3rd Person Pronominal Clitics in Dialects of Catalan

3RD PERSON PRONOMINAL CLITICS IN DIALECTS OF CATALAN Euldia Bonet Universitat Autbnoma de Barcelona This paper provides a description and an analysis of combinations of two third person clitics, accusative and dative, in four varieties of Catalan. While in one of these four varieties, the output forms are straightfonvard, in the other three there is an increasing presence of opaque outputs, that is outputs that differ from the form the clitics have when they appear in isolation. Both the linear order among the clitics and the presence of opaque forms is argued to take place in the Morphology Component, between s-structure and PF (a syntactic approach to clitic order being rejected). In this component, the internal structure of the clitics, a hierarchical structure of features, can be altered by several types of processes (deletion, insertion): insertion is argued to apply in one of the varieties under discussion. Linear order among pronominal clitics is determined by the mapping onto a template. This mapping can be based on either whole clitics or individual terminal features. All possibilities exist in the four varieties, and in two cases they coexist in a given dialect. The rest of the opaque forms in the four varieties arise during the mapping operation, when two instances of a given feature are mapped onto one and the same slot. An additional source of variation lies on tbe feature on which [feminine] depends. When pronominal clitics are combined, they often adopt an opaque form; that is, they do not appear with the shape they have when they appear in isolation. There are several kinds of modifications clitics can undergo, and there is a very considerable amount of dialectal variation in this respect. Third person pronominal clitics are particularly prone to suffer modifications when they are combined with each other. In this paper, four different third person (accusative and dative) pronominal clitic systems of Catalan will be examined, which differ with respect to Catalan Worquing Papers in Linguistics (CWPL) 3.1 (1993): 8511 1 Universitat Autbnoma de Barcelona the output of clitic combinations but hardly with respect to the forms those clitics have in isolation, and a proposa1 will be made to account for the variation. One of these systems belongs to Central Catalan, and it is shared by a large number of speakers. This is the dialect with the most opaque forms. Another system that will be taken under consideration is Majorcan, in one of its varieties, not spoken by younger people nowadays. The other two systems to be examined correspond especially to standard varieties of Catalan. One of them is Standard Valencian, which is also spoken in parts of Valbncia and Castelló. The other one is the standard variety of Catalan for the rest of the Catalan speaking area, also known as Standard Catalan. For convenience, I will concentrate only on this variety as spoken in Central Catalan. The clitic system of this variety is part of the colloquial language spoken in certain, very limited, areas. Throughout this article I will use letters to refer to each system in order to avoid any possible confusions. The terms will be: variety A (Standard Valencian), variety B (Standard Caialan), variety C (Majorcan), and variety D (Central Catalan). The paper is organized as follows: in section 1 I present the relevant data conceming the four varieties of Catalan mentioned in the previous paragraph, organized in increasing degrees of opacity. In section 2 a proposa1 is made to account for all the opaque forms, which is based on morphology. Syntactic and phonological approaches will be rejected in view of the characteristics of the data being analyzed. The paper also contains a last section with conclusions. 1. The data I .I. Variety A This variety of Catalan, spoken in certain areas of Valbncia and Castelló, has no opaque forms in combinations of two third person clitics, accusative and dative. As can be seen in most cases, the dative clitic precedes the accusative clitic. These facts are illustrated in (1) and (2). In (la-b), and (2a-b), the clitics appear in isolation. The examples in (lc) and (2c) contain combinations of the previously shown clitics. In all the examples throughout this paper I use the Clitic-Left Dislocation construction in order to clearly show the inflectional characteristics of the antecedent. All clitics and clitic combinations appear in boldface. (1) a. Al xiquet, li donaré la poma més tard. to-the boy 3sgDat give-lsg-FUT the apple later b. La poma, la donaré al xiquet més tard. the apple(fem) 3sgfemAcc give-lsg-FUT to-the boy later c. Al xiquet, la poma li la donaré més tard. to-the boy the apple 3sgDat 3sgfemAcc give-lsg-FUT later 'I will give the apple to the boy later.' (2) a. Als xiquets, els donaré les pomes més tard. to-the boys 3plDat give-lsg-FüT the apples later b. Les pomes, les donaré als xiquets més tard. the apples(fem) 3plfemAcc give-lsg-FUT to-the boys later c. Als xiquets, les pomes, els les donaré més tard. to-the boys the apples 3plDat 3plfemAcc give-lsg-FUT later 'I will give the apples to the boys later.' As can be seen in (3), below, the third person dative plural clitic is phonologically identical to the third person accusative plural (masculine) clitic. This is so in all the varieties of Catalan that will be taken into consideration, except in one (variety D). As will be shown later, the identity between the third person accusative plural clitic and the third person dative plural clitic is only phonological, not morphological. The forn of the third person dative clitic in isolation can be seen in (2a). (3) Als xiquets, els préssecs, els els donaré més tard. to-the boys the peaches(masc) 3plDat 3plmascAcc give-lsg-FüT later 'I will give the peaches to the boys later.' 1.2. Variety B The third person pronominal clitic system of this variety of Catalan is identical to that of Valencian in all respects except when a third person accusative clitic is combined with a third person dative singular clitic. In this case opaque outputs arise given that the third person dative singular clitic, li in isolation, seems to adopt the form /i/ (spelled hi). (4b) illustrates the combination of a third person accusative clitic with a third person dative plural clitic. As can be seen by comparison with (2c) the outputs in these two varieties are identicai: (4) a. Als nois, els donaré les pomes més tard. to-the boys 3plDat give-lsg-FUT the apples later b. Les pomes, als nois, els les donaré més tard. the apples(fem) to-the boys 3plDat 3plfemAcc givelsg-FUT later 'I will give the apples to the boys later.' (5a) shows the form of the third person dative singular clitic in isolation. In (5b) this clitic appears combined with a third person accusative clitic, and adopts the form li1 (hi). As shown, the non-opaque outputs les li and li les are non-existent. (5) a. Al noi, li donaré les pomes mCs tard. to-the boy 3sgDat give-lsg-FUT the apples later b. Al noi, les pomes, les hi donar6 més tard. to-the boy the apples(fem) 3plfemAcc 3sgDat (/i/) give-lsg-FUT later (*les li 1 *li les) 'I will give the apples to the boy later.' It is also important to note that, while in combinations with a third person dative plural clitic the surface order is dative > accusative, when the third person dative clitic is singular (when the opaque forms arise), the order is accusative > dative. 1.3. Variety C This variety of Catalan, spoken in Majorca, has no transparent combinations; all the outputs in combinations of two third person clitics are opaque. Like in variety B, the third person dative singular clitic, Ili1 in isolation, adopts the form li/ when it is combined with an accusative clitic. Moreover, the third person accusative clitic never shows feminine agreement, even though a feminine marker is present when this clitic appears in isolation and the antecedent is feminine. These two aspects are illustrated in (6a-c). I also give, in parentheses, the non-existent transparent forms: (6) a. Ses pomes, les donaré a s'al.lot més tard. the apples(fem) 3plfemAcc give-lsg-FüT to the-boy later b. A sfal.lot, li donaré ses pomes més tard. to the-boy 3sgDat give-lsg-FLJT the apples later c. Ses pomes, a sfal.lot, [alai donaré més tard. the apples to the-boy 3sgDat+3plfemAcc (Ilzil) give-lsg-FUT later (*les li I *li les) 'I will give the apples to the boy later.' If one were to say that the forn /i/ corresponds to the dative clitic, the surface order between the two clitics in (6c) would be accusative > dative, like in variety B and unlike variety A. I give the opaque ihird person dative output in its phonological forn only because there is no standard spelling for ihis variety. Othenvise li1 would be spelled hi. I will give only phonological forns in such cases. In combinations of two third person clitics in which the dative clitic is plural other opaque forms arise. In this case, the dative plural clitic, els (/lz/) remains transparent but the accusative clitic becomes /u/ in all cases, regardless of the gender and number of the antecedent. This "change" to /o/ is also existent in certain varieties of Valencian, as described in Todoli (1992). Catalan has an independent clitic Iol (or /u/), ho, the so-called neuter clitic, and later it will be claimed that, through a specific morphological process, the third person accusative clitic becomes a neuter clitic. This "change" is illustrated in (7c). The clitics that show up in an opaque form in this combination appear in isolation in (7a,b). As done earlier, I include, after (7c) and in parentheses, the non-existent transparent output forms for the combination: (7) a. Ses pomes, les donaré a ets al.lots més tard. the apples(fem) 3plfemAcc give-lsg-FLJT to the boys later b. A ets al.lots, els donaré ses pomes més tard. to the boys 3plDat give-lsg-FUT the apples later c. Ses pomes, a ets al.lots, [alzo] donaré mCs tard. the apples to the boys 3plDat+3plfemAcc [ah] give-lsg-FUT later (*els les / *les els) 'I will give the apples to the boys later.' If we compare (7a) with (7c) we will see that, even though the third person accusative clitic in (7a) is femine and plural, those markers do not show up in the combination in (7c). Instead, the invariable form /o/ appears. While the surface order could be said to be accusative > dative when the dative clitic was singular (cf. (6c)), the surface order would be dative > accusative whenever the dative clitic is plural (cf. (6c)). 1.4. Variety D Variety D corresponds basically to Central Catalan, and has no transparent outputs in combinations of two third person clitics. A basic difference between variety D and the other varieties previously described is that the form of the third person dative plural clitic is Ilzil, while in the other varieties this clitic has the phonological form Ilzl. As in Majorcan, no feminine marking ever appears in such combinations. Moreover, whenever the two clitics are singular, the output form of the combination is Ilil, while whenever one of the two clitics or both of them are plural, the output form is IIziI. These aspects are illustrated in (8) and (9), below. (8) a. La poma, la donaré als nois més tard. the apple(fem) 3sgfemAcc givelsg-FUT to-the boys later b. Als nois, [alai donaré la poma més tard. to-the boys, 3plDat give-lsg-FüT the apple later c. Als nois, la poma, [hi donart més tard. to-the boys the apple 3plDat+3sgfemAcc (Ilzil) give-lsg-FUT later (* [l*ll *[Mal) 'I will give the apple to the boys later.' In (8b) we see the form of the third person dative plural clitic in isolation, IIziI. As can be seen in (8c), the feminine marker that was present in (84 does not appear in this last example. From examples like (8c), we might conclude that what is really happening in combinations of two third person clitics is that only the dative clitic shows up in such combinations. Recall that in this variety the phonological form of the third person dative plural clitic in isolation is Ilzil, like the forn that appears in (SC). The omission of the third person accusative clitic is in fact common in certain dialects of French and also in very restricted areas of the Catalan speaking domain. However, this cannot be the case in variety D, as example (9c) shows: (9) a. Les pomes, les donaré al noi mts tard. the apples(fem) 3plfemAcc give-lsg-FUT to-the boy later b. Al noi, li donar6 les pomes m6s tard. to-the boy 3sgDat give-lsg-FUT the apples later c. Al noi, les pomes, [ah] donaré més tard. to-the boy the apples 3plfemAcc+3sgDat give-lsg-FUT later (*li les /*les li) 'I will give the apples to the boy later.' As can be seen in (9c), the output form of the combination is Ilzil, like the third person dative plural clitic in isolation, but in this case the antecedent of the dative is singular. Therefore we cannot conclude that the ouput of the combination simply ignores the accusative clitic. The plural marker in (9c) precisely has to come from the accusative clitic. In (9c) it can also be seen that the feminine marker coming from the accusative clitic does not show up in the combination. (IOC), below, illustrates the fact that the form llzil is also the ouput form in combinations of two third person plural clitics. As usual, in (10a,b) I give the form of the clitics involved in (10c) as they appear in isolation. I also include the non-existent transparent possible forms: (10) a. Les pomes, les donar6 als nois m6s tard. the apples(fem) 3plfemAcc givelsg-FUT to-the boys later b. Als nois, [al3 donar6 les pomes m6s tard. to-the boys 3plDat give-lsg-FUT the apples later c. Als nois, les pomes, [alai donar6 m6s tard. to-the boys the apples 3plDat+3plfemAcc (Ilzil) givelsg-FUT later (* [alzilas] I * [laaahi) 'Donar6 les pomes als nois m6s tard.' In this variety of Catalan it is impossible to tell what the order is between accusative and dative. 1.5. Summary of the facts In (1 l), below, a chart is given with the output phonological forms of the combinations of two third person dative clitics, accusative and dative, in the four varieties of Catalan previously described. In many cases, some epenthetic vowels will be added (d=dative; a=accusative; s=singular; p=plural; m=masculine; f=feminine):2 (11) var. A var. B var. C var. D d.s. + a.s.m. Ai11 Ai1 Ai1 Ai1 ds. + a.s.f. Aild Ilail Ai1 Ai1 d.s. + a.p.m. Aild Azil Azil ~zil ds. + a.p.f. Ailed Ilazil nzil nzil d.p. + a.s.m. Azll Azll /lzo/ Azil d.p. + a.s.f. nzld Ilzlal Ilzol Azil d.p. + a.p.m. nzld nzld I~ZOI nzi/ d.p. + a.p.f. Azled Ilzlad /lzol llzil In (12), below, I give the forms of all the third person clitics as they appear in isolation. All varieties have basically the same forms in all the cases, except when the third person dative clitic is plural. In this case, variety D has a different form. Another difference, which does not have any bearing on the topic of this paper, is the phonological form of the feminine marker. In most varieties it has the form la1 (or some /VI from which la1 can be derived by Vowel Deletion), while in variety 1 it has the form /a/ when feminine cooccurs with singular, and /e/ when it cooccurs with plural:3 2 The forms given in (11) are the ones the clitics have in preverbai position. In postverbai position there are some modifications not related to the topic of this paper. 3 This pattern can aiso be found in Standard Catalan (variety B) as spoken in the north-eastenl region of the Catalan area (18) - - [ARG] [OBL] In combinations of two third person clitics, two nodes [ARG] are always present, one of them coming from the dative clitic and the other one coming from the accusative clitic. Both instances of [ARG] will be mapped onto the template. (19) illustrates the mapping of two third person clitics, accusative and dative, both of them singular and masculine (that is, with no agreement features). Recall that the clitics themselves are not linearized; the dative clitic could have also been written before the accusative clitic: Resulting output forn: /li/ As indicated, [ARG] will be spelled out as /l/, through rule (15b). and [OBL] will be spelled out as li/, through rule (1%). When one of the two clitics (or both of them) is plural, the feature [plural] will be a dependent on [ARG] and will be mapped, with it, onto the slot. Since all [ARG] features will be mapped onto that slot, [plural] will also be always mapped, regardless of the clitic it comes from. [plural] will be spelled out as /z/ through rule (15e). Two exarnples appear in (20), below: a. CL b. CL A CL A n CL n ARG 3RD ARG 3RD ARG 3RD ARG 3m -- [ARG] [OBL] -- [ARG] [OBL] Resulting output form: Ilzil Resulting output form: llzil A plural marker 121 will, then, show up whenever at least one of the two clitics is plural Notice that the template in (18) has no slot for [3RD]. This means that this feature will never be mapped. Because of the setting (16a') the feature [fem], depending on [3RD], will then not be mapped; therefore, it will never surface.10 This is illustrated in (21): l0 1n Bonet (1991). the absence of a feminine marker in combinations of two third person clitics is amunted for through the application of a morphological de, instead of a parameter with two settings with respect to the location of the feature [fem]. The morphological de responsible for the absence of [fem] is shown in (i). [fem] always depends on [ARG]: (i) CL I CL I ARG I 3RD + [fem1 The de in (i) accounts for the fact that the feminine marker never shows up, not only with third person clitics, but with all kinds of clitics with the feature [3RD]. This also includes the genitive or partitive clitic and the locative clitic. Moreover, this rule also ensures that the feature [fem] is not lost when the other clitic is first person, second person or the impersonal or reflexive; these cl~tics do not have the feature [3RD]. The proposal presented in this paper is, in principle, better than the altemative sketched in this footnote because it emiches the representations instead of putting the burden on rules. However it is not clear at this point whether the representational option can amunt as well for the latter fact noted here: [fem] would always be a dependent on [3RD], and would never have a spell-out because there would never be a slot for it in the template. Therefore [fem] would also be wrongly lost in combinations with non-[3RD] clitics, like first person clitics. This problem might be solved by having a slightly different notion of template. I leave this issue for further research. CL CL n ARG 3RD n y/>L - - [ARGI [OBLI Resulting output form: Ili1 When the clitics appear in isolation, the template does not hold and all terminal features are spelled out. In this variety, [fem] is always -whenever licensedspelled out as la1 through rule (1%). 2.2. VarietyA While variety D has the most opaque third person pronominal clitic combinations, variety A has only transparent outputs. In order to derive these transparent forms, the mapping of the clitics onto the template has to be based on whole clitics (not on terminal features), as shown in (16b'). [femenine], like [plural], will be a dependent on [ARG] (cf. (16a)). The template for this variety appears in (22): (22) [ARG OBL] [ARG 3RD] (22) mentions only the relevant terminal features of the clitics involved, under each slot. It has to be understood that the minimal constituent containing the features mentioned (CL) is mapped. Thus, the first slot will be occupied by the third person dative clitic (which has both (unordered) [ARG] and [OBL]), while the accusative clitic (with [ARG] and [3RD]) will occupy the second slot (the dative clitic precedes the accusative clitic in this variety). (23), below, illustrates how the different third person clitics are mapped onto the template. Recall, again, that the two clitics could have been written in the opposite order (the accusative clitic before the dative clitic) because there is no linear order at that stage. The template (underneath the clitics) is responsible for the linear order. CL CL A ARG 3RD A ARG 3RD I I I\ ([PI]) OBL ([fem]) ([PI]) , - \ / [ARG OBL] [ARG 3RDl Resulting output forns: Ili+ll, /li+la/, /li+lz/, Ili+lezl, 11z+11, Ilz+lal, /Iz+lzl, Ilz+lezi The dative clitic, which is mapped onto the first slot, will be spelled out as lIl+li/, if this clitic is singular (through the application of rules (15b,c)), and as 111+12/ if the clitic is plural (through rules (15a,e)). In the latter case, no li1 will appear because in the context of [plural], [ARG] and [OBL] are spelled out together as 111. The accusative clitic will be spelled out as 111 if the clitic is masculine and singular (rule (15b)), as /I/+/a/ if [feminine] is present (rules (15b,f1)), and as /I/+/el+lz/ if both [feminine] and [plural] are present (rules (15b,f,e)). Rule (15f) will apply instead of (l5f') because of the presence of [plural].ll 11 As shown in (23). the output forn of the dative singular clitic in variety A is Ilil. 111 is the spell-out of [ARG] and li/, the spell-out of [OBL] (cf.(lSb,c)) However, nothing in the mechanisms proposed predicts the order between [ARG] and [OBL] in this variety, since the mapping onto the template picks up whole clitics, and the features interna1 to a clitic structure are unordered. One could probably assume that in this variety [ARG] and [OBL] do not have independent spell-outs (as shown in (15b,c)), but that the {[ARG], [OBL]) clitic has a specific spell-out Ilil. The fact that this variety lacks an independent [OBL] clitic (there is no locative clitic) makes this suggestion fairly plausible. I leave this issue for further research. 2.3. Variety B This variety is basically identical to variety A, but only when the third person dative clitic is plural. Under these circumstances, the combination of two third person clitics is not opaque, and the dative clitic precedes the accusative clitic. The analysis of such combinations will, thus, be basically identical to the one proposed for variety A in section 2.1, and, therefore, the corresponding template will be the one shown in (24), based on whole clitics. When the dative clitic is singular, variety B becomes very similar to variety D, with only opaque outputs. The template for variety B will, then, be identical to the template proposed for variety D, based on terminal features, not on whole clitics. This template is repeated in (24b), below: (24) a. - b. - - [pl OBL] [ARG 3RD] [ARG] [OBL] (24a) will be interpreted in the following way: A clitic with the terminal features [ARG] and [3RD], that is a third person accusative clitic, will be preceded by a clitic with the terminal features [pl] and [OBL]. The only clitic with these terminal features is the third person dative plural clitic. The Elsewhere Condition, invoked earlier, determines that the application of (24) will always precede the application of (24b), because (24a) is more specific (more complex) than (24b). This condition is necessary because otherwise we could generate output forms like *lzi instead of the existent els el in combinations of a third person accusative singular masculine clitic with a third person dative plural clitic (which is what happens in variety D).l2 12 Since in variety B the dative singular clitic, lli1 in isolation, adopts the forn li/, spelled hi, identical to the forn of the locative clitic, one wuld suggest that the dative clitic is really a locative clitic to start with, in the syntax. This would aiso be the case for other varieties. This possibility, plausible in principle, faces some problems, though. One of them, conceptual, would be that, intriguingly. the locative clitic would only be generated in case the source were singular. not plural. There is also an empirical problem: it is impossible to have two (independent) locative phrases in a given sentence, as shown in (i): (i) *Anira a Reus a I'estació. go-3sg-FUT to Reus to the-station 'Slhe will go to Reus to the station.' In variety B, the combinations with a third person dative singular clitic are identical to those in variety D except in one respect: in variety B, [feminine] is never lost; it surfaces as a schwa (through rule (15g)). In this variety, then, [feminine] will be a dependent on [ARG], like [plural] is (cf. the parameter setting in (16a)), instead of being a dependent on [3RD]. (25), below, shows how the mapping works for a combination of a third person accusative clitic and a third person dative singular clitic: If the dative clitic is considered to be really a locative clitic, we predict it to be impossible having a third person dative clitic and a locative clitic in a given sentence. This prediction is not bome out, as (ii) shows: (ii) En Pere no li hi (Ilil) anir8, a Reus the Pere not 3rd-dat(ben) loc go-3sg-FLIT to Reus 'Pere will not go to Reus for himlher.' Under the present proposal the fact in (ii) is accounted for in exactly the same tems as the cases with two instances of [ARG], seen so far: both the dative clitic and the locative have a feature [OBL]; the two instances of this feature are mapped onto one and the same slot, and, later, that feature is spelled out as a single li1 through de (1%). The intuition that the dative clitic and the locative clitic are in some way the same is also captured under the present proposal: both clitics have the feature [OBL]; the only difference is the presence versus the absence of the feature [ARG]. The structure of these two clitics can be compared in (iii) ((iiia) corresponds to the third person dative clitic, while (iiib) corresponds to the locative clitic): (ui) a. CL A b. CL ARG 3RD I 3RD I OBL I OBL Once the morphological rule in (27a) has applied, the output form and the structure corresponding to the dative plural clitic are mapped onto a template based on whole clitics (cf. (16b')). The dative clitic ([pl OBL]) will precede the neuter clitic ([NEUT 3RD1). When the third person dative clitic is singular, the output forms do not differ at all from variety D: the form Ili1 appears whenever the two clitics involved are singular; Ilzil appears, instead, when the accusative clitic is plural (recall that in this case the dative clitic is singular). No feminine marker appears. Under the present proposal, the lack of a feminine marker indicates that the feature [feminine] depends on [3RD], like in variety D. Likewise, the mapping to a template will be based on features (cf. (16b)). The two templates valid for variety C appear in (28), below: (28) a. - b. - - [pl OBL] [NEUT 3RDl [ARG] [OBL] Variety C constitutes, then, a mixed system with respect to mapping to a template, like variety B: for certain combinations the mapping is based on whole clitics (in both varieties, this looses its feature [NEUT] when it woccurs with any [3RD] clitic, as shown below: I CL ARG 3RD I 3RD In variety C, de (i) has to precede de (27a). Otherwise, the feature [NEW] introduced by de (i) would be lost by the application of rule (27a). The ordering between these two rules could follow from some structure prese~ation wndition. Jim Harris @.c.) suggests that Iol in this variety wuld be interpreted as a class marker (as the one that appears in carro), rather than the spell-out of the feature [NEUT]. This is certainly a possibility worth exploring, even though it would be more difficult to rule out the forn */lzozl as the spell-out for a wmbination with an accusative plural clitic. happens whenever a third person dative plural clitic is involved); for others, the mapping is based on features.14 The mapping of whole clitics is shown in (29a), while the mapping of features appears in (29b): a. CL 2. b. CL CL A A A ARG 3RD ARG 3RD ARG 3RD ARG 3RD I I I , [pll OBL, , NEüT OBL ([pi]) ([fem]) [pl OBL] PEüT 3RD] [ARG] [OBL] In (29a) the neuter clitic that appears on the right s~de (it could equally have been wr~tten on the l4 We have seen that when two features "fight" for a given slot in the template, no confiict anses and both of them are mapped. Later on each such slot will receive a single spell-out, which will be interpreted as having two different sources. Among the combinations taken into consideration in this paper, there are no cases of two whole clitics that compete for a single slot. These cases do exist, the potential combination of two instances of es being an example of them. First and second person clitics and the clitic es (all non-[3RD] clitics) are mapped as whole clitics in all dialects of Catalan. Two instances of es could in principle cooccur if one of them were an impersonal clitic and rhe other one were a reflexive third person clitic. However, these combinations do not have a grammatical output, as can be seen in (i): (i) a. *En aquesta casa es es renta. in this house refl+imp washes b. *En aquesta casa es renta. in this house refl+imp washes 'In this house, one washes oneself.' As (ia) shows. two instances of es cannot cooccur; in (ib) we see that a single instance of es cannot represent both a reflexive clitic and an impersonal clitic. The facts in (i), compared to the facts taken into account in the text with respect to the competltlon for a given slot, lead us to conclude that two whole clitlcs can never be mapped onto a single slot, whlle th~s IS not the case w~th ln&vldual features '%s cond~hon would amunt for the lntegnty (or recoverabillty, in tems taken from Bonet (1991)) that nou-[3RDl drtl~s seem to shoW (Mascaró (1986 135-6) presents a more complicated plcture of the facts relat cl1tlcs) left side) is the accusative clitic after the application of the morphological rule in (27a). The first slot in the corresponding template can only be occupied by third person dative plural clitics, which is the case in (29a). In (29b), the feature [feminine] is not mapped onto the template because there is no slot for [3RD] and, therefore, it cannot be spelled out. 3. Conclusions In this paper, four different varieties of Catalan have been analyzed with respect to their different outputs in combinations of two third person clitics. It has been shown that both linear order and the presence of opaque forms follow, except for one single aspect, from the way the mapping onto a template works. The exception is the change from an accusative clitic to a neuter clitic in one of the varieties. This type of opaque form, far from being an exception, is very common in all Romance languages, and has been argued to be derived from the application of a morphological rule. These rules, by their own nature, cannot generate random outputs, but give as a result a structure corresponding to an independently existing clitic. The results provided by the proposal made in this paper could hardly be obtained if a syntactic approach were adopted, which would, at most, provide an account of linear order -not of the presence of opaque formsby making obligatory use, at least for varieties B and C, of certain stipulations (because of the existence of a different ordering depending on the number of the dative clitic). It has to be taken into account, also, that the amount of dialectal and language variation with respect to pronominal clitics and their combination is extraordinary, while variation in other aspects is much smaller. Under a minimalist view of the grammar it does not make much sense to try to account for all these differences in syntactic tems. A post-syntactic morphological approach to this variation seems more plausible. I would like to thank Jim Harris, Joan Mascar6, an anonymous reviewer, and the audience at the 2nd Workshop on the Syntax of Central Romance Languages for insightful comments and help. All errors are my own. 110 References Anderson, S. R. (1992) A-Morphous Morphology, Cambridge University Press (Cambridge Studies in Linguistics, 62), Cambridge. Bonet, E. (1991) Morphology aper Syntax: Pronominal Clitics in Romance, doctoral dissertation, Massachusetts Institute of Technology, distributed by MIT Working Papers in Linguistics, Cambridge, MA. Chomsky, N. A. (1993) 'A Minimalist Approach for Linguistic Theory' in K. Hale and S. J. Keyser (eds) The Viewfrom Building 20, MIT Press, Cambridge, MA. Halle, M. (1990) 'An Approach to Morhology', Proceedings of the Northeastern Linguistics Society, 20, 150184. Halle, M. (1991) The Latvian Declension' in G. Booij and J. van Marle (eds) Yearbook of Morphology 1991, Kluwer, Dordrecht, 33-47. Halle, M. and A. Marantz (1992) 'Distributed Morphology and the Pieces of Inflection' in K. Hale and S. J. Keyser (eds) The Viewfrom Building 20, MIT Press, Cambridge, MA. Hanis, J. (1991) 'The Exponence of Gender in Spanish', Linguistic Inquiry, 22, 27-62. Kiparsky, P. (1982) 'Lexical Morphology and Phonology' in I.-S. Yang (ed) Linguistics in the Morning Calrm, Hanshin, Seoul, 3-91. Otero, C. (1976) 'The Dictionary in a Generative Grammar', paper presented at the 91st. Annual Convention of the MLA, New York. Pranka, P. (1983) Syntax and Word Formation, doctoral dissertation, Massachusetts Institute of Technology. Todoli, J. (1992) 'Variants dels pronoms febles de 3a persona al País Valencia: regles fonosintActiques i morfolbgiques subjacents', Zeitschriftjiir Katalanistik, 5, 137-160. Departament de Filologia Catalana Universitat Autbnoma de Barcelona E-O8193 Bellaterra e-rnail: [email protected]