No "double plurals" in Dominican Spanish: an optimality-theoretic account *.

Colina, Sonia

Abstract

In addition to standard methods of pluralization, Dominican Spanish has an alternative plural formation mechanism, normally referred to as the "double plural," in which -(e)se [(e)se] is adjoined to the base, libro > librose 'book-books', mujer > mujerese 'woman-women' (Jimenez Sabater 1976; Nunez-Cedeno 1980; Harris 1980; Terrell 1986; Nunez-Cedeno 2003). Extant analyses of double plural formation in Dominican Spanish suffer from framework-specific problems and more general ones, having to do with overgeneration of forms and inability to reveal the true nature of the process. Most analyses postulate a separate plural morpheme for Dominican "double plurals"/(e)se/.

This article proposes an optimality-theoretic analysis that demonstrates that there is no "double plural" in Dominican in the sense that the "double plural" is based on the regular plural There is only one plural morpheme, the traditional /s/. The apparently redundant attachment of the plural morpheme results from general restrictions on coda obstruents in combination with the need for overt realization of morphemes in prominent positions. The difference between the regular plural and the double plural (focused) is that in the latter case, a highly-ranked constraint realize morpheme-focus requires that the plural in focused positions have morphological exponence. In more general terms, [se] is the output realization of plural /s/ followed by epenthetic [e] in intonationally prominent (focus) positions. The present analysis also shows that the plural form is in an output-to-output relation to the singular. Epenthesis of [e] in the plural reflects the emergence of the unmarked (McCarthy and Prince 1994), with respect to the constraint against coda consonants (*Coda), whose effects can be seen in the output-to-output phonology of the plural, but not in the language as a whole, for example, mujer vs. mujere.

1. Introduction

Dominican Spanish has what appears to be quite an odd method of plural formation. More specifically, in addition to regular pluralization mechanisms, Dominican Spanish resorts to what has been dubbed in the literature as the "double plural" in which it appears that the plural morpheme has been attached twice by adding the additional allomorph es after the traditional (e)s: mujer, mujer-es [mu.he.re], mujerese /muher-s-s/ [mu.he.re.se] 'woman/women' (coda /s/ is deleted). All existent analyses of the phenomenon resort to some form of stipulative mechanism: separate, underlyingly-specified morphemes, prespecifled templates, language-specific, or stipulatory rules, etc. Harris (1980), for instance, proposes a language-specific prosodic template [[...] VCV] (realized as [[ese] or [se]) in addition to that of the regular plural; Nunez-Cedeno (1980) resorts to an optional, language-specific rule that inserts an epenthetic [e] after plural [s] in word-final position; and Nunez-Cedeno (2003) argues for two separate underlying allomorphs, one for the regular plural and one for the double plural. I show that all these language-specific, stipulative mechanisms are unnecessary; furthermore, their stipulative nature obscures the more general nature of the process.

This article demonstrates that there is no real "double plural" in Dominican Spanish in the sense that the "double plural" is based on the regular plural. There is only one plural morpheme, the traditional /s/. The apparently redundant attachment of the plural morpheme results from general restrictions on coda obstruents in combination with the need for overt realization of morphemes in prominent positions. The optimalitytheoretic (OT) analysis proposed reveals that the difference between the regular plural and the double plural (focused) is that in the latter case, a highly-ranked constraint realize morpheme-focus (RM/FOC) requires that the plural in focused positions have morphological exponence. In more general terms, [se] is the output realization of plural /s/ followed by epenthetic [e] in intonationally prominent (focus) positions.

An important contribution of the present analysis is that it shows that the plural form is in an output-to-output (OO) relation to the singular. More specifically, epenthesis of [e] in the plural reflects the emergence of the unmarked (McCarthy and Prince 1994), with respect to the constraint against coda consonants (*Coda), whose effects can be seen in the OO phonology of the plural, but not in the language as a whole, for example, mujer vs. mujere. This has important implications for the debate on the nature of [e] epenthesis and plural epenthesis in Spanish (phonological vs. morphological [Harris 1999]).

From a theoretical standpoint, the analysis proposed here highlights the superiority of a correspondence-theoretic account over serial analyses. An OO account reveals the natural and universals aspects of Dominican plural formation generally obscured by derivational formalisms. Theoretical implications can also be drawn with regard to a theory of prominence (Beckman 1997).

The article is organized as follows. After the presentation of the data in Section 2, Section 3 reviews and evaluates the existing analyses of "double plurals" in Dominican. The analysis proposed appears in Section 4: Section 4.1 summarizes the main points of the proposal in pretheoretical terms, in preparation for the formal optimality-theoretic analysis in Section 4.2. The analysis in Section 4.2 focuses on nonplural coda obstruents (Section 4.2.1), realize morpheme and OO-constraints (Section 4.2.2), plurals of V-final bases (Section 4.2.3), plurals of C-final bases (Section 4.2.4), and stress restrictions on pluralization (Section 4.2.5). Some conclusions are drawn in Section 5.

2. Data

Standard plural formation in Spanish seemingly consists of adding -s [s] to nonverbs ending in unstressed vowels, as in libro, libros 'book, books', and -es [es] to those ending in consonants, mujer, mujeres 'woman, women'. (1) Existing exceptions to this surface generalization, for example, crisis, crisis 'crisis, crises' (vs. regular lapiz[s], lapic[s]es 'pencil, pencils'), are reported in the standard sources (Foley 1967; Saltarelli 1970; Contreras 1977; Harris 1980).

Dominican Spanish, however, has developed an alternative plural formarion mechanism, normally referred to as the "double plural," in which -(e)se [(e)se], supposedly /s-s/, is adjoined to the nonverb base, libro, librose 'book, books', mujer, mujerese 'woman, women' (Jimenez Sabater 1975; Nunez-Cedeno 1980; Harris 1980; Terrell 1986; and Nunez-Cedeno 2003). This can be seen in (1) and in (3)-(4) below. Double-plural usage is found in informationally prominent positions, but is socially conditioned, being more prevalent amongst those with little formal education.

(1) Base Double plural Gloss (2)

 gallina gallinase 'hens'
 lata latase 'cans'
 pintura pinturase 'paints'
 esto estose 'these'
 eso esose 'those'
 muchacho muchachose 'boys'
 arriba arribase 'ups'
 mujer mujerese 'women'
 pan panese 'breads'
 papel papelese 'papers'
 vudu vuduse 'voodoos'
 aji ajise 'peppers'

Dominican Spanish does not normally allow coda obstruents (Jimenez Sabater 1975; Pineros 2003). These are often realized as zero (usted, ustedes [u.te] [u.te.De] 'you-polite, sg. and pl.') (aspiration of /s/ is also possible although less common in popular speech). Consequently, [s] is affected by coda restrictions in word-internal and word-final positions (Jimenez Sabater 1975; Pineros 2003). (3)

(2) mosca [moka] 'fly'
 basta [bata] 'enough'
 asno [ano] 'donkey'
 gas [ga] 'gas'
 virus [bi.ru] 'virus'

In accordance with the ill-formedness of coda [s], plural /s/ is realized as zero in popular speech, thus surfacing as zero [o] in vowel-final forms (3) and [e] in consonant-final ones (4). (3) and (4) also contain the corresponding double plurals.

(3) Bases ending in unstressed final vowels:

Singular Plural Double plural Gloss

gallina gallina[o] gallinase 'hens'
lata lata[o] latase 'cans'
pintura pintura[o] pinturase 'paints'
e(s)to e(s)to[o] e(s)tose 'these'
eso eso[o] esose 'those'
muchacho muchacho[o] muchachose 'boys'
arriba arriba[o] arribase 'ups'

(4) Bases ending in consonants or stressed vowels:

Singular Plural Double plural Gloss

mujer mujere[o] mujerese 'women'
pan pane[o] panese 'breads'
papel papele[o] papelese 'papers'
vudu vudu[o] vuduse 'voodoos'
aji aji[o] ajise 'peppers'

Nunez-Cedeno (2003) introduces data that indicate that double-plural selection (vs. [o] or [e]) is governed by pragmatic factors, being selected in informationally-prominent, focus positions. The double plural is realized only on the lexical head of a NP in focus position, indicated here by the use of CAPITALS, for example, la AMARILLASE vs. * lase AMARILLASE (double plural on the entire NP). The examples in (5) from Nunez-Cedeno (2003) illustrate the use of double plurals in context (the corresponding nonfocus forms are marked in [5] by the use of italics).

(5) (Ramon, male, age 28)

Researcher: Ramon, bueno y que va a hacer con esas pinturas? 'Well, Ramon, what are you going to do with those paints?'

Ramon: Pue la AMARILLASE la subo pa la azotea. No la BLANCASE; ESASE se quedan aqui porque hay que pintal toa la PAREDESE, y usar la pintura. Si yo le digo que dona Remi na ma me da a mi pa comprar lata de PINTURASE y fijese lo grande que son la parede, eso no se pinta con tan chin pintura. 'I am bringing the yellow ones up to the roof. Not the white ones; those will stay here because I got to paint all walls and use the paints. I tell you, dona Remi just gives me enough to buy two cans of paint and look how big these walls are, they cannot be painted with so little paint.'

(6) Researcher: Me imagino que se comen la poca grama que queda, esta muy seco, pero tambien hay gallina. 'I think they are eating up the little there is, it is dry but there are some hens as well.'

Don Otilio: No, GALLiNASE no, esa son PALOMASE. 'No, they're not hens, those are pigeons.'

Finally, Nunez-Cedeno (2003) also notes stress-related restrictions. The double plural is not documented in proparoxytonic words: platano, *platanose 'plantain', where it would result in preantepenultimate stress (not permitted in Spanish); no double plural exists either in the case of C-final paroxytonic forms, util, *utilese 'useful', virgen, *virgenese 'virgin', agil, *agilese, 'limber' (regular nonfocused plurals are, however, possible, utile, virgene, agile). Stress shift in a double plural does not seem to be a viable possibility.

3. Previous analyses

Harris (1980: 19) proposes that Spanish plurals require that the autosegmental prosodic template of noun and adjective plurals conform to the templatic ending [[[...] VC].sub.[alpha]] where [alpha] = noun or adjective, and to a rule that supplies phonological materials for noun and adjective plurals ("In noun and adjective plurals the segment s is associated with the rightmost C position in the prosodic template"). For instance,

(7) [ILLUSTRATION OMITTED]

Harris explains that before the application of the morphological rule, the V position of the prosodic template is associated with the class marker o, in accordance with the constraint [[[...] VC].sub.[alpha]], and the C position has no segment associated with it. The plural rule associates an s with this position. In the case of consonant final nouns and/or adjectives, a phonological epenthesis rule inserts e into the empty V of the prosodic template, as in (8).

(8) [ILLUSTRATION OMITTED]

Within the context of this proposal, Harris accounts for Dominican double plurals (e.g. muchacho, muchachose) by means of an additional prosodic template--[[[...] V C V].sub.[alpha]]. The derivation would proceed as in (9).

(9) [ILLUSTRATION OMITTED]

This proposal can be seen as a significant improvement over earlier accounts under the assumption that templates are independently justified in phonological systems; however, as Harris himself admits, his analysis also suffers from one important weakness: "the account just sketched gives no genuine explanation.... Since we simply stipulated the existence of the alternate template ..." (Harris 1980: 24).

Older analyses (Nunez-Cedeno 1980) propose a series of language-specific rules to account for [e] in plural formation. In example (10), (10a) accounts for [e] in standard plural formation ([es] allomorph) and (10b) for [e] in the double plural; in both cases [e] is considered epenthetic. (10b) is an optional rule.

(10) [ILLUSTRATION OMITTED]

Sample derivation

/pintura/ plural /papel/ plural /tarea/ plural
/pintura s papel s tarea s plural adjunction
-- papal e s -- (10a)
pintura s e papel e s e tarea se (10b)
[pinturase] [papelese] [tarease]

The proposal in (10) faces difficulties similar to those of Harris (1980, 1986) due to the stipulative nature of the rules as well as to their language-specific character. Harris (1980) and Nunez-Cedeno (1980) also miss an important generalization: that the form of double plural results from the interaction of Dominican phonology (in particular, syllabification) with standard plural formation mechanisms common to most dialects. This will become clear in the context of the current proposal. (4)

Nunez-Cedeno (2003) applies Harris' (1999) distributed-morphology account of Spanish plurals to Dominican Spanish. In his analysis the plain plural for words ending in consonants (class IIIa in Harris 1999) is /e/ or /es/; all other forms take the zero-morpheme. The double plurals are [ese] for C-final words and [se] for V-final ones as shown in (11).

(11) [root] Af [root] Af
 morphology
 Mujer ese gallina se
 phonology
 Mu.je.re.se ga.lli.na.se syllabification

This analysis also fails to capture the connection between double plurals and syllabification mechanisms in Dominican Spanish, as the presence of [e] after [s] is related to the ill-formedness of coda /s/. In addition, I argue that there is no need for postulating two separate underlying allomorphs as Nunez-Cedeno (2003) does.

In sum, extant analyses of double plural formation in Dominican Spanish suffer from framework-specific problems (e.g. language-specific nature of rules, stipulatory nature of rules and templates) and more general ones, such as their inability to show that double plurals result from the interaction of standard plural formation with the phonology of Dominican, thus obviating the need for special mechanisms such as language-specific templates or rules. In this article I propose an OT analysis that solves the difficulties encountered by previous ones; in addition, the OT account proposed here provides for a more adequate understanding of plural formation in Dominican, by showing that double plurals are not a language-specific oddity. Furthermore, in addition to offering an adequate formal account that ensures good phonological output, the current analysis incorporates pragmatic factors, thus accounting for the selection of the "double plural" versus standard [e]/zero realization.

4. Analysis

4.1. Basic proposal

I start by presenting the basics of the proposal in pretheoretical terms (derivationally-sounding terminology--e.g. repair, epenthesis, deletion--is used out of convenience and should not be read as an endorsement of a serial analysis of the data).

Attachment of the plural morpheme results in a coda [s] in a dialect that does not normally allow coda obstruents. Given that the general repair mechanism for coda obstruents in Dominican is deletion (usted, ustedes [u.te] [u.te.De] 'you-polite, sg. and pl.'), plural 's' is realized as zero (from [s]) or [e] (from [es]).

Despite the non-overt realization of a grammatical morpheme in pluralization, deletion is still the preferred mechanism as the plural meaning can be recovered either from [e] in consonant-final bases or through other marks (verb agreement, article allomorph, etc.). A phonological zero, however, is not always the preferred outcome for plural [s]. Since focus position is a prominent, strong position, deletion is no longer the best strategy for the realization of the plural morpheme on focused NP heads; as a result, coda [s] is avoided through epenthesis, [se]. Consideration of the pragmatics and semantics of double plural selection thus reveals the presence of the alternation [o] ~ [s], which in turn serves as evidence in favor of /s/ as the underlying plural allomorph in Dominican.

(12) Plural allomorph output forms in Dominican:

 Standard plural Plural in focus position
 [o] [e] [se] [ese]
casa casa casase
mujer mujere mujerese

The examples in (12) demonstrate the existence of this alternation: as explained above, plural /s/ surfaces as zero in regular pluralization and as [se] when followed by epenthetic [e] in focused positions. The plural allomorph is often zero because /s/ will normally be in coda position and therefore deleted like other coda obstruents; yet when deletion is no longer acceptable, and epenthesis is preferred, [e] allows [s] to be syllabified in an onset position and thus to be realized as [s].

In this analysis, I claim that the plural allomorph is /s/ and that [-es] can be derived from /s/. Most analyses of pluralization in Spanish consider plural [e] predictable and thus not underlying. Since the focus of this article is not regular pluralization, but the so-called "double plurals" in Dominican, I will not specifically argue for this particular position, but simply adopt the standard view that the V in [es] is derived (Saltarelli 1970; Contreras 1977; Harris 1980; Moyna and Wiltshire 2000) and show how this works in conjunction with the analysis proposed for "double plurals." (5) I do, however, agree with recent proposals (Harris 1999) that the motivation for -e is not "phonological" in a traditional sense, as speakers of Dominican, as well of other dialects, do in fact produce clusters such as [rs] as in, for instance, hypercorrect forms (e.g. una flo[rs] 'a flower') (see Harris [1999] for the proposal that -e epenthesis is morphological). The analysis presented in this article offers a solution to the problem of the "phonological vs. morphological" nature of plural epenthesis: plural epenthesis reveals the emergence of the unmarked (CV syllables) in the OO phonology. In other words, -e is inserted in plurals to provide a vocalic terminal element thus replacing a consonant-final syllable with its preferred vowel final correspondent. This is not the case for the phonology in general (IO) where faithfulness constraints (DEP-IO) dominate markedness (*Coda).

An important theoretical implication of the data on "double plural" realization and the account proposed here has to do with the role of pragmatic factors, often assumed to be outside the purview of a formal theory of phonology. As Nunez-Cedeno (2003) shows, "double plural" selection is not optional or random, but it is governed by informational prominence (focus). Any formal account of the data that aims at explanatory adequacy needs to capture these facts. Existing alternatives, such as two underlying plural morphemes (/s/ and /se/) or double attachment of the same morpheme, are neither descriptive nor explanatorily adequate as they cannot explain what motivates selection of one form versus the other. However, it is unclear whether, in order to account for the facts, a theory of phonology needs to incorporate pragmatic factors as such or whether it would suffice to refer to the phonetic and/or phonological features that identify informationally-prominent positions, thus excluding reference to pragmatics in formal phonology. Despite much recent work on the topic (Truckenbrodt 1995; Selkirk 2002 and references therein), the role and formalization of focus in phonology remains to be understood. Furthermore, the complexity of the matter places it well beyond the scope of this article. Therefore, and for the purposes of the analysis proposed in Section 4.2, I use "focus" as a cover term for phonetic/ phonological markers of focused NP heads (FOC constraint).

An alternative to the analysis proposed here would be to assume, as most of the literature does, that there is no underlying /s/ in plain Dominican plurals and that [se] is inserted in plurals in focus position (/(e)/ ~ /se/). This account needs to specify an additional morpheme for focused plurals /se/ (or [s] epenthesis, otherwise nonexistent in the language) along with nonfocused /e/ for C-final words. On the other hand, the /s/-plus-epenthesis proposal can account for the facts through a single plural morpheme /s/ and general mechanisms of syllabification already existing in the language; it also captures the connection between [s] and plural and [e] as the default epenthetic vowel in Spanish (yugoslavo vs. eslavo). Note also that this type of analysis is at least partly imposed by the limitations of a serial account. A rule-based account cannot obtain both the plain and double plural outputs from a unique plural morpheme /s/. In such a model, trying to derive both [oreha] and [OREHASE] from /oreha + s/ would require a rule of epenthesis and another of deletion. No matter the order of application of the rules, it is impossible to obtain both output forms:

(13) UR /oreha + s/ /oreha + s/
 deletion [oreha] epenthesis [orehase]
 epenthesis -- deletion --
 output [oreha] [orehase]

As a result, a rule-based analysis needs to specify a separate underlying morpheme for double plurals.

No underlying /s/ is also the standard analysis for Dominican coda 's' in nonplural forms like mas /ma/ 'more' (see also [16] and [17]) (Terrell 1986). Terrell argues that there is no evidence for an underlying coda /s/, given the lack of alternations and the fact that [s] does not normally surface in coda position. Terrell further argues that when [s] does surface, it is in cases of hypercorrection, often appearing in forms with no /s/, e.g. [fisno] ~ Standard Spanish fino [fino], a fact that would remain unexplained under a deletion analysis. Although Terrell (1986: 126) includes plurals amongst the forms for which he postulates no underlying /s/, for example, /kasa/ 'casa, casas', it is easy to see that the restructuring of coda /s/ as zero affects only lexical coda /s/ (underived forms) and not plural /s/ (plural /s/ does in fact have alternations, [o] ~ [se], plain and double plural). In other words, the plural /s/ account presented here is entirely compatible with the absence of underlying coda 's'. The apparent duplication in the analysis is not problematic at all beating in mind the different nature of [s], lexical in one case, underlyingly-specified plural allomorph, in the other.

4.2. Formalization: an OT analysis

In this section I show that the analysis presented in Section 4.1 can be expressed in a natural way in an optimality-theoretic framework. In addition, the OT account has the following advantages: (i) universality of constraints (vs. language specific rules or templates); (ii) greater empirical coverage, as it explains the "apparent" optionality of the so-called "double plurals" and the stress facts mentioned by Nunez-Cedeno (2003).

4.2.1. Lexical (nonplural) coda obstruents. Coda obstruent deletion is the result of the domination of *CodaObst over the faithfulness constraint MAX-IO, as seen in (14b). That epenthesis is not the preferred strategy indicates that only MAX-IO, not DEP-IO, is dominated by *CodaObst.

(14) a. *CodaObs: No obstruents in coda position (McCarthy 2002: 106).

MAX-IO: Every segment present in the input must have a correspondent in the output (Benua 1995; McCarthy 1995).

DEP-IO: Every segment present in the output must have a correspondent in the input (Benua 1995; McCarthy 1995).

b. *CodaObs, DEP-IO >> MAX-IO

The markedness constraint *CodaObs refers to coda obstruents to capture the fact that in Dominican different coda consonants behave differently: thus while obstruents tend to be deleted, sonorants are faithfully retained (or vocalized in some dialects [Pineros 2003]). *CodaSon (here referred to as *Coda, see Note 11 and [32]-[34]; see also McCarthy [2002: 106]) would therefore be dominated by MAX-IO (not shown in the tableaux).

The tableaux in (15) contains candidate evaluation for coda obstruent /d/. Candidates (15a) (final epenthesis) and (15b) (coda obstruent) violate DEP-IO and *CodaObs respectively, both more highly ranked than MAX-IO. (15c), the candidate with coda obstruent deletion, is preferred because it only incurs a violation of MAX-IO. In addition, all candidates in (15) incur a violation of MAX--indicated by the use of parentheses --due to the deletion of /s/ (alternative candidates involving /s/ are not considered for presentational reasons).

(15) /usted/ [ute]

 * CodaObs DEP-IO MAX-IC

a. utede * ! (*)
b. uted * ! (*)
[??] c. ute (*)

(16) and (17) examine /s/-final forms. They show evaluation of
candidates for two possible input forms for [ga] 'gas', one in which
the input contains a lexical /s/ (16) and one in which it does not (17).

(16) /gas/ [ga] (preliminary)

 * CodaObs DEP-IO MAX-IO
a. gase * !
b. gas * !
[??] c. ga

(17) /ga/ [ga] (final)

 * Coda0bs DEP-IO MAX-IO
a. gase ** !
b. gas * ! *
[??] c. ga

While both input forms lead to the same correct output, the preferred input by lexicon optimization is /ga/. Lexicon optimization is a learning strategy that assists the language learner in setting up an underlying representation when the data provide no evidence for it through alternation (Prince and Smolensky 1993). As McCarthy (2002: 77) puts it, learners "[c]hoose the underling representation that gives the most harmonic mapping." As (16) and (17) show, the phonetic output for /gas/ and /ga/ is the same. Nonetheless, /ga/ is selected as the input since there is a more harmonic mapping for the output [ga] from /ga/ (no constraint violation marks as shown in [17c]) than from /gas/ (one violation mark as seen in [16a]).

4.2.2. Realize morpheme and output-to-output constraints. Pluralization consists of the attachment of the plural allomorph /s/ to a singular base. A question that arises is that if there is only a single plural morpheme /s/, how is it that it is realized differently depending on whether it is in an informationally prominent position, e.g. [muherese] or not, e.g. [muhere]? In this context the crucial constraints are two morphemerealization constraints--RM and RM/FOC, (18) and (19)--regarding the relationship between the plural morpheme and overt phonological realization. In the OT literature morpheme-realization constraints have been conceptualized in a narrow fashion (MORPH-REAL) in which case satisfaction is dependent on whether a particular morpheme surfaces or not (Samek-Lodovici 1993; McCarthy 2000: 124) or in a more general way (realize morpheme [RM]) in which any change on the form of a base could be interpreted as the realization of a morpheme and thus satisfaction of the RM constraint in (18) (Kurisu 2001). In this article, I adopt the RM conceptualization of morpheme realization due to its broader empirical application. (6)

(18) RM: All morphemes must be realized. A morpheme is realized iff the outcome has some phonological property which distinguishes it from the base (Kurisu 2001)

Kurisu (2001) shows that the RM formalization has broad empirical coverage in that it can account for both concatenative and nonconcatenative morphology. Under this approach, morpheme realization in concatenative morphology is a consequence of faithfulness constraints (and their ranking) on the underlying form of the affix. More specifically, concatenative morphology is an epiphenomenon of the ranking RM >> DEP-IO, where an affix is inserted in order to realize a morpheme. On the other hand, in nonconcatenative morphology the ranking DEP-IO >> RM >> IDENT. B(ase) (7) captures the lack of any lexically-specified affix and the preference for realizing morphemes by means of phonological modifications to the base. Thus, while Kurisu's model incorporates the possibility of satisfying RM through base modification (nonconcatenative morphology), it also accounts for morphological processes in which morphemes can only be realized through affixation (not through modification of the base). This distinction is important for the current analysis dealing with concatenative morphology.

In addition to a general RM constraint, I propose a second RM constraint--RM/FOC (19)--that requires that morphemes be realized overtly in prosodically emphatic, prominent positions, such as a focus. RM/FOC accounts for plural realization (double plurals) in focus position in Dominican (see [28]-[30] and [34]).

(19) RM/FOC: All morphemes must be realized overtly in focus position (intonationally or otherwise strong position).

RM/FOC can be easily motivated on universal grounds as prominence is intimately connected with explicit marking (e.g. intonation, duration, stress, etc.). In this respect it is worth noting that intonationally prominent positions favor epenthesis and not deletion (cf. Martinez-Gil [1997], who shows that optional epenthesis in Galician is favored in metrical heads of intonational phrases). Additional justification can be found if one views this constraint as part of a universal hierarchy that relates linguistic prominence and overt phonological expression (and/or faithfulness). It is well-known that within the phonological realm, some positions (onset, metrical heads, etc.) are more resistant to epenthesis/deletion than others (coda, nonheads of metrical feet). Within an OT framework, Beckman (1997) proposes a universal hierarchy of constraints (positional faithfulness) in which faithfulness constraints on strong positions (e.g. faith/onset) are ranked higher than faithfulness to weak positions (faith/ coda). Similarly, in the realm of morphophonology, the requirement of morpheme realization can be seen as the result of faithfulness constraints on morphemes being ranked higher than faithfulness to prosodic positions which are not morphemes (e.g. faith/morpheme >> faith/prosodic position). Faithfulness constraints affecting morphemes which are also in focus position (i.e. intonationally prominent, marked by different pitch or duration) would occupy an even higher place in the hierarchy. A prominence hierarchy in which different types of linguistic prominence are related to phonological expression deserves much more detailed study than permitted by the confines of the current study, and therefore (20) is included only as an illustration of what this type of hierarchy might look like. For the purposes of this article, a general constraint relating overt expression (thus no deletion) with morphemes in strong positions such as focus (19) (RM/FOC) will suffice.

(20) Faithfulness to strong, prominent positions: Faith/morpheme/Foc >> Faith/morpheme

In addition to the ranking IDENT. B(ase), RM >> DEP-IO, proper of concatenative morphology, an adequate account of Spanish plural formation needs to refer to OO faithfulness constraints (Benua 1995; McCarthy 1995). As is well-known, the domain of application of pluralization in Spanish is the morphological word, as demonstrated by the fact that the plural /s/ is attached after all other derivational and inflectional morphemes, including terminal elements, -a, -o, and -e (cas-a-s 'house[s]', camin-o-s 'path[s]', chocolat-e-s, 'chocolate[s]'). That the plural morpheme is attached to the morphological word means that OO faithfulness constraints, which evaluate correspondence relations between two output forms, become relevant in the evaluation of plural outputs. For the sake of convenience, I use the abbreviations DEP-OO and MAX-OO to refer to the OO faithfulness constraints specific to pluralization, that is, DEPSgPl and MAX-SgPl. Spanish plural forms contain by definition a violation of DEP-IO with respect to the input of the singular on account of a segmentally specified plural morpheme /s/ (concatenative morphology) [given their definitional nature and for reasons of presentation, these violations of DEP-IO in the plural forms will not be the focus of attention and will not be shown in the tableaux]. In addition and more importantly, under OO correspondence, [s] and [e] in the plural outputs violate DEPOO (see [26a] and [26b]), since [s] and [e] are not present in the output of the singular. [e] does not incur a DEP-IO violation given that the relevant correspondence relation is output to output (the plural is formed on the output of the singular). DEP-IO is vacuously satisfied. Domination of *Coda over DEP-OO makes plural epenthesis possible, while word-final epenthesis is ruled out by DEP-IO >> *Coda (see Section 4.2.4).

Finally, morphological epenthesis (/s/, the DEP-IO violations incurred to satisfy RM) introduces a new set of IO correspondence relations--[IO.sub.Morpheme]. These are the correspondence relations established between the underlyingly specified form of the morpheme /s/ and its output correspondent, usually zero in Dominican Spanish. One way of integrating this into the constraints and constraint-ranking that is formally appropriate, but also convenient on presentational grounds, is to define RM in concatenative morphology as consisting of two conjoined constraints RM & [Faith-IO.sub.Morpheme]. [Faith-IO.sub.Morpheme] requires that an underlyingly specified morpheme have an output correspondent ([MAX-IO.sub.Morpheme]) and vice versa ([DEP-IO.sub.Morpheme]) (see [21]). It also requires preservation of the featural specification of the input ([IDENT-IO.sub.Morpheme]). For RM to be satisfied, [Faith-IO.sub.Morpheme] also must be satisfied; in other words, RM must be satisfied through the surface correspondent of the underlying-specified morpheme, and not through the violation of IDENT. B(ase) or LINEARITY (constraint against metathesis). For instance, a plural like [muhere] (singular [muher]), despite the change to the base, does not satisfy RM in a concatenative language like Spanish (IDENT. B(ase), RM >> DEP-IO) because the underlyingly specified plural morpheme /s/ is not realized, violating [Faith-IO.sub.Morpheme] ([MAX-IO.sub.Morpheme] or [IDENT-IO.sub.Morpheme], for a candidate in which the plural morpheme is in correspondence with [e]). Alternatively, RM satisfaction could be formalized as the combined effect of separate, nonconflicting RM and [Faith-IO.sub.Morpheme], constraints.

(21) [MAX-IO.sub.Morpheme]: A morpheme present in the input must have a correspondent in the output (adapted from Benua 1995; McCarthy 1995).

[DEP-IO.sub.Morpheme]: A morpheme present in the output must have a correspondent in the input (adapted from Benua 1995; McCarthy 1995).

[IDENT-IO.sub.Morpheme]: The featural specification of an input morpheme must be preserved in the output (adapted from Benua 1995; McCarthy 1995).

Although unranked with respect to RM, [Faith-IO.sub.Morpheme], constraints exhibit the same ranking as RM with regard to other constraints.

(22) ... >> RM, [MAX-IO.sub.Morpheme], [DEP-IO.sub.Morpheme], [IDENT-IO.sub.Morpheme] >>...

I will use the cover term RM to refer to satisfaction of RM and [Faith-IO.sub.Morpheme], regardless of whether this is formalized as conjoined constraints or unranked separate constraints. Under the separate constraint account a candidate could violate RM on the basis of a violation of [Faith-IO.sub.Morpheme] and not necessarily of RM. Under the conjoined account a violation of [Faith-IO.sub.Morpheme] means a violation of RM. The distinction has no bearing upon the data being considered here. Both types of outcomes will be represented in the tableaux as RM violations.

4.2.3. Pluralization of V-final words. The system of correspondence relations proposed for regular V-final plurals is shown schematically in (23).

(23) [ILLUSTRATION OMITTED]

(23) shows correspondence relations for [oreha], the optimal candidate for /oreha/ + /s/ as seen in (26c). *CodaObs must dominate RM because the plural morpheme surfaces as zero in (23) and (26c). Thus (26c) incurs a violation of RM (through [Max-IO.sub.Morpheme]). (26b) violates DEP-OO, since [s] is not present in the output of the singular, as well as the more highly ranked markedness constraint *CodaObs. Therefore, (26b) (one *CodaObs mark) loses to (26c) (one RM violation). An additional constraint, Align Plural in (24), accounts for the ill-formedness of a candidate with epenthesis: Align Plural must dominate RM, since a violation of RM is preferred to the misaligned candidate orehase (26a). Align plural is motivated by the fact that the plural morpheme is attached after all other derivational and inflectional morphemes. The data examined so far offer no evidence to indicate that *CodaObs and Align Plural are crucially ranked with respect to each other. However, the "double plural" data (e.g. muherese) will demonstrate that *CodaObs dominates Align Plural (see Section 4.2.4).

(24) Align (Pl, R, Wd, R) (Align Plural): the right edge of the plural morpheme must be aligned with the right edge of the word.

(25) *CodaObs >> Align Plural >> RM

(26) /oreha + s/

& [oreha] [oreha] (8)

 * CodaObs Align Plural RM

a. orehase * !
b. orehas * !
[??] c. oreha * (MAX-
 [IO.sub.Morpheme]

 DEP-OO

a. orehase * (s) * (e)
b. orehas * (s)
[??] c. oreha

An additional plural candidate [orehae] would violate DEP-OO in addition to RM. The correspondence relations obtaining for the losing candidates (26a) and (26b) are shown schematically in (27a) and (27b).

(27a) [ILLUSTRATION OMITTED]

(27b) [ILLUSTRATION OMITTED]

For "double plurals," such as OREJASE (CAPS = focused element), the constraint RM/FOC in (19) becomes relevant. Recall that RM/FOC requires that morphemes be realized overtly in prosodically emphatic, prominent positions, such as a focus. (9)

RM/FOC dominates Align Plural because (28a) OREHASE is preferred to OREHA (28c). As shown in (25) in *Coda Obs dominates Align Plural; furthermore this is shown by the selection of (28a) over (28b). Thus, domination of RM/FOC over Align Plural results in Align Plural and DEP-OO violations (a.k.a. epenthesis), even though epenthesis is not the strategy normally employed in Dominican to avoid coda obstruents. Note that nonplural codas are still affected by deletion in emphatic positions, thus the relevant constraint must refer to the morpheme.

(28) /OREHA + s/ [OREHASE] &
[oreha]

 RM/FOC * CodaObs Align RM DEP-
 Plural OO

[??] a. OREHASE * * (s) * (e)
b. OREHAS * ! * (s)
c. OREHA * ! *

The tableaux in (29) shows evaluation for a form with lexical coda [s]
(nonplural [s]) in focus position; (30) shows the plural of this same
form.

(29) /GA/ &[GA] (singular)

 RM/FOC * CodaObs DEP-IO MAX-IO

a. GASE * (s) * (e)!
b. GAS * ! * (s)
[??] c. GA

(30) /GA + s/ [GASE] &[GA]

 Align
 RM/FOC * CodaObs Plural RM DEP-00
[??] a. GASE * * (s) * (e)
b. GAS * ! * (s)
c. GA * ! *

In (29), given that the final [s] is not part of the plural morpheme, RM/ FOC is vacuously satisfied. Whereas (29a) and (29b) violate DEP-IO and *CodaObs, (29c) incurs no marks and is therefore the winner. In (30), however, pluralization makes RM/FOC relevant, ruling out (30c). (30a) is selected as the output because it only violates Align Plural (vs. a *CodaObs violation for [30b]).

4.2.4. C-final bases. The last set of data in need of consideration consists of C-final bases that end in consonants other than /s/. As explained above, I claim, in agreement with most of the literature, that in nonfocus positions the plural morpheme is /s/; /s/ has two allomorphs, [s] -[es] in standard varieties, and [o] [e] in Dominican. Thus in regular (nonfocus) plurals, (3lc) [muhere] is preferred over (31a) [muherse] due to the high ranking of Align Plural and over [muheres] (31b) because (31b) violates *CodaObs. (31c) violates RM through a violation of [Faith-IO.sub.Morpheme] ([Max-IO.sub.Morpheme] or [IDENT.sub.Morpheme] if [e] is in correspondence with /s/).

(31) /muher + s/ [muhere] & [muher]

 * CodaObs Align Plural RM

a. muherse * !
b. muheres * !
[??] c. muhere * (Faith-
 [IO.sub.Morpheme])

 DEP-OO

a. muherse * (s) * (e)
b. muheres * (s) * (e)
[??] c. muhere * (e)

Yet (31) cannot be complete as inclusion of the additional candidate
[muher] produces incorrect results, as seen in (32).

(32) /muher + s/ [muhere] (preliminary) (10)

 * CodaObs Align Plural RM DEP-OO

a. muherse * ! * (s) * (e)
b. muheres * ! * (s) * (e)
c. muhere * * (e)
[??] d. muher *

However, one crucial constraint is missing in (32). Although a high-ranked constraint on coda obstruents (*CodaObs) is necessary to account for deletion of obstruents, a *Coda constraint (= *CodaSonorant) is still necessary, as coda sonorants are retained in the dialect under consideration. (11) *Coda dominates DEP-OO. Thus while sonorant codas are not repaired by deletion or epenthesis in the singular (muher *muhe *muhere 'woman'), indicating that DEP-IO, MAX-IO >> *Coda, this is not the case for plural forms, where the constraints are *Coda >> DEP-OO. (33c) and (33d) are therefore tied with respect to RM, passing the decision onto *Coda, which correctly chooses (30c) [muhere]. *Coda is not crucially ranked with respect to RM.

(33) /muher + s/ [muhere] (final) & [muher]

 * CodaObs Align Plural RM * Coda DEP-00

[??] c. muhere * *
d. muher * * !

This account reveals the relevance of OO constraints in plural formation in Spanish and also serves to explain plural epenthesis, seen now as the result of favoring CV structure (or in morphological terms, "words are V-final in Spanish") in a morphologically derived, word-level context (output-to-output). In other words, epenthesis of [e] in the plural reflects the emergence of the unmarked (McCarthy and Prince 1994): *Coda effects are seen in the OO phonology of the plural, but not in the language as a whole, as an effect of the ranking DEP-IO >> *Coda >> DEP-OO. The issue of the morphological versus phonological status of plural epenthesis in Spanish appears now much clearer. While plural epenthesis may still be considered phonological, it does not belong to the input-to-output phonology, but to the OO phonology. This accounts for plural epenthesis after codas that would be well-formed in nonplural outputs. At the same time, the connection with the morphological structure of Spanish becomes obvious as the morphology reveals a bias toward CV structure evidenced by the preference for vowel-final words (word markers).

Evaluation for focused forms proceeds as in (34). Candidates (34c) and (34d) are eliminated on account of RM/FOC violations, while (34b) violates the highly ranked *CodaObs. Despite the ties on Align Plural and DEP-IO, (34e), MUHERESE, is better than (34b), MUHERES, because of the additional *Coda violation incurred by the latter. (34e) is thus selected as the optimal candidate. (12) An additional candidate MUHESE (not shown in the tableaux) is also ruled out, indicating that the antideletion constraint MAX-OO dominates *Coda.

(34) /MUHER + S/ [MUHERESE] & [muher]

 RM/ * CodaObs Align RM * Coda DEP
 FOC Plural -OO

a. * * ! * (s)
MUHERSE * (e)
b. * ! * * (s)
MUHERES * (e)
c. MUHERE * ! * * (e)
d. MUHER * ! * *
[??] e. * * (s)
MUHERESE * (e)
 * (e)

4.2.5. C-final paroxytones. An OO analysis like the one proposed here finds further support in that it explains the ill-formedness of "double plurals" with C-final paroxytones, e.g. agil, agile, ?agilese, 'quick'; util, utile, ?utilese 'useful' (Nunez-Cedeno 2003). The data reveal that double plurals in which the output would have to bear stress past the antepenultimate syllable (notice that [se] results in one additional syllable) are not possible. An OO account can explain this through the domination over RM/FOC of the stress constraint(s) that penalize the placement of stress to the left of the antepenultimate syllable.

(35) 3 [sigma] win: Stress cannot be placed to the left of the antepenultimate syllable (cover term for stress alignment constraints) (Harris 1983). [For reasons of simplicity, the relevant stress constraints are referred to by the cover term three-syllable window (3 [sigma] win.)] Stress-OO: The head of a foot in the output must be preserved in the output strings it stands in correspondence with.

3 [sigma] win and Stress-OO must both dominate RM/FOC given that a candidate without plural realization (37c) (RM/FOC violation) is preferred to one with stress past the antepenultimate (37a) or with stress shift (37b). (13)

(36) 3 [sigma] win, Stress-OO >> RM/FOC

(37) /ahil + s/ [aHILE] & [ahil]

 3 [sigma] win. Stress-OO RM/FOC

a. aHILESE * !
b. AHI'LESE * !
[??] c. aHILE *

Candidate evaluation in (37) demonstrates that (37c) is the optimal candidate because (37a) and (37b) incur violations of more highly ranked constraints--3 [sigma] win. and Stress-OO.

To conclude, it is important to emphasize the analysis of the stress data as it constitutes a strong argument in favor of the OO approach to Dominican plurals and to pluralization in Spanish in general. Furthermore, it shows that prosodic constraints can outrank morphological constraints such as RM/FOC.

5. Conclusions

This article presents an analysis of the so-called "double plurals" of Dominican Spanish. An important contribution of the proposal put forth here is the insight that there are no "double plurals" per se in Dominican Spanish: the apparently redundant attachment of the plural morpheme is only the result of general restrictions on coda obstruents in combination with the need for overt realization of morphemes in prominent positions. An optimality-theoretic account in which an entire system of universal constraints must, in principle, be considered is essential in reaching this conclusion. Furthermore, that constraints are universal and always present allows for the possibility of avoiding coda obstruents through deletion as well as epenthesis. This is something which could not be easily derived in a serial model in a nonstipulatory way, but which falls out naturally from the constraints and constraint ranking in OT. In the case at hand, deletion of coda obstruents is usually preferred due to the ranking *CodaObst >> MAX-IO. However, when an output with a MAX violation is not acceptable because it would also violate a more highly ranked constraint (i.e. RM/FOC), a candidate with epenthesis (DEP-OO) violation and no RM/FOC marks is preferred. That plural formation is an OO process and is also crucial to account for epenthesis in regular and double plurals.

The same analysis captures the connection between obstruent coda deletion, plural /s/, and double plurals without having to resort to separate rules or templates. Although the low-ranking of MAX and RM obscure the effects of DEP-IO (/s/ deletion rather than epenthesis in word-final position in singulars and plain plurals), such effects can be seen in double plurals when a more highly ranked constraint, RM/FOC, rules out a candidate with a MAX violation (deletion). In other words, word-final epenthesis is resorted to when other alternatives (deletion) are no longer optimal, thus revealing the actual presence of a process which may have appeared absent in word-final position (since aspiration and deletion are usually preferred). (14)

An additional advantage of the current proposal is the elimination of a separate, additional underlying representation/morpheme (also templates) postulated for the sole purpose of explaining the double plural. This move was required due to the nature of rules and rule ordering in a serial approach.

Finally, the present account of Dominican double plurals also advances some new proposals that are of relevance for phonological theory in general and for Spanish phonology. In the first place, the analysis presented here argues for the extension of a positional faithfulness hierarchy (Beckman 1997) beyond the syllable and the foot. In other words, I argue that a universal hierarchy that relates linguistic prominence and overt phonological expression (i.e. faithfulness) needs to refer to additional prominent positions/units, such as morphemes, intonational phrase heads, etc. This introduces the issue of the nature of linguistic prominence and how that relates to a theory of phonology.

Secondly, the analysis of Dominican plurals sheds new light into the nature of pluralization and the so-called "plural epenthesis" in Spanish. Epenthesis is allowed in the context of pluralization because it is no longer a matter of IO faithfulness, but of OO faithfulness. Thus, "plural epenthesis" is not simply a matter of phonotactics as many supposedly ill-formed clusters can be pronounced in nonplurals, but also of improving word-level well-formedness (emergence of the unmarked in OO phonology). In addition, this analysis brings out the different nature of syllable repair mechanisms in ill-formed codas and coda clusters (final coda deletion) and in the supposedly ill-formed clusters resulting from morpheme concatenation in pluralization (epenthesis), and can therefore explain the presence of deletion in one case, but epenthesis in the other. This insight is also valid for analysis of pluralization in other dialects of Spanish for which the nature (epenthetic vs. underlying) of the mid-vowel preceding the plural allomorph has been a matter of debate for decades. A more detailed analysis of standard Spanish pluralization (non-aspirating varieties) as an OO process is therefore warranted (Colina forthcoming).

Received 22 April 2003

Revised version received 8 June 2004

References

Beckman, Jill (1997). Positional faithfulness. Doctoral dissertation, University of Massachusetts at Amherst.

Benua, Laura (1995). Identity effects in morphological truncation. In University of Massachusetts Occasional Papers in Linguistics 18: Papers in Optimality Theory, Jill Beckman, Laura Walsh, and Suzanne Urbanczyk (eds.), 77-136. Amherst, MA: Graduate Linguistics Student Association.

Colina, Sonia (2003). The status of word-final [e] in Spanish. Southwest Journal of Linguistics 22, 87-108.

--(forthcoming). Output-to-output correspondence and the emergence of the unmarked in Spanish plural formation. In New Analyses in Romance Linguistics. Selected Papers from the 35th LSRL, Jean-Pierre Montreuil (ed.). Amsterdam: John Benjamins.

Contreras, Heles (1977). Spanish epenthesis and stress. Working Papers in Linguistics 3, 9-33.

Foley, J. (1967). Spanish plural formation. Language 43, 486-493.

Harris, James W. (1980). Nonconcatenative morphology and Spanish plurals. Journal of Linguistic Research 1, 15-31.

--(1983). Syllable Structure and Stress in Spanish. Cambridge, MA: MIT Press.

--(1986). El modelo multidimensional de la fonologia y la dialectologia caribena. In Estudios sobre la fonologia del espanol del Caribe, Rafael A. Nunez-Cedeno, Paez Urdaneta, and Jorge M. Guitart (eds.), 41-51. Caracas: Ediciones La Casa de Bello.

--(1999). Nasal depalatalization no, morphological wellformedness si; the structure of Spanish word classes. MIT Working Papers in Linguistics 33, 47-82.

Jimenez Sabater, Max A. (1975). Mas datos sobre el espanol de la Republica Dominicana. Santo Domingo: Ediciones Intec.

Kurisu, Kazutaka (2001). The phonology of morpheme realization. Unpublished doctoral dissertation, University of California, Santa Cruz. [ROA 490, http://roa.rutgers.edu/].

Martinez-Gil, Fernando (1997). Word-final epenthesis in Galician. In Issues in the Phonology and Morphology of the Major Iberian Languages, Fernando Martinez-Gil and Alfonso Morales-Front (eds.), 235-267. Washington, DC: Georgetown University Press.

McCarthy, John (1995). Faithfulness and reduplicative identity. In University of Massachusetts Occasional Papers in Linguistics 18: Papers in Optimality Theory, Jill Beckman, Laura Walsh, and Suzanne Urbanczyk (eds.), 249-384. Amherst, MA: Graduate Linguistics Student Association.

--(2000). A Thematic Guide to Optimality Theory. Cambridge and New York: Cambridge University Press.

--(2002). A Thematic Guide to Optimality Theory. Cambridge and New York: Cambridge University Press.

--; and Prince, Alan (1994). The emergence of the unmarked. In Proceedings of the North East Linguistic Society 24, Merce Gonzalez (ed.), 333-379. Amherst, MA: Graduate Linguistics Student Association.

Morin, Regina (1999). Spanish substantives: How many classes? In Advances in Hispanic Linguistics, Javier Gutierrez-Rexach and Fernando Martinez-Gil (eds.), 214-230. Somerville, MA: Cascadilla Press.

Moyna, Irene; and Wiltshire, Caroline (2000). Spanish plurals: Why [s] isn't always optimal. In Hispanic Linguistics at the Turn of the Millenium. Papers from the 3rd Hispanic Linguistics Symposium, Hector Campos, Elena Herburger, Alfonso Morales-Front, and Thomas J. Walsh (eds.), 31-48. Somerville, MA: Cascadilla Press.

Nunez-Cedeno, Rafael A. (1980). La fonologia moderna y el espanol de Santo Domingo. Santo Domingo: Taller.

--(2003). Double plurals in Dominican: a morpho-pragmatic account. In Theory, Practice, and Acquisition: Papers from the 6th Hispanic Linguistics Symposium and the 5th Conference on the Acquisition of Spanish, Paula Kempchinsky and Carlos-Eduardo Pineros (eds.), 68-82. Somerville, MA: Cascadilla Press.

Pineros, Carlos-Eduardo (2003). An alignment-based account of coda effects in a Caribbean Spanish dialect. [ROA 531, http://roa.rutgers.edu].

Prince, Alan; and Smolensky, Paul (1993). Optimality Theory: Constraint Integration in Generative Grammar. Report no. RuCCS-TR-2. New Brunswick, NJ: Rutgers University Center for Cognitive Science.

Roca, Iggy (1996). Phonology-morphology interface in Spanish plural formation: an optimality analysis. In Interfaces in Phonology, Ursula Kleinhenz (ed.), 210-230. Berlin: Akademie Verlag.

Saltarelli, Mario (1970). Spanish plural formation: apocope or epenthesis. Language 46, 89-96.

Samek-Lodovici, Vieri (1993). A unified analysis of crosslinguistic morphological gemination. Unpublished doctoral dissertation, Rutgers University.

Selkirk, Elizabeth (2002). Contrastive FOCUS vs. presentational focus: prosodic evidence form right-node raising in English. Paper presented at the Speech Prosody Conference, Aix-en-Provence. [http://www.lpl.univ-aix.fr/sp2002/pdf/selkirk.pdf].

Terrell, Tracy D. (1986). La desparicion de /s/ posnuclear a nivel lexico en el habla dominicana. In Estudios sobre la fonologia del espanol del Caribe, Rafael A. Nunez-Cedeno, Paez Urdaneta, and Jorge M. Guitart (eds.), 117-134. Caracas: Ediciones La Casa de Bello.

Truckenbrodt, Hubert (1995). Phonological phrases: their relation to syntax, focus, and prominence. Unpublished doctoral dissertation, MIT.

Notes

* I would like to thank Rafael Nunez-Cedeno for making available to me a prepublication copy of his 2003 article. My gratitude also goes to three anonymous reviewers for their helpful comments and suggestions. Any remaining errors are my own. Correspondence address: Dept. of Languages and Literatures, Arizona State University, Main Campus, P.O. Box 870202, Tempe, AZ 85287-0202, USA. E-mail: scolina@asu.edu.

(1.) I purposely avoid mention of the underlying representation(s) of the plural morpheme at this point, referring to the plural allomorphs in orthographic and phonetic form only. The issue of the underlying representation(s) will be discussed later in the analysis.

(2.) The data in (1)-(4) are from Nunez-Cedeno (2003).

(3.) The facts concerning consonants in Dominican are more complicated. Obstruents may some times vocalize (instead of deleting). While sonorants are usually retained, liquid sonorants often vocalize, and nasals assimilate to a following consonant or velarize. The interested reader is referred to Pineros (2003) for a more detailed analysis of the coda facts. For the purposes of this article, I simply refer to the general asymmetry between obstruents and sonorants: "whereas the former may delete regularly, the latter are normally preserved" (Pineros 2003: 27).

(4.) An alternative analysis in which the plural morpheme is attached twice /s - s/ also misses this generalization.

(5.) That the issue remains unresolved, however, is evidenced by the deletion proposal of Roca (1996); an earlier deletion proposal can be found in Foley (1967). One of Roca's arguments against epenthesis has to do with the reliance of the epenthesis analysis on lexical epenthesis ("which borders on logical inconsistency, since epenthesis is by its very nature non-lexical"). While I agree that word-final -e is not predictable in Spanish and that as a result lexical epenthesis is not justified, I also believe that this in itself does not preclude an epenthesis analysis. In any case, these issues, including a more detailed response to Roca (1996), deserve separate treatment, especially in light of an OO approach to the plural as the one proposed here.

(6.) The Dominican data have the potential for providing insights on how the RM conceptualization of morpheme realization would work for concatenative morphology (Kurisu 2001 focuses on nonconcatenative and subtractive morphology).

(7.) Faithfulness constraint that refers to nonsegments in the base.

(8.) In (26) and the tableaus that follow, '&' indicates the output of the singular. The output of the plural is indicated in brackets without '&'.

(9.) The observations made with regard to RM and conjoined or separate [Faith-IO.sub.Morpheme] constraints apply to RM/FOC as well.

(10.) [??] indicates a candidate incorrectly selected as optimal.

(11.) Pineros (2003) examines the Northern Rustic dialect of Dominican Spanish in which sonorants are instead vocalized and incorporated into the nucleus in order to avoid a *Coda sonorant violation. Under an alignment formulation of coda constraints and a consonant alignment hierarchy (ALIGN-L (stop, [sigma]) >> ALIGN-L (fricative, [sigma]) >> ALIGN-L (nasal, [sigma]) >> ALIGN-L (liquid, [sigma])), deletion of obstruents only is the result of ranking MAX(seg) below ALIGN-L (stop,[sigma]), ALIGN-L (fricative,[sigma]) (*Coda Obs) and above ALIGN-L (nasal,[sigma]), ALIGN-L (liquid,[sigma]). Vocalization of sonorants is obtained through the domination of ALIGN-L (nasal,[sigma]), ALIGN-L (liquid,[sigma]) over the relevant IDENT constraints. Dialects without vocalization rank the IDENT constraints over ALIGN-L (nasal, [sigma]), ALIGN-L (liquid, [sigma]). In the current analysis, *CodaObs comprises Pinero's ALIGN-L (stop,[sigma]) and ALIGN-L (fricative,[sigma]) and *Coda is the equivalent of ALIGN-L (nasal, [sigma]) ALIGN-L (liquid, [sigma]).

(12.) Notice that (34e) incurs three DEP-OO violations because of two epenthetic [e].

(13.) An anonymous reviewer mentions the singular/plural pair regimen, regimenes as evidence against an identity account. However, this plural could not be found in any of the standard sources for Dominican. Given the Greek etymology of the word, regimenes may be a lexicalized form, possibly even entirely absent from the lexicon of the speakers of this dialect of Dominican (generally uneducated). Furthermore, informal data collected by the author of this article suggest that regimenes occurs in child language in the standard dialects. In any case, plural forms of standard dialects (regimenes) per se do not constitute evidence against the current analysis of nonstandard Dominican dialects. Additional data need to be obtained for Dominican.

(14.) Notice also that most cases of phonological epenthesis in Spanish are in word-initial position. For recent discussion on other types of epenthesis in Spanish, see Morin (1999) and Colina (2003).

SONIA COLINA

Arizona State University