1 Introduction
Clitics in general, and clitic pronouns specifically, raise a number of research questions at various levels of analysis. One of these questions is how these elements interact with prosodic units such as the foot or the prosodic word, and how from this interaction a particular stress configuration is attained.
At the empirical level, this paper aims to present an exhaustive description of the typology of stress patterns of verb–enclitic groupings found in Catalan varieties, with the special emphasis on those that are stress-shifting. At the theoretical level, this paper is intended to address the longstanding problem of how grammatical elements, in this case pronominal enclitics, are organised prosodically. The existing formal analyses of the phonology of pronominal enclitics in Catalan are limited in scope and far from exhaustive, in the sense that they do not consider the totality of stress patterns found in this linguistic area (see, for instance, Grimalt (Reference Grimalt2002, Reference Grimalt2004), restricted to Mallorca and Menorca Catalan). A comprehensive picture of the kind presented here, we argue, is crucial in order to provide an accurate and unified formal analysis.
The focus of the present research is drawn from two main Catalan varieties that show a distinct stress behaviour in encliticisation: Barcelona Catalan and Balearic Catalan (spoken in an archipelago in the western Mediterranean sea). The latter is of particular interest because it has developed a set of stress patterns that diverge in many respects from other Romance varieties (including the rest of Catalan dialects, with the exception of North-Eastern Catalan). The prosodic properties of pronominal enclitics in Balearic Catalan are indeed intriguing, as they are stress-shifting. Although stress-shifting or stressable pronominal enclitics have already been reported in other Romance varieties, mainly in Italo-Romance (Kenstowicz Reference Kenstowicz1991; Bafile Reference Bafile1994; Peperkamp Reference Peperkamp1997; Moyna Reference Moyna1999; Loporcaro Reference Loporcaro2000; Ordóñez & Repetti Reference Ordóñez and Repetti2006, Reference Ordóñez and Repetti2014), the patterns found in Balearic Catalan constitute a rich new empirical field for theoretical exploration, because of both their singularity within the Romance space and their internal regularity, and also because of the stress microtypology they embody.
In order to contextualise these patterns and their theoretical relevance, we briefly present the stress facts related to encliticisation. As illustrated in (1), Formentera Catalan shows stress shift to the penultimate and the last syllable of the whole sequence (1a), Barcelona Catalan is characterised by stress stability with respect to the isolated verbal forms (1b), and Mallorca and Menorca Catalan present stress shift to the last syllable (1c), although in Menorca Catalan paradigmatic pressure effects can lead to penultimate stress in some specific combinations (1diii).
In this paper, we show that all patterns of stress shift in Balearic Catalan are in fact motivated by the choice of the unmarked foot type of Catalan, which is a right-aligned moraic trochee. Specifically, we argue that stress shift in Formentera Catalan conforms directly to the unmarked nominal stress system of Catalan, that is, a right-aligned moraic trochee, with the constraint All-Feet-Right undominated, leading to final and penultimate stress. In Barcelona Catalan, the effects of this constraint are inhibited by an anti-alignment constraint prohibiting the right edge of a foot from coinciding with the right edge of a clitic. In Mallorca and Menorca Catalan, stress shift is also understood as a strategy to meet the unmarked stress system of Catalan but in which the anti-alignment constraint is also undominated, leading to catalexis, in this case a melodically empty mora, and thus to final stress.
The theoretical model for this research falls within metrical stress theory as developed in Hayes (Reference Hayes1995) in its application to Optimality Theory (OT) (see McCarthy & Prince (Reference McCarthy and Prince1993a) and subsequent work). The formal analysis presented in the paper covers only and exclusively second singular imperatives and infinitives, but the same analysis of infinitives can be extended to second plural imperatives when followed by an enclitic. This is so because both second plural imperatives and infinitives end in a consonant and behave the same with respect of stress assignment.
The outline of the paper is as follows. In Section 2 the stress shift data under encliticisation is presented. The analysis of both the basic stress pattern in Catalan nominal forms and of verb–enclitic groupings is developed in Section 3. Section 4 is devoted to discussing alternative and previous analyses of the Balearic Catalan data. Section 5 concludes the paper.
2 Data
The empirical interest of this paper lies in the stress patterns concerning verb–enclitic groupings in the varieties of Catalan spoken in Barcelona and on the islands of Formentera, Mallorca and Menorca (to which we refer as Balearic Catalan, in the event of coincidence of behaviour). In most Catalan varieties, pronominal encliticisation is found with positive imperative, infinitive, and gerund forms of verbs of any conjugational class (e.g. Bonet Reference Bonet2002). However, Mallorca and Menorca Catalan show an apparent reluctance to enclisis with gerund verbal forms, for which speakers resort to alternative syntactic structures, or with 3sg imperative forms, for which they resort to proclisis (e.g. Dols Reference Dols2000, Grimalt Reference Grimalt2002; and personal observations). This is why we exclude these kinds of combination from the general analysis; we also exclude 1pl imperative forms, which are difficult to elicit in the same varieties. For expository reasons, we concentrate on enclisis with first conjugation verbal forms, whose infinitive ends in -ar, because this is the most regular class. Encliticisation with verbs of the second and the third conjugational classes shows the same stress patterns, but in most varieties the stems of these verbs are exposed to an allomorphic behaviour and to some irregularities, both unrelated to stress, that might slow down the overall analysis unnecessarily (e.g. Bonet Reference Bonet2002; Bonet & Torres-Tamarit Reference Bonet and Torres-Tamarit2010, Reference Bonet, Torres-Tamarit, Lloret and Pons-Moll2011). It is beyond the scope of this paper to discuss the underlying representation of pronominal clitics and the shape they acquire depending on their position and the phonological context in which they appear, features which have been extensively described in previous work by Wheeler (Reference Wheeler1979: 141–190), Viaplana (Reference Viaplana1980), Mascaró (Reference Mascaró1986: 131), Bonet (Reference Bonet1992, Reference Bonet1995, Reference Bonet2002), Bonet & Lloret (Reference Bonet and Lloret1998, Reference Bonet and Lloret2005), among many others.
In the three following subsections, we present the basic stress patterns found in Barcelona Catalan (Section 2.1), Formentera Catalan (Section 2.2), and Mallorca and Menorca Catalan (Section 2.3). Our data are drawn from two basic sources. The first source is the previous literature devoted to encliticisation in Catalan varieties, which includes work by Bonet (Reference Bonet1992, Reference Bonet1995, Reference Bonet2002), mainly focused on Central Catalan, work by Torres-Tamarit (Reference Torres-Tamarit2008), focused on Formentera Catalan and based on personal inquiries on the ground, and work by Moll (Reference Moll1932, Reference Moll1934), Bibiloni (Reference Bibiloni1983) and Grimalt (Reference Grimalt2002, Reference Grimalt2004). The second source comprises our own personal interviews with three speakers from Mallorca, Menorca and Formentera. These data were elicited through a translation task from Spanish to Catalan and in some cases through a questionnaire including daily life situations in which we asked speakers to give an order to a relative.
2.1 Stress patterns in Barcelona Catalan encliticisation
In the variety of Catalan spoken in Barcelona, as in most Catalan varieties, the adjunction of pronominal enclitics to 2sg imperative and infinitive verbal forms involves a preservation of the position the stress has in the isolated verbal form (e.g. Bonet Reference Bonet2002); namely, there is no stress shift.Footnote [3] With 2sg imperative forms, which have a paroxytonic structure ending in a vowel when occurring in isolation, the adjunction of the clitic leads to proparoxytonic (2ai) or preproparoxytonic structures (2aii). With infinitive forms, which have an oxytonic structure ending in a vowel when occurring in isolation, the adjunction of the clitic may lead to paroxytonic structures ending in a vowel (2bi), unless there is a combination of two clitics that form two syllables, in which case the result is a proparoxytonic structure ending in a vowel (2bii).Footnote [4] In Romance languages, this behaviour of stress stability is documented in Spanish and in Standard Italian (e.g. Loporcaro Reference Loporcaro2000, Ordóñez & Repetti Reference Ordóñez and Repetti2006).
2.2 Stress patterns in Formentera Catalan encliticisation
In Formentera Catalan, encliticisation triggers stress shift. In the combinations of verb plus enclitic(s) ending in a consonant, the stress falls on the last syllable (3aii, vi–viii). In the combinations of verb plus enclitic(s) ending in a vowel, the stress falls on the penultimate syllable (3aiii–v). When two enclitics are attached to the verbal form, stress always falls on the first one; again, final stress is obtained if the second enclitic is consonant-final (3aix) and penultimate stress is obtained if the second enclitic is vowel-final (3ax, xi). Encliticisation after infinitive forms does not cause stress shift if one enclitic is attached (3bii–viii); in the presence of two enclitics, final or penultimate stress is obtained depending on whether the last enclitic is consonant- or vowel-final (3bix–xi). This pattern consistently applies throughout all conjugations and irrespective of the clitic pronoun (see Torres-Tamarit Reference Torres-Tamarit2008; Bonet & Torres-Tamarit Reference Bonet and Torres-Tamarit2010, Reference Bonet, Torres-Tamarit, Lloret and Pons-Moll2011). In the Romance language area, penultimate stress has been identified in some varieties of Lucanian, and in some varieties spoken in Naples, although in the latter case penultimate stress is circumscribed to combinations of two clitics and the host preserves secondary stress (Peperkamp Reference Peperkamp1997). Note, in this respect, that in Formentera Catalan the occurrence of verb–enclitic groupings with a final consonant (3aii, vi–ix) also gives rise to oxytonic structures and that the verbal base does not carry stress, either primary or secondary, as evidenced by vowel reduction of the mid back vowel /o/ to the high back vowel [u] ([ˈkom.pɾə] ‘buy-2sg.imp’
${\sim}$
[kum.ˈpɾa] ‘buy-inf’), a general feature of Eastern Catalan dialects.
2.3 Stress patterns in Mallorca and Menorca Catalan encliticisation
In most varieties of Mallorca Catalan, encliticisation also triggers stress shift, but in this case the stress always falls on the last syllable of the whole verb–enclitic grouping (4a, b), either when the verb–clitic grouping ends in a vowel, a consonant, or when there is a combination of two clitics (Moll Reference Moll1934, Bibiloni Reference Bibiloni1983, Grimalt Reference Grimalt2002). Menorca Catalan shows the same stress patterns as Mallorca Catalan, although in the combinations of 1sg and 2sg.dat plus 3sg.f.acc clitics, stress unexpectedly falls on the penultimate syllable ([kum.ˈpɾə.lə], [kum.pɾə.ˈmə.lə]) (Moll Reference Moll1932). Interestingly enough, some Mallorca Catalan varieties (for instance, Llucmajor and Sóller) show the same behaviour as Menorca Catalan (Grimalt Reference Grimalt2002) under the same kind of combinations. The behaviour of Mallorca and Menorca Catalan, with stress on the last syllable, is analogous to that found in North-Eastern Catalan, where, however, enclisis is only possible with imperative forms (Veny Reference Veny1982; Gómez Duran Reference Gómez Duran2011), in Gascon Occitan, in some varieties of Sardinian (see, however, Kim & Repetti Reference Kim and Repetti2013), and in Viozene Piedmontese (e.g. Ordóñez & Repetti Reference Ordóñez and Repetti2006). Note, finally, that the verbal base does not carry stress, either primary or secondary, although in some varieties of Mallorca Catalan a lack of vowel reduction has been documented (the [a] in [mi.raw.ˈmə] does not reduce to schwa ‘look-2pl.imp at me’) (Bibiloni Reference Bibiloni1983).Footnote [5] The general situation, though, is vowel reduction whenever a verbal form is followed by enclitics. In the examples in (6), /o/ vacuously reduces to [o] because in Mallorca Catalan /ɔ, o/ reduce to [o] but not to [u].
3 Analysis
In this section, we develop a proposal for the stress patterns found in verb–enclitic groupings in Catalan varieties framed within OT. Since the proposal is an attempt to conform the encliticisation stress patterns to the unmarked nominal stress system found in Catalan, first we introduce the basic mechanisms that rule stress in this language and their implementation inside the OT machinery (Section 3.1). We then expand this implementation so it is able to account for the stress patterns found in encliticisation (Section 3.2).
3.1 Basic stress patterns in Catalan
In Catalan, words can bear stress on the last, the penultimate or the antepenultimate syllable. Although the position of stress is not always predictable, since additional morphological and lexical factors can interfere, the patterns summarised in (5) are the ones identified as the unmarked in the Catalan stress system for nominal elements (see, among others, Serra Reference Serra1996, Reference Serra2002; Vallverdú Reference Vallverdú1997). As can be seen in (5), the general tendency is that words ending in a vowel are paroxytonic (5a), and that words ending in a consonant are oxytonic (5b), regardless of whether the vowel belongs to the stem or is an inflectional gender marker. According to Serra (Reference Serra1996), this pattern, which is found in all varieties under study, accounts for almost 86% of the entire lexicon (inflected word forms), and applies to monomorphemic and inflected nominal elements. For the sake of simplicity, the phonetic transcriptions in (5) and (6) correspond to Barcelona Catalan. The stress patterns illustrated, though, are valid for all the varieties under analysis. In the examples below, final vowels are either part of the root (ciri ‘altar candle’ cf. ciri-et ‘altar candle-dim’) or gender markers (bigot+i ‘moustache’ cf. bigot-et ‘moustache-dim’); in both cases, stress is assigned identically.
There are, however, some deviations from these general patterns. First, some words ending in a consonant are paroxytonic (6a); of these, a significant proportion corresponds to derived words with prestressing suffixes (6aiii, iv) (see Mascaró Reference Mascaró1976, Reference Mascaró2003; Pons-Moll Reference Pons-Moll2014). Second, some words ending in a vowel are unexpectedly oxytonic (6b); among these, it is important to distinguish between words ending in an underlying word-final /n/ or /ɾ/ (which do not surface because of the application of the general processes of /n/ and /ɾ/ word-final deletion after stressed vowels in most varieties (6bi–iv)), and words strictly ending in a vowel, which in most (though not all) cases correspond to loanwords and learned words (6bv–viii) (note the lack of derived forms, which reinforces their status as loanwords). Third, there are words with antepenultimate stress, ending either in a vowel or in a consonant (6c), and which are either monomorphemic (6ci–vi) or heteromorphemic (6cvii, viii). With respect to this last group, however, it is important to highlight that there are no proparoxytonic words whose penultimate syllable is closed, except for a few loanwords, such as Manchester or Washington.
Since our proposal is an attempt to conform the encliticisation stress systems to the unmarked stress patterns found in Catalan, here we will focus on these unmarked patterns, and we will put the marked stress patterns on hold, unless they shed some crucial light on the data under study. It is important to note, in this regard, that the deviations found in the Catalan stress system are taken by Mascaró (Reference Mascaró1976) as an argument to consider that stress is specified underlyingly for each lexical item; it cannot be derived from the grammar, but has to be memorised instead. If our predictions are correct, and the encliticisation stress patterns indeed conform to the unmarked nominal stress system depicted above, this would constitute further evidence for a grammar-oriented approach to stress in Catalan.
Leaving this debate aside, previous work by Serra (Reference Serra1996) and Vallverdú (Reference Vallverdú1997) framed within OT has argued that it is indeed possible to draw a regular mechanism of stress assignment in Catalan, governed by the grammar (that is, by a specific constraint hierarchy). The authors, following Cabré (Reference Cabré1993) and Cabré & Kenstowicz (Reference Cabré and Kenstowicz1995), claim that the unmarked foot in Catalan is trochaic (which derives from the ranking Trochee
$\gg$
Iamb, and which represents that the head of the foot is initial and not final); that feet are constructed from right to left (a consequence of the activity of the constraint Align(Foot, Right, PWd, Right)); that the integration of syllables into feet is not exhaustive in that there are no alternating stresses (by means of the mentioned alignment constraint outranking Parse-
$\unicode[STIX]{x1D70E}$
, according to which all syllables must be parsed into feet); that feet must be minimally and maximally bimoraic (as a consequence of the high ranking of Foot-Binarity-
$\unicode[STIX]{x1D707}$
); and that stress in Catalan is sensitive to weight (for which they resort to the crucial ranking of the constraint Weight-to-Stress, according to which heavy syllables are metrically prominent, and which yields final stress in consonant-final words like enemic ‘enemy’ – a behaviour which can also be taken as evidence for mora-defined feet). The study of stress in lexical categories other than nominal forms has received much less attention. For verbal forms, in which stress assignment is morphologised, see Oltra-Massuet (Reference Oltra-Massuet1999, Reference Oltra-Massuet2000).
Our assumption for stress assignment in Catalan follows, grosso modo, these approaches. In what follows, we present the constraints involved, with their definitions, grouped according to the constraint family they belong to (7–10). In (11) we present their ranking in Catalan as a Hasse diagram. (Disjunctive rankings, in which either constraint A or constraint B dominates constraint C, are not indicated.)
In tableau (12) we illustrate the evaluation of a two-syllable vowel-final noun, ciri ‘altar candle’, to show the ranking Trochee
$\gg$
Iamb. Parsing only one syllable into a foot, as in candidate (12c), fatally violates Foot-Binarity
$\unicode[STIX]{x1D707}$
. Candidate (12a) has initial stress, whereas candidate (12b) has final stress. The former is selected as the most harmonic candidate because it satisfies the top-ranked constraint Trochee at the expense of violating the dominated constraint Iamb. The two candidates perform equally well with respect to the two alignment constraints. Only candidate (12c) violates All-Feet-Right; with two-syllable words it is impossible to know whether Foot-Binarity
$\unicode[STIX]{x1D707}$
or All-Feet-Right dominate Iamb, but see tableau (13). (Parentheses around exclamation marks indicate that there is a disjunctive constraint ranking, that is, that either one or the other constraint must dominate the lower constraint violated by the winning candidate. Parentheses around candidates indicate foot boundaries.)
In tableau (13) we illustrate the evaluation of a three-syllable vowel-final noun, canari ‘canary’. This type of forms are relevant for establishing the ranking All-Feet-Right
$\gg$
All-Feet-Left. Parsing only one syllable into a foot, as in candidate (13e), fatally violates Foot-Binarity
$\unicode[STIX]{x1D707}$
; this constraint dominates Iamb. Candidate (13d) has a left-aligned iambic foot. Although this metrification produces penultimate stress in three-syllable words, this candidate is ruled out because it violates both Trochee and All-Feet-Right. Candidate (13c), although it aligns the foot with the right word edge, still violates Trochee. Candidates (13a) and (13b) both satisfy Trochee. However, only candidate (13a) satisfies All-Feet-Right. This tableau, therefore, illustrates two ranking arguments: Trochee
$\gg$
Iamb and All-Feet-Right
$\gg$
All-Feet-Left. These two dominated constraints are violated by the winning candidate (13a).
The tableau in (14) shows the evaluation of a vowel-final noun in which the penultimate syllable is closed. The ranking of *C-
$\unicode[STIX]{x1D707}$
above Weight-by-Position is characteristic of languages in which coda consonants do not contribute to syllable weight. This ranking implies that a coda consonant does not project a mora, as shown in candidate (14a), the optimal candidate. Both candidates (14b) and (14c) include a weight-contributing coda consonant. Besides their violation of *C-
$\unicode[STIX]{x1D707}$
, candidate (14b) also incurs a violation of All-Feet-Right, because the last syllable of the word is not parsed into the foot, and candidate (14c) also violates Foot-Binarity-
$\unicode[STIX]{x1D707}$
because the foot includes more than two moras. The winning candidate aligns the foot perfectly with the right word edge and satisfies Foot-Binarity-
$\unicode[STIX]{x1D707}$
. No direct ranking arguments can be deduced from this evaluation; it can only be inferred that Weight-by-Position is dominated by either Foot-Binarity-
$\unicode[STIX]{x1D707}$
or *C-
$\unicode[STIX]{x1D707}$
, and by either All-Feet-Right or *C-
$\unicode[STIX]{x1D707}$
. The reason why the ranking *C-
$\unicode[STIX]{x1D707}$
$\gg$
Weight-by-Position is necessary for Catalan will become clear when analysing words with stress on the third-to-last syllable and in which the penultimate syllable is closed (tableau 16). At this point, we assume that aligning the foot with the right word edge and restricting moras to only dominate vowels (the segments with the highest level of sonority) takes priority over Weight-by-Position.
In words ending in a consonant, however, stress is not penultimate, but final. We interpret this pattern as the result of positional markedness. Only word-final coda consonants contribute to syllable weight. The ranking of the positional markedness constraint Weight-by-Position]
$_{PWd}$
(relativised to the right edge of the Prosodic Word) above *C-
$\unicode[STIX]{x1D707}$
, which militates against moraic coda consonants, explains this pattern. The effect of Weight-by-Position]
$_{PWd}$
is illustrated in the tableau in (15).
The reason why we claim that only word-final coda consonants are weight-contributing lies in the stress pattern found in loanwords with antepenultimate stress in which the penultimate syllable is closed, as in Manchester or Washington. We suggest that antepenultimate stress is driven by the satisfaction of a lexically indexed version of the constraint Non-Finality(foot)
$_{I}$
(based on McCarthy (Reference McCarthy2008)), which penalises parsing word-final syllables into feet (see Pater (Reference Pater2000) for lexically indexed constraints). Let us consider the tableau in (16). Both candidates (16d) and (16c) violate the top-ranked lexically indexed constraint Non-Finality(foot)
$_{I}$
because the last syllable is integrated into a foot. Only candidate (16b) incurs more violations of *C-
$\unicode[STIX]{x1D707}$
because the word-internal coda consonant is also moraic. It is precisely the ranking *C-
$\unicode[STIX]{x1D707}$
$\gg$
Weight-by-Position that derives antepenultimate stress in these exceptional cases. If the opposite ranking would have been postulated, that is, Weight-by-Position
$\gg$
*C-
$\unicode[STIX]{x1D707}$
, no constraint could stop assigning penultimate stress in such forms. Candidate (16a) parses the first two syllables of the word into a foot because these two syllables are light, although closed. If all coda consonants were moraic in Catalan, Non-Finality(foot)
$_{I}$
would be insufficient for deriving antepenultimate stress. This is the rationale behind the proposed ranking Weight-by-Position]
$_{PWd}$
$\gg$
*C-
$\unicode[STIX]{x1D707}$
$\gg$
Weight-by-Position.
As said in (6a), another exceptional stress pattern is found in words that end in a consonant and have penultimate stress, such as capítol ‘chapter’, llapis ‘pencil’, or numèric ‘numerical’. For the first two cases, which are monomorphemic, making use of an indexed version of the constraint *C/
$\unicode[STIX]{x1D707}_{I}$
(to which these words are associated), ranked above Weight-by-Position]
$_{PWd}$
, is enough to prevent them from having a final heavy syllable. A syllabic trochee is thus built, and penultimate stress is derived. A form like numèric contains a prestressing suffix-ic, which forces the stress to fall on the previous syllable. For these cases, it is the suffix that carries the lexical index, and which explains the exceptional pattern. For vowel-final nouns with final stress, see Section 3.3.2.
Once the general analysis of stress in Catalan nominal elements has been presented, in the next subsection we can focus on the analysis of enclitic-induced stress shift.
3.2 Analysis of stress patterns under encliticisation
As we argue in depth in Section 4, the stress patterns found in Catalan cannot be accounted for in a unified way if they are exclusively attributed to different strategies regarding the incorporation of clitic elements into the prosodic hierarchy, as pursued in Selkirk (Reference Selkirk and Dickey1995) and later in Peperkamp (Reference Peperkamp1997) for Italian dialects; neither they can be explained because of their specific morphosyntactic structure, as proposed in Ordóñez & Repetti (Reference Ordóñez and Repetti2006, Reference Ordóñez and Repetti2014) for various Romance varieties. In this paper we show that in fact the stress patterns in Catalan verb–enclitic groupings follow from the interaction of the constraint hierarchy that governs nominal stress in this language (see Section 3.1) with an anti-alignment constraint that expresses the tendency for certain morphosyntactically defined elements, in our case pronominal clitics, to lie outside the metrical structure (as proposed in Buckley (Reference Buckley1998), among others). Before developing our proposal (Section 3.2.3), we introduce a full justification of the anti-alignment constraint on which our analysis pivots (Section 3.2.2).
3.2.1 The anti-alignment constraint *Align-Right(Foot, Clitic)
In our analysis we make use of an anti-alignment constraint that prohibits the presence of feet whose right edge coincides with the right edge of a clitic, *Align-Right(Foot, Clitic) (17).
Unlike standard alignment constraints (McCarthy & Prince Reference McCarthy and Prince1993a, Reference McCarthy and Princeb), which demand the alignment of two categories in a defined edge, anti-alignment constraints (such as *Align-Right(Foot, Clitic)) prohibit the alignment of two categories, also in a defined edge. This kind of anti-alignment formulation was proposed first by Downing (Reference Downing1994) and Inkelas (Reference Inkelas and Kager1999), and applied by Buckley (Reference Buckley1998) to stress in Manam. Let us consider very quickly Buckley’s analysis. Stress in Manam is final if the last syllable of the word is heavy; otherwise, stress is penultimate, a behaviour that also applies to suffixed forms. However, there is a pre-defined set of suffixes that idiosyncratically entail final extrametricality, in the sense that if they occur word-finally and yield a heavy penultimate syllable, stress will be penultimate, and if the penultimate syllable is light, stress will retract to the antepenultimate syllable. This is why these suffixes are referred to as AP (antepenultimate) suffixes. This morphologically driven type of final extrametricality is understood as the effect of an anti-alignment constraint (Buckley Reference Buckley1998) (18).
We argue that the stress patterns found in Catalan encliticisation are driven by a version of these anti-alignment constraints, namely *Align-Right(Foot, Clitic). The benefit of invoking a constraint such as this one, with respect to alternative analyses (see Section 4), is that it is capable of deriving not only the unstressed and non-shifting character of pronominal enclitics in Barcelona Catalan, but also their stressability in Mallorca and Menorca Catalan. As we will see, in some cases satisfying this constraint will entail having the clitic or the combination of clitics unfooted (i.e. Barcelona Catalan). In some other cases, the very same constraint will imply not having phonological material with diverse morphological affiliation parsed within the same foot, as in Barcelona Catalan, and also as in Mallorca and Menorca Catalan, in this latter case without the need to leave the clitic(s) unparsed, as we will see in detail.Footnote
[7]
Of course, the satisfaction of the anti-alignment constraint (Barcelona, Mallorca and Menorca Catalan vs. Formentera Catalan) or the way it is satisfied (Barcelona Catalan vs. Mallorca and Menorca Catalan) will depend on its crucial interaction with other constraints: as we will see, the constraint responsible for right alignment of feet, All-Feet-Right, and also *Empty-
$\unicode[STIX]{x1D707}$
, whose violation will entail catalexis, the addition of a melodically empty prosodic unit, which we will identify as a mora in our analysis of Mallorca and Menorca Catalan.
3.2.2 Analysis of the stress patterns in Formentera Catalan encliticisation
In Formentera Catalan, the stress pattern of verb–enclitic groupings directly matches the unmarked nominal stress system of Catalan depicted in Section 2.1. (see Torres-Tamarit (Reference Torres-Tamarit2008) and Bonet & Torres-Tamarit (Reference Bonet and Torres-Tamarit2010, Reference Bonet, Torres-Tamarit, Lloret and Pons-Moll2011)). Recall that an isolated imperative form like compra has penultimate stress ([ˈkompɾə]). This is why the attachment of a vowel-final enclitic triggers stress shift to the penultimate syllable and why the attachment of a consonant-final enclitic triggers stress shift to the final syllable of the whole sequence. A perfectly right-aligned moraic trochee is thus built considering the whole verb–enclitic grouping.Footnote
[8]
In this variety, in which the achievement of a right-aligned moraic trochee has priority, feet need to be right-aligned with the whole verb–clitic sequence, and this is a consequence of the ranking All-Feet-Right
$\gg$
*Align-Right(Foot, Clitic). Therefore, clitics can either form a foot with the preceding verbal material ([kum(ˈpɾə.lə)], [kum(ˈpɾəl)]), or be footed alone, if more than one clitic is attached ([kum.pɾə(ˈmə.lə)]). Both situations are illustrated in the tableaux (20–21). For expository reasons, in the following tableaux no candidates violating Trochee are considered. Note, on the other hand, that all enclitics are considered to be internal to the prosodic word, a configuration that follows from the undominated character of Exhaustivity, according to which no syllable can be parsed directly by a phonological phrase.
Tableau (19) illustrates the evaluation of a 2sg imperative form followed by a 3acc.f.sg enclitic, which ends in a vowel. In these cases we expect to find stress shift one syllable to the right, which coincides with the last syllable of the imperative form. The vowel-final enclitic is parsed as the foot’s dependent, and a perfectly right-aligned moraic trochee is built in order to satisfy undominated All-Feet-Right, which dominates the anti-alignment constraint *Align-Right(Foot, Clitic).Footnote [9]
Tableau (20) illustrates the evaluation of a 2sg imperative form followed by a 3acc.m.sg enclitic, which has no vowel; this means that the verb plus enclitic sequence ends in a consonant. In such cases, stress shifts to the last syllable of the imperative form. The enclitic, which contributes a mora because it is syllabified as a word-final coda, is also parsed into the foot, and All-Feet-Right is again satisfied at the expense of violating *Align-Right(Foot, Clitic). This tableau also demonstrates that Weight-by-Position]
$_{PWd}$
dominates *C-
$\unicode[STIX]{x1D707}$
.
As shown in tableau (21), when two enclitics are adjoined to an imperative form, the clitic sequence provides enough segmental material to absorb a whole right-aligned moraic trochee.
The following tableaux illustrate why there is no stress shift in sequences of infinitive forms followed by one enclitic. Isolated infinitive forms have final stress (/kompɾ+a+ɾ/
$\rightarrow$
[kum.ˈpɾa]). A perfectly right-aligned moraic trochee can be constructed thus when a clitic containing a vowel is attached without the need of stress shift (22). When two enclitics are attached, as in tableau (23), however, stress shift occurs, as in imperative forms, because shifting the stress is the only way to obtain a right-aligned moraic trochee.
3.2.3 Analysis of the stress patterns in Barcelona Catalan encliticisation
In Barcelona Catalan, pronominal enclitics never trigger stress shift. This follows from the need to satisfy *Align-Right(Foot, Clitic), which has priority over reaching an unmarked metrical structure, that is, the right-aligned moraic trochee. The satisfaction of *Align-Right(Foot, Clitic) always implies that all clitics remain unfooted, but still parsed into the prosodic word (we assume that Exhaustivity is undominated for all Catalan dialects under consideration). In tableau (24), satisfying the top-ranked anti-alignment constraint implies the selection of candidate (24a), with the clitic unparsed, as the most harmonic. In tableau (25), the clitic has no vowel, and it necessarily has to be parsed as a coda consonant. As a consequence of this, if the right edge of the clitic cannot coincide with the right edge of the foot as demanded by *Align-Right(Foot, Clitic), the last syllable of the sequence containing the clitic must be left unparsed. This results in selecting candidate (25a), which contains a degenerate, monomoraic foot; the anti-alignment constraint *Align-Right (Clitic, Foot) thus dominates Foot-Binarity-
$\unicode[STIX]{x1D707}$
. In sequences containing two enclitics that correspond to two syllables, as in (26), the foot ends up being located two syllables away from the right word edge.
With infinitive forms, stress always falls on the last syllable of the verbal base, that is, on the thematic vowel preceding the infinitive morph /ɾ/. In order to rule out candidate (c) in tableau (27) 
, with initial stress, one must assume that the thematic vowel is underlyingly stressed.Footnote
[10]
A faithfulness constraint Ident(stress) is therefore satisfied in encliticised infinitives in Barcelona Catalan. The constraint Ident(stress) dominates Foot-Binarity-
$\unicode[STIX]{x1D707}$
because it enforces the creation of a degenerate, monomoraic foot.Footnote
[11]
It should be shown that underlying stress on the theme vowel does in fact derive the correct surface forms for all Catalan verb forms. We follow the morphological analysis of verbal forms in Mascaró (Reference Mascaró1986). If we consider verbs of the first conjugation type, like comprar ‘to buy’, assuming an underlyingly stressed thematic vowel does in fact derive the infinitive form: /kompɾ-ˈa-ɾ/
$\rightarrow$
[kum.ˈpɾa], where /-ɾ/ is the infinitive morph. In the indicative mode, the thematic vowel also surfaces as stressed in the imperfect: /kompɾ-ˈa-bə-m/
$\rightarrow$
[kum.ˈpɾa.βəm] ‘we used to buy’, where /-bə-/ is the tense/mode morph, imperfect, and /-m/ corresponds to person/number features, first person plural. In the future and in the conditional, however, the thematic vowel surfaces as an unstressed schwa. This is not a problem since the tense/mode morph of the future and the conditional surfaces as stressed, meaning that they must also be underlyingly stressed: /kompɾ-ˈa-ˈɾɛ-m/
$\rightarrow$
[kum.pɾə.ˈrɛm] ‘we will buy’; /kompɾ-ˈa-ˈɾiə-m/
$\rightarrow$
[kum.pɾə.ˈri.əm] ‘we would buy’. As it happens in derivation, the rightmost underlyingly stressed morpheme is the one that surfaces as such: /nəsion-ˈal/
$\rightarrow$
[nə.sju.ˈnal] ‘national’; /nəsion-ˈal-ˈiʣ-u/
$\rightarrow$
[nə.sju.nə.ˈli.ʣu] ‘I nationalise’; /nəsion-ˈal-ˈiʣ-ˈa-ɾ/
$\rightarrow$
[nə.sju.nə.li.ˈʣa] ‘to nationalise’. The thematic vowel also appears as stressed in first and second person plural forms in the present indicative and subjunctive, which are homophonous: 
‘webuy(-subj)’. In these forms, the thematic vowel of first conjugation type verbs is instead /ɛ/. In the past subjunctive, the thematic vowel also surfaces as stressed, which implies that the tense/mood morph that follows is underlyingly unstressed: /kompɾ-ˈe-si-m/
$\rightarrow$
[kum.ˈpɾe.sim] ‘we bought(-subj)’. In these forms, the thematic vowel is instead /ˈe/. Comparing [kum.ˈpɾe.sim] ‘we bought(-subj)’ with [kum.pɾə.ˈri.əm] ‘we would buy’, for instance, shows that stress assignment in Catalan verbal forms is morphologised and needs to be a lexical property of some morphs. In Balearic Catalan, the thematic vowel in first and second person plural forms in the present indicative and in all forms of the past subjunctive is /ˈa/: [kum.ˈpɾam], [kum.ˈpɾa.sim], a further evidence that both /ˈɛ/ and /ˈe/ are also thematic vowels in Barcelona Catalan (see Mascaró Reference Mascaró1986 for more details). The alternation of vowel quality can be attributed to paradigmatic pressures originated in second conjugation verbs.
In sequences of an infinitive followed by more than one enclitic, stress is also stable compared with the non-encliticised forms. In tableau (28) the evaluation of an infinitive followed by two enclitics that correspond to two syllables is shown. Candidates (28b, c) are ruled out because they fatally violate the anti-alignment constraint. Candidate (28d) is better than candidate (28a) because a bimoraic trochee is built, but only (28a) stresses the thematic vowel as demanded by Ident(stress).
3.2.4 Analysis of the stress patterns in Mallorca and Menorca Catalan encliticisation
So far, the two crucial constraints that we have put forward to account for the differences between Formentera and Barcelona Catalan, *Align-Right(Foot, Clitic) and All-Feet-Right, were in conflict; when one of the two constraints was satisfied, the other was violated. It is legitimate, though, to explore whether the selection of a candidate that satisfies both constraints is possible. In other words, can GEN generate candidates in which the right edge of a foot does not coincide with the right edge of a clitic but, at the same time, have this foot perfectly aligned with the right word edge? Our answer to this question is affirmative and we claim that this situation is attained if catalexis is considered.
Catalexis is a concept borrowed from metrics and, since Kiparsky (Reference Kiparsky1991), it has been understood as the operation of inserting a prosodic constituent that is melodically empty and the logical counterpart of extrametricality. Whereas extrametrical constituents are melodically full but invisible to prosodic rules and principles, catalectic elements are melodically empty but accessible to prosodic rules (Kager Reference Kager1995). Although the original theoretical justifications for catalexis were the elimination of degenerate feet and the satisfaction of word-minimality, in our case there is a different reason: the need to avoid a clitic right-aligned with the foot, but at the same time satisfying the alignment constraint demanding feet to align with the right edge of the word (for a catalectic account of marked stress in Spanish, see Meinschaefer forthcoming, which has inspired part of our proposal).
At a more precise level, we claim that a catalectic mora – Catalan is a weight-sensitive language – is inserted to the right word edge in encliticised forms so that both *Align-Right(Foot, Clitic) and All-Feet-Right are satisfied. This catalectic mora indeed intervenes between the right edge of the clitic and that of the foot, and, at the same time, allows the foot to be placed at the right word edge. This explains why stress is consistently shifted to the last syllable of the whole verb–enclitic grouping in Mallorca and Menorca Catalan. Of course, though, the insertion of this prosodic element and its parsing with the preceding syllable in the foot also entails the satisfaction of Foot-Binarity-
$\unicode[STIX]{x1D707}$
.
The prosodic representation in (30) shows an encliticised imperative with final stress (
‘give-2sg.imp tohim/her’). The superscript mora indicates it is catalectic, that is, melodically empty. This representation shows that a right-aligned bimoraic foot can be built at the right word edge and still misalign the foot with the right edge of the clitic. The insertion of the catalectic mora introduces a violation of the faithfulness constraint *Empty-
$\unicode[STIX]{x1D707}$
, defined in (29).
A candidate in which the inserted mora is linked to the vowel belonging to the clitic would be ruled out by the constraint against bimoraic long vowels *V:, which is undominated in Catalan, a language that bans long vowels completely. It should be noted that the catalectic mora in (30) counts as metrical structure because it is part of the syllable and the foot, although it dominates no melodic element.
The tableau in (31) shows that the optimal candidate in Mallorca and Menorca Catalan is the one with a catalectic mora and final stress. This candidate satisfies both *Align-Right(Foot, Clitic) and All-Feet-Right at the expense of violating *Empty-
$\unicode[STIX]{x1D707}$
, which is dominated.
The fact that the right edge of the clitic cannot coincide with the right edge of the foot forces clitics parsed as coda consonants not to project a mora, that is, a violation of Weight-by-Position]
$_{PWd}$
. By comparing candidate (32c), in which the clitic right-aligns with the foot, with candidate (32a), the winner, we see that *Align-Right(Foot, Clitic) dominates Weight-by-Position]
$_{PWd}$
.Footnote
[12]
There are no more ranking arguments in Mallorca and Menorca Catalan, so the above tableaux are enough to illustrate the system. In sum, for all encliticised forms, the candidate with final stress and a catalectic mora is selected as the optimal one because both *Align-Right(Foot, Clitic) and All-Feet-Right must be satisfied. This is achieved at the expense of inserting a melodically empty mora that has no correspondent in the input. The need for catalexis also renders final codas non-moraic. If they were moraic, feet would end up being trimoraic, which would cause a violation of Foot-Binarity-
$\unicode[STIX]{x1D707}$
.
3.3 Interim remarks
In this section we present the crucial aspects of the analysis of the stress patterns in each dialect (Section 3.3.1). We also give some additional evidence for catalexis in Mallorca and Menorca Catalan (Section 3.3.2) and look closer at exceptional penultimate stress in Menorca Catalan verb–enclitic groupings (Section 3.3.3).
3.3.1 Main constraint-ranking differences between the three varieties
In the previous sections, we have shown that the encliticisation stress patterns conform to the unmarked stress system for nominals in Catalan. This fact should not be surprising as long as enclitics are in fact pronouns.
Let us revise now the main differences between the three varieties. We have seen that one enclitic in Formentera Catalan triggers stress shift in imperative forms, and a sequence of two enclitics triggers stress shift in both imperative and infinitive forms (2pl imperative forms in isolation also have final stress, and therefore they pattern with infinitive forms). Stress is final when the whole verb–enclitic grouping ends in a consonant, and penultimate if the whole sequence ends in a vowel. This pattern found in verb–enclitic groupings gives additional support to the argument that the unmarked foot in (Formentera) Catalan is the moraic trochee, and that clitics are prosodically internal, that is, are always parsed into a foot, which necessarily entails that clitics are internal to the prosodic word. This pattern is achieved through satisfaction of All-Feet-Right at the expense of violating *Align-Right(Foot, Clitic), the anti-alignment constraint against aligning the right edge of clitics and the foot. In Barcelona Catalan, the absence of stress shift is the result of satisfying the anti-alignment constraint *Align-Right(Foot, Clitic), which takes priority over building a right-aligned moraic trochee. Finally, in Mallorca and Menorca Catalan, the last enclitic in a sequence of two enclitics is stressed, as it is the only one in one-enclitic sequences. This is the result of satisfying both *Align-Right(Foot, Clitic) and All-Feet-Right thanks to the insertion of a catalectic, melodically empty mora.
In (34) we present the factorial typology that derives from the ranking of the three crucial constraints involved in pronominal enclitic-induced stress shift in Catalan. These three constraints are enough to derive the microvariation found in Catalan dialects with respect to the stress patterns found in verb–enclitic groupings.
3.3.2 Independent evidence for catalexis in Catalan
Catalexis is put forward by Kiparsky (Reference Kiparsky1991) and Kager (Reference Kager1995) to explain a robust cross-linguistic correlation between the occurrence of peripheral stresses at the end of metrical parsing, which need to be analysed as stemming from degenerate feet, and the occurrence of subminimal words in the same language. These authors claim that if a language allows for peripheral stresses, then subminimal words will also be possible. If degenerate feet are not allowed, therefore subminimal words should not exist, meaning that catalexis is not possible in such language.
Some independent evidence suggests that catalexis is available in the Balearic varieties spoken in Mallorca and Menorca. In these varieties, there is a higher frequency of “minimal words” (i.e. with the surface structure CV) than in other Catalan dialects due to a widespread process of word-final posttonic /ɾ/ deletion. Some examples, which are circumscribed to Mallorca and Menorca Catalan, appear in (35).
Catalan in general, though, is a fair contender for a catalectic approach, at least as a potential strategy, since it is possible to find subminimal words either without any underlying word-final consonant (36a) or containing an underlying /n/ or /ɾ/, which delete in word-final posttonic position (36b) (see 9b). Catalan also permits polysyllabic oxytonic words ending in a vowel both in the native vocabulary (36c) and in loanwords (36d). (For expository reasons, the phonetic transcriptions in the following examples correspond to Central Catalan.)
In principle both subminimal words and oxytonic vowel-final words violate Foot-Binarity-
$\unicode[STIX]{x1D707}$
. By resorting to catalexis, however, we can assume that all the above cases in fact conform to a bimoraic foot. We claim that it is precisely the fact that Catalan allows catalectic moras that subminimal words and vowel-final oxytonic forms exist in the language, following the argumentation in Kiparsky (Reference Kiparsky1991) and Kager (Reference Kager1995). For the cases in which there is an underlying /n/ or /ɾ/ that deletes when it occurs in posttonic word-final position, it can be claimed that the catalectic mora has originated from that consonant (this is in fact a case of a cross-level opaque interaction in which word-final /n/ or /ɾ/ deletion counterbleeds final stress assignment).Footnote
[13]
For the other cases in which there is no trace of an underlying word-final consonant, one can assume that there is an underlying bimoraic final syllable in which the last mora is catalectic, melodically empty. Exceptional stress in these cases is solved by postulating underlying prosodic structure, in this case just a bimoraic final syllable with a second catalectic mora as in (37). The grammar is then responsible for building a right-aligned foot on top of that syllable.Footnote
[14]
In sum, we think that the advantages of catalexis against assuming degenerate feet are: (i) the catalectic mora is not only a strategy to comply with well-formed metrical structures, but when it is underlying it yields marked stress (i.e. oxytonic words that end in a vowel); and (ii) it correlates with the possibility of admitting apparently subminimal words (recall that the correlation between peripheral stress and subminimal words is the argument behind catalexis in Kager (Reference Kager1995)).
3.3.3 Stress paradigm uniformity in Menorca Catalan
As we have shown, in Menorca Catalan stress shifts to the last syllable of the encliticised sequence, as in Mallorca Catalan, except for certain forms with 3.f.acc pronominal enclitics (see the data in Section 2 for more details).Footnote [15] It is interesting to note that the same pronominal enclitics can be stressed or not depending on whether the verbal form they attach to is an imperative or an infinitive. 3.f.acc pronominal enclitics are the only ones that are inflected by gender and number, as opposed to 3.dat pronominal enclitics, which are only inflected by number. We can therefore assume that encliticised forms involving 3.acc pronominal enclitics enter into paradigmatic relations governed by inflection. If we compare the paradigm in (38) involving 2sg.imp forms plus 3.acc enclitics with the paradigm in (39), involving infinitive forms plus the same enclitics, we see that the location of stress is uniform within all the forms in the paradigm. If the stress in (38) were always final, as in Mallorca Catalan, it would be non-uniform across the paradigm because the masculine forms have no desinential ending and it would fall on the last vowel of the verbal stem except for forms with a feminine enclitic. The infinitive–enclitic groupings in (39), in contrast, end all in a vowel, which allows both final stress and stress uniformity within the paradigm (Grimalt (Reference Grimalt2002, Reference Grimalt2004) also attributes these cases to paradigmatic pressures).
In order to analyse paradigm uniformity effects, we make use of McCarthy’s (Reference McCarthy2005) Optimal Paradigms approach. Below we list the principles of Optimal Paradigm.
Verb–enclitic groupings can be seen as complex stems (enclitics, which interact transparently with some word-level phonological processes, are equivalent to derivational affixes). In fact, more than one syntactic terminal, in this case the verb plus the enclitic(s), can be assigned a single morphological word terminal by the post-syntactic morphological component (as defended in Bermúdez-Otero & Payne Reference Bermúdez-Otero and Payne2011), if it is assumed that morphology takes place after syntax. Therefore, it makes sense to postulate an Optimal Paradigms constraint OP
$_{V+Cl}$
-Ident(stress), which guarantees stress uniformity among the members of a paradigm consisting of a verb plus accusative enclitics, which are the only enclitics that are inflected for both gender and number.
The constraint OP
$_{V+Cl}$
-Ident(stress), ranked above the anti-alignment constraint in Menorca Catalan, derives the data presented above, as illustrated in the following tableaux (only catalectic moras are now included). Candidate (b) in (44) violates eight times OP
$_{V+Cl}$
-Ident(stress) because for each member of the candidate, there are two other candidates with a different stress pattern (4x2). Candidate (a), however, satisfies this constraint because stress is regular throughout the paradigm. The winning candidate, however, fairs worse with respect to the anti-alignment constraint because all four members of the candidate align the right edge of the clitic with the right edge of the foot.
When accusative enclitics all end in a vowel (after infinitives, and also after second plural imperative forms), both candidates satisfy the Optimal Paradigms constraint. It is then *Align-Right(Foot, Clitic) the responsible constraint for choosing the most harmonic candidate, candidate (a) in tableau (50), with stress falling on all the enclitics.
In the light of all these analyses, in the next section we discuss alternative approaches in order to frame our proposal into a broader picture and highlight its advantages.
4 Alternative analyses and discussion
In this section we discuss the analyses for the stress patterns related to encliticisation in Romance varieties (with a particular focus on the ones in Catalan) that have been put forward in the literature and we compare them with the proposal pursued in this paper.
4.1 Clitics and their incorporation into the prosodic hierarchy
According to Selkirk (Reference Selkirk and Dickey1995), the most influential work on clitic phonology within OT, clitics can be incorporated prosodically into their hosts in different ways. They can be parsed as dependents of the phonological phrase (44a), as dependents of a recursive prosodic word (44b), as internal clitics within a foot (44c), and as independent prosodic words (44d). As the prosodic word is the domain for stress assignment, in (44a) the clitic is unstressed and does not affect the position of stress in the host; in (44b) the host–clitic sequence receives one primary stress and one secondary stress; in (44c) the clitic is unstressed (or not) but affects the location of stress, and in (44d), the clitic receives primary stress, as well as the verbal base.
Peperkamp (Reference Peperkamp1997), following Selkirk (Reference Selkirk and Dickey1995), claims that the stress patterns found in Standard Italian (with stress stability, as Barcelona Catalan), Lucanian (with stress on the penultimate or final syllable, as in Formentera Catalan), and Neapolitan (with stress on the penultimate syllable in combinations of two clitics), can be attributed to a divergent incorporation of clitics into the prosodic hierarchy. This is determined by a specific constraint hierarchy of the constraints on the prosodic hierarchy: Exhaustivity (no C
$_{i}$
directly dominates a C
$_{j}$
, where j
${<}$
i
$-$
1 (e.g. no PWd directly dominates a
$\unicode[STIX]{x1D70E}$
).), Non-Recursivity (no C
$_{i}$
directly dominates another C
$_{i}$
), and Prosodic Faithfulness (prosodic structure in the input should be preserved in the output.).Footnote
[16]
In Standard Italian clitics are incorporated into the Phonological Phrase as a consequence of the ranking Prosodic Faithfulness, Non-Recursive(PWd)
$\gg$
Exhaustivity(PPh); in Lucanian, clitics are incorporated into the prosodic word, obeying the ranking Exhaustivity(PPh), Non-Recursive(PWd)
$\gg$
Prosodic Faithfulness, and in Neapolitan, they are incorporated recursively into the prosodic word, following the ranking Prosodic Faithfulness, Exhaustivity(PPh)
$\gg$
Non-Recursive(PWd).
Trying to apply this schema to the Catalan patterns faces some obvious difficulties. The varieties without stress shift (Barcelona Catalan) could accommodate to the first schema (Standard Italian), although, given this external incorporation of the clitic(s) into the Phonological Phrase, at the border between the verb and the enclitic, one would expect the application of word-final processes such as word-final obstruent devoicing, posttonic /ɾ/ deletion, consonant cluster simplification, etc., and this is not the case (see Mascaró Reference Mascaró1986). The varieties with stress shift on the penultimate or final syllable (Formentera Catalan) could accommodate to the second schema (Lucanian), but it would still be unexplained why stress is final in Mallorca and Menorca Catalan, given the constraint hierarchy that regulates stress when no clitics are involved. The crucial argument against a Peperkampian account of our data is, indeed, final stress in Mallorca and Menorca Catalan. The important point here is that for these varieties, assuming that clitics are internal to the prosodic word, as for Formentera Catalan or Lucanian, is not enough. Note how an account with clitics associated to its own prosodic word, independent from the one of the verbal base, as in (44d), would not account for Mallorca and Menorca behaviour either, because, as seen, the verbal base does not preserve the stress.Footnote [17]
4.2 Clitics and the cycle
Kenstowicz (Reference Kenstowicz1991), and later Bafile (Reference Bafile1994), propose that the Neapolitan data can be accounted for by assuming that stress assignment proceeds cyclically. First, stress is assigned to the verbal base. Pronominal enclitics are incorporated postlexically, but the metrical structure built in the previous cycle cannot be altered. Loporcaro (Reference Loporcaro2000) also argues that the crucial difference between dialects showing stress stability, like Standard Italian, and those showing stress shift, like Lucanian and Neapolitan, lies in a parametric choice that determines whether or not stress can be reassigned postlexically. Again, although a cyclic view of stress assignment is useful to account for stress stability versus stress shift, it is insufficient to account for the difference found between Formentera and Mallorca and Menorca Catalan, for instance, which are two dialects with stress shift but with different stress patterns. This means that an accurate account of stress shift under encliticisation transcends any derivational approach and requires a closer look at the specific metrification of the sequences as a whole. Cyclic effects have also been explained in OT by means of output–output faithfulness relations. Such an approach would easily account for the difference between stress stability (in which the stress location in the encliticised form must be faithful to the position of stress in the non-encliticised, isolated verbal form) and stress shift (in which output–output faithfulness is sacrificed in order to avoid a marked stress structure). Still, this approach would again leave unexplained the differences in stress patterns associated with those systems displaying stress shift.
4.3 Specific grammar for Mallorca and Menorca encliticisation and an iambic approach
Grimalt (Reference Grimalt2002) is the first formal study that we are aware of that focuses on the stress patterns related to encliticisation in Mallorca and Menorca Catalan. He concludes that the verb plus enclitic(s) combinations ending in a consonant ([do(ˈnəm)] ‘give me’, [do.nə(ˈmos)] ‘give us’, [do.nəm(ˈmos)] ‘to give to us’) can be analysed following the unmarked stress pattern of Catalan already described in Serra (Reference Serra1996) and Vallverdú (Reference Vallverdú1997). As for the verb plus enclitic(s) combinations ending in a vowel ([do.nə(ˈli)] ‘give to him/her’, [do.nəm(ˈmə)] ‘to give me’, [do.nə(ˈɾo)] ‘to give it’), the author acknowledges that they escape from the unmarked stress pattern of Catalan, and this is why he resorts to a constraint Align-Left(Clitic, Foot), according to which the left edge of a clitic must coincide with the left edge of the foot. This constraint, ranked above Foot-Binarity, ensures the selection of candidates with monomoraic feet over otherwise more harmonic candidates with a trochaic structure and satisfying Foot-Binarity, like the forms found in Formentera Catalan. In order to account for final stress in sequences containing more than one enclitic as in [do.nəm.mon.ˈnə] ‘to give some to us’, the author assumes that prosodic words with more than one foot are built ([do.nəm(ˈmon)(nə)], [do.nəm(mon)(ˈnə)]), and that the constraint Align-Right(Head, PWd) is responsible for selecting the one with the second clitic parsed as the prosodic word’s head. This approach, as the author acknowledges, has a clear drawback – the generation of marked structures, that is, monomoraic feet. On the other hand, the other alignment constraint that is put forward by the author, Align-Left(Clitic, Foot), just explains the behaviour of Mallorca and Menorca Catalan – and still with difficulties for cases with clitics whose underlying representation is just a vowel and the clitic is not left-aligned with the foot [do.nə(ˈɾo)] ‘to give it’–, and has nothing to say about the behaviour of the rest of Catalan varieties, including Formentera Catalan. In order to avoid the generation of a typologically marked pattern with monomoraic feet, Grimalt (Reference Grimalt2004), following Wheeler (Reference Wheeler2004), assumes (i) that Catalan has default iambic stress and (ii) the existence of the colon, a prosodic category that stands between the foot and the prosodic word. It is not our purpose in this section to explain the details of Grimalt’s (Reference Grimalt2004) analysis, but two general concerns arise from this perspective. First, the colon is not a widely accepted prosodic category. Second, assuming that Catalan has iambic feet contradicts well known facts of truncation patterns in Catalan (Cabré Reference Cabré1993), among other aspects. But the problem with assuming that encliticised sequences in Mallorca and Menorca Catalan apply an iambic pattern is the existence of the perfectly right-aligned moraic trochee pattern found in Formentera Catalan. It is precisely this pattern found in Formentera Catalan that gives additional support to the hypothesis that Catalan is a weight-sensitive trochaic language.
5 Final remarks
In this paper, we have shown that our prediction that the encliticisation stress patterns conform to the unmarked nominal stress system of Catalan is correct. Beyond that, the fact that verb–enclitic groupings behave as nominal forms with respect to stress assignment and differ from verbal forms is not as idiosyncratic as it could seem, since enclitics are pronouns.
Let us summarise what we see as the advantages of our analysis of pronominal enclitic-induced stress shift in Catalan. First, there is no need to postulate different prosodic incorporation sites for different dialects (cf. Peperkamp Reference Peperkamp1997); we have claimed that in all Catalan dialects under study pronominal enclitics are not parsed outside the prosodic word, although in Barcelona Catalan enclitics misalign with the foot right edge. We could still allow enclitics to directly attach to the Phonological Phrase node for Barcelona Catalan, as Peperkamp (Reference Peperkamp1997) suggests for Standard Italian. However, the difference between Formentera Catalan on the one hand and Mallorca and Menorca Catalan, on the other, would still need to rely on an additional mechanism, which our analysis identifies as catalexis. Therefore, resorting to different prosodic incorporation strategies does not account for the whole set of the facts. Second, by resorting to catalexis, the hypothesis that stress shift is always driven by the avoidance of marked stress patterns (penultimate stress in consonant-final verb–enclitic groupings and antepenultimate stress) can be maintained for all dialects (in contrast with Grimalt (Reference Grimalt2004), who resorts to iambic feet to account for final stress in Mallorca and Menorca Catalan); both in Formentera Catalan and in Mallorca and Menorca Catalan, the result is a right-aligned moraic trochee, which is the unmarked foot in Catalan. Third, the use of the anti-alignment constraint *Align-Right(Foot, Clitic) not only accounts for the Barcelona Catalan data, but is precisely the constraint that also drives final stress in Mallorca and Menorca Catalan. No third constraint is needed to account for either Barcelona or Mallorca and Menorca Catalan. Finally, the different stress shift pattern found in Formentera vs. Mallorca and Menorca Catalan finds a strictly phonological explanation in our analysis, and we have demonstrated that the distinction between weak pronouns and true clitics is not enough to account for the differences found between Balearic Catalan dialects.
Overall, with this study, we hope to broaden the understanding of the phonology of cliticisation in Romance in general and in Catalan in particular.
