On the Functional Properties of AGR

ON THE FUNCTIONAL PROPERTIES OF AGR Gemma &gau Universitat Autbnoma de Barcelona This paper is concemed with the properties of the Agreement node: Person and Number. We analyze the role that each of these elements plays in the sentence. It will be claimed that Person is the Nominative Case assigner, whereas Number marks the most prominent argument of the predicate. Consequently, number agreement between the verb and an argument does not necessarily mean that the argument receives Nominative Case. The data we will discuss belongs to two Catalan dialects Northwestern and Central. It will be proposed that number agreement is autonomous with respect to person agreement in Northwestern Catalan. In Central Catalan, however, lack of specification for Person entails lack of specification for Number. O. Introduction Functional elements have traditionaily been considered the expression of the relation that holds between the subject and the predicate in a clausal structure. It has long been accepted that, in many languages, the functional category AG~UBJE~ may contain two elements: PERSON and NUMBER. In this paper, I analyze the role that each of these elements plays in the sentence. It will be claimed that PERSON (PERS) is the Nominative Case assigner, whereas i NUMBER (NUM) marks the most prominent argument of the predicate. In the Catalan exarnples (1 a,b) the prominent argument is the external en Pere: (1) a. En Pere menja pomes 'Pere eats apples' b. En Pere plorava 'Pere cried' Catalan Working Papers in Linguistics (CWPL) 1991: 235-260 Universitat Autbnoma de Barcelona VO to 10 movement allows the element PERS to formally relate the verb and its externa1 argument via Nominative Case assignment, while NUM agreement manifests the syntactic relation between the verbal element and the most prominent argument in the sentential stnicture. This relation is implemented by checking NUM specification with that argument. NUM agreement appears to be independent of PERS agreement in some Catalan dialects. In other dialects, however, NUM and PERS agreement interact with each other: lack of PERS specification results in lack of NUM specification.The data discussed in this paper belongs to two Catalan dialects: Northwestern and Central. 1, Case Restrictions on the Syntactic Nature of Verbal Arguments My analysis will adopt a series of hypotheses. First, I assume that AGR [-PERS] is not a Case assigner. In this I will follow Kayne (1989:fn 1). This proposal is compatible with Raposo's (1987) suggestion that for AGR to assign Case, it must itself receive Case from Tense. Second, Nominative and Accusative Case restrict the syntactic nature of the receptor argument, the same way that Partitive assignment does. Arguments that receive Partitive Case are NPs, whereas arguments that receive structural (Nominative and Acusative) Case are DPs (see Torrego (1983) and Belletti (1988)). I will claim that an argument can only receive one Case. As opposed to Belletti (1988), I argue that agreement between an unaccusative verb and its internal argument is not obtained by the conjoined assignment of inherent Partitive at Dstructure and structural Nominative at S-structure. On the contrary, I will suggest that ari argument receives either inherent or structural Case, but not both. Third, AGR [+NUMI is not necessarily related to Nominative Case assignment. Thus, the NP autobusos 'buses' in (2 a) and the clitic en '(00 it/themf in (2 b) receive only Partitive Case: 236 (2) a. Dem& no circularan autobusos tomorrow will not circulate buses 'Buses will not circulate tomorrow' b. Demh no en circularan tomorrow will not (of) thema circulate The interna1 argument autobusos in (2 a) is an NP. Therefore, it is not compatible with structural Case. However, in cases like (3) below, the only type of Case compatible with the interna1 argument of the unaccusative verb --i.e. els autobusos 'the busest-- is structural (Nominative). This is so because it is a DP: (3) Demh no circularan els autobusos tomorrow will not circulate the buses The buses will not circulate tomorrow' 1 .I. Quantified Arguments It has been proposed that a DP is incompatible with Partitive Case, whereas NP is incompatible with structural Case. Under this assumption, the grammaticality of (4) below, needs to be accounted for. Should the argument tres autobusos 'three buses' have an NP or a DP analysis?: (4) Dem& circularan tres autobusos tomorrow will circulate three buses Three buses will circulate tomorrow' It should be noted that tres autobusos can have either a specific reading or a nonspecific reading. The specific interpretation can be paraphrased as in (5): (5) Demh circularan tres dels autobusos tomorrow will circulate three of the buses Three of the buses will circulate tomorrow' I will propose that tres autobusos in (4) may have two different analyses. In one, it is a quantified NP, with a structure roughly represented in (6): (6) [QUANT tres [~p autobusos]] three buses This structure, which does not contain a DP is assigned Partitive Case, and receives a nonspecific interpretation ('some three buses or others'). The phrase can alternatively be analyzed as a quantified DP, possibly with an empty head, and a numeral in a specifier position. This analysis is infonnally represented in (7), and corresponds to the specific interpretation. The argument will be assigned structural Case: (7) [QUANT tres [~p e [Np autobusos I]] three buses We will now discuss some data in support of our claim that the argument in (4) can receive two alternative analyses, and two types of Case (structural and inherent). Consider the following: (8) a. Tres autobusos semblaven espatllats three (of the) buses seemed damaged b. Dues pomes s6n d'aquell pomer two (of the) apples are from that apple tree The derived subjects receive structural (Nominative) Case. According to our proposal, they must be analyzed as DPs, and their structural representation should correspond to (7). Note that the derived subjects in (8) can only receive a specific interpretation, i.e. 'three of the buses', 'two of the apples' respectively. We should also note that the arguments that receive Accusative from a ECM verb are always DPs. For that reason, the argument una aquarel.la 'a watercolor' in a sentence like (9) below has always a specific interpretation, a reading that makes possible to paraphrase una aquarel.la by 'a particular watercolor', 'a certain watercolor', or 'one of the watercolors': (9) Considero molt impressionant una aquarel.la (I) consider very impressive a watercolor Turkish, which expresses Case morphologically, offers us an interesting proof of the relation that exists between the nature of the verbal argument and the type of Case that it is able to receive. According to Eng (1991), interna1 arguments showing a morphological mark for Accusative are obligatorily interpreted as specific in Turkish. Internal arguments without an explicit morphological mark for Case --i.e. the ones receiving Partitive--, are interpreted as nonspecific. The following examples are from Eng (1991): (10) a. Ali bir kitab-l al& Ali one book-Acc bought 'A book is such that Ali bought it' b. Ali bir kitap al& Ali one book bought 'Ali bought some book or other' 1.2. The Clitic ENINE At this point, it is necessary to briefly discuss the clitic en Ine 'of itlthem'. If we argue that Accusative and Nominative Case are only assigned to DPs, while claiming that Partitive is assigned to NPs, we must clarify which type of Case the clitic en Ine may or must receive. We will distinguish between a Partitive en lm and a Geni tive en Ine. According to Belletti (1988), en lne receives Partitive. This is so in the example (2 b) above, where the clitic pronominalizes the NP autobusos 'buses'. However, not all cases of en Ine cliticization instantiate the Partitive Case assigned by a verb, because en lne doesn't necessarily receive Case from a verb. Consider the following examples: ( 1 1) a. Tots en recordem algun, dels seus poemes all us (of)them remember some, of hislher poems 'We all remember some, of hislher poems' b. Vaig trobar els teus amics i només en vaig reconiixer dos: en Pere i en Pau (I) have met your friends and only (of)them have recognized two: Pere and Pau 'I have met your friends and I have only recognized two (of them): Pere and Pau' Here, the clitic pronominalizes a part of the direct object. This object is semantically partitive in each case (see En$ (199 1)). The object en..algun Ien. .dos 'CL.. somelCL.. two' denotes a part of a definite specific set (hislher poemslyour friends). The verb in (1 1) assigns Accusative to the direct object, a quantified DP. Thus, the clitic en lne does not receive Case from the verb, but from the quantifier. It is the same Case that the clitic en Ine receives from an N. Consider the following examples from Bartra (1987): (12) a. Se me n' escapen els detalls CL (to)me (of)it escape the details The details of it escape me' (12) b. N' han cremat totes les proves (of)it have bumed all the evidence They have burned all the evidence' These examples show that the presence of en Ine does not imply that the verb has assigned Partitive to its interna1 argument. The clitic can express Genitive Case assigned by a quantifier or a Noun. 2. The Properties of PERS and NUM in Northwestern Catalan We will now turn to discussing the properties of PERS and NUM. Consider the sentences exemplified in (13) from Northwestem Catalan (see Soli (1973), (1987)), and compare them with the ungrammatical (14) from the same dialect: (13) a. Falta més bra~os is needed more arms 'More arns are needed' b. Arriba parents arrives relatives '(some) relatives amve' c. Ha vingut més turistes has come more tourists 'More tourists have come' d. Passa molts cotxes is passing many cars 'Many cars are passing by' (14) a. *Falten més bragos are needed more anns (14) b. *Ambenparents arrive relatives c. *Han vingut més turistes have come more tourists d. *Passen molts cotxes are passing many cars The grammatical examples in (13) show that there is no agreement between the argument and the verb. In these examples the verbs are unaccusative and able to assign Partitive to their interna1 argument, an NP, which is a receptor of Partitive Case. These are impersonal sentences, in the sense that AGR is not specified for PERS or NUM. The Northwestern Catalan sentences in (15) below are also impersonal (see Sola (1973), (1987)): (15) a. Hi ha molts hbmens there is many men There are many men' b. Enguany es plantarh molts arbres this year CL(=people) will-~lant3,d/~i~~ many trees 'People will plant many trees this year' c. Plou rains 'I t rains' The examples in (13) and (15) show a defective verbal morphology. The verb is inflected for 3rd/sing., which is the morphological expression of absence of Person and Number in Catalan. Our analysis predicts that a sentence can not be impersonal if the interna1 argument of an unaccusative verb is a DP. This prediction is borne out: (16) Vindrh 10 pare will-come the father The father will come' The interna1 argument, a DP, is incompatible with Partitive Case. The only available Case for this DP is Nominative, given that the verb is unaccusative. AGR must then contain [+PERS]. The Northwestern Catalan examples (17) show that the most prominent argument (i.e. the internal DP) must agree with the verb: (17) a. Vindran els parents will-c0me3~d/~l~~ the relatives The relatives will come' b. Els parents vindran aviat the relatives will-~0me3~d/~l~~ soon c. * VindrA els parents will-come3,d/,ing the relatives d . Nosaltres arribarem aviat we will-come~rst~p~ur soon The ungrammatical sentence (17 c) shows that the internal DP requires Number agreement also, when it is the most prominent argument. Nominative Case is assigned to the internal DP subject in the grammatical examples above when the DP undergoes movement to a position where it can be governed by AGR [+PERS]. Let us assume that this position is [Spec,VP]. The following represents the abstract configuration of exarnple (16), where details irrelevant to the present discussion have been ommitted: referential content. Pro is allowed in a given structure if it agrees with a functional category able to assign features to pro. In (30), NUM agreement with the verb via AGR will be enough to recover and identify Accusative pro: (30) a. Les finestres ja es poden obrir the windows already SE may open 'The windows may already be opened' b. [[Les finestresli [ja es podenpiur obrir proi]] There are other facts showing that the Case assigned to the internal argument in (27) and (28) is not Nominative. The evidence can be found in sentences similar to those showing SEARB: those with unaccusative pronominal verbs. Sola (1987) notes that there is a contrast in grammaticality between (3 1 a) and (3 1 b) in Northwestern Catalan: (3 1) a. Amb aquelles nevades es van enfonsar totes les teulades del poble with these snows SE ~ollapsed3,d/~l, al1 the roofs of the town That snow caused all the roofs of town to collapse' b. *Amb aquelles nevades es va enfonsar totes les teulades del poble with these snows SE collap~ed3~d/~i,~ a11 the roofs of the town The pronominal verb enfonsar-se 'to collapse' is unaccusative, therefore it is unable to assign Accusative Case. The only (structural) Case available to the DP is Nominative. Therefore, these types of sentences are personal, like the grammatical Jo m'enfonso, tu t'enfonses, etc. 'I collapse, you collapse', etc. We will now discuss an interesting fact in Central Catalan. There is no NUM agreement when the internal argument of a sentence with SEARB is a definite clitic, as in (32 a) and (33 a) below. NUM agreement is, however, possible in (34), where the clitic is Partitive (i.e. non definite): (32) a. No se us aprecia gaire not SE youp~~~ appreciatqrdlsing very much 'Y ou are not appreciated very much' b. *No se us aprecien gaire not SE youp~~~ appre~iate3~dl~l~~ very much (33) a. No se'ls ha tractat bé not SE them havqrdlsing treated well 'They have not been well treated' b. *No se'ls han tractat bé not SE them havqrd/,lW treated well (34) a. No se'n . veu, de fum not SE EN(=of it) sees, of smoke 'Smoke, it is not seen' b. No se'n veuen, de miracles not SE EN(=of them) see, of miracles 'Miracles, they are not seen' The behavior of definite and non definite clitics can be explained if we accept the hypothesis that definite clitics are themselves AGROBJECT (see Fernández-Soriano (1989), Cordin (1990)). Hence, they may not establish a new agreement relation. That is, the clitics us /els 'youlthem' in (32) and (33) respectively may not express AGRSUBJECT and AGROBJECT agreement at the same time. Consider now the data in (35), which belongs to a subdialect of Central Catalan, and where the definite clitic expresses AGROBJECT. The Past participle agrees with the clitic in Number and Gender: (35) a. No se les ha tractades bé not SE them~~~ has treated~~~g~m well TheyFEM have not been well treated' b. No se 1' ha tractada bé not SE her has treated~,~~~~ well 'She has not been well treated' Se les~~g constructions in Central Catalan show a strong preference for pro over a definite clitic when a [-human] DP is pronominalized. Pro, as opposed to the clitic la 'her' establishes NUM agreement, because pro needs to agree in order to be identified: (36) a. Aquestes copes, no es poden rentar pro amb aigua calenta these cups ,not SE can3,dlplur cleanpro with hot water These cups cannot be cleaned with hot water' b. ??I?~questes copes, no se les pot rentar amb aigua calenta these cups not SE thema Can3,dlsing clean with hot water 5. The Predicate HAVER-HI Impersonal sentences with the verb haver-hi (lit.:'to have-thereCL'; 'there be') show a behavior similar to those with SEARB: (37) a. No hi ha pa not therec~ has bread There is no bread' b. Hi havia un home thera had a man 'There was a mant The clitic hi 'there' seems to have been incorporated to the verb haver 'to have' in the lexicon, to form the predicate haver-hi. Formerly, haver 'to have' expressed possession in Catalan. The possessive meaning of haver still remains in some cases: (38) a. Aquesta noia, no 1' haurhs pas this girl, not herc~ will-ha~e2~d,~i~~ NEG. This girl, you won't have her' b. Aixb costa d'haver this is hard of to have Thls is hard to have' In sentences of the type (37), the clitic hi behaves like a locative subject (see Torrego (1989)). In other words, hi blocks the externa1 argument of haver 'to have' when i t incorporates into this verbal form in the lexicon. Hi also blocks Nominative Case assignment, preventing the realization of a subject. Evidence that hi is not a 'free' locative, but an incorporated clitic, is shown by the fact that the clitic may not be substituted by a real locative expression: (39) a. Hi havia un noi HI had a boy 'There was a boy' b. *All& havia un noi 'there had a boy' c. *Un noi havia a Banyoles a boy had in Banyoles The incorporated clitic hi impersonalizes the verb by bloclung PERS, which is blocking Nominative assigment according to our proposal. Therefore, it is impossible to have NUM agreement between the verb and its only argument in Northwestern Catalan: (40) a. Hi haur& pocs hbmens (Northwestern) HI will-ha~e3~dl~i~~ few men There will be few men' b. *HI hauran pocs hbmens HI wiii-ha~e3~d/~l~ few men c. Hi havia pastissos de xocolata HI had3rdlsing chocolate cakes There were chocolate cakes' d . *HI havien pastissos de xocolata HI had3rd/plur chocolate cakes Our hypothesis predicts that sentences (40 b, d) will be well forrned in Central Catalan, and this is the case. There is NUM agreement between the argument and the verb, even though the construction is an impersonal sentence. This is an interesting fact, given that the clitic hi blocks PERS and NUM in the majority of Catalan dialects. I will suggest that this phenomenon occurs because hi is incorporated into the verb at the lexical level, which is not the case with SEARB. This hypothesis appears to be confirmed when one considers Spanish equivalent constructions. The appearance of y (=hi ) blocks PERS and NUM: (41) a. Hay mosquitos has-Y mosquitoes There are mosquitoes' b. *Hany mosquitos have-Y mosqui toes If y is non overt, NUM may not be absorbed in Spanish. The following sentences are both gmmatical: (42) a. Había mosquitos had3rd/sing mosqui toes b. Habían mosquitos had3raplur mosquitoes There were mosquitoes' 6. NUM Agreement I have shown that NUM agreement between the verb and an argument doesn't necessarily mean that the argument receives Nominative Case. Nominative is assigned by PERS, not by NUM. However, when PERS is specified and assigns Nominative, it selects NUM obligatorily (i.e. (23 a)). NUM agreement obtains between the verb and its most prominent argument. Therefore, when there is a Nominative argument, it will also agree with the verb in NUM features. Hence, the ungrammaticality of (43) follows from the fact that the verb agrees in NUM with the object, but not with the subject: (43) *Els infants prefereix aquest joc the children prefers this game Now the question is the following: What causes the presencelabsence of number agreement between the verb and its interna1 argument in Catalan dialects when the sentence is [-PERS]? We want to propose that number agreement with an internal argument depends on how predication is licenced at LF in each dialect. Consider again the following examples: (44) a. Vindrii parents (Northwestern) will-~omg~d/~i~~ relatives b. Vindran parents (Central) will-come~r~p~w relatives 'Relatives will come' These examples satisfy thematic requirements and the internal argument satisfies Visibility requirements in (45 a, b) because it is assigned Case. Central Catalan appears to still require another condition for the sentence to be interpreted as a predicative structure. We suggest that sentence (45 a) can not be interpreted in Central Catalan. When a predicate selects one or more arguments, one of them has to be overtly marked as prominent in this dialect. Prominence is expressed by NUM agreement between the tensed verb and the argument. As was said, NUM agreement depends on the hierarchical relations between PERS and NUM, but also on the hierarchical relations between arguments. The thematic hierarchy needs also to be 'visible' in Central Catdan. At LF, the thematic condition in (45) excludes (44 a) in Central Catalan: (45) Prominence Condition When a predicate selects one or more arguments, one of the arguments has to be interpreted as the prominent argument of the predication. Prominence is expressed through number agreement between the verb and the argument. The thematic hierarchy is respected. This condition is always active in Central Catalan, even in impersonal ([-PERS]) sentences. By contrast, the Northwestern variety does not require (45) in impersonal sentences. NUM agreement is only required when an argument is assigned Nominative Case in Northwestern. We must recall that [+PERS] always selects [+NUMI in Catalan dialects, and that [-PERS] selects [-NUMI in Northwestem. Sentences with a 'self-sufficient' verb, that is, a verb that doesn't need an argument to express predication --like ploure 'to raint--, will not manifest [+NUMI: (46) a. Plou [-PERS,-NUM] rains b. *Plou [-PERS,+NUMsing] c. *Plouen [-PERS ,+NUMplur] What excludes sentences (46 b, c)?. Recall that positive specification for PERS and NUM must be checked with an argument and appear suffixed to the verb. A sentence will be ill-fonned if [+NUMI is not checked, the same way that the sentence is ill-fonned if [Tense] is not amalgamat4 to the verb. The function of [+PERS] and [+NUMI is that of marking agreement, hence, these elements neeú to be expressed in a verb as well as in an argument. A sentence like (47) will be specified [+PERS,+NUM]. The fact that it may have an arbitrary interpretation does not follow from the functional elements in the sentence, but follows from other factors: (47) a. pro truquen [+P~S,+NUMJ a la porta (they) are knocking on the door b. pro ~~~~~~~[+PERs,+NuMI a la porta (they=someone) are knocking on the door 7. Conclusion Suppose that there are two possible ways to assign [PERS] (or Nominative Case) and NUM to an argument: by government, if the argument is in [Spec,VP]; or by Spec-head agreement. The latter is possible for DPs, but not for NPs. Consider the schemata (48) and (49): IP n I' \ h I VP [+PERS] [+NUMI It has been claimed that the fact that a subject agrees with the verb does not imply that it must necessarily be assigned Nominative. It has been our contention that PERS and NUM trigger different types of relations between a subject and the verb. Evidence for this claim can be found in the two Catalan dialects we have discussed. It is possible, and tempting, to propose that these two functional elements project onto two separate syntactic categories. However, the question of whether PERS and NUM constitute separate X-bar projections or not is not crucial at this stage of our research. Here, our goa1 has simply been to acquire a better knowledge of the factors intervening in agreement relations. This research has been supported by a DGICYT grant PB 89-0324, awarded to the UAB. Some versions of this paper were read at a Workshop on Central Romance Languages, Université de Gen&ve, July 1990; at the Primer Coloquio de Gramática Generativa, UAM Madrid, March 1991; and at the UniversitA di Venezia, May 1991. I would like to thank these audiences for their questions and comments. I am also indebted to Joan Sola and to my colleages at the UAB Linguistics Seminar, to Carme Picallo in particular, for their suggestions to previous drafts of this manuscript. References Bartra, A. (1987) 'Encara n'hi ha més. (Entorn de en i alguns SNs genitius', Llengua i Literatura 2,377-427. Belletti, A. (1987) 'Los inacusativos como asignadores de caso', in Sintaxis de las lenguas románicas V. Demonte and M.F.Lagunilla (eds), El Arquero, Madrid, 167-230. See also Linguistic Inquiry 19.1, 1-34. Cordin, P. (1990) 'Dative Clitic Doubling and Visibility of Case in Trentino', unpublished ms. Univ. di Trento. En$, M. (1991) The Semantics of Specificity', Linguistic Inquiry 22.1 1-26. Fernández Soriano, 0. (1989) Reccidn y Ligamiento en español: Aspectos del parámetro del sujeto nulo, PhD dissertation, U.A.M., Madrid. Kayne, R. (1989) 'Facets of Romance Past Participle Agreement', in Dialect Variation and the Theory of Grammar, P. Beninca (ed), Foris, Dordrecht, 85-103. Picallo, C. (1991) 'El posesivo y el principio "Evítese el Pronombre"', paper read at the Primer Coloquio de Gramática Generativa, Mraflores de la Sierra, Madrid. Raposo, E. (1987) 'Case Theory and Inflto-Comp: The Inflected Infini tive in European Portuguese', Linguistic Inquiry 18, 85109. Solh, J. (1973) Estudis de Sintaxi Catalanal2, Edicions 62, Barcelona Sol&, J. (1987) Qüestions Controvertides de Sintaxi Catalana, Edicions 62, Barcelona.