Clitics : Lexicalization Patterns of the So-called 3rd Person Dative

Abstract Manzini and Savoia (1999, 2001, 2002, to appear) argue that the basic facts about the clitic string are best accounted for without having recourse to anything but a minimalist syntactic component, i.e. making no use of a specialized morphological component nor of optimality-type comparisons between derivations/ representations. In particular, they assume that clitics correspond to specialized inflectional categories, and are merged directly into the positions where they surface; such categories are furthermore ordered in a universal hierarchy, as we will detail below. The aim of the present paper is to consider datives in the light of this framework. We will conclude that there is no evidence for the category dative in the Romance dialects we shall consider, while in fact there is evidence for categorizations of so-called dative clitics as quantificational elements or as deictic elements (locatives). In all cases, the relevant categorization relies entirely on referential properties, or more generally on interpretive properties intrinsic to the lexical items involved, calling into question the traditional notion of Case itself. Key words: clitics, Case, mutual exclusion, suppletion, dative, locative, partitive, accusative, si, Italian dialects. *The research reported in this article has been financed partially through the Programma di ricerca cofinanziato of the MIUR ‘La cartografia strutturale delle configurazioni sintattiche e le sue interfacce con la fonologia e la semantica’for the years 1999-2001 and ‘Categorie linguistiche: dalla mappatura strutturale all’architettura delle rappresentazioni mentali’ for the years 2001-2003. The article is a result of the collaboration of the two authors in all respects. Nevertheless, for administrative purposes, Rita Manzini takes responsibility for sections 2. and 4. and Leonardo Savoia for sections 1. and 3. The data concerning Vagli, Olivetta, and Làconi are also discussed by Manzini and Savoia (to appear), though the emphasis of that article is entirely different from the present one, being concerned with mutual exclusion between 3rd person clitics, including the object for subject phenomenon. Catalan Journal of Linguistics 1, 2002 117-155 Clitics: Lexicalization Patterns of the So-called 3rd Person Dative* Rita Manzini Leonardo M. Savoia Università di Firenze. Dipartimento di Linguistica 50121 Firenze [email protected] [email protected] Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 117 1. Theoretical background Following Sportiche (1996), rather than Kayne (1975, 1989, 1991, 1994), we assume that clitics are inserted under specialized functional categories, i.e. in stricter minimalist terms their merger projects specialized functional categories. If clitics are generated in the ordinary argument positions and adjoined to verbal or inflectional positions it is hard to predict that they appear in a fixed number, in a fixed order and with fixed cooccurrence (or mutual exclusion) patterns which do not necessarily correspond to the number, order, cooccurrence (or mutual exclusion) patterns of corresponding arguments and adjuncts. To be more precise, the theory can derive the relevant properties of clitics in conjunction with a morphological component able to (re)order strings (Bonet 1995, Halle and Marantz 1993, 1994). To the extent that the (re)ordering operations match those of the syntax (Merge and Move), the resulting system is however highly redundant; viceversa to the extent that the two sets of (re)ordering operations do not match, the resulting system is considerably more complex. Therefore we assume that a purely syntactic account is to be preferred for reasons of simplicity of the theory. We specifically assume that clitics are generated directly in the position where they surface, hence that there are clitic positions between I and C. Adopting a universal hierarchy of functional positions of the type in Cinque (1999), though not necessarily one containing the positions postulated there, we are led to propose a universal clitic string, within which positions can neither be reordered (contra Ouhalla 1991) nor packed and unpacked (contra Giorgi and Pianesi 1998). A first set of clitic categories relevant for the hierarchy is motivated in relation to subject clitics by Manzini and Savoia (2002), who argue in favor of a category P(erson) for 1st and 2nd person clitics, a category N(oun) for 3rd person clitics, a category Q(uantifier) for plural clitics, and a category D(efiniteness) for otherwise uninflected clitics. In their conception, therefore, clitic categories correspond to denotational properties. Thus P implies reference to the speaker (1st person singular), the hearer (2nd person singular) and the sets including them (1st and 2nd person plural); in turn, N identifies the so-called 3rd person simply with the predicative property N. As for the Q and D categories, they are to be understood exactly as in the analysis of noun phrases, i.e. as encoding weak quantificational properties (corresponding to numerals, existentials, etc.) in what concerns Q, and definite denotation in what concerns D. To these categories, Manzini and Savoia (to appear) add three further categories characterized in broadly denotational terms, and specifically connected to their discussion of object clitics. In particular they individuate the need for a Loc(ative) category lexicalizing reference to the spatial coordinates of discourse and for a R(eferential) category corresponding to strong quantificational or specificity properties. They also add a category DOpassociated with modal/ intensional properties to serve as a nominal counterpart to modal/ intensional categories of the verb, generally represented by the complementizer system. A natural order for the categories individuated so far is suggested by the observation that many, if not all, such categories are independently postulated in cur118 CJL 1, 2002 M. Rita Manzini; Leonardo M. Savoia Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 118 rent generative analyses of the internal structure of the noun phrase. In particular, the sequence D - R - Q - N constitutes the basic skeleton of the noun phrase, where N is associated to the nominal head, hence to the predicative content of the phrase, Q to indefinite quantifiers, R to specific quantifiers and D to the definite article. As for Loc, this position can be identified with demonstratives, on the grounds of the general spatial interpretation of these elements, and more specifically of the fact that in Romance dialects they surface coupled with overt locative pronouns; on the basis of the position occupied by the latter the Loc position is relatively low within the noun phrase (Brugè 1996, Bernstein 1997), presumably between N and the quantificational projections. The P category, which like Loc is interpreted in terms of discourse-anchored reference, is naturally construed as occurring in the same area of the nominal tree; there it can correspond to the merger position for possessives. Finally DOpcan host prepositional elements which can equally well introduce a noun phrase or serve as complementizers of a sentence, such as Italian di (‘of’), a(‘to’); they close off the entire noun phrase appearing in the position immediately above D. Therefore the hierarchy of nominal categories within the noun phrase takes the shape in (1), with the content of the different categories briefly summarized in (2). (1) [DOp[D [R [Q [P [Loc [N (2) a. N is associated with the head of the Noun phrase b. Q is associated with Quantifiers (indefinite quantification) c. R is associated with Referentiality (specific quantification) d. D is associated with Definiteness e. P is associated with Person, i.e. reference to speaker and hearer (possessives) f. Loc is associated with Locative, i.e. reference to the spatial coordinates (demonstratives) g. DOpis associated with a nominal counterpart to modal/ intensional categories of the sentence (‘of’, etc.) Manzini and Savoia (to appear) argue that the string in (1) also defines the basic order of clitic categories within the sentential string, and assign the various descriptive classes of clitics to the categories in it. In keeping with the observation that subject clitics generally appear before object clitics, it seems natural to reserve the higher positions of the string for them; more specifically, we identify the subject clitic position with D. If Manzini and Savoia (forthcoming) are correct, furthermore, the highest DOpposition is not associated with clitics at all, but rather hosts complementizers, including those of the that type (Italian che, etc.) as well as prepositional ones. As we anticipated, the basic aim of our discussion is to arrive at a characterization of datives, which we shall therefore leave for later discussion; here Lexicalization Patterns of the So-called 3rd Person CJL 1, 2002 119 Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 119 and throughout, reference to Case categories such as dative (as well as to person, number and gender ones) is purely descriptive. Obviously enough, the P category is lexicalized by 1st and 2nd person (non-subject) clitics. The observed behavior in Italian clitic systems (Manzini and Savoia 1998) supports the position already taken by typological approaches, which sharply differentiates the status of 1st and 2nd person (singular, and eventually plural) from that of the so-called 3rd person. In our grammar, therefore, the total membership of the Person category is constituted by the speaker and the hearer, whose denotation is fixed directly by the universe of discourse, and by the sets including them. When it comes to the dative, this entitles us to consider 3rd person only, abstracting from what may be described as 1st and 2nd person datives. Crucially, in contrast to what happens with 3rd person, there is no morphological differentiation between P forms used as accusatives and datives; nor do P forms have a different position in the clitic string depending on the dative/ accusative divide. We note next that the characterization of Loc in terms of spatial reference must be conceived in wide enough terms to include a whole series of possible interpretations associated with the locative clitic. Thus in a language like Italian, ci can have a strictly locative meaning, an instrumental one, a comitative one, etc. The purely locative interpretation itself can be seen to be internally articulated in several different meanings. Thus the locative can be associated with a stative interpretation or with a motion intepretation, under which the locative typically refers to the coordinates of the final point of the event. In general, the Loc category does not correspond to a point on the aspectual contour of the event. Rather it must be understood deictically in connection with other elements whose denotation is fixed by the universe of discourse, namely the speaker (‘I’), the hearer (‘you’) and the temporal coordinates of the discouse (‘now’). A language like standard Italian provides evidence that as suggested in (1), the P clitic is ordered before the Loc one, as in (3a). The relevant substructure is as in (3b); note that, here and througout, a linear format rather than a tree one has been used for structural representations for purely practical reasons. (3) a. Mi ci vede. me there he.sees ‘He sees me there.’ b. ... P Loc ... || mi ci We already indicated that N corresponds to 3rd person; the observation that accusative clitics appear, in many languages, in the lowest position of the string leads us to assign them to the N category. Thus in a language like standard Italian they follow both P and Loc clitics, as indicated in (4)-(5). In this framework, the total content of the so-called accusative clitic is reduced to its N intrinsic content; the interpretation of the N clitic as the internal argument of the verb is the result of 120 CJL 1, 2002 M. Rita Manzini; Leonardo M. Savoia Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 120 Lexicalization Patterns of the So-called 3rd Person CJL 1, 2002 121 the application of some interpretive principle. In other words, interpretive categories such as ‘theme of’, or ‘Measure of’ in aspectual terms, are interpretations available for the relevant syntactic structures at the LF interface, essentially along the lines adopted by Chomsky (1995). In particular N is never lexicalized but by the internal argument, though it is obvious that the reverse does not hold. Thus in inaccusatives, the internal and only argument of the verb corresponds to the D clitic in subject clitic languages in virtue of some version of Chomsky’s (1995) EPP. (4) a. Me lo dà. to.me it he.gives ‘He gives it to me.’ b. ... P Loc N ... || me lo (5) a. Ce lo mette. there it he.puts ‘He puts it there.’ b. ... Loc N ... || ce lo The partitive clitic, ne in standard Italian, does not directly denote an argument either in the event structure, or in the domain of discourse, but it contributes to the denotation of one such arguments. For instance in (6a) the denotational content of ne enables us to fix the reference of the internal argument of the verb, represented here by the numeral quantifier tre ‘three’. Since in (2f) we analyze the di ‘of’ element introducing partitive noun phrases as belonging to the DOpcategory, we tentatively assign the partitive clitic to the category DOpas well. In other words standard Italian ne is of category DOpon the evidence of the fact that it doubles phrases headed by an DOpelement such as di. On the other hand, the evidence relating to the position of ne in the string indicates that it is lower than DOpand in fact corresponds to N in a language like standard Italian, where the partitive appears lower after P and Loc clitics, as in (6b). (6) a. Ce ne mette tre. there of.them he.puts three ‘He puts three of them there.’ b. ... P Loc N ... || ce ne We take it that the N position of the DOpelement corresponds to the fact that an DOpelement cannot but be interpreted as a specification of a predicative N Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 121 head. In particular in an example like (6) merger of ne in the N position of the string corresponds its interpretation as an DOpspecification of the internal argument independently lexicalized in the string by the quantificational head tre ‘three’. We shall return to this property of ne in the discussion in section 2.3 below. We tentatively assume that not only it licences the N merger position of ne, but it also prevents it from surfacing in a position corresponding to DOpof the sentential string, where it would not be possible to assign to it the relevant relation to a predicative N head. The one remaining clitic in a language like standard Italian is at this point si, associated with reflexives, impersonals, and passives. The only relevant point is that we take its merger position to be normally Q, in virtue of its denotational properties, which are essentially those of a free variable (Manzini 1986). As detailed by Manzini and Savoia (1999, 2001) an analysis along these lines derives the different construals for si. Interestingly, a Q merger point would predict that si precedes not only accusative and partitive clitics, but also P and Loc ones, as it does in a large number of Italian dialects (Manzini and Savoia 2001). In standard Italian on the contrary si normally precedes partitive and accusative, but follows P and Loc clitics, as in (7a). Our conception of the R position helps in this respect, since its specificity properties make it a potential host for the whole series of object clitics. As in (7b) we may assume therefore that R hosts the locative clitic, preceding si in Q. (7) a. (Lui) ci se le lava (le mani). he there to.himself them washes (the hands) ‘(His hands) he washes them there.’ b. ... R Q P Loc N ... || | ci se le Naturally, the conception of R as a specificity category predicts that we should be able to find in the same R position not only a Loc clitic such as ci in (7), but also other types of clitics. Indeed in section 2.1 we shall propose that the accusative series of the dialect of Olivetta can be hosted in R, and the same holds for the si-type clitic, i.e. a Q clitic, in the dialect of Vagli. For the dialects of Piobbico in section 2.2 and of Celle di Bulgheria in section 3.2 we shall propose that the dative is hosted by R. For the dialects of Nocara in section 2.3, of Làconi in section 3.1 and Nociglia in section 3.2 we shall associate R with the partitive. In some of the cases just reviewed, it is interesting to note that R is a possible point of merger, rather than a necessary one. This is evident already from the comparison of standard Italian (7), where the locative is in R, with (3), where it is in Loc. We take it that in the case of (3) and (7), the Loc merger point is straightforwardly justified by the denotational content of the clitic; we conceptualize the R merger point in terms of scopal properties of the clitic itself. In other words the specific nature of the locative denotation allows for the scopal R position as well for the Loc one. On the 122 CJL 1, 2002 M. Rita Manzini; Leonardo M. Savoia Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 122 basis of a scopal conception of R, we may equally expect that in some languages one or more clitics necessarily appear in the higher scopal position, as we shall see in particular for Olivetta in section 2.1. From a purely empirical point of view, of course, R serves as the one major source of reordering within the clitic string. On the basis of the discussion at the outset, the clitic string in (1), hence the partial structures in (3)-(7), occupy the area of the sentence between the I position, where the finite verb normally appears in declarative sentences, and the C position where it appears in main clause interrogatives. Empirical evidence, relating in particular to the doubling of clitics on either side of the verb in C, strongly argues in favor of a conception in which the clitic string repeats itself identical above C as well, as discussed in particular by Manzini and Savoia (1999). Generalizing this conclusion, we assume that a clitic string is generated above each of the three main verbal domains, i.e. immediately above V, I and C. This gives rise to the organization of the sentence in (8), where the dotted space is to be filled by the string in (1). Evidence for the lower string is provided by Manzini and Savoia (forthcoming); intuitively, however, it is clear that it corresponds to the main argumental domain of the sentence, so that we may provisionally assume that lexical arguments are merged in the (Spec of) the relevant positions. (8) ... [C ... [I ... [V As well as an analysis of the overall structure of the clitic string and of the categories it consists of, an account of cliticization in Romance dialects presupposes an analysis of the internal structure of clitics themselves. Previous approaches in the literature include both morphological and syntactic ones. Among the former, we find James Harris’s (1994) account of the internal make-up of Spanish clitics, which recognizes a lexical basis lfor the 3rd person series as well as nominal class morphemes such as -a(traditionally the feminine) and a number suffix -s. A syntactic, rather than morphological, characterization of the internal structure of clitics is attempted in a few recent papers, including Kayne (2000) on 1st and 2nd person clitics as opposed to 3rd person ones, and Cardinaletti and Starke (1999) on clitics compared to weak and strong pronouns. The general idea of Cardinaletti and Starke (1999) is that clitics have the internal structure of a DP, albeit an impoverished one with respect to lexical DP’s or even non-clitic pronouns. In their terms the latter are associated with a full structure, equivalent to a sentential CP; on the other hand clitics are characterized by a deficient structure, reducing to the equivalent of a sentential IP projection. According to Kayne (2000), on the other hand, 1st and 2nd person clitics and pronouns lack full DP structure, while the latter characterizes 3rd person clitics, as revealed by the presence of full agreement features. The approach that we take to the internal structure of clitics relies on the idea that clitics are just ordinary noun phrases. As for the structure of the latter we have already seen that (1) corresponds to the basic organization of nominal categories not only within the sentence, but also within the noun phrase itself. This idea needs to be made more precise in just one respect. Following work by Abney (1986), Lexicalization Patterns of the So-called 3rd Person CJL 1, 2002 123 Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 123 Szabolcsi (1994), the structure of the noun phrase is organized along similar lines as the structure of the sentence. On the model of the sentence, the lowest position in the noun phrase, associated with predicative content, can be taken to coincide with N in (1); but we also need to identify an I and a C position. What we propose is that the string in (1) as a whole repeats itself within the noun phrase as within the sentence, yielding the basic noun phrase skeleton in (9). I and C are just labels for the N positions of higher strings, reflecting their scopal properties. The dotted space in (9) therefore encloses the functional subsequence of the string in (1), namely DOp - D - R - Q - P - Loc - N. (9) ... [C ... [I ... [N We are now in a position to turn to the internal structure of clitic forms. We can translate the morphological analysis of Romance clitics proposed in works such as James Harris (1994) into the present syntactic model by identifying nominal class (gender) morphemes such as o, a which accompany the llexical base in many Romance dialects with the I projection of a noun phrase. As for number morphemes, a natural analysis within our framework identifies them with the Q category. This analysis applies in particular to number morphemes added to nominal class ones, as is the case for sin Spanish. We then obtain structures of the type in (10), which account for instance for the as , ɔs observed in the plural nominal inflections of typical Sardinian dialects.This analysis introduces an asymmetry between the conceptual status of nominal class (gender) morphemes, which lexicalize the nominal category I, and that of number morphemes, which lexicalize the functional category Q. This result appears to be on the right track, since number is indeed a functional specification of the noun, while gender is an intrinsic property of a nominal category. (10) (Sardinia) a. [I ɔ [Qs[ N ]]] b. [I a[Qs[ N ]]] Inflections of the type in (10) can be added to adjectival or nominal bases, but what interests us here directly is that they can be added to the lmorpheme of 3rd person clitics. Precisely the observation that nominal constituents of the type in (10) have independent existence as agreement morphemes suggests that 3rd person clitics involve a nominal head l. As illustrated in (11), we take it that llexicalizes the normal inflectional position, i.e. I, within its own noun phrase, embedding the separate noun phrases in (10). (11) (Sardinia) a. [I l... [I ɔ [Qs[ N ]]] b. [I l... [I a[Qs[ N ]]] 124 CJL 1, 2002 M. Rita Manzini; Leonardo M. Savoia Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 124 The internal structures of clitics are directly relevant to an important question concerning the hierarchy in (1). In rejecting in particular the morphological model of Halle and Marantz (1993), we have adopted the point of view of the minimalist grammar of Chomsky (1995), where syntactic structures are directly projected by the insertion of lexical material. Thus there cannot be structures such as (1) produced by the syntactic component, to which lexical material is matched by lexical insertion. Rather, if hierarchies such (1) hold, it must be because of independent constraints which as suggested by Manzini and Savoia (to appear) can ultimately be thought of in Full Intepretation terms. This puts a heavy constraint on our grammar, since we cannot simply insert a clitic in an already given position as a default lexicalization, not presenting any mismatch with the syntactic category. On the contrary, we must be able to show that in each case, it is an internal category of the clitic that projects the relevant category of the sentential string. The discussion that follows will uphold this general conclusion; in many cases we shall be able to show that the category projected by the clitic on the sentential tree corresponds exactly to the category of the internal head of the clitic itself. Thus a clitic series such as (11) will typically project N on the sentential tree. For pure ease reading, exactly as we describe clitics in terms of accusative, dative, 1st and 2nd person i.e. of features that do not correspond to our actual categories, so we will speak of their insertion points. In all cases we will understand by insertion point of a clitic, the category that the clitic itself projects on the basis of its internal constituency. 2. Types of datives 2.1. Morphologically 3rd person datives Manzini and Savoia (forthcoming), in considering the lexicalization of the socalled dative argument in several dozens of Italian dialects, note that it is the exception, rather than the rule, that they should present a morphologically 3rd person clitic form for the 3rd person dative. Standard Italian is among the dialects which possess such a form, which furthermore combines with the accusative, preceding it. This pattern is normally found in Central Italian dialects, and emerges in the dialects of Lucania as well as in Tuscan dialects, including Vagli di Sopra in (12). The (a) example displays the isolation form of the dative, i.e.  i. The (b) example shows the combination of dative and accusative, in this order. The ə morpheme that surfaces in the feminine plural l ə is a phonological alternant of e, which surfaces for instance in sentence-final position as in the enclitic cam ə -le ‘call them!’; this yields an accusative clitic series l/la/  i/le. As shown in (c), the dative is also followed by other clitics, such as the partitive. Note that to economize on glosses we have given the meaning of  i as ‘to him’; in fact, this is short for ‘to him/ to her/ to them’. We have followed the same general principle throughout the article. Lexicalization Patterns of the So-called 3rd Person CJL 1, 2002 125 Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 125 (22) Piobbico (Marche) a. i da 'kwest to.him he.gives this ‘He gives this to him.’ b. m el/ la/ (l)i/ lə'da to.me it-m./it-f./them-m./them-f. he.gives ‘He gives it/them to me.’ c. i li 'da to.him it/them he.gives ‘He gives it/them to him.’ d. i si da 'sεmprəra'do:nə to.him one gives always reason ‘One always approves of him/ he is always approved of.’ Let us begin by considering the internal structure of the accusative and dative series. Taking up again the analyses proposed in section 2.1 we assume that the l morpheme lexicalizes the I head of a nominal constituent, while a vocalic morpheme corresponding to inflectional class specifications such as a/ ə appears in the I head of a separate nominal constituent embedded under l, yielding structures of the type in (23b)-(23c). The case of el in (23a) is analyzed on the other hand on a par with the pure lforms of section 2.1, i.e. as the I head of a nominal not embedding any inflectional specification. The iclitic in turn, corresponding to both the descriptive accusative masculine plural and the isolation form of the dative, is amenable to a Q categorization as in (23d). Finally nothing in the grammar bars the combination of the Q morpheme, i.e. i, with the l base denoting definiteness. It is therefore natural to propose the internal structure in (23e) for the li form, which corresponds to the internal argument in the context of a dative, as in (22c), but also to the internal argument with (masculine) plural interpretation in other contexts, as in (22b). (23) Piobbico a. [Iel [ N ]]] b. [Il... [I a[ N ]]] c. [Il... [I ə [ N ]]] d. [Q i[I [ N ]]] e. [Il... [Q i[I [ N ]]] The lexical properties of all clitics in (23) are compatible with insertion in the N position of the clitic string, where the el/la/l ə set receives the ordinary interpretation as internal arguments of the verb, as does the (l)iform, whose quantificational properties also induce a plural reading. As argued at length in section 2.1, the iclitic can furthermore lexicalize the Q position, or more correctly in this case 132 CJL 1, 2002 M. Rita Manzini; Leonardo M. Savoia Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 132 the R position, since it precedes si in Q in examples such as (24a), receiving there a distributive interpretation, which corresponds to the descriptive label of dative. Remember that the iclitic of Olivetta could analogously lexicalize R where however it could be interpreted non only as a dative (distributor) but also as a masculine plural accusative. We correlate this parameter between the two languages to the obvious fact that it is the whole 3rd person clitic series that merges in R in Olivetta, but only i in Piobbico. We maintain the proposal developed above that merger in the scopal R position corresponds to the specificity properties of the Olivetta 3rd person series; on the other hand the exclusively dative interpretation of the i clitic in R of Piobbico is explained, if what motivates it is specifically its scopal sentential properties as a distributor. The problem that we need to consider is that in the case of a 3rd person argument distributing over a 3rd person internal argument, the latter is lexicalized by li. We note that the internal structure suggested for li in (23e) in fact consists both of a Q morpheme, i.e. in our terms a potential distributor, and of the lmorpheme, which in terms of the present proposal lexicalizes 3rd person reference under the form of a pure definiteness property. Therefore it is reasonable to assume that li is the specialized lexicalization precisely for a 3rd person object in the scope of a distributor. The insertion position of li can in turn coincide with N, given the presence of the nominal lhead; in this position it is of course preceded by i, which we can equally well associate with Q as with R, as indicated in (24b). (24) Piobbico a. DOpDRQPLoc N I || | isi da b. DOpDRQPLoc N I ||| ilida 2.3. The types ‘ci’, ‘ne’, ‘si’ for 3rd person dative In many Italian dialects the so-called dative is represented not by a morphologically 3rd person form, but one which coincides with a clitic of the language independently associated with the locative denotation, Loc, or with the partitive denotation, i.e. DOp, or finally with the impersonal/reflexive one, i.e. a Q element of the si type. At least this third type of lexicalization of the dative is known in the literature for contexts including a 3rd person accusative. In fact, in a language like Spanish, the incompatibility of 3rd accusative and 3rd dative leads to the apparent substitution of the dative with the se clitic (Perlmutter 1972, Bonet 1995, and in an optimality framework Grimshaw 1997). It is important to realize, however, that in this section we shall present cases where the lexicalization of the so-called 3rd person dative by a si-type clitic, or by a locative or by a partitive, is totally independent of the syntactic context; thus it holds in all clitic combinations, and in isolation as well. Lexicalization Patterns of the So-called 3rd Person CJL 1, 2002 133 Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 133 To begin with, the dative coincides with the locative in the majority of Northern Italian dialects, as well as in many dialects of Central and Southern Italy. In a Northern Italian dialect such as Modena, glexicalizes the 3rd person dative in isolation as in (25a) and in combination with other clitics, as in (25b); at the same time it represents the locative form of the language, as in (25c). We describe the data in terms of a lexicalization of the dative by the locative, rather than the other way round, because forms such gare unconnected to object or subject 3rd person morphology, contrary to the so-called datives considered in the previous sections. The gclitic not only precedes the accusative, as in (25b), and the partitive, but also follows all other clitics it can cooccurr with, i.e. the P and si clitic, as in (25d)- (25e). Note that gis glossed ‘there’ or ‘to him’ (meaning ‘to him / to her/ to them’ as above) in accordance with the translation; the same principle is followed in glosses throughout this section. (25) Modena (Emilia) a. a g 'dag kwas-'kε I to.him give this ‘I give this to him.’ b. a g al/la/i/li 'dag I to.him it-m./it-f./them-m./them-f. give ‘I give it/them to him.’ c. a g 'mεt kwas-'kε I there put this ‘I put this there’ d. a m g la 'mεt I myself there it put ‘I put it there (for myself).’ e. a se g 'mεt dla 'rɔba it one there puts some stuff ‘One puts some stuff there/ some stuff is being put there.’ The relative position of the gclitic with respect to P clitics, to the si clitic in Q and to accusative clitics confirms that it is associated with the locative denotation and inserted in the Loc position of the string, as in (26), which illustrates the position of the gclitic relative to the accusative clitic in N. (26) Modena DOpDR QPLoc N I |||| | amgla m ε t We can assign to the gclitic a lexical entry which directly reflects both its locative interpretation, and its insertion point in Loc, by associating the gmorpheme with the Loc position within the clitic noun phrase as in (27). 134 CJL 1, 2002 M. Rita Manzini; Leonardo M. Savoia Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 134 (27) Modena [Loc g[I [ N]]] While in sections 2.1-2.2 we have analyzed dialects in which the descriptive category of dative corresponds to a Q element, i.e. a distributor, in dialects of the Modena type the descriptive dative corresponds to locative properties, connected to the spatial coordinates of the discourse and the event. Thus if possession is a sort of location (cf. Freeze 1992), the classical idea of Kayne (1984) that double object verbs embed a small clause, where the dative is the possessor of the accusative argument, amounts to a locative interpretation of the dative. More generally, typical dative-accusative verbs such as give can be described in terms of a change in the spatial location of the internal argument; thus John gave a book to Peter implies that the book, located at John at the beginning of the event, changed its location to Peter at the end of the event. To complete our discussion it is worth mentioning that in several Italian dialects the identification of dative and locative involves an iclitic which like those considered in previous sections is morphologically a 3rd person form, coinciding with the accusative (masculine) plural. A case in point is the Lombardy-type dialect of Casaccia, where in descriptive terms iis the masculine plural accusative, as in (28b), as well as the dative in (28a) and (28c), and the locative in (28d). (28) Casaccia (Grisons) a. a i 'di:ʃvar'got he to.him says something ‘He says something to him.’ b. a i 've he them sees ‘He sees them.’ c. a i al 'di:ʃ he to.him it says ‘He says it to him.’ d. a i an 'met ent 'doi he to.it of.them puts inside two ‘He puts two of them inside it.’ On the basis of the discussion in sections 2.1-2.2 we are led to analyze the i clitic of the Casaccia dialect as a pure Q form, as in (29), merged in the Q position of the string. This means that a dialect like Casaccia, though superficially similar to dialects like Modena, turns out to be specular to them with respect to the lexicalization of dative and locative. A dialect like Modena never lexicalizes a distributor, but inserts a locative in contexts where other languages may have a distributor. Viceversa, we are led to claim that a dialect like Casaccia never lexicalizes a locative (not of the ‘there’ type in any event), but rather inserts a distributor in contexts where other languages have a locative. Lexicalization Patterns of the So-called 3rd Person CJL 1, 2002 135 Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 135 (29) Casaccia [Qi[ I [ N]]] If our construal of the data is correct, we expect several consequences to follow. In particular, we predict that in languages that have both distributors and locatives, they should be licenced in the same eventive environments and therefore alternate to a large extent. This prediction is verified in standard Italian by examples of the type in (30a). The occurrence of locative hi and dative li in many of the same contexts is discussed for Catalan by Rigau (1982), who accounts for the alternation in terms of animacy. Both the distributor gli and the locative ci are however interpreted as inanimate in (30a). Another potential prediction of the account introduced here for Modena and Casaccia is that there will be eventive environments that though compatible with Modena’s gwill not be compatible with Casaccia’s i or viceversa, leading to no lexicalization of one of the two forms. It is indeed frequently the case that dialects of the Casaccia subgroup (including in particular Piedmontese dialects) will present a reduced occurrence of iin locative contexts, where a specialized locative is lexicalized by other dialects. An example for Casaccia is provided in (30c), which represents the native speaker’s translation of standard Italian (30b): (30) a. Gliene/ Ce ne attacco due to.it-of.them /there of.them I.stick two ‘I stick two of them to/on it.’ b. Mi ci manda me there he.sends ‘He sends me there.’ c. Casaccia a m 'manda he me sends ‘He sends me (there).’ That the same event (or state) can support different argumental series is wellknown from the literature on phenomena such as the locative alternation whereby I loaded the wagon with hay alternates with I loaded the hay on the wagon. The approach often taken in the literature (Levin and Rappaport 1995) is that there are underlying arrays of arguments which can be differently linked to syntactic structures. Here we rather take the view that the superficially seen array is the only real one; thus it must be admitted that a verb such as load is compatible with a construal of the location as a location or as an internal argument (accusative). To take another example, Longa, Lorenzo and Rigau (1998) note that the same locative environments, with verbs of the ‘to be’ class, support in the different dialects of Iberian Romance either a locative clitic (Catalan) or an accusative neuter clitic (Galician, Asturian, Northwestern Spanish). The authors make a point that the appearence of the accusative neuter in the latter dialects is the implementation of 136 CJL 1, 2002 M. Rita Manzini; Leonardo M. Savoia Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 136 a default strategy, the Clitic Recycling Strategy, requiring ‘Use the (most) unmarked clitic to fill in gaps of the system’. Thus the appearence of the neuter accusative reflects lack of a locative in the system, given presumably an underlying argumental array. The point of view taken here is of course different, namely that there are no underlying abstract arrays corresponding to either optimal or default surface lexicalizations; but rather that the same verbal environments can truly support different argumental arrays. Several dialects of the Southern Lucania/ Northern Calabria area (the so-called Lausberg area) and of the Salento, again do not have a morphologically 3rd person form of the dative but lexicalize instead a ne clitic, which generally coincides with the partitive form. Thus for Nocara, (31a)-(31b) exemplify the lexicalization of the interpretation corresponding to a 3rd person dative by the n ə clitic, both by itself and in combination with an accusative clitic. The example in (31c) illustrates the case in which n ə lexicalizes the partitive; finally (31d) shows that n ə can occurr twice in the string, giving rise to a combination of its two possible interpretations. The alternation between ð aand ð a ðə is phonologically determined. (31) Nocara (Calabria) a. nə'ða stu 'kundə to.him he.gives this thing ‘He gives this to him.’ b. n u/a/i 'ðaðə to.him it-m./it-f./them he.gives ‘He gives it/ them to him.’ c. mənə'ðad'du:jə to.me of.them he.gives two ‘He gives two of them to me.’ d. nənə'ðad'du:jə to.him of.them he.gives two ‘He gives two of them to him.’ The fact that two n ə 's cooccur as in (31d) shows that there must be at least two points of insertion available to the clitic. On the basis of the discussion of partitives in section 1, the lower one can be identified with N; in such case n ə will receive what we call a partitive interpretation, connected to the internal argument of the verb, and will therefore appear after other clitics, such as P clitics in (32a). The n ə clitic that lexicalizes the 3rd dative interpretation in connection with accusative clitics can on the other hand correspond to a structure of the type in (32b), where n ə is inserted in the higher R position. The cooccurrence of the two n ə clitics is then predicted to be possible, as in (32c), with the lower n ə inserted in N and interpreted as the partitive, and the higher one in R. On the other hand, we already suggested in section 1, that the DOpposition of insertion is generally not available to a n ə clitic in that its DOpspecification must be interpreted in connection with a predicative N head and cannot be interpreted as an intensional specification of the sentence. Lexicalization Patterns of the So-called 3rd Person CJL 1, 2002 137 Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 137 (32) Nocara a. DOpDRQ P Loc N I ||| mə nə ða b. DOpDRQ P Loc N I ||| n u/a/i ðaðə c. DOpDRQ P Loc N I ||| n ə n əða If our characterization of the partitive in section1 is on the right track, n ə can be analyzed as in (33), where nrepresents a specialized DOpmorpheme, while ə , as in general the vocalic inflections of Italian dialects, is associated with the head position of the clitic noun phrase. For reasons discussed by Manzini and Savoia (forthcoming) we assume that ə does not lexicalize a nominal class specification, i.e. I, but rather corresponds to a N element. (33) Nocara [Op n[N ə ]]. In section 1, we provided an explanation for how the so-called partitive interpretation comes about. The question posed by dialects such as Nocara is how the ne-type clitic becomes associated with 3rd person dative interpretation. In particular we have proposed in section 1 that the ne-type clitic is not itself interpreted as an argument of the event but contributes to fixing the denotation of such an argument, namely the obligatory internal one. As we did in all of the cases that precede, we assume that the same basic characterization holds for contexts traditionally described in terms of a dative interpretation. Consider concretely the n u cluster in (32b). While uin N is associated with the internal argument of the verb, n in R concurs to the fixation of its reference, by introducing a partitive specification, or in traditional Case terms a genitive specification, of the N argument itself. This strategy is particularly close to the one we have just described for dialects of the type of Modena; in this latter case, the insertion of a Loc clitic fixes the coordinates of the internal argument of the verb, lexicalizing its possessor at the end of the event being described. Intuitively, languages like Nocara do the same thing, anchoring the reference of the internal argument of the verb at a possessor, which is lexicalized however as an DOpelement; thus Nocara’s (32c) corresponds roughly to ‘I give it (and it is) of his’. The final distribution of clitics to be considered in this section involves dialects of Calabria, where the si-clitic lexicalizes the impersonal and the 3rd person reflexive, exactly as described in section 1 for standard Italian (cf. Manzini and Savoia 2001), but also the 3rd person dative both in isolation, as in (34a) and in combination with other clitics, as in (34b). Because of its general properties, we predict 138 CJL 1, 2002 M. Rita Manzini; Leonardo M. Savoia Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 138 that given the right context, both a reflexive reading and a 3rd dative one are equally salient and available. This is indeed the case in an example such as (34c), which thus exemplifies also the reflexive reading of si, meaning both ‘he buys it for him’ or ‘he buys it for himself’. (34) S.Agata del Bianco (Calabria) a. si 'ðunanu 'kistu to.him they.give this ‘They give this to him.’ b. s u 'ðunanu (a'ð iʎʎu) to.him it they.give to.him ‘They give it to him.’ c. 'iʎʎu s u k'khattha he to.him/to.himself it buys ‘He buys it for him/himself.’ Taking up again the analysis of impersonal and reflexive si hinted at in section 1, we associate si of the relevant Calabrian dialects with a structure of the type in (35), where the sformative is associated with the Q specification internal to the clitic constituent. The imorpheme can in turn be identified with the I nominal head, since in the relevant dialects, which have a rather different inflectional structure from the other Italian dialects considered so far, it arguably corresponds to a nominal class specification, rather than to a Q morpheme itself. (35) S.Agata del Bianco [Qs[Ii[N]]] The Q categorization of the si clitic in (35) forms the basis for its insertion in the clitic string, which targets the Q position, preceding in particular accusative clitics in N, as illustrated in (36). (36) S.Agata del Bianco DOpDRQ PLoc N F ||| si w kkhattha Manzini and Savoia (2001) argue that the interpretive properties of impersonal and reflexive si can be naturally derived from its characterization as a quantificational variable (Manzini 1986). In particular the so-called impersonal reading is the result of the binding of the si variable by a generic operator (Chierchia 1995), while the reflexive reading implies a pronominal interpretation dependent on an antecedent. The discussion of morphologically 3rd person datives in sections 2.1 and 2.2 above as Q elements, lexicalizing a distributivity property, suggests a similar treatment for the cases involving dative si such as (36). In other words, we are Lexicalization Patterns of the So-called 3rd Person CJL 1, 2002 139 Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 139 led to propose that in appropriate environments the quantificational properties of si can equally well receive a distributive reading in the relevant languages, hence conventionally a dative one. 2.4. Summary It is worth stopping at this point to briefly summarize the conception of parametric variation emerging from the preceding discussion. In general both traditional and generative analyses imply that there is a common nucleus of syntactic and semantic properties that are properly labelled together as a distinctive category of dative. In this perspective, the parametrization between languages would have to do with the particular way in which these same properties are morphologically realized, for instance by a specialized form (say of the itype) or by suppletion, typically construed as replacement by an underspecified form (say si). As already noted at the outset, our theory programmatically avoids reference to what we consider to be theoretically expensive notions of underspecification or default; nor does it conceive of parameters in terms of the overt realization of the same underlying semantico-syntactic units by different lexical material. This is particularly evident if the proposals being advanced here are compared with the model of Halle and Marantz (1993), in which syntactic operations manipulate features and lexical insertion is Late, meaning at the end of the syntactic cycle. In the present model, as in the minimalist model of Chomsky (1995), syntactic structures are conceived as the result of applying the operation Merge to actual lexical material. Therefore, we exclude that there is a ‘dative’ category, or a predefined ‘dative’ set of features, which remains constant in the face of superficial variation. On the contrary, where a language like standard Italian (or Vagli or Olivetta or Piobbico) lexicalizes a morphologically 3rd person distributor, another language such as S.Agata may lexicalizes a si-type distributor. Another possibility is the lexicalization of an DOpspecification of the N argument, i.e. ne as in the Nocara language, or of a Loc clitic specifying the spatial coordinates of the N argument, as in the Modena language. This latter case is interesting also in that it is equally possible to find languages, such as Casaccia, where a morphologically 3rd person element can be used not only in traditional dative contexts but also in locative ones. 3. Mutual exclusion and suppletion 3.1. The ‘Spurious se’ pattern In sections 2.1 and 2.2 above we have considered several languages in which clusters of morphologically 3rd person datives and accusatives are possible in either order; while in section 2.3 we have illustrated several languages which lack a morphologically 3rd person dative independently of its clustering with other clitics. As we have already mentioned, in some Romance languages a morphologically 3rd 140 CJL 1, 2002 M. Rita Manzini; Leonardo M. Savoia Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 140 person dative is excluded by clusters including a 3rd person accusative, though it surfaces in isolation or in combination with other clitics. This mutual exclusion between 3rd dative and 3rd accusative clitics has received wide attention in the literature, as have the suppletion phenomena to which it apparently gives rise. The best known single instance of the dative-accusative mutual exclusion pattern in Romance languages is the so-called ‘Spurious se rule’ of Spanish. The discussion in section 2 is directly relevant to this complex question, in that the apparently suppletive pattern produced by the ‘Spurious se rule’ of Spanish, whereby se receives the dative interpretation in combination with an accusative clitic, is actually found in some languages (S.Agata) independently of any mutual exclusion. This amounts to saying that the pattern emerging from the apparent suppletion mechanism does not require any explanation beyond those provided above for languages where suppletion is not found. Among Italian dialects, the spurious se pattern is attested by Sardininian ones. The essential data are reproduced in (37) for the dialect of Làconi. The language has a specialized dative form which emerges in isolation, as in (37a), and a full accusative paradigm, illustrated in (37b). In combination with an accusative, the dative interpretation is however conveyed by the si clitic, as in (37c). Both the accusative clitic and the dative clitic appear to follow all other clitics, such as the P clitic or the partitive in (37d)-(37e). (37) Làconi (Sardinia) ai/ izi a k'kustu to.him/to.them gives this ‘He gives this to him/ them.’ b. u/ a/ uzu/ aza b'biu him/her./ them-m./them-f. I.see ‘I see him/her/them.’ c. si u'aða to.him it gives ‘He gives it to him.’ d. mi u'aða to.me it gives ‘He gives it to me.’ e. ndi i a d'duaza of.them to.him gives two ‘He gives two of them to him.’ The analyses that precede provide us with a basis for the systematization of both the clitic inventory and the insertion positions involved in a dialect like Làconi. The evidence concerning the position of both accusative and dative clitics is compatible with the conclusion that the insertion point of both clitics is N. This explains the fact that they surface to the right of all other clitics, including P clitics which preLexicalization Patterns of the So-called 3rd Person CJL 1, 2002 141 Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 141 The basic properties of the dialect of Nociglia, that account for the complementary distribution of morphologically 3rd person dative and accusative, are not unlike those already considered in section 3.1 for Làconi. We associate the clitics of the accusative series lu/la/li/l ε with the lexical entries in (46), where the lmorpheme corresponds to a nominal I head while the vocalic morpheme that follows it occupies the I head of an embedded nominal or, in the case of u, its N head. The imorpheme, associated with the plural or distributive (dative) interpretation is inserted in Q. (46) Nociglia a. [Il... [ I [Nu]]] b. [Il... [Ia[ N ]]] c. [Il... [I ε [ N ]]] d. [Il... [Qi[ I [ N ]]]] Because of their nominal properties, the clitics in (46) are inserted in the N position of the string. This holds in particular for the li form, which even as a distributor follows all other clitics it can cooccur with, including the partitive. If a high position was available to the dative, say Q, we could expect the partitive to occur in N, hence to the right of the dative. Instead the partitive is presumably inserted in R, as in (47). As discussed more than once, the availability of R to the partitive clitic but not to the dative depends on an independent parameter. Thus nε admits of lexicalization in the scopal specificity position corresponding to R, while the lseries, including li, is constrained to the N position. The insertion of all clitics in (46) in N position means of course that they are in complementary distribution, excluding in particular the combination of li as a distributor with another clitic of the series. (47) Nociglia DOpDR Q P Loc N I ||| nε li dajε The impossibility of combining two clitics of the set in (46) in a dative - accusative cluster gives rise to the apparent substitution of the dative by nε . According to the discussion in section 2.3 above concerning Nocara, nε can be analyzed as a DOp element, whose interpretation contributes to fixing the reference of the internal argument N of the event. This is true both of the partitive interpretation of nε and of its so-called dative one, whereby (45c) informally corresponds to ‘they give it (and it is) of his’. As in the case of ‘spurious se’, the problem is why the partitive takes on this particular interpretation only in the presence of an accusative clitic. In the traditional perspective, taken up and theorized by optimality accounts, inserting the more specialized form of dative i.e. li is necessary when possible; insertion of nε in its place is just a last resort option for those contexts where insertion of li is not possible. The approach taken here however sees the alternation between the lexicalization of li and nε in a radically different light, since the two clitics effectively lexicalize different interpretive contents, which can only descriptively be imputed to a common label of dative. In this respect, it is crucial that li 148 CJL 1, 2002 M. Rita Manzini; Leonardo M. Savoia Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 148 does not in any way represent a specialized dative, as optimality treatments would imply, since it is also the accusative masculine plural; it is not obvious therefore that li has more features in common with the genderand numberneutral dative than nε . In this sense neither li nor nε represents an optimal solution to some underlying 3rd person dative feature; or more precisely, both of them are equally optimal solutions, if as Chomsky (1995) puts it syntax is an optimal solution to the problem of interfacing LF and PF. We must therefore conclude that the the child learning the Nociglia language learns a slightly more complex system than those considered so far, in which distributivity is lexicalized only in contexts where the internal argument is not a definite clitic pronoun. When the internal argument is such a clitic, what is lexicalized in the same contexts is an DOpspecification. The last typology to be considered here involves dialects of Central and Southern Italy where a morphologically 3rd person form of dative emerging in isolation and in non-accusative contexts, alternates with a Loc clitic in clusters with accusatives, as illustrated in (48). As before, (a) provides the isolation form of the dative, while (b)-(c) illustrate the accusative paradigm; note that the accusative plural form differs from the dative form in that the former but not the latter triggers gemination of the following consonant (a type of ‘raddoppiamento fonosintattico’). As can be seen in (d)-(e), it is the Loc clitic that combines with the accusative in dative contexts, exactly as in locative ones, cf. (g). The combination of dative and partitive furthermore gives rise to the order li - ne, as illustrated in (f). (48) Celle di Bulgheria (Campania) a. li 'danu 'kistu to.him they.give this ‘They give this to him.’ b. lu/la 'viðinu him/her they.see ‘They see him/her.’ c. li b'biðinu them they.see ‘They see them.’ d. di lu 'danu to.him it they.give ‘They give it to him.’ e. di li d'danu to.him them they.give ‘They give them to him.’ f. li si nni 'daji 'trɔppu to.him one of.it gives too much ‘One gives too much of it to him./ He is being given too much of it.’ g. di'mittu 'kistu there I.put this ‘I put this there.’ Lexicalization Patterns of the So-called 3rd Person CJL 1, 2002 149 Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 149 The analysis of the clitic set in the dialect of Celle cannot abstract from the fact that the relative order of clitics is compatible with a high insertion position for the dative itself. The latter in general precedes the clitics it cooccurs with, including the partitive and the si clitic; this suggests an R insertion position, as illustrated in (49). (49) Celle di Bulgheria DOpDR Q P Loc N I || | | li si nni daji An immediate consequence of the high insertion position of the dative in (49) is that the complementary distribution between datives and accusatives cannot be explained simply by the fact that they compete for the same N position. This situation, though not considered so far, is far from rare in Italian dialects; in other words, there are many dialects where mutual exclusion patterns are found even when two or more different positions in the string are available for insertion of the relevant clitics. Several such cases are considered in detail by Manzini and Savoia (to appear) who provide an explanation depending on the lexical properties of the clitics themselves. The idea is that the lmorpheme of clitics whose insertion excludes that of other clitics of the same series lexicalizes all properties it is associated with for the whole clitic string. In particular then the insertion of an l clitic prevents the re-lexicalization in the string of the nominal properties associated with the lmorpheme, and interpretively connected in our model to 3rd person reference. Let us then consider the accusative series in the Celle di Bulgheria dialect, i.e. lu/ la/ li . As discussed above, we associate the lmorpheme with the I head of a nominal constituent; in turn the uand amorphemes can be associated with their own nominal head, namely an I head in the case of aand an N head in the case of u. Since the plural li provokes phonosyntactic gemination of the following consonant, we are led to assume that its lexical entry includes an abstract final consonant. Following previous discussion, i is a Q formative; the fact that an abstract consonant enters into the interpretation of the clitic as plural suggests however an overall analysis on the model of Sardinian (39) in which it is the consonant that morphologizes plural. This yields a clitic paradigm of the type in (50). (50) Celle di Bulgheria a. [I l... [ I [Nu]]] b. [I l... [Ia[ N ]]] c. [I l... [Qi[I [Q C[ N ]]] d. [I l... [Qi[I [ N ]]] The idea that insertion in any position of the clitic string of an lclitic of the type in (50) succeeds in lexicalizing the relevant nominal properties of the whole 150 CJL 1, 2002 M. Rita Manzini; Leonardo M. Savoia Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 150 string means that if the li clitic in (50d) is inserted in R, it prevents the insertion of an accusative clitic in N. Viceversa insertion of an accusative clitic in N, excludes that of the distributor li in the higher R position for the same reason. This explanation does not touch the possibility of combining the clitics of the lseries in (50) with other clitics, which do not have the relevant l-type properties; hence li can be combined with si and the partitive as in (49), and the accusative can of course be combined with the locative, as in the suppletion pattern in (48d). As for the apparent suppletion patterns itself, its explanation follows already familiar lines. On the one hand we have indicated in some detail in section 2.3 how the Loc clitic can come to lexicalize contexts which in other languages may be lexicalized by a 3rd person distributor (the so-called specialized dative). On the other hand in a language like Celle it is only in combination with a 3rd person accusative that the Loc clitic takes on the so-called dative interpretation, i.e. one in which it provides the possessor coordinatines of the internal argument of the verb. As in the case of ne suppletion patterns, we will assume that this relatively complex distribution is learned by the native speaker. The discussion concerning Celle di Bulgheria, in basing the mutual exclusion between dative and accusative on a lexical property of the lmorpheme implies that it is independent of the status of the clitic as a dative or as an accusative. Indeed Manzini and Savoia (to appear) show that many Northern Italian dialects with subject clitics do not allow for the combination of a 3rd person subject clitic with a 3rd person object clitic. Interestingly in the simplest case this mutual exclusion leads to the lexicalization of only one of the two clitics, namely the accusative; in many dialects the accusative takes a fixed form, reminiscent of the form taken by the accusative in the dative-accusative pattern of the Piobbico type. In no cases that we know of, one of the two clitics can actually be substituted by a different form altogether (ci, ne, si etc.). This further clinches the argument in favor of the conceptual and empirical separation of mutual exclusion and suppletion. 4. General summary On the evidence of our discussion of the so-called dative, the traditional morphological category of Case is a spurious one. In some languages, indeed, reference to the dative reduces to reference to the spatial (Loc) or other (DOp) coordinates of the internal argument of the event. In other languages, reference to the dative is introduced by means of a category which appears to be associated with quantificational properties, interpreted both as distributivity and as plurality (as with the type i/li), or as genericity (in the case of si). What interests us directly is that all empirical elements are in place for concluding that crucially dative is a descriptive category and does not correspond to a syntactic category. What is more, the categories that we adopt in alternative to dative characterize intrinsic denotational content; thus Loc is interpreted with reference to the locative coordinates of discourse, Q is interpreted as plurality, distributivity, genericity, etc. In no case are the relevant categories characterized by relational properties such as Case would be. Our discussion suggests furthermore that the conclusions just drawn for the Lexicalization Patterns of the So-called 3rd Person CJL 1, 2002 151 Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 151 dative hold for Case categories of traditional grammar in general. Thus traditional accusative reduces to the internal argument interpretation (forcing a reanalysis of ECM), while nominative can be construed as another name for the EPP property. The observation that in many dialects the dative coincides with the accusative plural (masculine) is in fact directly relevant not only to the status of the traditional feature of Case but also to that of number, which traditionally represents the distinction between singular and plural. In fact, the discussion that precedes supports the conclusion that there is no independent number category, but rather an all-purpose Q category underlying weak quantification, which encompasses plurality, as it does numeral quantification and more. Nor do the other traditional phi-features survive a careful analysis. The gender category is in fact problematic even within the framework of Chomsky (1995), at least if we want to enforce the idea that agreement features are interpretable on Nouns; for gender corresponds to a property with referential import (roughly feminine sex) only in a small subset of cases in the Romance languages. On the contrary a characterization of gender that will hold true in all cases is that it corresponds to a nominal inflection class, as we assumed throughout this article. Thus in a language like standard Italian the so-called masculine (-o) and feminine (-a) coincide with two separate inflectional classes, to which must be added a third (-e) class which can combine with either of the preceding (i.e. it is either feminine or masculine in traditional terms). Concerning Person, it is of course a category of our grammar; but its content is not that of traditional (and generative) treatments opposing speaker (1p), hearer (2p) and others (3p). Rather we take it that P(erson) coincides with 1st and 2nd person, whose distribution and general behavior differs from those of traditional 3rd person. Another respect in which the present approach differs from the others found in the literature is that it does not introduce any form of comparison between derivations in the grammar to account for the ‘preference’ of one clitic over another according to context. That comparison between derivations (or representations) is involved is particularly evident in the recent optimality treatment of Romance clitics by Grimshaw (1997, 1999). In essence according to Grimshaw (1997, 1999) lexical insertion takes place on the basis of the need to safisfy the maximum possible amount of constraints defined by the grammar. This means that in isolation the closests match to a 3rd person dative, in some languages a dedicated form, is inserted. If for some reason the dedicate form is unavailable, the grammar provides for the insertion of a severely underspecified element such as se, other positively specified elements necessarily violating more constraints than it does. Essentially the same conceptual schema, based on the implicit or explicit comparison between derivations or representations is in fact implied by morphological theories that use Elsewhere as the basic lexical insertion principle, effectively the main line of generative morphology down to current Distributed Morphology frameworks (Halle and Marantz 1993). Our account of the relevant phenomena makes use of no Elsewhere principle, with the allied notions of underspecification or default, nor of comparison between derivations/representations. It seems to us that to the extent that such notions rep152 CJL 1, 2002 M. Rita Manzini; Leonardo M. Savoia Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 152 resent an enrichement of the grammar the present account has an edge over its competitors. As for notions of comparisons of derivations or representations, recall that though they play some role in the earlier minimalist framework of Chomsky (1995) they have shown to be not only unnecessary, but to effectively derive the wrong results in more recent statements of the theory (Chomsky 2000, 2001) where they are altogether abandoned. As for notions of Elsewhere, and the attending concepts of underspecification and default, we note that these notions have been discounted in the very phonological domain in which they have first arisen (cf. the government phonology literature, e.g. John Harris 1994). References Abney, S. (1986). The English noun phrase in its Sentential Aspect. MIT, doctoral dissertation. Beghelli, F. (1997). «The syntax of distributivity and pair-list readings». In: Szabolcsi, A. (ed.). Ways of Scope Taking. Dordrecht: Kluwer, p. 349-408. Beghelli F.; Stowell T. (1997). «Distributivity and negation: The syntax of each and every». In: Szabolcsi, A. (ed.). Ways of Scope Taking. Dordrecht: Kluwer, p. 71-107. Bernstein, J. (1997). «Demonstratives and reinforcers in Germanic and Romance languages». Lingua 102: 87-113. Bonet, E. (1991). Morphology after syntax: Pronominal clitics in Romance. MIT, doctoral dissertation. Bonet, E. (1995). «Feature structure of Romance clitics». Natural Language and Linguistic Theory 13: 607-647. Brody, M. (1995). Lexico-Logical Form. Cambridge, Mass.: The MIT Press. Brugè, L. (1996). «Demonstrative movement in Spanish: A comparative approach». University of Venice Working Papers in Linguistics, 6.1:1-53. Calabrese, A. (1997). «Distributed Morphology and sentential complementation in the Salentino dialect of Italian». Harvard Working Papers in Linguistics 6: 47-85. Cardinaletti A.; Starke, M. (1999). The typology of structural deficiency: A case study of three classes of pronouns. In: van Riemsdijk, H. Clitics in the Languages of Europe. Berlin: Mouton de Gruyter, p. 145-233. Chierchia, G. (1995). «Impersonal Subjects». In: Bach, E.; Jelinek, E.; Kratz, A.; Partee B. H. (eds.). Quantification in natural languages. Dordrect: Kluwer, p. 107-143. Chomsky N. (1995). The Minimalist Program, Cambridge Mass: The MIT Press. Chomsky, N. (2000). «Minimalist inquiries: The framework». In: Martin, R.; Michaels, D.; Uriagereka, J. Step by step. Essays on Minimalist Syntax in Honor of Howard Lasnik. Cambridge, Mass: The MIT Press, p. 89-155. Chomsky, N. (2001). «Derivation by phases». In: Kenstowicz, M. Ken Hale. A Life in Language. Cambridge, Mass: The MIT Press, p. 1-52. Cinque, G. (1999). Adverbs and Functional Heads - A Cross-Linguistic Perspective. Oxford: Oxford University Press. Freeze, R. (1992). «Existentials and other locatives». Language 68: 553-595. Giorgi A.; Pianesi, F. 1998. Tense and Aspect. From semantics to morphosyntax. New York: Oxford University Press. Lexicalization Patterns of the So-called 3rd Person CJL 1, 2002 153 Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 153 Grimshaw, J. (1997). «The best clitic: Constraint conflict in morphosyntax». In: Haegeman, L. (ed.). Elements of grammar. Dordrecht: Kluwer, p. 169-196. Grimshaw, J. (1999). «Optimal clitic positions and the lexicon in Romance clitic systems». Rutgers University, unpublished manuscript. Halle, M.; Marantz, A. (1993). «Distributed morphology and the pieces of inflection». In: Hale K.; Keyser, S. J. (eds.). The view from Building 20. Cambridge, Mass.: The MIT Press. Halle, M.; Marantz, A. (1994). «Some Key Features of Distributed Morphology». In: Carnie, A.; Harley, H.; Bures, T. (eds.). Papers on Phonology and Morphology. MIT Working Papers in Linguistics 21:275-288. Harris, James (1994). «The syntax-phonology mapping in Catalan and Spanish clitics». In: Carnie, A.; Harley, H.; Bures, T. (eds.). Papers on Phonology and Morphology. MIT Working Papers in Linguistics 21: XXX. Harris, John (1994). English sound structure. Oxford: Blackwell. Kayne, R. (1975). French Syntax: The transformational cycle. Cambridge, Mass: The MIT Press. Kayne, R. (1984). Connectedness and Binary Branching. Dordrecht: Foris. Kayne, R. (1989). «Null subjects and clitic climbing». In: Jaeggli, O.; Safir, K. (eds.). The Null Subject Parameter. Dordrecht: Kluwer, p. XXX. Kayne, R. (1991). «Romance Clitics, Verb Movement and PRO». Linguistic Inquiry 22: 647-686. Kayne, R. (1994). The antisymmetry of syntax. Cambridge, Mass: The MIT Press. Kayne, R. (2000). Parameters and Universals. New York: Oxford University Press. Levin, B.; Rappaport, M. (1995). Unaccusativity. Cambridge, Mass.: The MIT Press. Longa, V.M.; Lorenzo, G.; Rigau, G. (1998). «Subject clitics and clitic recycling: locative sentences in some Iberian Romance languages». Journal of Linguistics 34: 125-164. Manzini, M.R. (1986). «On Italian si». In: Borer, H. (ed.). The Syntax of Pronominal Clitics (Syntax and Semantics vol. 18). New York: Academic Press, p. 241-262. Manzini M.R.; Savoia, L. (1998). «Clitics and auxiliary choice in Italian dialects: Their relevance for the Person ergativity split». Recherches linguistiques de Vincennes 27:115-138. Manzini M.R.; Savoia, L.M. (1999). «The syntax of middle-reflexive and object clitics: a case of parametrization in arbëresh dialects». In: Mandalà, M. (ed.). Studi in onore di Luigi Marlekaj. Bari: Adriatica, p. 283-328. Manzini M.R.; Savoia, L.M. (2001). «The syntax of object clitics: si in italian dialects». In: Cinque, G.; Salvi, G. (eds.). Current Studies in Italian Syntax. Essays to Honour Lorenzo Renzi. Amsterdam: North Holland, p. 234-264. Manzini M.R.; Savoia, L.M. (2002). «Parameters of Subject Inflection». In: Svenonius, P. (ed.). Subjects, Expletives and the EPP. Oxford: Oxford University Press. Manzini M.R.; Savoia, L.M. (to appear). «Clitics: Cooccurrence and mutual exclusion patterns». In: Belletti, A.; Rizzi, L. (eds.). The structure of CP and IP. New York: Oxford University Press. Manzini M.R.; Savoia, L.M. (forthcoming). I dialetti italiani: Morfosintassi. Alessandria: Edizioni dell’Orso. May, R. (1985). Logical Form. Cambridge, Mass.: The MIT Press. 154 CJL 1, 2002 M. Rita Manzini; Leonardo M. Savoia Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 154 Ouhalla, J. (1991). Functional categories and parametric variation. London: Routledge. Perlmutter, D. (1972). Deep and surface constraints in syntax. New York: Holt, Rinehart and Winston. Reinhart, T. (1983). Anaphora and semantic interpretation. London: Croom Helm. Rigau, G. (1982). «Inanimate indirect object in Catalan». Linguistic Inquiry 13: 146-150. Sportiche, D. (1996). «Clitic constructions». In: Rooryck, J.; Zaring, J. L. (eds.). Phrase structure and the lexicon. Dordrecht: Kluwer, p. 213-276 Szabolcsi, A. (1994). «The noun phrase». In: Kiss, K.; Kiefer, F. (eds.). The syntactic structure of Hungarian (Syntax and Semantics 27). New York: Academic Press, p. 179-274. Lexicalization Patterns of the So-called 3rd Person CJL 1, 2002 155 Cat.Jour.Ling. 1 001-259 26/2/03 15:53 Página 155