High vowel distribution and trochaic markedness in Québécois

2.3.1 Déchaine (1991)

At the core of the proposal contained in Déchaine (1991) is the posited length distinction between high tense and high lax vowels. Tenseness is a derivative of length, or the number of root nodes to which the high vowel is associated. Short (lax) high vowels are mapped to a single root node, while long (tense) high vowels, to two.

In (7a), the vowel is mapped to a single root node underlyingly, while in (7b), the vowel is mapped to two root nodes as a result of weight gain, represented by a dotted line. Weight gain is obligatory in strong rhymes, or rhymes exhaustively dominated by [s], but is optional in rhymes non-exhaustively dominated by [s]. No weight gain may take place in weak rhymes, [w], under this analysis. The proposed metrical template is as follows:

The rightmost strong rhyme, exhaustively dominated by an s-node, represents the position of primary stress. In this position, a high vowel associated to a single root node must gain weight, as shown in (7b). The leftmost strong rhyme is not exhaustively dominated by an s-node, which means that the weight gain is optional, and the vowel may remain associated to a single root node, i. e. surface as lax. The vowel in the weak rhyme, i. e. the medial one, may not augment and surface as tense. Under this analysis, a word like “vie” can only surface as [vi], whereas a word like “numéro” can surface with a lax high vowel [nʏmero] and with a tense vowel [nymero]. However, for Déchaine, a word like “camouflage” cannot surface with a tense vowel: [kamᴜflaʒ], but *[kamuflaʒ], unless such a vowel is word-medial but morpheme-final. In Déchaine’s analysis, the final syllable – a rhyme exhaustively dominated by [s] – is the most prominent metrical position in the word: it is the location of primary stress where weight gain is obligatory, and no segmental weakening can take place.

For a high vowel to surface as lax, i. e. without branching the root node, the rhyme must contain an additional, distinctive root node. This root node may be occupied by a consonant:

This analysis appears to imply that there is a size minimality restriction on the final metrical constituent: [s]-rhyme under [s]-node at the word level. This restriction seems to be two skeletal slots – x x, whether under a branching root node resulting from weight gain, or from two underlying root nodes. An alternative interpretation of the minimal size imposed on the final syllables, in terms of weight units, is undertaken in Montreuil (2004a, b, 2005).

2.3.2 Montreuil (2004a, b, 2005)

Montreuil (2004a,b, 2005) establishes a framework in which the surface realization of high vowels is derived from syllabic weight. Montreuil’s analysis is based on a proposal for a relationship between inherent segmental sonority and phonological weight, amply explored in in the literature (cf Gordon 2002b, 2004b, Morén 2003, Zec 2003^[5]). Québécois vowels are divided into three tiers of height with corresponding degrees of sonority: high vowels (Deg1), mid vowels (Deg2), and low vowels (Deg3). ^[6] Deg2 and Deg3 vowels are associated to µ. The split between Deg1 vowels and Deg2 and Deg3 vowels is viewed as phonologically significant. To represent this distinction within Moraic Theory (Hyman 1985, Hayes 1989b), Montreuil uses a half-mora with a notation of γ to encode the inherent weight of high vowels (Montreuil 2004a,b).

According to Montreuil, consonants are also weight-bearers in Québécois. All consonants outside the class of voiced continuants are inherently weightless, while voiced continuants are associated to the half-mora γ. This sonority distinction between voiced continuants and all other consonants is based on the fact that the former constitutes a group of lengthening consonants: high vowels preceding them surface as tense and lengthened or lax and diphthongized.

In coda position, consonants’ inherent weight is augmented to the next weight value, contributing weight to the total weight of the rhyme – positional weight. Thus, voiced continuants become fully moraic, while all other consonants become associated to γ. Montreuil posits a quantitative minimum for word-final syllables: they must be heavier than γ.

Lax high vowels, linked to a weight that is too low, are unable to constitute a PrWd on their own, illustrated in (12c). In syllables containing a coda, (12b), the minimum is achieved and the vowel surfaces as lax. In (12a), a high vowel in an open syllable meets the weight minimum via association to a full mora, which on the surface corresponds to tenseness.

To sum up, the current analysis exploits the previously explored proposals of quantity-sensitivity and the possibility of positing a branching foot to account for a series of phenomena in French. The sections that follow develop a case for a quantity-sensitive foot in Québécois, primarily motivated by the distribution of high vowels in this dialect and of critical import in analyzing the behavior of these vowels.

3.1 High vowel moraicity: The hypomora

Following Montreuil’s work, I adopt the sonority-based weight distinction for Québécois vowels, and, in line with Scullen’s view, the MinPrWd in Québécois is posited as a light syllable L. Given that high vowels are not able to constitute sufficient rhyme material, in the current proposal, they are analyzed as superlight: SL. This means that for moraic encoding, lax high vowels must be associated to weight value that is less than a mora. I propose to designate this value of fragmented weight as a hypomora, λ. This captures the key assumption from which this analysis departs: anything less than a mora is insufficient, given the minimally monomoraic prosodic requirement, thus re-interpreting for the context of Québécois, where the MinPrWd is L, the conventional view of minimality based on the principle of binarity. While a mora is traditionally viewed as the elementary, basic unit of weight in traditional Moraic Theory (esp. Hayes 1989b) this study follows previously developed analyses of reducing a mora to a smaller unit. ^[7]

Based on Morén (2003), I assume that Québécois follows the cross-linguistic pattern of displaying two types of weight – distinctive and coerced. Distinctive weight is defined as “underlying moraicity reflected in a surface contrast”, while coerced weight is defined as “a restriction on minimal or maximal surface moraicity in some phonological contexts” (Morén 2003: 284). Distinctive moraicity is said to be exemplified by underived geminates for consonants and underived vowel length for vowels. Under the current analysis, vowels are not associated to weight underlyingly – weight is assigned by the grammar based on sonority. Therefore, this is not distinctive weight, given that distinctive weight results from an interaction of markedness and faithfulness constraints. As such, the only distinctive weight in Québécois is the underlying bimoraicity of inherently long vowels: these vowels are underlyingly associated to two moras.

To formalize the proposed view of the foot in Quebecois, centered on the interplay between the underlying form and the surface form mediated by the grammar, I will make use of the Classical version of Optimality Theory (Prince and Smolensky 1993/2003/2004, Archangeli and Langendoen 1997, Kager 1997). In order to encode distinctive moraicity in OT terms, the following constraints are necessary: a familiar faithfulness constraint preserving the moraic affiliation from the input to the output form, and a markedness constraint requiring vowels to be associated to a mora. The latter constraint, adopted here following Montreuil (2004a), is essentially a counterpart to the markedness constraint prohibiting long vowels, *Long V. Unlike the constraint against bimoraic vowels, although present in the grammar, the proposed constraint allows to distinguish among three types of moraic association for vowels under the current view: hypomoraic, monomoraic and bimoraic.

FAITH-µ: The moraic association of a segment in the input is identical to that in the output. One violation per segment that fails to satisfy this condition. (After McCarthy et al. 1995a)

V-µ: The moraic association of a vowel must be a single mora. One violation per vowel that is associated to something other than a single mora.

FAITH-µ>>V-µ: Faithfulness to input vs. Unmarked weight µ

Unlike distinctive weight, coerced weight is not underlying but is a result of weight alteration due to some prosodic restrictions. Under Morén’s coerced moraicity, a vowel augments in weight beyond the traditional monomoraic minimum of syllable nuclei. High vowels, under the current analysis, are assigned by the grammar a weight value that is less than one mora. However, I argue that due to prosodic pressures, high vowels may surface with a monomoraic association – coerced weight.

Morén further delineates three major patterns for coerced moraicity: “no vowels are forced to be moraic in some environment, only the more sonorous vowels are forced to be moraic in some environment, or all vowels are forced to be moraic in some environment” (Morén 2003: 286). Québécois constitutes the third pattern of coerced weight: all vowels are forced to be moraic in some environment. Namely, all vowels, including high, are coerced to be moraic when the vowel is the nucleus of a non-branching syllable in a simple foot. In OT terms, coerced moraicity is expressed in the following constraints: a constraint requiring a grammatical word to constitute a PrWd and a constraint proposed here following Montreuil (2004a), requiring that the vowel’s moraic association be based on its degree of sonority, i. e. for high vowels, the proper weight association is hypomoraic.

GrWd=PrWd: A grammatical word must be a prosodic word. One violation per output that violates this condition. (Kager 1999: 152)

SON-Weight (after Son in Montreuil 2004b): A vowel’s surface weight association corresponds to its degree of sonority, i. e. mid and low vowels are monomoraic, high vowels are hypomoraic. A vowel that surfaces with an improper weight association incurs a violation.

SON-Weight, as defined above, does not militate for vowels to be associated to some weight – it assigns appropriate weight associations for vowels. This means that a vowel that has no weight does not violate SON-Weight. That vowels have weight is compelled by a different constraint (see further discussion.) As such, SON-Weight penalizes high vowels that are monomoraic, their low sonority requiring them to be associated to a hypomora. This may appear problematic for the markednes-based analysis of weight: if the unmarked association for vowels is µ, then SON-Weight, a markedness constraint, calls for a marked, hypomoraic association for a subset of vowels. To clarify, in the current view, the monomoraic association is unmarked, while the hypomoraic is clearly marked (later expressed in a constraint against hypomoras, *λ). SON-Weight serves as the means for the grammar to mediate between the two markedness statements. In other words, while monomoraic is generally unmarked, the low-sonority high vowels are unable to meet the monomoraicity criterion, referred to here as L (as discussed earlier in 2.2) and must thus be lower in weight in the present analysis. What is unmarked for high vowels is marked for non-high vowels, and vice versa.

The view of vocalic moraicity retained here is that all vowels are weight-bearing, ^[10] but not all vowels are moraic, if moraic is defined as being associated to exactly one mora. Long vowels are inherently bimoraic in Québécois, per distinctive weight, non-high vowels are monomoraic, and high vowels, hypomoraic grammatically, are coerced into a monomoraic surface association, per coerced weight.

To sum up in Morén’s terms, who represents coerced weight with the following ranking – *MORA[Seg1] >>BEMORAIC >>*MORA[SEG2] (Morén 2003: 290), grammatical weight assignment in Québécois can be captured as follows, with an addition of BEWEIGHT-BEARING, by analogy with BEMORAIC:

BEWEIGHT-BEARING>> *MORA[+high] >> BEMORAIC >> *MORA [-high]

where BEWEIGHT-BEARING must be defined as “All vowels must be associated to some weight.”

If a high vowel is said to be weight-bearing, but less than moraic, a weight value that is less than one mora is necessitated for the analysis: to represent insufficient weight to satisfy minimal prosodic requirements, but weight nonetheless, to motivate the presence of underweight vowel nucleus (and possible foot head) positions.

To sum up, the surface weight associations for high vowels in Québécois discussed up to this point are as follows:

In order for the high vowel to be associated to λ on the surface, the weight assigned by the grammar, (SON-Weight), must outrank the requirement for vowels to be monomoraic (V-µ). This is demonstrated in (20) below using the forms [vi] “vie” and [vɪt] “vite”, fast. Only vocalic weight is evaluated at this stage.

(20a) and (20c) both win in the absence of context, given that they both respect the weight association assigned by the grammar to high vowels – a hypomora. According to (20), high vowels are realized as lax regardless of the type of syllabic closure. However, high vowels in open monosyllables do not surface as lax, (20a). At this stage of the analysis, let us ensure that this condition is satisfied by invoking the positional markedness constraint cited in (16): GrWd=PrWd.

In (21), the tense variant is selected: while violating the sonority-based weight assignment, it respects the higher ranked GrWd=PrWd. In monosyllables that contain a coda, the high vowel surfaces as lax, i. e. with a hypomoraic association. This means that GrWd=PrWd is satisfied. The presence of a coda, e. g. [vɪt], however, enables a rhyme to satisfy L, endorsing the current assumption that codas in Québécois are moraic. This is assured by the undominated Weight-by-Position constraint (Hayes 1989b):

The tableau shows that assuming coda moraicity leads to the satisfaction of the GrWd=PrWd constraint by (23a): the combined weight of the rhyme is sufficient to avoid the SON-Weight violation that would lead to a monomoraic association of the high vowel, as in (23c). When WbyP is not invoked, the minimal size constraint cannot be satisfied, as in (23b).

For the sake of simplicity and consistency with the Moraic Theory, the weight value to which a coda consonant is to be associated is µ. Given that a hypomora is a construct used to represent a relational weight contrast between two sets of vowels, and not an elementary unit of weight, it is avoided in the representation of consonantal weight. Furthermore, it allows us to avoid adding two hypomoras for candidates with a hypomoraic nucleus: lɪ_λt_λ, which would force a re-conceptualization of the hypomora as a precise unit of weight, unnecessarily complicating the analysis. In the current view, a closed syllable with a lax high nucleus, structurally corresponding to (23a) on the surface, constitutes L, given that it is minimally monomoraic but is short of bimoraic to constitute H.

3.2 High vowel distribution in the quantity-sensitive unary foot

To follow the previously discussed assumptions about the French foot, this simple non-derived foot corresponds to a surface syllable. This type of foot is cross-linguistically marked: feet must be binary at the syllabic or moraic level of analysis. This means that the well-known FTBin constraint, calling for all feet to be binary, is low-ranked in French. To reflect this language-specific scenario in an OT grammar, a pair of constraints of the Alignment family must be introduced. These constraints, first proposed in McCarthy and Prince (1993a), militate for edge-alignment of prosodic constituents: ALIGN(PCat, PCat). The constraint ALIGN(F,R,σ,R) can assure that the right edge of a foot is aligned with the right edge of a syllable, while ALIGN(F,L,σ,L) serves as its left-edge counterpart. In order to achieve a monosyllabic foot, the two alignment constraints must act together on the same syllable. I propose to conjoin the two constraints into a single constraint, referring to the same syllable: ALIGN (F,R,σ,R) & ALIGN(F,L,σ,L). ^[13] The precise formulation is shown below in (24). Additionally, to represent the condition that the final syllable is the the head foot of the PrWd, RIGHTMOST is used. Exhaustive footing is assured by Parse-σ.

ALIGN(F,R,σ,R) & ALIGN(F,L,σ,L): The right edge of a foot corresponds to the right edge of a syllable, the left edge of a foot must correspond to the left edge of the same syllable (based on McCarthy et al. 1993a). One violation per foot that violates this condition.

PARSE-σ: Syllables are parsed by feet. One violation per syllable not dominated by a foot node. (Prince and Smolensky 1993/2002/2004)

RIGHTMOST (Align Hd-Ft, Right, PrWd, Right)

The head foot is rightmost in PrWd. One violation per candidate whose Head foot is not final in the word. (McCarthy et al. 1993a)

ALIGN (F,R,σ,R) & ALIGN (F,L,σ,L) is a constraint specific to the French language, representing the low functionality of the metrical foot in the prosody of the language. This constraint is in competition with the binarity principle, yielding the following constraint ranking: ALIGN (F,R,σ,R) & ALIGN (F,L,σ,L) >> FTBin. Let us recall that in Québécois, a simple foot is not binary: the MinPrWd is set at L in this language. All syllables are dominated by feet, via PARSE-σ. Finally, the final foot, resistant to variation and weakening is the head foot (underlined) of the PrWd.

The winning candidate (25d), fully parsed with four unary feet, only violates FTBin, which is a very low ranked constraint in French. (25a) is eliminated due to an incomplete parse, (25b) due to the violation of the alignment constraint requiring a foot to be no bigger than a single syllable, (25d) has an incorrect word-level dominance structure, with the leftmost head foot.

The PrWd-level constraint GrWd=PrWd, requiring a minimal word in French to correspond to a light syllable, applies at the level of the Prosodic Word and needs only to reference the level of the syllable – the content of the rhyme must be equal to or greater than L. Two critical facts arise that put into question the sufficiency of a minimality requirement imposed by the syllable on the one hand, and the PrWd on the other. First, word-internally, superlight rhymes are permitted: [nʏ.me.ro], [a.mɪ.tje], [dɪ.fɪ.sɪl]. This indicates that a purely quantitative syllable-level requirement is not sufficient to account for the insufficient size of final rhymes. Second, to satisfy the L-requirement in words longer than one syllable, the non-final syllable is not combined with the final syllable into a prosodic unit for the satisfaction of the weight requirement: trɪ_λ.by_µ, but *trɪ_λ.bʏ_λ. This indicates that the size minimum is not imposed by the PrWd, in which case a form like trɪ_λ.bʏ_λ, with two hypomoraic syllables, would be perfectly grammatical. The final syllable stands out as prosodically distinct. I propose to move from the level of the syllabic rhyme to the level of the foot.

As such, the canonical foot of French is a minimally monomoraic, light syllable (L). Any monosyllable constitutes a well-formed foot. The final syllable constitutes the head foot, corresponding to the assumptions about the prominence of the final syllable in French.

Given the current assumption about the size of the MinPrWd, the quantitative minimum at the level of the foot in Québécois is a single mora, resulting in another language-specific deviation from cross-linguistic prosodic generalizations in regards to binarity. It has already been established that FTBin, which is a constraint combining two conditions – minimal and maximal binarity – is of low functionality in the prosody of Québécois. In order to represent the proposed monomoraic minimality in this language, I propose to divorce the minimality stipulation from FTBin and set its value to a single mora.

Given the cross-linguistic understanding of unmarked foot structure, i. e. binary, FTMin-µ as defined in (27) appears to be calling for a marked prosodic constituent. While this is an accurate cross-linguistic insight, it must be remembered that in Québécois, the minimal word size is unary at both the syllabic and the moraic level of analysis. This means that the unary nature of the foot is sanctioned by the prosody of this language and can be viewed as unmarked in this language.

The parameters of a simple foot are analyzed in an OT tableau below. The tableau shows that foot projection is not a factor in the realization of non-high vowels.

The winning candidate (28a) satisfies all the constraints except FTBin, which is very low ranked in French, as discussed previously. Monomoraic underlyingly, low vowels surface as monomoraic without the pressure exerted by FTMin-µ. The picture is dramatically different for high nuclei: FTMin-µ is violated by a hypomoraic vowel, but satisfied by a monomoraic vowel, leading to a violation of SON-Weight.

When a moraic association is inserted in the input, DEP is violated. It is designated here as DEP-W(eight), to include the insertion of both λ and µ.

This formulation allows us to treat a hypomora the same as a mora when possible, to minimize deviation from the traditional understanding of weight, as claimed at the outset of the hypomora proposal. Additionally, given that no segment is associated to λ in the input, there is no need to determine what is being inserted when λ in the input goes to µ in the output. Once again, the analysis remains consistent with the assumption that setting an exact value of λ is an unnecessary step.

The realization of high vowels in open syllables is analyzed with the following constraint ranking:

Candidate (30a) is not footed, and therefore cannot augment under the pressure of FTMin-µ. Candidate (30b) is footed but fails to satisfy FTMin-µ with a hypomoraic vowel. Candidate (30c), (30d) and (30e) incur two violations of DEP-W. Candidate (30f) incurs only one violation of DEP-W, and is thus more harmonic with the ranking than the three previous forms. Comparing the winner with (30b), which also has only one violation of DEP-W, the tableau shows that although both candidates are equally penalized for the insertion of different moraic material – µ vs λ, the candidate with the former satisfies FTMin-µ, while the candidate with the latter fatally violates it. In other words, although DEP-W is not specific enough to distinguish between the two, it does not need to do so – the ranking of FTMin-µ >> SON-Weight accomplishes this objective.

The evaluation of FTBin in the proposed hypomoraic system merits a special mention: given the goal of treating a hypomora the same as a mora when possible, a hypomora is treated equal to a mora in terms of constituting a weight value analyzed separately for the evaluation of binarity. In other words, candidates like (30c-e) are to be viewed as binary at the foot level. This being said, the low-ranking status of FTBin renders this distinction as functionally irrelevant.

For (30b) to be selected as the winner, i. e. for the high vowel to surface as hypomoraic, i. e., when both FTMin-µ and SON-Weight are satisfied, another rhyme constituent must contribute to the total weight of the rhyme. The grammar producing a lax vowel in a closed syllable is shown in (31):

In this tableau, candidate (31d), with the hypomoraic association and in compliance with WBYP, and thus, FTMin-µ, is now the winner.

In polysyllabic forms, FTMin-µ can potentially be satisfied by footing the hypomoraic rhyme with the syllable to the left.

Candidates (32a) and (32b) satisfy foot minimality but crucially violate a highly-ranked alignment constraint. Crucially, (32b) violates SON-Weight gratuitously – it can satisfy FTMin-µ without violating SON-Weight. Candidate (32c) has the correct foot structure, but fails to augment to satisfy FTMin-µ. Candidate (32d) is the winner: the correct foot structure motivates augmentation to satisfy FTMin-µ.

The current proposal for the segmental weight in Québécois and the role of the unary foot projection in the realization of high vowels will now be extended to the non-final environment, which lends itself to a branching foot projection.

4.1 Tense-lax alternation

In the non-final environment, the behavior of high vowels falls short of the complementary distribution in the final syllable. In the closed syllable, the vowel can indeed surface as lax. In fact, Déchaine (1991) goes so far as asserting that laxness is obligatory instead of optional in this environment: [kᴐmɪk|mɑ̃] “comiquement”, [anʏl|mɑ̃] “annulement”, [ʃatᴜj|mɑ̃] “chatouillement.” (Déchaine 1991: 109). While Déchaine does not supply forms in which there is no possibility of a morphological boundary accounting for obligatory laxness, Walker (1984) supplies analogy-independent forms: [bᴜl.var] “boulevard”, [vʏl.gɛr] “vulgaire”. In regards to position, the lax variant can appear in an initial or a medial closed syllable: [kɔ̃.strʏk.sjɔ̃] “construction”, [re.pʏl.siv] “répulsive”. In sum, it is sufficient for a non-final syllable to be closed in order to surface with a lax high vowel, aligning with the distribution in the final syllable, in which laxness is obligatory.

The behavior of high vowels in open syllables is considerably less systematic. The only possible source of traditionally-assumed assimilatory laxness is another high vowel – in the most straightforward cases, the obligatorily lax high vowel in the final closed syllable. Poliquin (2006) puts forth three types of harmonic assimilation of laxness: local non-iterative – [ilɪsɪt], non-local – [ɪlisɪt], and across-the-board – [ɪlɪsɪt] (“illicite”). Non-local harmony also includes cases with transparent vowels: [ɪnedɪt] “inédite”. However well-motivated for the subset of forms containing a laxness trigger, the harmony analysis runs into a problem when the trigger is not present. Two distinct configurations are analyzed: the final syllable contains a tense high vowel: (33), the final syllable does not contain a high vowel: (34).

With forms such as (33a), Poliquin (2006) invokes a derivation in which Final Closed Syllable Laxing and Laxing Harmony take place before Voiced Continuant Lengthening. With forms like (33b), Dumas (1976) invokes harmony via a morphologically related form that does contain a trigger: fɪ.nɪs “finisse”, pres. subj. Forms such as (33c) are accounted for via dissimilation of tenseness (Dumas 1976: 166). For no high vowels in the final syllable, (34a) can also be treated with morphological analogy: [bᴜ.de] ~ [bᴜd] (“boude”, “boudes”, “boudent”). This account, however, would have nothing to say for [bɪ.dɔ̃], in (34b), a noun for which no morphologically-related form can supply a laxness trigger.

Québécois contains a plethora of polysyllabic words which contain only one high vowel in an open syllable, with the rest of the vowels non-high. In sum, three principal types of accounts can be extended to these forms. First, high vowels are assumed to be lax underlyingly, as in Déchaine (1991), and surface as such. Second, lax vowels are tense underlyingly, and the non-final laxness is accounted for via Open Syllable Laxing in Walker (1984). Walker, however, limits this process to initial syllables of trisyllabics only, with medial syllables accounted for by adjacent Laxing Harmony (thus requiring a morphological relationship with a form containing the appropriate trigger). Third, the lax variants in such words can simply be rejected, as in Poliquin (2006). Poliquin excludes the possibility of a lax high vowel in a pre-tonic open syllable, if there is no lax high vowel (or a high vowel) to trigger laxing by a type of harmony. Thus, /mitɛn/, has only one possible realization: [mi.tɛn], at the exclusion of [mɪ.tɛn]. However, [mɪ.tɛn] is perfectly acceptable for Déchaine (1991: 111). ^[15]

An analysis that fully accounts for the behavior of high vowels in the non-final environment must include all reported realizations of these vowels in a unified and systematic fashion. The same analysis must be able to treat the tense/lax alternation in both open and closed syllables, independent of the presence of a harmony trigger or a possibility of establishing morphological analogy. Crucially, forms like [mɪ.tɛn] must not only be included into the analysis but become its epicenter. I will assume that any high vowel in a non-final open syllable can surface as lax, as can any high vowel in a non-final closed syllable.

An important aspect of the distribution that remains to be examined is whether there is a distributional distinction to be made between initial and medial syllables. Previous studies supply contradictory evidence in this regard. Walker (1984: 57–58) states that the process of high vowel laxing is favored in initial syllables. Déchaine (1991: 107) favors the medial syllable for the lax realization: the distributional generalization she provides allows for both tense and lax variants in the initial position, and for only lax variants in the medial position, i. e. only weak vowels can appear in a metrically weak position. The only instance of quantitative findings is presented in Poliquin (2006: 58–73). In his study of harmonic patterns in words in which all vowels are high, e. g. /ʒyridik/ “juridique”, Poliquin finds that two out of 12 speakers opt for the across-the-board pattern, [ʒʏrɪdɪk], three give higher acceptance rates to a non-local pattern, preferring the initial syllable, [ʒʏridɪk], and four speakers prefer the adjacent non-iterative pattern, [ʒyrɪdɪk]. The remaining three speakers are said to not display a distinctive pattern that can be classified. These numbers do not clearly point to one pattern being overwhelmingly more favored over others. Interestingly, however, although inconclusive, more speakers prefer the adjacent non-iterative pattern, the least marked instantiation of harmony. Dumas (1987) confirms the existence of the three patterns, but states that the preferred pattern is the alternating one, corresponding to Poliquin’s non-local. These facts do not easily lend themselves to a generalization setting the medial or the initial syllable apart as a more likely host of laxness for high vowels. In light of this discussion, I will assume that all things being equal, lax variants are as plausible in medial as they are in initial positions.

4.2 Optional deletion and devoicing

These two phenomena share a crucial prosodic restriction: the medial syllable is by far the more frequent locus, with the initial syllable largely excluded. Deletion takes place primarily in open syllables with a simple onset, preferably also followed by a syllable with a simple onset, making re-syllabification possible. A representative data sample for devoicing and deletion in the medial syllable is presented below:

The first study to relate the locus of deletion to abstract phonological structure is Cedergren and Simoneau (1985). For these authors, the two processes of vowel reduction – deletion and devoicing, are a result of the interaction between underlying phonological structure and phonetic surface realization (Cedergren and Simoneau 1985: 57–58). This view strongly resonates with the prosodic account proposed here – the interaction between abstract phonological structure and phonetic realization. The findings in regards to the position favoring deletion are at first glance surprising: the initial position of the rhythmic group is the most likely to host deletion of a high vowel. However, a closer look reveals a critical distinction between function and lexical words. That is, of the 18 items exemplifying syncope (Cedergren and Simoneau 1985: 70–71), ten contain an unrealized high vowel in the initial syllable, but eight of the ten are function words, for the most part not presented in isolation. Of the nine lexical items, eight contain the syncopated vowel in a medial syllable. These facts point in the direction of the validity of the generalization that deletion is largely limited to medial syllables of PrWds. Interestingly, the analysis foregoes incorporating this generalization into the proposal for prominence structure. Instead, the authors adopt the notion of the metrical grid, developed for French in Selkirk (1978). This grid is based on the prominence relationship among syllables, expressed in rhythmic beats: full vowels receive a demi-beat and a basic beat, while schwas – only a demi-beat. Cedergren and Simoneau extend this analysis to high vowels in Québécois. High vowels only receive a demi-beat, unless they are located in the last syllable of a lexical word, or in a closed syllable, in which case they are assigned a full beat. Under this view, the following rule of high vowel deletion is proposed: a high vowel deletes when it is not assigned a basic beat. A quantitative account of high vowel deletion that sheds light on the locality constraints is found in Couturier (1998), a phonetic study. In a corpus of isolated words, Couturier (1998: 159) finds that the rate of realization of high vowels under stress, i. e. the final syllable, is 100 %. Of the high vowels in the non-final environment, 37 % are syncopated. The distribution of these vowels across the syllabic positions reveals a pattern: the higher rates of syncope vs retention are found in the second syllable of tri-, tetra-, and pentasyllabic words, while the rates of retention are higher in the first syllable of di-, tri-, and tetra- and pentasyllabic words (Couturier 1998: 160). The clear trends revealed by these findings are as follows: 1) final syllables are completely resistant to deletion, and 2) initial syllables are strongly resistant to deletion. This provides a basis for establishing a prominence relationship for adjacent syllables, to be grouped into a branching disyllabic foot.

4.3 A proposal for foot structure in the non-final environment

Previous proposals for branching foot structure in Québécois include a single iamb at the right edge (Charette 1991), right-branching metrical structure at the level of the Phonological Phrase and the foot (Phinney 1980), and both iambic and trochaic feet: a left-edge trochee and a right-edge iamb (Goad and Buckley 2006). These proposals are based on the location of primary and, for some, secondary stress in Québécois. ^[17] In the same manner, the type of foot proposed here also posits prosodic headship on the basis of the initial and final phonological prominence: foot heads are aligned with these positions. However, unlike these proposals, the present work deals with more abstract, phonological prominence rather than rhythmic stress, and the current foot proposal is more aligned with the schwa-based trochees developed in Selkirk (1978), further developed in Montreuil (1993), and Bullock (1994), discussed earlier in 2.1. As such, the grouping of the syllables into feet is independent of the location of these prominence positions. Namely, I propose to group non-final hypomoraic syllables with an adjacent syllable to form a left-dominant quantity-sensitive foot. A hypomoraic rhyme will attach itself to the syllable on the left, if available, then to the syllable on the right. The foot configuration is trochaic: the strong left branch prosodically stabilizes the high vowel in the nucleus. The weak branch corresponds to the site of weakening – devoicing and deletion of high vowels. The branching foot allows high vowels to surface as hypomoraic, i. e. lax, in both the strong and the weak branches of the foot, without resorting to obligatory weight augmentation characterizing the final syllable. The current section develops and motivates the proposed foot projection.