spanish has, and the partitive clitic of Catalan have different features and by no means can be related to this kind of NP. Then, given the absence of the clitic, the above requirement for topicalized constructions is not fulfilled -there is an empty A-position and the Projection Principle is not preservedand the sentence is ruled out. Under this view the most puzzling case is the difference between indefinite and bare NPs in Spanish. We have stated that, in Spanish, dislocated bare NPs produce a null object construction. So we could ask ourselves why the same strategy is not allowed with indefinite NPs. We are not going to pursue this matter further; we will simply note that this null object strategy seems to be the Spanish counterpart of the Catalan constructions with the clitic en.6 Under these assumptions, we can easily capture the strong parallelism between Spanish and In fact this seems to be the case, at least as far as dislocated structures are involved, as shown by the following examples with PP complements of a verb (i a) and genitive arguments (i b), the other uses of en in Catalan: (i) a. De política. ya hablaremos mañana. Spanish of politics. already (we) talk-FUT tomomw We will talk about politics tomorrow.' a'. De política, ja en parlarem dema. Catalan b. De este libro, me he leído la primera parte. Spanish of this book, lpDat (I) have read the first part 'I have read the first part of this book! b' D'aquest llibre, me n'he llegit la primera part. Catalan Catalan concerning these structures and the character of DO clitics as opposed to I0 clitics, that can be related to any kind of NP and, in this sense, seem to behave like the subject agreement morphemes, which are not restrictive on the nature of NPs either. Turning to the features of the DO clitic, there are sentences like the ones of (29) that offer further evidence for the claim that the DO clitic is inherently marked as [+defJ: (29) a. Los vimos a todos. 3pA (we) saw to all 'We saw all of them.' b. *Los vimos a algunos. 3pA (we) saw to some 'We saw some (of them).' The quantified phrase todos 'all' in (29a) is interpreted as definite and the clitic-NP doubling structure is allowed. However, in (29b) the indefinite quantifier algunos 'some' does not admit the presence of the DO clitic. We think that this is a clear proof for the [+defl feture of this clitic and against the possibility of linking it to indefinite phrases. Before leaving this, let me point up two more structures that show that DO clitic forms are sensitive to the nature of the argument they are referring to. 2.3.1. Cliticization of Nominal Predicates in Catalan. According to normative Catalan, in nominal predicates (that is, constructions where two noun phrases and the copulative verb ser 'to be' are involved) two different pronominal clitics can be used: the neutral forrn ho 'it', used also for the direct object when it is neuter or a subordinate clause, and the usual DO clitic forms. The distribution of these two clitics is clearly defined: while the neutral form can pronominalize any instance of nominal predicate -an adjective, a prepositional phrase or a noun phrase-, the DO clitic forn must be used when this nominal predicate is a definite NP. Let us consider the following examples, from Fabra (1956): (30) a. En Pere 6s mestre. the P. is teacher 'Pere is a teacher.' a ' En Pere ho 6s. b. En Pere 6s el mestre dlEspot. the P. is the teacher of Espot 'Pere is the teacher of Espot.' b ' En Pere 1 '6s. In (30a) the nominal predicate mestre 'teacher' is an NP, but given that it is not interpreted with a [+defl value, it cannot be pronominalized by the [+defJ clitic and the neutral clitic is used. On the other hand, in (30b) the nominal predicate is a clearly definite NP and the DO clitic is required. This would show again that the DO clitic really has a [+defJ feature and that it can only refer to definite NPs. 2.3.2. DO Clitic-NP Doubling constructions in Porteh Spanish. As is well known in the literature (see Jaeggli (1982, 1986) and Suñer (1988, 1989)), there are some American varieties of Spanish, like Porteño, that allow some instances of DO clitic-NP doubling. The possibility of having this kind of doubled structures seems to be related to the definite nature of the NP in the direct object position, as the following examples show: (31) a. La vi a Mafalda her (I) saw to M. 'I saw Mafalda.' b. *Lo vi un carni6n. it (I) saw a truck Moreover, Suñer claims that these constructions are regulated by a Matching Principle (MP), that ensures that the features of the clitic and the doubled NP agree, and puts fonvard that the DO can only be doubled if it is interpreted as [+spec]. She gives the following exarnples, that involve DOS that are not preceded by the preposition a 'to' and Wh-phrases: (32) a. Yo la tenia prevista esta muerte. I it had foreseen this death 'I foresaw this death.' b. ¿A cuántas de ellas las interrogaron? to how-many of them them (they) questioned 'How many of them did they question?' Usually, Wh-phrases in DO position are not doubled by a clitic, but she points out that they can be doubled if they are interpreted as [+spec], as in (32b). Briefly, her idea is that any instance of [+spec] direct object can be doubled by a clitic because the DO clitic is lexically marked as [+spec]. On the contrary, a [-spec] NP cannot be doubled because this feature will clash with the [+spec] of the DO clitic and the MP would be violated. The doubling of an I0 is always possible because the I0 clitic is not inherently marked as [+spec] and, consequently, the MP is satisfied. To sum up, in this section we have seen that, concerning the kind of NPs they can be related to, the DO clitic is more restricted than the I0 clitic. More precisely, we have put fonvard that the DO clitic, but not the I0 clitic, is inherently marked as [+defJ -or, maybe, [+spec] in Suñer's viewand that this is the reason why they behave differently in this respect. Evidence for this comes from left-dislocation and clitic-NP doubling structures in peninsular and Porteño Spanish. 2.4. Pronominalization of Other Arguments While the I0 clitic can refer to dative complements that are not arguments of the verb -i.e., the non-subcategorized datives-, the DO clitic is clearly restricted to express this verbal argument. We think that this clear difference is actually interesting and that it can be enforced, at least in Spanish, by the ability of the I0 clitic to pronominalize an intemai argument of the verb that by no means can be considered a dative. Look at the following sentences, borrowed from Hernanz-Brucart ( 1987): (33) a. Las aiumnos se rien de Maria. the students se laugh of M. The students laugh at Maria.' b. El ladr6n se escap6 de la policia. the thief se escaped of the police The thief escaped from the police.' c. Se apiadaron de 61. se (they) took-pity of him They took pity on him.' In these cases the internal argument of the verb is a PP headed by the preposition de 'of. None of these complements can be believed to be a dative argument; they are rather instances of what traditional grammars call prepositional complements selected, or governed, by the verb. However, when these arguments pronominalize, the dative clitic is always used, there is no altemative: (34) a. Lm alumnos se le rien. b. El ladr6n se le escap6. c. Se le apiadaron. The pronominalization of these arguments is very different in Catalan. In this language, the partitive clitic en is used to express bare NPs in object position and internal arguments headed by the preposition de. This is precisely the case of the constructions we are dealing with, and, as expected, the partitive clitic is required and the dative form is wmpletely out: (35) a. Els alumnes se n'e~uen (de la Maria). the students se of-her laugh (of the M.) The students laugh at her.' b. El lladre se n'ha pogut escapar (de la policia). the thief se of-it has wuld to-escape (of the police) The thief could escape from it.' c. Se n'han penedit (d'en Pere). se of-him (they) have taken pity (of the P.) They have taken pity on him.' (36) a. *Els alumnes se li enriuen. b. *Se li han penedit.7 The case of the verb escapar-se 'to escape' is especial since it allows both kinds of pronominalization: partitive en (see (54b)) and I0 clitic (see (i)). (i) El lladre se'[lzi] ha escapat (als policies). the thief se -3pplDat has escaped (to-the policemen) The thief escaped from them.' As the bracketed phrase indicates, the 'a + NP forn is allowed. Note also that this behaviour of the dative clitic is only possible with pronominal verbs both in Spaaish and Catalan. As shown in (ii), the non pronominal verb escapar does not allow the I0 clitic: The main difference between these examples and the non-subcategorized datives lies in the fact that now the I0 clitic seems to refer to a PP with the form 'de + NP', whereas in the other cases it reproduces a phrase with the form 'a + NP, the usual form of dative arguments. This seems to suggest that the Spanish dative clitic is doing the work of the partitive clitic in Catalan, at least with these PP internal arguments. However, this is not true because, leaving aside the fact that this dative cliticization strategy is also possible in certain cases in Catalan, there are similar PP arguments that do not allow the dative clitic in Spanish. Thus, we will prefer to state simply that the presence of the I0 clitic in these structures shows its ability to express severa1 kinds of arguments. This has severa1 interesting consequences concerning the way how this arguments can be represented. As Hernanz-Brucart (1987) point out, the presence of this clitic interacts with the possibility of having an 'a + NP' complement, as in (37): (37) a. A Maria se le ríen en clase. b. Se le escap6 el ladr6n a la policia. c. ?A Juan se le apiadaron. This possibility can be seen as a 'dative-like' feature that precisely correlates with a property of non-subcategorized datives that we have seen above: the presence of the clitic is obligatory in these cases. As (38) shows, if the 'a + NP' phrase is maintained, the absence of the clitic leads to ungrammaticality. (38) a. *A María se nen en clase. b. *Se escap6 el ladr6n a la policía. c. *A Juan se apiadaron. (ii) a. El ladr6n escap5 de la policia. the thief escaped of the police The thief escaped from the police.' b. *El ladr6n le escap6. Probably, this presence of the dative clitic is related to the properties of this pronominal se. 2.5. Invariant le This phenomenon consists of the use of the singular form le to refer to a plural dative, that is, when the plural les is expected. This can be seen in the following constructions: (39) a. NO lei tenia miedo a las balasi. not 3psgDat (he) had fear to the bullets 'He was not afraid of bullets.' b. No darlei importancia a 10s detalles; ... not to-give-3psgDat relevance to the details 'Not to give relevance to details ...I This property of I0 clitics has been already noted by various traditional grammarians (see Marca Marín (1978) and references cited there), and they all agree that this use of the singular form is a widespread phenomenon. All the occurrences of le in (39) are instances of clitic-NP doubling structures. This means that the defective clitic is simply advancing the presence of a dative argument immediately represented as a full NP, and that, probably, this is the reason why this defectiveness is allowed. In other words, as long as the dative argument is immediately identified, the specification of all its features does not appear so necessary. However, there is a certain controversy because, while there are sentences whose indirect object is expressed only by the invariant form (see (Na-b)), it seems that when the indirect object goes at the begiming of the sentence the defective forn is not allowed (4Cc): (40) a. Ellos; ... peen frente al desprecio que Cste lei inspira ... they ... have before to-the scorn that this-one 3pD incites They ... have before the scorn that this one causes in them ...I (40) b. (ellosi) ... Por temor a que nuestro contacto con 10s indios lej acarrease (they) ... by fear to that our contact with the Indians 3pD cause algdn tip de enfermedad ... some kind of illnes I... because they were afraid that our contact with the Indians could cause them some illness ...I c. A los niños les /*le dije que ... to the boys 3pplD/3psgD told that 'I told them that ...' Although there is no clear explanation for this, we think that this property can be seen as a step towards a stage where this pronominal clitic will have lost all its argumental properties and will appear as an element that simply announces the presence of certain arguments, as a sort of dative marker. However, no matter what analysis is the right one, we would like to point out that this behaviour is possible only with the I0 clitic. There is no occurrence of invariant DO clitics in Spanish, not even in some laísta dialects, which use the DO clitic form to express both accusative and dative arguments. If both object pronominal clitics were basicaly the same kind of element, we would expect them to behave exactly alike in this respect. Since this expectation is not borne out, we have another piece of evidence to distinguish the status and properties of the two clitics. In conclusion, in this section we have explored some clear differences between I0 and DO pronominal clitics in Spanish and Catalan. These differences concern the overt and inherent features of the clitics, that determine the class of NPs they can denote, the clitic-NP doubling structures, and the relation with certain internal arguments and non-argumental complements. In the preceding sections, we have observed that the DO clitic has more features and undergoes I more restrictions than the I0 clitic. These restrictions can be related to the properties of the I definite determiner, which heads a DP. Consequently, we can assume that the DO clitic is a DP very close to the definite determiner. In its turn, the I0 clitic seems to be freer concerning these aspects: it is not specified for so many features as the DO clitic -sometimes, it even appears as a defective fom-, and its relations with severa1 kinds of NPs and arguments are not so restricted. This could lead us to assume that it is an AGR head. However, we must be very careful with such a statement, because this would mean that the I0 clitic is closer to a subject morpheme than to a pronominal clitic; but we should not forget that there are several syntactic properties shared by the two object clitics such as its position at S-structure, the behaviour in coordinate and clitic climbing structures, etc. Maybe these properties of the I0 clitic follow from the fact that it is a DP or an AGRP with very special properties that can provide an answer for the relations between a true pronominal element and an inflectional morpheme with pronominal features. This is what we will explore in the next section. 3. An Analysis In this section we would like to suggest a possible analysis of object clitics that accounts for its main properties and for the differences between the two object clitics seen in the preceding section. What we can conclude so far is that, at least in Catalan and peninsular Spanish, the DO clitic is much closer to the definite determiner el, la 'the', while the I0 clitic seems to exhibit a different behaviour that keeps it away from the typical properties of definite deteminers. Bearing this in (46) b. *(Le) rompi el brazo a Maria. 3pDat (I) broke the arm to María 'I broke Maria's arm.' In (46b) we have a non-subcategorized dative and the sentence without the clitic is out. On the other hand, the subcategorized dative of the verb devolver 'to give back' in (46a) can be doubled by the clitic or not with no consequences in grammaticality. Probably, we could account for this contrast if we assume, as Torrego (1991) suggests, that oniy verbs that subcategorize for an indirect object can assign case to this argument. If this is true, we can explain the preceding contrast in the following way: in (46a) the verb devolver 'to give back' subcategorizes a dative and, consequently, it assigns case to the indirect object a Luis 'to Luis' and the presencelabsence of the clitic is not relevant from the point of view of case assignment; however, in (46b) the verb romper 'to break' does not subcategorize a dative argument and cannot give case to the dative phrase a Maria 'to Maria', that in order to avoid a Case Filter violation should get case from some other element, narnely the dative clitic. An interesting issue is to state the way the dative clitic can transmit the case to these non-subcategorizeú datives. In this point, we follow again Torrego and suggest that the NP moves to the specifier of the projection headed by the clitic, and that it gets case through specifier-head agreement; exactly in the same way as Nominative case is assigned to the subject. In the last paragraph, we have seen what happens in the instances of clitic-NP doubling and when the dative clitic is not present. Now, let us explore another of the properties of the clitic: the identification of pro when there is no overt NP in the A-position. We will assume that the dative clitic also identifies apro via specifier-head agreement. In this case the pro would be base-generat4 in the A-position where it receives its 8role and would move to the specifier of the head occupied by the clitic in order to be identified: This derivation is very close to the derivation of a subject pro, with the further parallelism that the +features are person and number in both cases. Then, we could explore the similarities between the dative clitic and the AGR,,bj head in the sense that both assign case to an argument. If this is true, the Dative case would be similar to Nominative and, since they are assigned under the same structural configuration, both could be considered structural cases. In this respect, it would be interesting to study the properties of the Person and Number features of this functional head, as Rigau (1991) does within the AGR projection. Rigau concludes that the Person feature is the Nominative Case assigner. We are not going to develop this question, but we simply put fonvard that we can expect the Person feature to be the structural Case assigner also in this case-remember that the presence of the Number feature is not always necessary in this 'dative marker' (see section 2.5). Another interesting property that distinguishes DO and I0 clitics is the ability to express indefinite arguments. In the case of direct object clitics this has been accounted for by assuming that they are true definite determiners. In the same way, since indirect object clitics can be related to any kind of indefinite, we should expect them not to be specified for any definiteness feature. This is precisely what happens in agreement systems such as subject agreement, for instance. Moreover, as Uriagereka (1992) points out, clitic systems never start with indefinites, but they can evolve and get grammaticalized into agreement systems where indefinites are perfectly possible. If this is true, then we could consider the DO clitics a typical clitic system and the I0 clitics a step towards a paradigm of agreement. This idea can be enforced by the similarities between subject agreement and I0 clitics, but we should not forget that an indirect object clitic is not an agreement morpheme, and that it has several properties, from its morphological form to its syntactic behaviour, that make it closer to a pronominal clitic than to an affix. So, we must make sure that, whatever these dative markers are, they behave first of all as pronominal clitics, and that all the properties related to agreement systems are compatible with this behaviour. 4. Summary In this paper we have suggested an analysis that distinguishes the two pronominal object clitics in languages such as Spanish and Catalan. The need for this distinction follows from the fact that they have different properties and that, consequently, they behave differently in certain cases. In section 2 we have explored these differences and we have concluded that the DO clitic has more overt and inherent features than the I0 clitic and that this is the reason why the DO clitic can only denote a restricted set of NPs: those with a [+def] or [+spec] interpretation. We have also pointed up that the I0 clitic appears in some syntactic configurations, such as clitic-NP doubling structures, in which the DO clitic is usually impossible. In section 3 we have tried to attribute these restrictions of the DO clitic to the different status of each element. In agreement with this, we have proposed that direct object clitics are actually determiners that head a DP. We have also put forward that this D is the same definite determiner that we find preceding NPs, APs, PPs or CPs with the only difference that it takes a pro as its complement. In its turn, the I0 clitic is analyzed as the head of a functional category different from the DP of direct objects. The properties of this functional head will be the responsible for the clitic-NP doubling structures, the inherent features, and the relation with indefinites and several kinds arguments, three possibilities of the I0 clitic that the DO clitic does not have. For the time being, we do not go deeper into the definition of the syntactic derivation that such a distinction between the two object clitics involves. We leave this question open, but we would like to point out that, although at first glance this analysis seems a bit strange, a derivation according to some of the conditions on clitic movement of Kayne (1991) would probably give the right results. Obviously, we will need to modify some of his criteria and to make some new assumptions. Many thanks to my colleagues at the UAB Syntax Seminar for their comments and suggestions. This research has been supported by a DGICYT grant PB89-0324 awarded to the UAB. References Abney, S. (1987) The English Noun Phrase in Zts Sentential Aspect, Ph. D. dissertation, MIT. Borer, H. (1983) Parametric Syntax, Foris Publications, Dordrecht. Branchadell, A. (1992) A Study of Lexical and Non-lexical Datives, Ph. D. dissertation, UAB. Fabra, P. (1956) Grarmitica catalana, Teide, Barcelona. FemAndez Soriano, 0. (1989) Reccidn y Ligamiento en Espaiiol: Aspectos del Parámetro de Sujeto Nulo, Ph. D. dissertation, UAM. Hemanz, M.L. and J. M. Brucart (1987) La sintaxis, Crítica, Barcelona. Jaeggli, 0. (1982) Topics in Romance Syntax, Fons Publications, Dordrecht. Jaeggli, 0. (1986) Three Issues in the Theory of Clitics: Case, Doubled NPs and Extraction', in H. Borer (ed) Syntax and Sernantics 19, Academic Press, Orlando, Florida. Kayne, R. (197'5) French Syntax. The transformational Cycle, MIT Press. Kayne, R. (1991) 'Romance Clitics, Verb Movement, and PRO' Linguistic Znquiry 22.4., 647-686. Laenzlinger, C. (1990) Le déplacement des clitiques et l'analyse-DP, travail de certificat de specialisation, Un. de Gen2ve. Marcos Marín, F. (1978) Estudios sobre el pronombre, Madrid, Gredos. Picallo, C. (1991) Funcions dobles: tres estudis de sintaxi catalana, Ph. D.dissertation, UAB. Rigau, G. (1988) 'Strong Pronouns', Linguistic Znquiry 19.3., 503-510. Rigau, G. (1991) 'On the Functional Properties of AGR', Catalan Working Papers in Linguistics 1991, U.A.B., 235-260. Roberts, I. (1992) The Syntax of Inversion', course at the GISSL 1992. Sportiche, D. (1992) 'Clitics, Voice and SpeclHead Licensing', conference held at GLOW 1992, Lisbon. Suñer, M. (1988) The Role of Agreement in Clitic Doubled Constructions', Natural Lunguage and Linguistic Theory 6., 391-434 Suñer, M. (1989) 'Dialectal Variation and Clitic-Doubled Direct Objects', in C. Kirschner & J. Decesaris (eds) Studies in Romance Linguistics. Selected Papers pom the Linguistic Symposium on Romance Lunguages XVIZ, 377-395. Torrego, E. (1990) 'Particular Syntactic Constructions in Castilian Spanish', course at the GISSL 1990. Torrego, E. (1991) 'Experiencers and Raising Verbs in Spanish', ms. UMass. Uriagereka, J. (1992) 'Some Issues on Clitic Placement in Western Romance', ms. Un. of Maryland. Departament de Filologia Catalana Edifici B Universitat Autdnoma de Barcelona E-O8193 Bellaterra (Barcelona) E-mail: i@[email protected]