OCP Effects in Catalan Cliticization

Abstract In Catalan, sequences of sibilants are never pronounced as such. In most contexts all varieties coincide in the «strategies» used to avoid these sequences, namely epenthesis or deletion. Variation is only found in the domain of pronominal clitics (but not with other types of clitics). One source of variation is accounted for by decomposing a general constraint into two specific ones, which implies partial constraint reranking. The other source of variation, which involves a case of apparent opacity, is explained through an Output-Output constraint that makes reference to paradigmatic relations. Keywords: OCP, clitics, epenthesis, deletion, fusion, opacity, Optimality Theory. 1. Introduction In Catalan, sequences of sibilants are never pronounced as such (in normal speech). In the cases where two adjacent sibilants would potentially occur, two main ways of avoiding this adjacency can be found: (a) reduction of the two sibilants to one, and (b) insertion of an epenthetic vowel between the two sibilants. As we shall see, reduction is by far the most common strategy while epenthesis is restricted to very specific morphophonological environments.1 *. We would like to thank Joan Mascaró, Clàudia Pons, and two anonymous reviewers for their helpful comments. This work is supported by the grants BFF2000-0403-C02-02 from the Ministerio de Ciencia y Tecnología, and 2001SGR 00150 and 2001SGR 00004 from the Departament d’Universitats, Recerca i Societat de la Informació de la Generalitat de Catalunya. 1. There is an additional strategy, with a dissimilatory effect, which is found only in Majorcan and Minorcan Catalan. In these varieties a sequence of sibilants is pronounced as a lengthened affricate in the contexts where other varieties show reduction (a sequence like dos sons ‘two sounds’, for instance, is pronounced [dot.ts´ɔns] in these two varieties but [do.s´ɔns] in the other dialects). In this paper we disregard these data because we focus on the differences of distribution between epenthesis and reduction, especially in the domain of clitics (in some varieties belonging to Central Catalan). An analysis of the different strategies (including dissimilation) used to avoid sibilant sequences in Catalan within autosegmental phonology appears in Palmada (1994a, b). Catalan Journal of Linguistics 1, 2002 19-39 OCP Effects in Catalan Cliticization* Eulàlia Bonet Universitat Autònoma de Barcelona. Departament de Filologia Catalana 08193 Bellaterra (Barcelona). Spain [email protected] Maria-Rosa Lloret Universitat de Barcelona. Departament de Filologia Catalana Gran Via de les Corts Catalanes, 585. 08007 Barcelona (Spain) [email protected].es Cat.Jour.Ling. 1 001-259 26/2/03 15:51 Página 19 All varieties of Central Catalan (i.e., the dialects spoken in eastern Catalonia, including the Barcelona area) coincide in inserting an epenthetic schwa between two adjacent sibilants when they belong to the same word and the second sis a suffix (e.g., cuses /kuz+z/, pronounced [kú.zəs] ‘(you) sew’), except in masculine nominals, in which [u] is most commonly inserted (cf. casos /kaz+z/, pronounced [ká.zus] ‘cases’). In the other contexts where all varieties of Central Catalan coincide, only reduction of the two sibilants to one (fusion or deletion) is found. Reduction appears, for instance, when a definite determiner in the plural, els (masc.) or les (fem.) (both being clitics), is adjacent to a word starting with a sibilant (e.g., els sons [əl.s ´ɔns] ‘the sounds’ or les sopes [lə.só.pəs] ‘the soups’) or when the sibilants belong to different words (e.g., portes sacs [p`ɔr.tə.sáks] ‘(you) bring sacks’, or pis senzill [pì.sən.zíʎ] ‘simple apartment’). Within words, deletion/fusion of sibilants is found in compounds (e.g., dos-cents [do.séns] ‘two hundred’) and in prefixed words (e.g., dessalar [də.sə.lá] ‘to desalt’, with the prefix des and the verb salar).2The data presented so far are identical (with slight differences in pronunciation) in almost all varieties of Catalan. There is one fact, though, that has passed unnoticed: within Central Catalan (more concretely, in the Barcelona area) three types of variation can be found as to the choice of reduction or epenthesis in clitic–verb sequences.3The main goal of this paper is to explain why this variation occurs only in the domain of pronominal clitics.4 The clitics that end in sin the varieties of Catalan under discussion are the following: es /s/ third person reflexive/impersonal clitic; ens /nz/, first person plural; us /uz/, second person plural; els /l+z/, third person accusative masculine plural, and les /l+a+z/, third person accusative feminine plural. The examples in (1) show the phonetic realization of these clitics in these varieties when the verb starts with a non-sibilant consonant. Some clitic forms show an initial epenthetic schwa for syllabification purposes; for instance, the clitic els in (1d), with an underlying form /l+z/, needs an initial epenthetic vowel in proclitic position because the sequence [lz] does not constitute a possible onset. (From now on, we underline all occurrences of epenthetic schwas for expository reasons.) (1) a. Es trenca. [əs.tɾε´√.kə] itself breaks ‘It breaks.’ 20 CJL 1, 2002 Eulàlia Bonet; Maria-Rosa Lloret 2. Even though we refer simply to sequences of sibilants, all the cases we contemplate in this paper are sequences of voiced or voiceless anterior coronal sibilants. We ignore sequences of sibilants involving a different place of articulation (/ʃ/ or //) because they are fairly rare and because their behavior implies complications that escape the scope of this paper. 3. It is impossible to know what the facts would be in verb–clitic sequences; that is, in enclisis: the only clitic starting with sis the third person reflexive/impersonal clitic es (/s/, realized [sə] postverbally after a consonant: fer-se [fér.sə]), and this clitic can never cooccur with a verbal form ending in a sibilant (only the imperative second person forms vés ‘go!’ and fes ‘do!’ end in a sibilant, and they cannot combine with the third person clitic /s/). 4. We use the term clitic for convenience; it does not necessarily have to be understood as a primitive of the theory (for a proposal, within Optimality Theory, in which the different types of clitics are derived from differences in constraint ranking, see Selkirk 1995). Cat.Jour.Ling. 1 001-259 26/2/03 15:51 Página 20 b. Ens parla. [əns.pár.lə] to-us talks ‘(S/he) talks to us.’ c. Us creu. [us.kɾε´w] you (pl.) believes ‘(S/he) believes you all.’ d. Els porta. [əls.p´ɔr.tə] them (masc.) brings ‘(S/he) brings them.’ e. Les compra. [ləs.kóm.pɾə] them (fem.) buys ‘(S/he) buys them.’ As mentioned, variation is found only when the verb that follows one of these clitics starts with a sibilant. One of the varieties, let us call it variety A, systematically inserts a schwa in this context, as shown in (2). The presence of this schwa has its reflex in the orthography only in the case of the third person reflexive/impersonal clitic es (i.e., se), illustrated in (2a). (2) a. Se sap. [sə.sáp] impers. knows ‘It is known.’ b. Ens sent. [ən.zə.sén] us hears ‘(S/he) hears us.’ c. Us citarà. [u.zə.si.tə.ɾá] you (pl.) will-quote ‘(S/he) will quote you all.’ d. Els sé. [əl.zə.sé] them (masc.) know ‘(I) know them.’ e. Les supera. [lə.zə.su.pé.ɾə] them (fem.) overcomes ‘(S/he) overcomes them.’ The clitics that end in a consonant other than sdo not cause the appearance of a schwa after the clitic even when the verb does start with a sibilant. If vowel insertion is required for syllabification purposes, a schwa appears, as mentioned, in initial position (cf. em sent /m#sent/, [əm.sén] ‘(s/he) hears me’; el simula /l#simul+ə/, [əl.si.mú.lə] ‘(s/he) simulates it (masc.)’). The epenthetic schwa between the pronominal clitic and the verb appears only to break the contact between two sibilants. OCP Effects in Catalan Cliticization CJL 1, 2002 21 Cat.Jour.Ling. 1 001-259 26/2/03 15:51 Página 21 It is important to recall that, in variety A, the behavior of the third person accusative plural pronominal clitic is very different from the otherwise identical definite determiner. Before a host starting with a non-sibilant consonant they both surface, in their masculine forms, with an initial schwa, inserted for syllabification purposes (cf. els portes ‘(you) bring them’: [əls.p´ɔr.təs], from an underlying form /l+z#pɔɾt+ə+z/, and els porcs ‘the pigs’: [əls.p´ɔrks], from an underlying form /l+z#pɔɾk+z/). Before a sibilant, the behavior of both clitics differs. In the case of the pronominal clitic, as illustrated in (2d,e), epenthesis takes place (cf. els sé ‘(I) know them’: [əl.zə.sé], from an underlying form /l+z#se/), while in the case of the definite determiner, the «strategy» used to avoid the contact of sibilants is deletion/fusion (cf. els sons ‘the sounds’: [əl.s´ɔns], from an underlying form /l+z#sɔn+z/). A different variety, variety B, inserts a schwa after a clitic only when the first sibilant belongs to the third person reflexive/impersonal clitic, which has the underlying form /s/ (shown in (3)). With all other clitics ending in a sibilant, fusion/deletion is found when the verb starts with a sibilant (as shown in (4)). (3) Se sap. [sə.sáp] impers. knows ‘It is known.’ (4) a. Ens sent. [ən.sén] us hears ‘(S/he) hears us.’ b. Us citarà. [u.si.tə.ɾá] you (pl.) will-quote ‘(S/he) will quote you all.’ c. Els sé. [əl.sé] them (masc.) know ‘(I) know them.’ d. Les supera. [lə.su.pé.ɾə] them (fem.) overcomes ‘(S/he) overcomes them.’ Finally, what we can call variety C systematically presents fusion/deletion when a clitic ending in sis adjacent to a verb starting with an s. This is shown in (5). (5) a. Se sap. [ə.sáp] impers. knows ‘It is known.’ b. Ens sent. [ən.sén] us hears ‘(S/he) hears us.’ c. Us citarà. [u.si.tə.ɾá] you (pl.) will-quote ‘(S/he) will quote you all.’ 22 CJL 1, 2002 Eulàlia Bonet; Maria-Rosa Lloret Cat.Jour.Ling. 1 001-259 26/2/03 15:51 Página 22 d. Els sé. [əl.sé] them (masc.) know ‘(I) know them.’ e. Les supera. [lə.su.pé.ɾə] them (fem.) overcomes ‘(S/he) overcomes them.’ The table in (6) summarizes all the facts concerning the realization of underlying sequences of sibilants in the varieties A, B, and C of Central Catalan. We exclude from the table and the analysis the insertion of [u] in masculine nominals, previously mentioned, due to the interference of gender allomorphy (/u/ being a marked masculine morph), an issue that lies beyond the scope of this paper. In (6) all epenthetic vowels appear underlined. (As said, those include not only the schwas that break sequences of sibilants but also the initial schwas that are needed for syllabification purposes.) We use the notation «~Verb» to reflect the fact that the definite determiner can appear with any category other than verb (cf. els sempre disposats a … ‘the always ready to…’, els de la Maria ‘those of Mary’, els que et dic ‘those that I tell you’). OCP Effects in Catalan Cliticization CJL 1, 2002 23 (6) Epenthesis Reduction Epenthesis & Reduction X+suffix cuses A, B, C /kuz+z/ [ku.zəs] cl # Verb se sap A, B C /s#sab/ [sə.sáp] [ə.sáp] els sé A B, C /l+z#se/ [əl.zəsé] [əl.sé] Les supera A B, C /l+a+z#supeɾə/[lə.zə.su.pé.ɾə][lə.su.pé.ɾə] cl # ~Verb els sons A, B, C /l+z#sɔn+z/ [əl.sɔ´ns] prefixed words dessalar A, B, C /dəs##sal+a+ɾ/[də.sə.lá] compounds dos-cents /doz##sent+z/ [do.séns] between words portes sacs /pɔɾt+ə+z##sak+z/ [pɔ`r.tə.sáks] Cat.Jour.Ling. 1 001-259 26/2/03 15:51 Página 23 Epenthesis and reduction (deletion or fusion) are two ways of avoiding the contact of sibilants, which would constitute a violation of the Obligatory Contour Principle (OCP), a principle originally proposed in Leben (1973) for tones, but later extended to other phonological and morphological domains. Myers (1997) argues convincingly for an Optimality-theoretic account of the OCP and the whole range of «repair strategies» that are used to avoid adjacent similar elements in tone phonology. In the next sections, we provide an account for the choice between epenthesis and deletion/fusion in the different contexts where a potential OCP violation would occur in the three Central varieties considered in this paper. We first present an analysis of variety A, including all the cases in which this variety does not differ from the others. Section 2.2 is devoted to variety B, and section 2.3 deals with variety C. Most of our assumptions are based on Bonet and Lloret (in press), which provides a detailed analysis of the phonology of pronominal clitics, both in contact with the verb and in clitic sequences. As previously mentioned, a crucial claim in that paper, assumed here, is that the schwa that appears associated to some clitics and in different positions is the product of epenthesis, and that it is not necessary to resort to allomorphy to account for all the shapes a clitic might surface with. Our analysis is framed in Optimality Theory, a framework that has proved to be more adequate than others in accounting for the phonology of Catalan clitics. 2. The analysis 2.1. Variety A The fact that sibilant sequences are systematically avoided in Catalan shows that the constraint OCP is very high-ranked in Catalan, in the version of it that makes reference to sibilants. We give an informal formulation of the constraint in (7). (7) OCP-SIBILANT (OCPS): Adjacent sibilant segments are forbidden. The high ranking of the Markedness constraint OCP-SIBILANT in Catalan forces outputs to be less faithful to their inputs. Both the addition of an epenthetic vowel and the deletion of a consonant constitute violations of Faithfulness constraints. In the case of epenthesis, the constraint that punishes it is DEP-IO. (8) DEP-IO (DEP): «Every segment of the output has a correspondent in the input»; i.e., epenthesis is prohibited (McCarthy and Prince 1995: 264). In the case of consonant reduction, the appearance of a single sibilant instead of two adjacent ones can, in principle, be interpreted in one of three ways: as deletion of the second s, shown in (9a), as deletion of the first s, shown in (9b), and as fusion, in which an output [s] corresponds to two input identical segments /s s/, shown in (9c) (‘s’ stands for any sibilant regardless of its voiced/voiceless realization). 24 CJL 1, 2002 Eulàlia Bonet; Maria-Rosa Lloret Cat.Jour.Ling. 1 001-259 26/2/03 15:51 Página 24 (9) a. /s1s2/b./s 1s2/c./s 1s2/input ||\/ [s1][s 2][s 1,2]output The representations in (9a,b) constitute violations of the constraint MAX-IO, which bans the deletion of a segment. (10) MAX-IO (MAX): «Every segment of the input has a correspondent in the output»; i.e., deletion is prohibited (McCarthy and Prince 1995: 264). In (9c), the case of fusion, MAX is not violated because the two instances of /s/ do have a correspondent in the output; it just happens to be the same one. The constraint that is violated in (9c) is UNIFORMITY-IO: (11) UNIFORMITY-IO (UNIF): No segment of the output has multiple correspondents in the input; i.e., fusion is prohibited (see McCarthy and Prince 1995). With the constraints given so far, the presence of an epenthetic schwa between a clitic and a verb in the context of two input sibilants could be obtained with the following ranking: (12) Provisional ranking: OCP-SIBILANT » UNIFORMITY, MAX » DEP The tableau in (13) shows how the correct output [əl.zə.sé], from an input /l+z#se/, is obtained through this ranking. (13) /l+z#se/: [əl.zə.sé] ‘(I) know them (masc.)’ All the candidates in (13) violate the constraint DEP because all of them have at least one epenthetic vowel (the first one being needed for syllabification purposes). The output [əl.zə.sé] is the optimal candidate in spite of the fact that it is the only one with a double violation of DEP. This is so because the other candidates all violate more highly ranked constraints. All the candidates in (13), and in the rest of the tableaux in this paper, appear with the voicing specifications (voiced or voiceless) of the relevant sibilants according to what would be expected. For instance, OCP Effects in Catalan Cliticization CJL 1, 2002 25 /l+z1#s2e/ OCPSUNIF MAX DEP a. əls1.s2é*! * b. əl.s1,2é*! * c. əl.s2é*!* d. əl.z1é *! * ☞e. əl.z1ə.s2é** Cat.Jour.Ling. 1 001-259 26/2/03 15:51 Página 25 the candidate in (13a), [əls1.s2é], appears with a first voiceless sibilant because in that context voicing assimilation takes place; the candidate in (13d), [əl.z1é], appears with a voiced sibilant because we assume, with Mascaró (1986) and others, that (pro)clitics keep their voicing specification intact when the host starts with a vowel, which is the case in [əl.z1é] (the remaining sibilant belonging to the clitic). The ranking given in (12), however, would give the wrong result when applied to a non-pronominal clitic like the determiner els ‘the (pl.)’ followed by initial s. Parallel to els sé in (13), els sons ‘the sounds’, from an input /l+z#sɔn+z/, would have the ungrammatical output *[əl.zə.s´ɔns] instead of the grammatical [əl.s´ɔns]. The solution to this difference cannot be related to prosodic domains (like the clitic group, as in Nespor and Vogel 1986 and later work), because both the determiner els and the pronominal clitic els are clitics, more specifically proclitics; therefore they belong to the same type of prosodic domain. The relevant difference here is the fact that in one case the host is a verb (V) while in the other it is a non-verbal form (~V). Alignment constraints, which demand that constituent-edges coincide, can subsume this difference by distinguishing the constraint that ensures the contiguity relation between the pronominal clitic and the verb (the constraint ALIGN(CL-V) given in (14)) from the one that ensures the contiguity relation between the definite determiner and its host (the constraint ALIGN(CL-~V) given in (15)). (14) ALIGN(CL-V) (AL(CL-V)): Align the left edge of V(erb)[+tense] with the right edge of a pronominal clitic. (15) ALIGN(CL-~V) (AL(CL-~V)): Align the left edge of ~V with the right edge of a determiner. The constraint ALIGN(CL-V), which demands the configuration CL)(V, has been argued for in Bonet and Lloret (in press) to explain the presence of peripheral epenthesis in clitic–verb sequences (cf. en tira: [ən.tí.ɾə] vs. *[nə.tí.ɾə], from an underlying form /n#tí.ɾ+ə/, ‘(s/he) throws some’: ALIGN(CL-V) is violated in candidates like *[nə.tí.ɾə] due to the presence of the epenthetic vowel between the clitic and the verb but it is not violated when epenthesis is peripheral, that is, when the schwa precedes the proclitic, as in the grammatical form [ən.tí.ɾə]).5Parallel to ALIGN(CL-V), the constraint ALIGN(CL-~V) is violated whenever the configurationCL)(~V is not obtained (e.g., it is violated in *[əl.zə.s´ɔns], from the underlying form /l+z#sɔn+z/ ‘the sounds’). As we shall see next, in variety A the different behavior of determiners and pronouns in sibilant contexts (els sé [əl.zə.sé] vs. els sons [əl.s´ɔns]) is explained by the different ranking of these two Alignment constraints (with other constraints 26 CJL 1, 2002 Eulàlia Bonet; Maria-Rosa Lloret 5. A parallel constraint, ALIGN(V-CL), accounts for enclisis, and the presence of peripheral epenthesis in that context, that is, the presence of a schwa following the clitic (as in tirem-ne ‘let us throw some’ /tiɾ+ε+m#n/, [ti.ɾε´m.nə]). For variety A, there is no evidence for a different ranking of the two constraints; for this reason they can be collapsed under the term ALIGN(CL/V), as proposed in Bonet and Lloret (in press). Cat.Jour.Ling. 1 001-259 26/2/03 15:51 Página 26 intervening between them). The ranking of ALIGN(CL-~V) (but not that of ALIGN(CL-V)) just below OCP-SIBILANT causes a candidate like *[əl.zə.s´ɔns], for els sons ‘the sounds’, to be discarded as the optimal candidate, as shown in (16). (16) /l+z#sɔn+z/: [əl.s´ɔns] ‘the sounds’ The candidate in (16e), with an epenthetic vowel between the determiner and the noun —and with a configuration CL)ə(~V—, is not the only one to violate ALIGN(CL-~V). In (16b) fusion also causes a violation of this constraint, given that the left edge of the noun is before s1,2, while the right edge of the determiner is after s1,2 (the configuration being [əl(~V s1,2 CL)´ɔns]). In (16), then, both the candidate with epenthesis and the candidate with fusion are ruled out. However, two ouputs fare even: (16c), with deletion of the first sibilant, and (16d), with deletion of the second sibilant and a faithfully voiced first sibilant, which does not correspond to the grammatical phonetic form, [əl.s´ɔns], with a voiceless sibilant. An additional constraint, justified below, is needed to undo the tie. Words like esport ‘sport’ or especificar ‘to specify’ surface, in Catalan, with an initial epenthetic vowel [ə]: [əs.p´ɔrt], [əs.pə.si.fi.ká]. When a proclitic is added to these words, this epenthetic vowel remains: /l#spɔɾt/ ‘the sport’ surfaces as [ləs.p´ɔrt] (cf. *[əls.p´ɔrt], also with a well-formed syllabification in Catalan), and /u#spəsifikə/ ‘(s/he) specifies it’ surfaces as [wəs.pə.si.fí.kə] (cf. *[us.pə.si.fí.kə], which would be more faithful to the input and would not cause any syllabification problems). Traditionally it has been assumed that epenthesis takes place first at the word level, and clitics are added later to the epenthesized word. Leaving aside some problems such a cyclic account would have to face in dealing with the phonology of clitics, Optimality Theory offers several alternatives to cyclicity that avoid having to resort to levels, one of them being Output-Output correspondence constraints (see, for instance, Benua 1995 or Kenstowicz 1996). In this type of Faithfulness constraints a correspondence relation is established between the base form (which has to be a possible free standing word) and an affixed (or cliticized) form. The Output-Output constraint stated in (17) makes reference to the initial segment of the base. The final segment of the base is more unstable (for instance, the final schwa of a verbal form like passa ‘pass!’, [pá.sə], might be deleted before the neuter clitic ho (/u/) in some of the varieties discussed in this paper (cf. passaOCP Effects in Catalan Cliticization CJL 1, 2002 27 /l+z1#s2ɔn+z/ OCPSAL(CL-~V) UNIF MAX DEP a. əls1.s2ɔ´ns *! * b. əl.s1,2ɔ´ns *! * * ☞c. əl.s2ɔ´ns * * d. əl.z1ɔ´ns * * e. əl.z1ə.s2ɔ´ns *! ** Cat.Jour.Ling. 1 001-259 26/2/03 15:51 Página 27 2.2. Variety B In variety B most pronominal clitics behave like determiners do; that is, when a proclitic ending in a sibilant is adjacent to a verb starting with a sibilant, a single [s] is present in the surface form. In this variety, then, no distinction is made between the two types of clitics. For this reason there is no need for the existence of a constraint ALIGN(CL-V) different from ALIGN(CL-~V); the two constraints can be collapsed into a more general one, ALIGN(CL-LEX).11 (30) ALIGN(CL-LEX)(AL(CL-LEX)): Align the left edge of a lexical word with the right edge of a clitic. ALIGN(CL-LEX)occupies the same position as ALIGN(CL-~V) in variety A, unordered with respect to OOINITIAL. The tableau in (31) shows how the output [əl.sé] is obtained for els sé ‘(I) know them’. (31) /l+z#se/: [əl.sé] ‘(I) know them’ Not all the pronominal clitics behave like els. The impersonal/reflexive clitic surfaces with an epenthetic vowel instead of showing deletion or fusion (cf. se sap [sə.sáp]). At first sight this different behavior among the clitics might seem surprising, but there is a crucial difference between the reflexive clitic and all the other clitics that end in a sibilant: the reflexive clitic is underlyingly a single segment, /s/, while the other clitics have more than one segment (cf. third person masculine plural, /l+z/; third person feminine plural, /l+a+z/; first person plural, /nz/; second person plural, /uz/). Therefore, if the sibilant of the reflexive clitic is deleted, the whole clitic is deleted, while this is not the case for any other clitic. The constraint that punishes the deletion of a whole clitic is REALIZE-CLITIC. 34 CJL 1, 2002 Eulàlia Bonet; Maria-Rosa Lloret 11. In footnote 5 it was mentioned that ALIGN(CL-V) and ALIGN(V-CL)have an identical ranking in variety A; for this reason they can be collapsed into a single constraint ALIGN(CL/V). In a similar way, one could wonder whether in variety B it is possible to collapse all proclisis and enclisis into a constraint ALIGN(CL/LEX). This might be the case, but a detailed study of enclisis in this variety is needed before jumping to such conclusions. /l+z1#s2e/ OCPSOOIN AL(CL-LEX)REALµMAXSUNIF MAX DEP a. əls1.s2é*! * b. əl.s1,2é*!** ☞c. əl.s2é**** d. əl.z1é*! *** e. əl.z1ə.s2é*! ** Cat.Jour.Ling. 1 001-259 26/2/03 15:51 Página 34 (32) REALIZE-CLITIC (REALCL): A clitic must have some phonological exponent in the output. As mentioned, this constraint will only be relevant when the clitic is a single segment (of course, in any complete tableau for other clitics there will be a candidate with deletion of all the consonants, thus violating REALIZE-CLITIC; but this candidate will also violate many other constraints and will not have a chance to survive). The tableau in (33) shows how se sap is obtained in variety B. (33) /s#sab/: [sə.sáp] ‘it is known’ REALIZE-CLITIC is a constraint also present in variety A, although it did not appear in the corresponding tableau in (26) for expository reasons. In a way parallel to what we see in (33), REALIZE-CLITIC fatally punishes the candidate with deletion of the clitic (*[s2áp]), a candidate that also violates REALIZE-µand MAX-SIBILANT, as can also be seen in (33). 2.3. Variety C Variety C is almost identical to variety B.12 The lack of epenthesis between the clitic and the verb in cases like els sé [əl.sé] ‘(I) know them’, which are treated like sequences with determiners (like els sons [əl.s´ɔns]), indicates that the relevant Alignment constraint for these cases is ALIGN(CL-LEX). We can assume the same ranking it has in variety B. The only difference between variety B and variety C lies in the behavior of the reflexive/impersonal clitic, the only clitic of the language that consists of a single sibilant. A sequence like se sap ‘it is known’ is realized, in variety C, as a sequence with sibilant deletion/fusion and initial epenthesis: [ə.sáp] (from an underlying form /s#sab/). This case raises one of the most difficult problems for Optimality Theory, namely the problem of opacity. The sequence [ə.sáp] is opaque OCP Effects in Catalan Cliticization CJL 1, 2002 35 12. Variety C, in comparison with varieties A and B, is spoken by few people in the Barcelona area. Most of the speakers of this variety have Catalan as a second language. /s1#s2ab/ OCPSOOIN REALCL AL(CL-LEX)REALµMAXSUNIF MAX DEP a. s1s2áp *! b. s1,2áp * *! c. s2áp * * *! * d. s1áp * *! * ☞e. s1ə.s2áp * * Cat.Jour.Ling. 1 001-259 26/2/03 15:51 Página 35 because there is no apparent need for the initial epenthetic vowel: the OCP problem is solved via deletion or fusion, and the initial schwa is not needed for syllabification (*[sáp] would be fine). Given our claim that the schwa is not present in the underlying form, the presence of an epenthetic vowel in the grammatical output cannot be explained straightforwardly. With the constraints we have presented so far, *[sáp], in any of the interpretations for the appearance of a single [s], will always violate a subset of the constraints violated by [ə.sáp], which violates DEP in addition to others. In a serial model of phonology one could stipulate that the rule inserting the epenthetic vowel applies before the rule that deletes one of the sibilants or fuses them (an instance of extrinsic ordering), a possibility not available here. Intuitively, the absence of the initial schwa (in a form like *[sáp]) would cause the sequence to become phonetically identical to the simple verb, [sáp] (the presence of the clitic could not be perceived). The schwa in [ə.sáp] is the clue for the presence of the clitic, even if the clitic itself cannot be identified (at least, apparently). Moreover, the presence of the schwa between the proclitic and the verb (like in [sə.sáp] for varieties A and B) would constitute the only instance of epenthesis between a clitic and its host, all other proclitics having initial epenthesis when needed (cf. en sap [ən.sáp] ‘(s/he) knows some’, em veu [əm.b´εw] ‘(s/he) sees me’, et truca [ət.tɾú.kə] ‘(s/he) calls you’). On the contrary, and leaving aside the OCP problem, initial epenthesis in [ə.sáp] causes the clitic /s/ to have exactly the same phonological behavior as all the other consonantal clitics, which drives them to have a final VC(C) shape in proclitic position (cf. [əm], [ət], [ən], [əl], [əls], [əns], and [əs]). Even though capturing these intuitions within the framework adopted here is not easy and the issue needs further investigation, we can assume, following the lines of Kenstowicz (2001) for paradigmatic uniformity phenomena, an Output-Output constraint between members of a paradigm, in the present case pronominal clitics.13 For the clitic es, this Output-Output constraint, let us call it OUTPUT-OUTPUTPARADIGM (OOPAR), will favor an output with a VC shape, namely [əs], parallel to the other consonantal proclitics of the language. This constraint can be ranked fairly low, just above UNIFORMITY. The tableau corresponding to [ə.sáp] is given in (34).14 Notice that in this case, but not in other cases of deletion /fusion (in which deletion of the first consonant was the «strategy» chosen), the optimal candidate is realized with fusion of the sibilants; therefore the clitic does have a final VC shape ([əs]) without causing, at the same time, a violation of OUTPUT-OUTPUTINITIAL in the verb ([sáp]): [ə.s1,2áp]. 36 CJL 1, 2002 Eulàlia Bonet; Maria-Rosa Lloret 13. Kenstowicz (2001) considers paradigmatic uniformity phenomena the cases where «the grammar strives to maintain the same output shape for pairs of inflected words that the regular phonology should drive apart». In this paper, we extend this notion to the set of pronominal clitics. 14. Another, obvious, way of avoiding the opacity problem raised in this section is to assume that the initial schwa in [ə.sáp] is not epenthetic but part of the clitic. If the underlying form of the clitic (at least in certain contexts) is /əs/, the problem disappears, and the realization [ə.sáp] for an underlying sequence /əs#sab/ is explained exactly like the other cases. However, this would be the only case in which an underlying schwa would have to be posited for pronominal clitics. Cat.Jour.Ling. 1 001-259 26/2/03 15:51 Página 36 (34) /s#sab/: [ə.sáp] ‘it is known’ The OUTPUT-OUTPUTPARADIGM constraint is obviously also present in the other varieties, although its effects are not noticeable because it is ranked lower (at least below UNIFORMITY and MAX).15 3. Conclusions In this paper we have examined the effects that the highly ranked constraint OCPSIBILANT has in different varieties of Catalan, especially in environments involving clitics. As we saw, the main difference among the three varieties under discussion is whether all clitics behave alike (as in varieties B and C) or a distinction is made between pronominal clitics and other types of clitics (as in variety A). In OT, dialectal and language variation has commonly been accounted for by constraint reranking. In our analysis, the different behavior of variety A with respect to varieties B and C is captured by the decomposition of a general constraint, ALIGN(CL-LEX), into more specific constraints, ALIGN(CL-~V) and ALIGN(CL-V), which are to be considered members of the same constraint family. The ranking ALIGN(CL-~V) » ALIGN(CL-V) cannot be attributed to the Paninian constraint relation because there is no subset relation between the two constraints.16 In addition, OCP Effects in Catalan Cliticization CJL 1, 2002 37 15. A reviewer pointed out to us that the slightly higher ranking of OUTPUT-OUTPUTPARADIGM in variety C, which corresponds to speakers that have Catalan as a second language (and thus are less competent in Catalan than native speakers), might not be a coincidence, since from the point of view of language acquisition it is well known that regular forms are learned before irregular ones. In this sense, the general low ranking of the constraint OUTPUT-OUTPUTPARADIGM in all varieties might not be language specific but derivable from a more general imperative. 16. The ranking ALIGN(CL-~V) » ALIGN(CL-V) in variety A might not be arbitrary though; it might be motivated by the fact that in sequences with a determiner the consonant eligible for deletion can only /s 1 #s 2 ab/ OCP S OO IN R EAL CL A L ( CL - LEX )R EAL µM AX S OO PAR U NIF M AX D EP a. s 1 s 2 áp *! * b. s 1,2 áp * *! * c. s 2 áp * * * *! * d. s 1 áp * * *! * e. s 1 ə.s 2 áp * *! * f. əs 1 .s 2 áp *! * ☞g. ə.s 1,2 áp * * * h. ə.s 2 áp * * * *! * * i. ə.s 1 áp * *! * * Cat.Jour.Ling. 1 001-259 26/2/03 15:51 Página 37 the decomposition of a general constraint into more specific ones implies that at least one constraint has to intervene between them (in the case at hand, three unordered constraints intervene: REALIZE-µ,ALIGN(µ−µ), and MAX-SIBILANT); otherwise there would be no evidence for such decomposition. Decomposition necessarily involves partial constraint reranking, as shown in (35). The other source of variation we have seen is restricted to the behavior of a specific clitic, namely the third person reflexive/impersonal clitic (/s/). In variety C, as opposed to varieties A and B, the output form of this clitic constitutes an apparent opacity case. As an alternative to levels or extrinsic rule ordering, we have proposed an approach in terms of an Output-Output constraint, OUTPUT-OUTPUTPARADIGM, which forces the clitic /s/ to adopt the same output shape (a VC structure) as all the other consonantal proclitics, establishing thus a kind of analogical relation. To conclude, in (35) we give the complete hierarchy for varieties A, B, and C. The constraints that constitute a source of variation appear in bold face. (35) a. Variety A OCPS» OOIN, REALCL, AL(W-W), AL(CL-~V) » REALµ, AL(µ−µ), MAXS» AL(CL-V) » UNIF, MAX » DEP, OOPAR b. Variety B OCPS» OOIN, REALCL, AL(W-W), AL(CL-LEX)» REALµ, AL(µ−µ), MAXS» UNIF, MAX » DEP, OOPAR c. Variety C OCPS» OOIN, REALCL, AL(W-W), AL(CL-LEX)» REALµ, AL(µ−µ), MAXS, OOPAR » UNIF, MAX » DEP 38 CJL 1, 2002 Eulàlia Bonet; Maria-Rosa Lloret be the plural morph s, whose conveyed information is always recoverable from other plural morphs within its phrase due to agreement (in els sons [əl.s´ɔns] ‘the sounds’, for instance, the last [s] of the noun would be sufficient to mark the plural character of the sequence). In clitic–verb sequences, however, the information conveyed by the sof the pronominal clitic (whether this information is grammatical, as ‘plural’ in /l+z/, or lexical, as part of the stem in /nz/) cannot be straighly recovered by other means precisely because they are pronouns (e.g., els sé ‘(I) know them (masc.)’ pronounced [əl.sé] is homophonous with el sé ‘(I) know it (masc.)’; ens sap ‘s/he knows for us’ pronounced [ən.sáp] is homophonous with en sap ‘s/he knows it’), not even in reflexive constructions (in ens sostenim ‘(we) hold ourselves’, for instance, the pronunciation with sibilant reduction, [ən.sus.tə.ním], is homophonous with en sostenim ‘(we) hold it’). As a reviewer pointed out to us, it will be interesting to look in greater depth at cases of sdeletion with respect to other dialects that show, in general, more instances of consonant deletion, and see to what extent sbehaves differently from other consonants. We leave this issue for further research (see, though, Bonet and Lloret in press for a first approximation to consonant loss with respect to deletion of morphs in verb–clitic sequences). Cat.Jour.Ling. 1 001-259 26/2/03 15:51 Página 38 References Akinlabi, Akinbiyi (1996). «Featural Affixation». Journal of Linguistic 32: 239-289. Benua, Laura (1995). «Identity Effects in Morphological Truncation». In: Beckman, J.; Walsh Dickey, L.; Urbanczyk, S. (eds.). Papers in Optimality Theory. Amherst, Mass.: Graduate Linguistic Student Association (University of Massachusetts Occasional Papers in Linguistics 18), p. 77-136. Bonet, Eulàlia; Lloret, Maria-Rosa (in press). «More on Alignment as an Alternative to Domains: the Syllabification of Catalan Clitics». Probus. Colina, Sonia (1995). A Constraint-based Analysis of Syllabification in Spanish, Catalan, and Galician. Doctoral dissertation, University of Illinois, Urbana-Champaign. Jiménez, Jesús (1997). L’estructura sil·làbica del dialecte valencià. Doctoral dissertation, Universitat de València. Kenstowicz, Michael (1996). «Base-Identity and Uniform Exponence: Alternatives to Cyclicity». In: Durand, J.: Lacks, B. (eds.). Current Trends in Phonology: Models and Methods, vol. 1. Salford: European Studies Research Institute (ESRI), University of Salford, p. 363-393. Kenstowicz, Michael (2001). «Paradigmatic Uniformity and Contrast». Handout, 3èmes Journées Internationales du GDR Phonologie, Université de Nantes, June 1, 2001. Leben, William (1973). Suprasegmental phonology. Doctoral dissertation, Massachusetts Institute of Technology, Cambridge, MA. Mascaró, Joan (1986). Morfologia, Barcelona: Enciclopèdia Catalana. McCarthy, John (2000). «The Prosody of Phase in Rotuman». Natural Language and Linguistic Theory 18: 147-197. McCarthy, John; Prince, Alan (1995). «Faithfulness and reduplicative identity». In: Beckman, J.; Walsh Dickey, L.; Urbanczyk, S. (eds.). Papers in Optimality Theory. Amherst, Mass.: Graduate Linguistic Student Association (University of Massachusetts Occasional Papers in Linguistics 18), p. 249-384. Myers, Scott (1997). «OCP Effects in Optimality Theory». Natural Language and Linguistic Theory 15: 847-892. Nespor, Marina; Vogel, Irene (1986). Prosodic Phonology, Dordrecht: Foris. Palmada, Blanca (1994a). La fonologia del català. Els principis generals i la variació. Bellaterra: Publicacions de la Universitat Autònoma de Barcelona. Palmada, Blanca (1994b). «Sobre la semblança excessiva en fonologia». Els Marges 49: 95-106. Prince, Alan; Smolensky, Paul (1993). Optimality Theory: Constraint Interaction in Generative Grammar. Rutgers University, New Brunswick; University of Colorado, Boulder. To be published, Cambridge, Mass.: MIT Press. Selkirk, Elisabeth (1995). «The prosodic structure of function words». In: Beckman, J.; Walsh Dickey, L.; Urbanczyk, S. (eds.). Papers in Optimality Theory. Amherst, Mass.: Graduate Linguistic Student Association (University of Massachusetts Occasional Papers in Linguistics 18), p. 439-469. Walker, Rachel (1998). Nasalization, Neutral Segments, and Opacity Effects. Doctoral dissertation, UC, Santa Cruz. To be published, New York: Garland. OCP Effects in Catalan Cliticization CJL 1, 2002 39 Cat.Jour.Ling. 1 001-259 26/2/03 15:51 Página 39