Prosodic Structure in Child French : evidence for the Foot

Abstract There is disagreement in the literature on whether French has stress and on whether it has a foot projection. The disagreement stems from the observation that French is unusual in that the phrase rather than the word is the domain of stress assignment, there is optional initial stress in addition to obligatory final stress, and there are rampant violations of word minimality. In view of these observations, this paper examines the outputs of a child learner of Québec French in an attempt to determine the conclusions she has arrived at concerning the status of the foot in the language being acquired. It is demonstrated that, in spite of the challenge that the facts of the target language present, from the onset of production, the child’s outputs are compatible with standard views on prosodic structure. Word minimality effects, the distribution of final lengthening, the emergence of word-final consonants, and the organisation of functional material into prosodic structure are all examined. The paper also provides a preliminary analysis of stress in target French which is, to the greatest extent possible, consistent with standard views on prosodic structure. Key words: French, Québec French, language acquisition, stress, foot binarity, word minimality, truncation, prosodic structure * This research was supported through grants from FCAR/FQRSC and SSHRC. We would like to thank Katherine Demuth, Anna Gavarró, Imme Kuchenbrandt, Conxita Lleó, Annie Lord, and Yvan Rose for comments and/or for answering our many questions by e-mail. We also gratefully acknowledge the comments of two anonymous reviewers. Catalan Journal of Linguistics 5, 2006 109-142 Prosodic Structure in Child French: Evidence for the Foot* Heather Goad Meaghen Buckley McGill University. Department of Linguistics 1085 Dr. Penfield. Montreal, Quebec (Canada) Table of Contents 1. Introduction 2. Stress in French 3. Earlier evidence for the foot in child QF 4. Overview of data source and stages in development 5. Word minimality in Clara’s grammar 6. The lexical/functional split 7. Conclusion References Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 109 1. Introduction There is disagreement in the literature on whether French has stress and, indeed, on whether it has a foot projection at all. If French has feet, the language is best analysed as involving a single iambic foot at the right edge (Charette 1991). However, virtually all of the properties which typically define iambic systems are absent from the language: the canonical iambic foot is quantitatively uneven, which is arguably not the case for French (see note 2), and iambic languages favour left-toright iterative foot construction (Hayes 1995). An additional complication is the observation that «stress» in French is assigned to the last syllable in the phonological phrase (PPh) (Dell 1984), not the prosodic word (PWd) as it is in the vast majority of languages (e.g., Hayes 1995). Given standard approaches to prosodic structure, where the foot is organised into the PWd, it is not immediately obvious how to ensure that only the rightmost PWd in the PPh bears stress, without including rampant destressing rules for words in non-phrase-final position (cf. Dell 1984) or without violating well-formedness conditions on prosodic constituency (Selkirk 1984, 1986, McCarthy and Prince 1986, Nespor and Vogel 1986). Equally problematic is the observation that French freely violates word minimality, the requirement that lexical words be minimally one binary foot to be well-formed (McCarthy and Prince 1986); subminimal (monomoraic) words abound in the language (e.g., [lε] lait ‘milk’, [ʁi] gris ‘grey-MASC’). If there were no foot in French and «stress» were instead phrase-final prominence, these observations would follow straightforwardly. From the point of view of early acquisition, the possibility that some adult languages without a foot projection exist may be desirable, as children’s earliest outputs across many languages show a strong preference for CV words (Jakobson 1941/68, Ingram 1978, Fikkert 1994). This preference is difficult to motivate on the basis of adult language behaviour as outputs of this shape are subminimal, and most adult languages that permit violations of word minimality do so only under limited conditions. Accordingly, if children’s grammars reflect what is cross-linguistically unmarked (Jakobson 1941/68, Stampe 1969, Gnanadesikan 1995/2004), then for monosyllabic targets, foot well-formedness should prefer outputs of the shape CVV, CVC or CVCV, not CV. If, however, there are adult languages for which the absence of a foot projection can be motivated, French being a logical choice to consider, then this constituent will not be projected in the initial state grammar1and syllable well-formedness will first favour CV outputs. In short, the possibility that French lacks feet is significant for all early grammars, as it may lead to a formal account of the otherwise intractable CV stage in acquisition. In light of this and the challenge that French «stress» presents for the linguist, this paper examines the outputs of one learner of Québec French, Clara (see Rose 110 CatJL 5, 2006 Heather Goad; Meaghen Buckley 1. This is in contrast to many works in Optimality Theory (OT), where it is assumed that all prosodic structure is present from the onset of acquisition, and where the effect of an impoverished prosodic hierarchy is left entirely to constraint ranking (e.g., Demuth 1995, Pater 1997; cf. Goad 1996). The problem in this case is that any constraint which favours CV words (e.g., alignment) cannot be motivated on the basis of adult language behaviour. Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 110 2000), in an attempt to determine what conclusions she has arrived at concerning the status of the foot in French. On the basis of the above discussion, we might have expected Clara’s outputs to display rampant violations of word minimality. Counter to expectation, we will demonstrate that, from the onset of production, Clara builds a binary iambic foot at the right edge of the PWd/PPh (the PWd and PPh are initially co-extensive). Indeed, her outputs are compatible with standard views on prosodic structure, including the organisation of constituents and constituent binarity. Given this finding, an important goal of the paper is to provide an analysis of stress in French which is, to the greatest extent possible, consistent with standard views on prosodic structure. We turn to this issue first. 2. Stress in French 2.1. Observations French is traditionally described as having prominence on the rightmost syllable containing a full-vowelled (non-schwa) nucleus in the phrase, where prominence involves both greater duration and higher pitch (Fouché 1934, Garde 1968). The precise formal nature of this «prominence», however, is disputed. Some scholars have interpreted prominence as evidence for foot structure: an iambic foot is built at the right edge of the phrase (Charette 1991, Scullen 1997). Others have assumed that French has no feet at all (Verluyten 1982, Mertens 1987, Jun and Fougeron 2000), and «stress» is instead formally a phrase-final effect. An additional complicating factor is the observation that prominence is not restricted to phrase-final position. Some researchers have argued that French is characterised by both initial and final prominence (e.g., Gendron 1966 for Québec French (QF); Mertens 1987 for European French (EF)); others have interpreted initial prominence as secondary stress (Paradis and Deshaies 1990, Scullen 1997). Part of the challenge involved in formally characterising the system stems from the observation that these accents differ in a number of ways. First, the initial accent is optional, while the final accent is always realised (Paradis and Deshaies 1990 for QF; Jun and Fougeron 2000 and Post 2003 for EF). Second, different phonetic cues are associated with each accent. As is characteristic of iambic systems, the final accent is principally realised through increased length on the vowel (Walker 1984 for QF; Delattre 1966 for EF); this vowel is also typically assigned a high tone (Ouellet and Thibault 1996 for QF; Jun and Fougeron 2000 for EF). More like trochaic systems, however, the initial accent does not involve lengthening, but is instead marked by a high tone on the first or second syllable of the syntactic constituent (Mertens 1987, Jun and Fougeron 2000). Representative examples are provided in (1).2 Prosodic Structure in Child French: Evidence for the Foot CatJL 5, 2006 111 2. We have not transcribed the final vowels in (1) as long as we do not believe that the increased duration observed on these vowels warrants the addition of a mora. This position is supported through a comparison of the relative duration values of final stressed open syllables and non-final unstressed open syllables in French in contrast to English where final stressed syllables are indisputably bimoraic. Although in non-final position, unstressed open syllables are somewhat longer Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 111 (1) (H) H || [ [telefɔne]PWd ]PPh téléphoner ‘to telephone’ (H) H || [lə [telefɔn]PWd ]PPh le téléphone ‘the telephone’ In short, while the two accents in French fulfil a similar prosodic function as domain-edge markers, it is also clear that they are fundamentally different in nature, and any formal characterisation of the system will need to take account of this. Before turning to the analysis we propose, we outline our assumptions about prosodic structure. 2.2. Prosodic structure A partial prosodic hierarchy, hypothesised to be universal, is given in (2) (Selkirk 1984, 1986, McCarthy and Prince 1986, Nespor and Vogel 1986). (2) Phonological Phrase (PPh) | Prosodic Word (PWd) | Foot (Ft) | Syllable (σ) Early work in prosodic phonology assumed that the organisation of material into prosodic constituency must respect the Strict Layer Hypothesis (SLH) (e.g., Selkirk 1984, Nespor and Vogel 1986), where each constituent is strictly dominated by the immediately higher category, as per (2). More recently, researchers have recognised that the SLH is untenable as an inviolable whole. Selkirk (1996), for example, has proposed that the SLH be decomposed into four optimality-theoretic constraints, as in (3). 112 CatJL 5, 2006 Heather Goad; Meaghen Buckley in duration in French (mean ratio French to English is 1:0.87), final stressed open syllables are considerably longer in English than in French (mean ratio 1:1.36) and the within language ratio for final stressed open syllables to non-final unstressed open syllables is 1:2.78 for English but only 1:1.79 for French. (The ratios for closed syllables are similar to those of open syllables. All ratios are based on the values provided in Delattre 1966: 186.) Thus, while both languages exhibit a phonetic final lengthening effect, only for English is there compelling evidence that final stressed syllables are bimoraic. Further evidence that final lengthening does not involve the addition of a mora in adult French is provided in section 5.1.2 when Clara’s outputs are examined. Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 112 (3) Constraints on prosodic domination (Selkirk 1996: 190): a. LAYEREDNESS: No Cidominates a Cj, j > i (e.g., «No σdominates a Ft») b. HEADEDNESS:Any C imust dominate a Ci-1 (except if Ci= σ) (e.g., «A PWd must dominate a Ft») c. EXHAUSTIVITY: No Ciimmediately dominates a constituent Cj, j < i-1 (e.g., «No PWd immediately dominates a σ») d. NONRECURSIVITY: No Cidominates Cj, j = i (e.g., «No Ft dominates a Ft») The constraints in (3a-b) are universally inviolable; that is, domination relations in the prosodic hierarchy are non-commutable (LAYEREDNESS), and every constituent must dominate at least one constituent of the immediately inferior category (HEADEDNESS). If the prosodic hierarchy is universal and if all structures must respect HEADEDNESS, then each phrase in French must contain at least one prosodic word, and each prosodic word must contain at least one foot. Drawing on evidence from Jun and Fougeron (2000) and Post (2003), we will show that there is clear evidence for the PWd level in French. We will suggest further that the initial high tone denotes the presence of a trochaic foot at the left edge of the PWd, while final lengthening denotes an iambic foot at the right edge of the PPh. Before turning to the details, we first consider the prosodification of functional material. In contrast to (3a-b), the constraints in (3c-d) are violable; indeed, they are frequently violated in constructions containing function words and inflectional morphology. There are a number of options for the prosodification of functional material, depending on both morpho-syntactic and phonological criteria. Here, we consider only the prosodification of morphologically-free functional items. As (4) shows, in the spirit of Selkirk (1996), we assume that in the unmarked case, such items are prosodified as free clitics; they link directly to the PPh, outside the PWd of the lex on which they depend. (4) Free clitic (Selkirk 1996: 188): Although this representation comes at some cost as concerns strict layering (EXHAUST(PPh) (3c) is violated), the benefit to the learner of a structure like this, where functional material is organised outside the PWd which contains lex, is that the syntactic and phonological representations are isomorphic. Accordingly, learners can use their knowledge of the functional-lexical split in syntax to bootstrap Prosodic Structure in Child French: Evidence for the Foot CatJL 5, 2006 113 PPh PWd fnc lex Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 113 into the prosodic representation, or they can use their understanding of differences in the phonological properties of functional versus lexical material to bootstrap into the syntactic representation.3In French, morphologically-free functional items which precede their host such as determiners, prepositions, subject clitics, and preverbal object clitics are prosodified as free clitics (Buckley 2004; see also Cardoso 2003 on closely-related Picard). Indeed, there is no phonological evidence (stress, segmental phenomena) that would lead learners to stray away from this analysis.4 In section 6, we will show that this position finds support in Clara’s grammar; it is thereby consistent with the view of an elaborated prosodic hierarchy (foot-PWdPPh) for child French. 2.3. Analysis In section 2.1, we summarised the debate concerning the status of stress in (Québec) French. Some have assumed that French has no feet; others maintain that it does have feet despite seeming to violate well-attested generalisations governing foot structure. As the prosodic hierarchy – including the foot projection – is hypothesised to hold universally (section 2.2), in the following lines, we will sketch an account of French stress which is largely consistent with the organisation of constituents in this hierarchy. We will then turn our attention to Clara’s early outputs which, we will argue, are consistent with this approach. We saw in section 2.1 that prominence in French is characterised by both an initial and final accent. Initial prominence is indicated by pitch, while the major perceptual cue for final accent is length. Cross-linguistic research reveals that 114 CatJL 5, 2006 Heather Goad; Meaghen Buckley 3. The position taken here is the inverse of that where functional material is initially prosodified internal to the PWd and is moved outside of this domain in the course of acquisition (see Lleó 1997, 2003 and Demuth 2001 for discussion consistent with this view). If functional material were initially prosodified internal to the PWd, we would expect parallel treatment of fnc+lex and lex strings which are the same length and have the same stress profile, for example (ðə)fnc + (œtejbəl)lex ‘the table’ and (pəœtejtow)lex ‘potato’ in English. As we will show in section 6, Clara’s grammar does not treat fnc+lex and lex strings in the same fashion which is consistent with the view that leftedge fnc is prosodified outside of the PWd which defines lex. 4. There is one fact which may lead learners to posit that fnc is PWd-internal. Since French permits subminimal words, it is conceivable that the determiner in phrases like [lə lε] le lait ‘the milk’ would be organised internal to the foot and PWd to respect foot binarity. The fact that determiners are almost always required in French is consistent with such an analysis; they could be obligatory before monosyllabic nouns for prosodic reasons and this pattern would then be generalised to nouns of all shapes on morphosyntactic grounds. There is some data in the literature which is compatible with this position. Veneziano and Sinclair (2000) and Tremblay (in press) report a higher incidence of putative function words being produced before monosyllabic lex than before longer lex for the French children they examine; if foot binarity places a lower bound on early outputs, these findings are consistent with these children having arrived at a PWd-internal analysis for fnc (note, however, that Veneziano and Sinclair interpret this asymmetry as support for the view that the putative function words are instead fillers). We do not consider this option further for two reasons. One, in the adult grammar, if function words were organised within the lowest PWd, we would incorrectly expect them to be candidates for left-edge stress (section 2.3). Two, Clara shows evidence of linking fnc directly to the PPh, regardless of the length of the following lexical word. Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 114 trochaic systems tend to be characterised by alternations in pitch and intensity, while iambic systems are marked by alternations in length (Hayes 1995). It would appear, then, that French is a language that has both iambic and trochaic feet. In parametric approaches to stress, it is difficult to formally express a system with both iambs and trochees. In OT, however, where a single constraint can take more than one argument, this scenario is predicted to occur, even if it is rare; that is, there should be some languages where both FOOTFORM(Trochaic) and FOOTFORM(Iambic) are highly ranked.5A conflict will, of course, arise when a language strives to build different types of feet at the same time. We suggest that in French, such a conflict is largely averted as exactly one foot of each type is aligned with opposite edges of the relevant domain. When a conflict does arise, the rightaligned iambic foot takes priority (see below). To formally express the view that the initial high tone marks the left edge of a trochee while the final accent denotes the right edge of an iamb, we propose the alignment constraints in (5). Since alignment constraints can take any prosodic constituents as arguments, ALIGNIAMB (5a) can align a right-headed foot with the right edge of the PPh without running afoul of HEADEDNESS (3b) or the organisation of constituents in the prosodic hierarchy (2). Importantly, however, we will demonstrate that there is a role for PWds in the assignment of stress in QF as well, as revealed by (5b) which aligns a left-headed foot with the left edge of the PWd. (5) a. Final accent: ALIGNIAMB: Align (Iamb,R,PPh,R): Align the right edge of every rightheaded foot with the right edge of some PPh b. Initial accent: ALIGNTROCH: Align (Troch,L,PWd,L): Align the left edge of every leftheaded foot with the left edge of some PWd In order for the patterns discussed in section 2.1 to be selected as optimal, ALIGNIAMB must dominate ALIGNTROCH. First, recall that the initial H tone, which marks the trochaic pattern, is optional, while final lengthening, which reflects the iambic pattern, is always present. Second, the dominance of ALIGNIAMB is revealed in examples where the only PWd in a PPh is bisyllabic. In constructions of this type, ALIGNIAMB is satisfied at the expense of ALIGNTROCH; see [lə[(stiœlo)Ft]PWd]PPh in (6a). (Both alignment constraints could be satisfied by the alternative [lə[(sti)Ft(œlo)Ft]PWd]PPh, a parse which violates *CLASH (no stressed syllables are adjacent) as well as FOOTBINARITY (feet are binary, σσ or µµ). Many speakers of QF do permit parses of this type, although the well-formedness of such parses Prosodic Structure in Child French: Evidence for the Foot CatJL 5, 2006 115 5. Languages which appear to have both types of feet have been discussed in the literature; see Gordon (2002) on several languages which have both word-initial and word-final stress, e.g. Armenian and Udihe (he classifies Canadian French in this way as well). See also McCarthy and Prince (1990) on broken plurals in Arabic which follow an iambic pattern whereas the stress system is trochaic, and van de Vijver (1998) on Tiriyó Carib where the stress pattern looks to be iambic except in bisyllabic words where it is trochaic. Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 115 depends in large part on the quality of the vowel in the penultimate syllable.) Turning to (6b), both alignment constraints can readily be satisfied when the PPh contains a single PWd of three or more syllables. Finally, the structure in (6c) reveals one of the reasons why the argument for the left alignment constraint is the PWd rather than the PPh: left-edge functional material never appears with high tone; it is the first lexical syllable that hosts the initial accent, not the functional material that precedes it (Jun and Fougeron 2000). This is consistent with the view that function words are organised in French as free clitics (4). As mentioned in note 4, if function words were instead within the lowest PWd, we would expect them to be candidates for left-edge stress. The formulation of the ALIGNTROCH constraint, in combination with the structure in (4), enables this stress pattern to be captured while also reflecting the syntactic constituency. The second reason why the argument for left alignment is the PWd stems from the results of a study conducted by Post (2003). Post set out to examine the conditions governing variability of stress marking in European French. Using the location of pitch accent as a diagnostic of phrasing, she found that there is variation in what may comprise a PPh. A PPh is minimally composed of one lexical word with its associated function words, but syntactically-dependent units consisting of more than one PWd, such as noun-adjective pairs, may also be parsed as one PPh. Examples of this optionality in phrasing are shown in (7). While the same sequence may be prosodified in a number of ways, each PPh is subject to the constraints provided in (5). Thus, the option in (7c) where vergers is parsed as trochaic rather than iambic is illicit as it incurs a fatal violation of highly-ranked ALIGNIAMB. (7) a. [ les [vergers]PWd [œverts]PWd ]PPh ‘the green orchards’ the orchards green b. [les [verœgers]PWd]PPh [ [œverts]PWd ]PPh c. *[les [œvergers]PWd ]PPh [ [œverts]PWd ]PPh 116 CatJL 5, 2006 Heather Goad; Meaghen Buckley a. PPh PWd Ft σσ σ lə sti œlo le stylo ‘the pen’ b. PPh PWd Ft Ft σσσσ te le fɔ œne téléphoner ‘to telephone’ c. PPh PWd PWd Ft Ft σσσσσ lə sti lo vjɔ œlε le stylo violet ‘the purple pen’ (6) Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 116 In short, while the domain of primary stress assignment in French is the PPh, prosodic structure internal to the PPh, in particular PWd-parsing, is also relevant. This view is consistent with the prosodic hierarchy provided in (2). 2.4. The challenge for the learner The preceding sections have revealed that the facts of French stress are complex. Indeed, the starting point of the debate on how to best capture the system is not what kind of feet French has, but, more fundamentally, whether the language has a foot projection at all. While we have offered a preliminary analysis which makes reference to the various levels of the prosodic hierarchy – including the foot – the fact remains that the system is complex on several dimensions: the presence of a tonal contour, the specifics of which may suggest to the child that the language is not a stress language; the presence of different types of feet at the left and right edges of higher prosodic domains, with different phonetic cues to foot-headedness and different domains for alignment; the observation that the trochaic foot can appear optionally; the alignment of the rightmost foot with the edge of the PPh rather than the PWd; and finally, the existence of rampant violations of word minimality. In spite of the complexity of the system, we will argue below that Clara’s grammar does provide early knowledge of the foot in French. Indeed, we will show that at the earliest stages in development, her grammar respects constraints on prosodic well-formedness which are not necessarily respected in the target grammar. 3. Earlier evidence for the foot in child QF In order to provide a context for an investigation of Clara’s prosodic development, we begin by reviewing some of the previous literature which discusses the status of the foot in the acquisition of French. We will demonstrate that, to date, the evidence available for or against the (binary) foot is not conclusive. We begin with Archibald (1996) and Archibald and Carson (2000). The general research question which these papers seek to address is whether children show an early trochaic bias, as initially proposed by Allen and Hawkins (1978), or whether the evidence for trochees over iambs in the languages studied thus far is an effect of the ambient input. Archibald (1996) observes that final lengthening is robustly present in adult QF outputs. Given the observation mentioned in section 2.3 that the principal cue to stress in iambic languages is length rather than pitch, he concludes that lengthening in the ambient language is a strong enough signal for learners to determine that QF is iambic, even if there may be an initial trochaic bias. The latter position is supported by his examination of the outputs from one child aged 2;4 where final lengthening was widely attested. The link that Archibald (1996) makes between the presence of lengthening and the building of iambs is not definitive, however, as the lengthening that is observed may simply be phrase-final lengthening (see Allen 1983 and Hallé, Boysson-Bardies and Vihman 1991 on final lengthening in child French). Thus, the findings from this paper are not conclusive as regards the status of the foot in French. Indeed, in Prosodic Structure in Child French: Evidence for the Foot CatJL 5, 2006 117 Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 117 Phonetically prominent syllables are underlined as in (10) above. (FAITH collapses both MAX-IO and DEP-IO; the latter disfavours epenthesis: every segment in the output has a correspondent in the input.) (12) Expected outputs with no foot projection The tableau in (12a) shows that one-syllable targets should not be augmented to two syllables (12a-ii), as an additional violation of EXHAUST(PWd) will ensue. Augmenting the form through vowel lengthening, (12a-iii), ties with the optimal output on EXHAUST(PWd), but it needlessly incurs a violation of FAITH (as well as violating a constraint against long vowels). The tableau in (12b) reveals that EXHAUST(PWd) will also be responsible for ensuring that two-syllable targets are truncated to one syllable. In short, with no foot projection, all outputs at early stages in development are predicted to be truncated to the final prominent syllable, regardless of the number of syllables in the input. Had the optimal outputs in (12) been footed, they would have violated FTBIN. Accordingly, under the option where the French learner’s grammar does have a foot projection, different outputs are predicted to be optimal, as can be seen in the tableaux in (13). To satisfy high-ranking FTBIN, monosyllabic outputs like Guy must be augmented to (C)VCV (13a-ii) or to CVV (13a-iii) (depending on the ranking of other constraints). Similarly, both syllables in bisyllabic forms like aussi must be preserved (13b-i) or, if there is truncation due to other markedness constraints, the surviving vowel must be lengthened. We will demonstrate below that Clara’s outputs are consistent with her grammar having a foot projection, as productions along the lines of the optimal forms in (13) are robustly attested, until Stage 4. 124 CatJL 5, 2006 Heather Goad; Meaghen Buckley EXH ALIGN STRESS ANCHOR a. /i/ Guy (name) FTBIN (PWd) FOOT FAITH RIGHT FAITH ☞i. [i]PWd * ii. [əi]PWd **! * iii. [i]PWd **! EXH ALIGN STRESS ANCHOR b. /osi/ aussi ‘also’ FTBIN (PWd) FOOT FAITH RIGHT FAITH i. [osi]PWd **! ii. [o]PWd **!*** ☞iii. [si]PWd ** Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 124 (13) Expected outputs with foot projection 5.1.1. The role of foot binarity Tables 2 and 3 summarise the patterns for the lower bound on Clara’s words for the five stages under consideration. Importantly, as [ʁ]-final targets virtually always trigger compensatory lengthening in Clara’s outputs (Rose 2000, 2002), all stimuli of this shape have been excluded from the counts in Tables 2 and 3, so as not to inadvertently inflate the number of forms which have been augmented for foot structure reasons. The final column in Table 2 shows that the overall pattern, until Stage 4, is that Clara’s outputs for one-syllable lexical targets are infrequently subminimal, consistent with undominated FTBIN. Prosodic Structure in Child French: Evidence for the Foot CatJL 5, 2006 125 EXH ALIGN STRESS ANCHOR a. /œi/ Guy (name) FTBIN (PWd) FOOT FAITH RIGHT FAITH i. [(œi)Ft]PWd *! ☞ii. [(əœi)Ft]PWd * ☞iii. [(œi)Ft]PWd * EXH ALIGN STRESS ANCHOR b. /oœsi/ aussi ‘also’ FTBIN (PWd) FOOT FAITH RIGHT FAITH ☞i. [(oœsi)Ft]PWd ☞ii. [(œsi)Ft]PWd ** iii. [(œo)Ft]PWd *! * * ** iv. [(œsi)Ft]PWd *! * Table 2. One-syll lexical targets, (A) not requiring and (B) requiring fnc (σis lexical; σis functional). (A) Outputs for 1σσ-lex targets (B) Outputs for 1σσ-fnc + 1σσ-lex targets Total No(σσµµ)( σσµµµµ)( σσσσ)N o(σσµµ)σσ(σσµµ)( σσµµµµ)σσ(σσµµµµ)( σσσσ)σσ(σσσσ) submin Stage attemp (submin) attemp (submin) (submin) outputs 1 0 -- -- -- 0 -- -- -- -- -- -- 0 2 15 2 11 2 1 0 0 0 0 1 0 2 (13%) (74%) (13%) (100%) (13%) 3 13 5 4 4 11 1 1 2 5 2 0 7 (38%) (31%) (31%) (18%) (64%) (18%) (29%) 4 32 19 11 2 29 9 13 2 2 0 3 41 (60%) (34%) (6%) (76%) (14%) (10%) (67%) 5 82 58 21 3 107 38 43 7 12 5 2 139 (71%) (25%) (4%) (76%) (18%) (6%) (74%) Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 125 The columns under (A) show the profile for those one-syllable lexical targets which do not require a proclitic. Until Stage 4, these forms are overwhelmingly augmented, primarily through vowel lengthening, σ→(σµµ)Ft, but also through the addition of a pretonic syllable, σ→(σσ)Ft (note that, for each of Stages 2-3 and 5, there is one example of augmentation to three syllables). These outputs are parallel to candidates (iii) and (ii) respectively in tableau (13a). Representative examples of all output patterns are in (14). (Some examples which follow the σ→(σσ)Ft pattern may surface as such because of constraints on segmental licensing or syllable complexity, a question which we leave to future research.) (14) One-syllable lexical targets which do not require a proclitic (1σ-lex): Pattern: Stage: Target: Clara’s output: Orthography: Gloss: σ→(σµ)Ft 1-- 2[nɔ˜][nɔ] non ‘no’ [i][i] Guy (name) 3[pɑ][pɑ] pas ‘not’ [i][i] Guy (name) 4[ki][ki] qui ‘who’ [nɔ˜][nɔ] non ‘no’ 5[bʁ˜][b] brun ‘brown’ [mε][mε] mets ‘put’ (imper) σ→(σµµ)Ft 1-- 2[i][i] Guy (name) [nɔ˜][nɔυ] non ‘no’ 3[wi][wi] oui ‘yes’ [ʁoz][lυs] rose ‘pink’ 4[pɑ][p] pas ‘not’ [wi][wəi] oui ‘yes’ 5[sɑ][sɑ] ça ‘that’ [pɑ][pa] pas ‘not’ σ→(σσ)Ft 1-- 2[wi][əœɥi] oui ‘yes’ [i][iiœi] Guy (name) 3[wi][iœji] oui ‘yes’ [i][əœi] Guy (name) 4[nɔ˜][˘œn] non ‘no’ [lɑ][eœjε] là ‘there’ 5[wi][əœi] oui ‘yes’ [blø][əbyœlø] bleu ‘blue’ The columns under (B) in Table 2 provide the profile for one-syllable lexical targets which obligatorily require a proclitic in the target grammar (i.e., 1σ-fnc + 1σ126 CatJL 5, 2006 Heather Goad; Meaghen Buckley Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 126 lex).13 Recall that such forms are subminimal in adult French, as left-edge function morphemes are organised as free clitics linking directly to the PPh, e.g. [lə [(œʃɑ)Ft]PWd]PPh le chat ‘the cat’ (see section 2.2). Parallel to the generalisations that emerge from the columns under (A), at Stage 4, there is a large increase in the percentage of subminimal forms attested. At Stage 3, strings of this shape are typically augmented through vowel lengthening, regardless of whether or not the clitic is produced by Clara. Examples of all output patterns for 1σ-fnc + 1σ-lex targets are provided in (15). (15) One-syllable lexical targets which obligatorily require a proclitic (1σ-fnc + 1σ-lex): Pattern: Stage: Target: Clara’s output: Orthography: Gloss: σσ→(σµ)1 -- 2-- 3[ləœne][«e] (le) nez ‘(the) nose’ 4[ləœpε˜][pε] (le) pain ‘(the) bread’ [ləœʃɑ][ç] (le) chat ‘(the) cat’ 5[ləœpε˜][pε] (le) pain ‘(the) bread’ [ləœʃjε˜][çj] (le) chien ‘(the) dog’ σσ→σ(σµ)1 -- 2-- 3[ə˜œliv][aœji] un livre ‘a book’ 4[dəœdɑ˜][dɵœdɒ] de dents ‘of teeth’ [ə˜œʃɑ][ɵ˘œtʃa] un chat ‘a cat’ 5[ə˜œpo][əœpo] un pot ‘a pot’ [ləœne][ləœni] le nez ‘the nose’ σσ→(σµµ) 1 -- 2-- 3[ləœne][nei] (le) nez ‘(the) nose’ 4[laœpɔm][bɔ] (la) pomme ‘(the) apple’ [leœdɑ˜][dɔ] (les) dents ‘(the) teeth’ 5[leœdwa][dw] (les) doigts ‘(the) fingers’ [ləœne][ne] (le) nez ‘(the) nose’ σσ→σ(σµµ) 1 -- 2-- 3[ləœʃɑ][ləœtɑ] le chat ‘the cat’ [əœljɔ˜][əœlɑ] un lion ‘a lion’ Prosodic Structure in Child French: Evidence for the Foot CatJL 5, 2006 127 13. As is evident from the text, the abbreviation 1σ-fnc + 1σ-lex refers to targets where the entire first syllable is functional. Targets like [lɔto] l’auto ‘the car’, where lex is vowel-initial, have not been included in this category, as the determiner in this case is organised internal to the PWd of the stem, to satisfy ONSET: [[lɔœto]PWd]PPh. No forms of the latter type are attempted until Stage 3, and they are not included in any of the counts in Table 2. Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 127 Pattern: Stage: Target: Clara’s output: Orthography: Gloss: 4[ləœliv][jəœji] le livre ‘the book’ [ləœliv][jœji] le livre ‘the book’ 5[εlœsot][œçot] elle saute ‘she jumps’ [dzyœy][dyœy] du jus ‘some juice’ σσ→(σσ) 1 -- 2[laœtait][taiœdɑ] (la) tête ‘(the) head’ 3[ləœbʁɥi][βuœji] (le) bruit ‘(the) noise’ [ləœʃjε˜][səœlε] (le) chien ‘(the) dog’ 4-- 5[ləœljɔ˜][œlεjɔ] (le) lion ‘(the) lion’ [ləœne][œnehe!] (le) nez ‘(the) nose’ σσ→σ(σσ) 1 -- 2-- 3-- 4[ə˜œliv][əjaœli] un livre ‘a book’ [ləœpε«][japɔœ«e] le peigne ‘the comb’ 5[ə˜œljɔ˜][y˘œlυjɵ˘] un lion ‘a lion’ [ynœpυl][ɑœpυlə] une poule ‘a hen’ Table 3 provides the profile observed in Clara’s grammar for two-syllable lexical targets which do not require a proclitic. (Note that this table includes eight outputs at Stages 3-5 labelled (σσ) which actually involve augmentation to three syllables, and one output at Stage 4 which involves augmentation to four syllables.) As can be seen, two-syllable lexical targets are rarely truncated to one syllable, in contrast to the findings of Archibald and Carson (2000). (There are a number of [ʁ]-final forms like [ɑ˜kɔʁ] encore ‘again’ and [də.ɔʁ] dehors ‘outside’ which surface as monosyllabic (typically as CVV), perhaps due to other marked128 CatJL 5, 2006 Heather Goad; Meaghen Buckley Table 3. Two-syll lexical targets. No(σσµµ)( σσµµµµ)( σσσσ) Stage attemp (submin) 17106 (14%) (86%) 2211 020 (5%) (95%) 3412 237 (5%) (5%) (90%) 4250 124 (4%) (96%) 5762 074 (3%) (97%) Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 128 ness constraints, as suggested by the higher proportion of such outputs during Stages 1 and 2; recall, however, that [ʁ]-final targets have not been included in the counts since they trigger compensatory lengthening for reasons independent of foot structure.) In short, outputs which respect FTBIN, along the lines of the optimal candidate in (i) in tableau (13b), are favoured by Clara’s grammar. Examples of all output patterns are provided in (16). (16) Two-syllable lexical targets which do not require a proclitic (2σ-lex) Pattern: Stage: Target: Clara’s output: Orthography: Gloss: σσ→(σµ)Ft 1[maœmɑ˜][m] maman ‘mum’ 2[paʁœtsi][tit.ʃ] parti ‘departed’ 3[maœmɑ˜][mm] maman ‘mum’ [maœmɑ˜][m] maman ‘mum’ 4-- 5[miœmi][mi] Mimi (name) [miœmi][me] Mimi (name) σσ→(σµµ)Ft 1-- 2-- 3[paœpa][p] papa ‘dad’ [maœmɑ˜][mə] maman ‘mum’ 4[kaœka][ka] caca ‘poop’ 5-- σσ→(σσ)Ft 1[katœlin][tœti] Kathleen (name) [maœmɑ˜][məœm] maman ‘mum’ 2[dəœdɑ˜][dəœdɑ] dedans ‘inside’ [dəœbu][bɑœbu] debout ‘standing’ 3[ʃεʁœʃe][ʃiœtʃi] chercher ‘to look for’ [paʁœtu][taœtu] partout ‘everywhere’ 4[iœsi][iœsi] ici ‘here’ [kaœju][taœj] Caillou (name) 5[aœsi][œsi] assis ‘seated’ [mεʁœsi][mεœçi] merci ‘thank you’ 5.1.2. Augmentation as phrase-final lengthening? We have observed that Clara’s outputs for one-syllable lexical targets display vowel lengthening to a great extent. That is, the native-speaker transcribers heard these forms as involving significantly greater length than how they would be produced by adult speakers and, thus, they transcribed these forms as CVV. We propose that this difference is formally reflected through the addition of a mora to Clara’s early outputs, yielding bimoraic forms for targets of this shape. An important question that must be addressed, however, is whether this lengthening involves augmentation to satisfy FTBIN, as suggested in section 5.1.1, or whether it is instead phrase-final lengthening. In apparent support of the latter position, at the stages when lengthening Prosodic Structure in Child French: Evidence for the Foot CatJL 5, 2006 129 Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 129 is particularly robust (Stages 1-3), most phrases contain only one PWd; indeed, recall from Table 1 that phrases of more than one PWd are not attempted until Stage 3. If the lengthening exhibited in Table 2 were phrase-final lengthening, it would provide us with no evidence one way or the other about the status of the foot in Clara’s grammar. A comparison of the proportion of lengthening observed in outputs for onesyllable lexical targets, (σµ) versus (σµµ), with the proportion of lengthening observed on the final syllable in outputs for two-syllable lexical targets, (σσµ) versus (σσµµ), reveals that lengthening is indeed motivated by foot well-formedness rather than being a phrase-final effect. If the latter were the source of lengthening, then at early stages in Clara’s development, lengthening should be as widely attested for final syllables in two-syllable lexical targets as for one-syllable lexical targets. A glance at Table 4, however, reveals that this is not the case. At Stages 1-3, the proportion of lengthened to non-lengthened outputs for onesyllable lexical targets is the opposite of that observed for two-syllable lexical targets. This is as expected if, for one-syllable lexical targets, lengthening is motivated by FTBIN. For two-syllable lexical targets, FTBIN is already satisfied by a two-syllable output; thus, there is no motivation for lengthening involving foot well-formedness and widespread lengthening is not expected to occur. At Stages 45, when the number of subminimal forms in Clara’s outputs greatly increases, the proportion of lengthened to non-lengthened outputs for oneand two-syllable lexical targets is expected to be the same; Table 4 shows that this is indeed the case. 5.1.3. The emergence of word-final consonants As mentioned above, Table 2 shows a sharp increase in the number of subminimal outputs for one-syllable lexical targets in the transition from Stage 3 to Stage 4, from an average of 21% at earlier stages to 71% at Stages 4 and 5. Why would this be the case? If Clara’s patterns of behaviour at earlier stages support the postulation of the foot, it is highly unlikely that this projection is eliminated from her representations at later stages. Not only is this inconsistent with continuity, it is in violation of HEADEDNESS (3b), in this case, the requirement that every PWd contain at least one foot; it is also not consistent with the position we have taken that adult French does indeed support a foot projection (section 2.3). By Stage 4, when lengthening is less robustly attested in Clara’s outputs and her grammar is beginning to tolerate subminimal feet, it would appear that satis130 CatJL 5, 2006 Heather Goad; Meaghen Buckley Table 4. Final lengthening in oneand two-syll lexical targets. 1σσ-lex targets 2σσ-lex targets (σσµµ)( σσµµµµ)( σσσσµµ)( σσσσµµµµ) Stage (submin) 1-3 7 15 47 16 (32%) (68%) (75%) (25%) 4-5 77 32 67 31 (71%) (29%) (68%) (32%) Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 130 faction of the faithfulness constraints that were earlier violated through augmentation is starting to take priority, as in the target grammar. One potentially confounding factor that must be addressed, however, is whether any relationship holds between the sharp increase in subminimal outputs at Stage 4 and Clara’s acquisition of word-final consonants. If final consonants were analysed by Clara as moraic codas, vowel lengthening in the case of consonant-final targets would have nothing to do with foot well-formedness, as we have argued, as it would instead be a compensatory lengthening effect; that is, when final consonants are deleted, the preceding vowel would lengthen to license the stranded mora (see Ota 1999 for evidence of compensatory lengthening in child Japanese). When final consonants emerge in Clara’s outputs at later stages, the motivation for lengthening would then no longer hold. In the following lines, we will show that this explanation cannot hold. First, word-final consonants are not robustly attested until Stage 5 (see Table 1), whereas Table 2 shows that the dramatic increase in subminimal outputs begins two months earlier, at the onset of Stage 4; there is also no change in the number of subminimal outputs from Stage 4 to Stage 5. Second, the proportion of vowel-final targets which undergo lengthening is approximately the same as the proportion of consonant-final targets which undergo lengthening at all five stages, thus, even after word-final consonants emerge in outputs.14 Finally, there are a number of arguments which support the proposal that word-final consonants (other than [ʁ]; see note 13) in Clara’s grammar are syllabified as onsets, not as codas, which makes them ineligible as triggers of compensatory lengthening (see Rose 2000, 2002, Goad 2002). In sum, there is no relationship between the acquisition of word-final consonants and the decrease in subminimal words in Clara’s outputs. In this section, we have argued that there is a lower bound of one binary iamb on Clara’s outputs at early stages in development and that this supports the presence of a foot projection in her grammar. We turn now to examine upper bound effects on her early productions. 5.2. Upper bound effects If Clara’s grammar at early stages is similar to the grammars of children learning trochaic languages, FTBIN, in combination with undominated ALIGNFOOT and EXHAUST(PWd), will conspire to yield an upper bound on the length of her PWds. That is, outputs should be no longer than one foot, and longer forms should be subject to truncation, as was seen in (11) for English. The tableau in (17) shows how this result is expected for three-syllable lexical words in French. (We will not consider violations of ANCHORRIGHT, so this constraint has been removed from the tableau.) Prosodic Structure in Child French: Evidence for the Foot CatJL 5, 2006 131 14. This, however, excludes [ʁ]. As mentioned earlier, word-final [ʁ], almost without exception, triggers compensatory lengthening at Stages 1-5. It is not mastered until 2;03,15, five months after the end of Stage 5, at the same point that word-internal codas are acquired (see Rose 2000, 2002 for details). Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 131 (17) Truncation in child French Before we discuss the candidates in (17), some comments on the input are necessary. Recall from section 2.1 that three-syllable targets like difficile optionally have an initial high tone (which we have analysed as secondary stress). Aside from the observation that secondary stress is optional, we have not included this information in the input in (17) because, if Clara were sensitive to the initial H, we would have expected two patterns of behaviour in her outputs which we do not find. First, as we will see below, three-syllable lexical words are commonly truncated to two syllables in Clara’s outputs through Stage 5 (see also Parsons 2005). If the initial H were particularly salient, we might have expected this syllable to be retained in truncation, yielding forms like [di œ si] for difficile. This type of pattern is observed for learners of other languages; for example, Lleó (2002) reports that Spanish-speaking children preserve the initial secondarily-stressed syllable in four-syllable targets (e.g., [ à maɾi œ posa] mariposa ‘butterfly’ →[pa œ bɔta] for José at 1;9.2). Clara’s truncated outputs, however, typically show preservation of the syllable adjacent to the final stressed syllable, yielding forms like [fi œ si]. Of all truncations of lexical material to bisyllabic for the five stages under consideration where a decision can be made as to which syllable is retained, the penult survives 71% of the time (cf. Parsons 2005 on Clara; see also Paradis 2001 on other French-speaking children). Second, to avoid unfooted syllables in early outputs, we might have expected Clara to produce two-foot parses for words like difficile relatively often, yielding forms like [ à difi œ si] (or [ à difi œ si]). Outputs of this shape are unattested until Stage 4 and are relatively infrequent at this point as well (see Table 5 below). With these considerations in mind, let us return to examine (17). Candidate (a) fatally violates EXHAUST(PWd) but it does not incur a violation of STRESSFAITH if the input is /difi œ sil/. The introduction of a second foot in candidate (b) results in a fatal violation of undominated FTBIN as well as ALIGNFOOT. Of the two remaining candidates, (d) wins over (c) if another highly-ranked constraint is introduced, I-CONTIGUITY (the portion of the input standing in correspondence forms a contiguous string); as mentioned in note 6, I-CONTIG disfavours outputs that violate the integrity of a morpheme through morpheme-internal deletion or epenthesis, as in candidate (c). Table 5 below shows that Clara’s early productions are largely compatible with a grammar like (17) where the constraint ranking yields an upper bound of onefoot on the shape of outputs. As the number of three-syllable lexical targets which 132 CatJL 5, 2006 Heather Goad; Meaghen Buckley EXH ALIGN STRESS /difiœsil/ difficile ‘difficult’ FTBIN (PWd) FOOT FAITH CONTIG MAX a. [di(fiœsi)Ft]PWd *! * b. [(Ãdi)Ft(fiœsi)Ft]PWd *! * * c. [(diœsi)Ft]PWd *! *** ☞d. [(fiœsi)Ft]PWd *** Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 132 do not require a proclitic is quite small in Clara’s outputs, the table includes three types of simuli: 3σ-lex targets (e.g., [atɑ˜ œ sjɔ˜] attention ‘watch out’); 3σvowel-initial lex targets which must be preceded by a single consonant fnc (e.g., [leky œ ʁj] l’écureuil ‘the squirrel’), as recall from note 12 that, in these cases, fnc is organised internal to the PWd to satisfy ONSET; and 4σtargets shaped 1σ-fnc + 3σ-lex where the proclitic is not produced (e.g., [pɑ˜ta œ lɔ˜] for target [lə pɑ˜ta œ lɔ˜] (le) pantalon ‘(the) trousers’). Although the numbers are still low, it is important to point out that, across stages, all truncations are to two syllables, supporting the view that a binary foot is what constrains the upper length of Clara’s PWds (cf. Archibald and Carson 2000 where truncation was typically to one syllable (section 3)). The speech samples available suggest that Clara did not attempt any three-syllable words until Stage 2; we take the absence of words of this shape at Stage 1 to be indicative of «selection and avoidance», a commonly-observed phenomenon where children avoid utterances which contain structures not compatible with their grammar (Schwartz and Leonard 1982). While there are only four three-syllable examples attempted at Stages 2 and 3, they all display truncation. The tableau in (17), where EXHAUST(PWd) is undominated, is thus consistent with Clara’s behaviour at Stages 1-3. At Stage 4, the target structure with a PWd-initial unfooted syllable appears, but it is not robustly attested until after Stage 5. At Stages 4 and 5, truncation is still the most common pattern, but the number of cases of truncation is roughly equal to those where all three syllables surface, either with the initial syllable as unfooted or with this syllable forming its own foot. This suggests that the ranking between EXHAUST(PWd) and FTBIN/ALIGNFOOT on the one hand and MAX-IO on the other is not stable during this six-month window. Representative examples of the unfooted σand truncation parses are provided in (18); two-foot parses are discussed below. (Note that there are seven examples of stress shift to the penult included in the numbers of unfooted σand truncation parses at Stages 4 and 5 in Table 5, a pattern which is attested in the adult grammar as well (e.g., Ouellet and Thibault 1996). See, for example, Clara’s output for écureuil in (18).) Prosodic Structure in Child French: Evidence for the Foot CatJL 5, 2006 133 Table 5. Three-syll lexical targets. NoUnfooted σσTruncation 2-ft Stage attemp σσ((σσœσσ))(σσœσσ) parses 1 0 -- -- -- 21 0 10 (100%) 33 0 30 (100%) 48 1 43 (12%) (50%) (38%) 5 25 6 13 6 (24%) (52%) (24%) Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 133 Demuth, Katherine; Fee, E. Jane (1995). «Minimal words in early phonological development». Ms., Brown University and Dalhousie University. Demuth, Katherine; Johnson, Mark (2003). «Truncation to subminimal words in early French». Canadian Journal of Linguistics 48: 211-241. Deville, Gérard (1891). «Notes sur le développement du langage II». Revue de linguistique et de philologie comparée 24: 10-42, 128-143, 242-257, 300-320. Echols, Catharine; Newport, Elissa (1992). «The role of stress and position in determining first words». Language Acquisition 2: 189-220. Fikkert, Paula (1994). On the Acquisition of Prosodic Structure. The Hague: Holland Academic Graphics. Fouché, Pierre (1934). «Évolution phonétique du français du XVI siècle à nos jours». Le français moderne 2: 217-236. Garde, Paul (1968). L’accent. Paris: Presses Universitaires de France. Gendron, Jean-Denis (1966). Tendances phonétiques du français parlé au Canada. Québec: Les Presses de l’Université Laval. Gerken, LouAnn (1994). «A metrical template account of children’s weak syllable omissions from multisyllabic words». Journal of Child Language 21: 565-584. Gnanadesikan, Amalia (2004). «Markedness and Faithfulness constraints in child phonology». In: Kager, René; Pater, Joe; Zonneveld, Wim (eds.), Constraints in Phonological Acquisition. Cambridge: Cambridge University Press, pp. 73-108. Available (1995) as ROA-67 from http://roa.rutgers.edu. Goad, Heather (1996). «Codas, word-minimality, and empty-headed syllables». In: Clark, Eve (ed.), Child Language Research Forum 28: 113-122. Stanford: Center for Language and Information. — (2002). «Markedness in right-edge syllabification: Parallels across populations». Canadian Journal of Linguistics 47: 151-186. Gordon, Matthew (2002). «A factorial typology of quantity-insensitive stress». Natural Language & Linguistic Theory 20: 491-552. Hallé, Pierre; Boysson-Bardies, Bénédicte de; Vihman, Marilyn (1991). «Beginnings of prosodic organization: Intonation and duration patterns of disyllables produced by Japanese and French infants». Language and Speech 34: 299-318. Hayes, Bruce (1995). Metrical Stress Theory: Principles and Case Studies. Chicago: University of Chicago Press. Ingram, David (1978). «The role of the syllable in phonological development». In: Bell, Alan; Hooper, Joan B. (eds.), Syllables and Segments. Amsterdam: North-Holland, pp. 143-155. Jakobson, Roman (1941). Child Language, Aphasia and Phonological Universals. The Hague: Mouton. Translated 1968 by Allan R. Keiler. Jun, Sun-Ah; Fougeron, Cécile (2000). «A phonological model of French intonation». In: Botinis, Antonis (ed.), Intonation: Analysis, Modelling and Technology. Dordrecht: Kluwer, pp. 209-242. Kehoe, Margaret; Stoel-Gammon, Carol (1997). «The acquisition of prosodic structure: An investigation of current accounts of children’s prosodic development». Language 73: 113-144. Lleó, Conxita (1997). «Filler syllables, proto-articles and early prosodic constraints in Spanish and German». In: Sorace, Antonella; Heycock, Caroline; Shillcock, Richard (eds.), Language Acquisition: Knowledge, Representation and Processing. 140 CatJL 5, 2006 Heather Goad; Meaghen Buckley Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 140 Proceedings of the GALA '97 Conference on Language Acquisition. Human Communication Research Centre, University of Edinburgh, pp. 251-256. — (2002). «The role of markedness in the acquisition of complex prosodic structures in German-Spanish bilinguals». International Journal of Bilingualism 6: 291-313. — (2003). «Child prosody and filler syllables: Looking into Spanish through the optimal window of acquisition». In: Montrul, Silvina; Ordóñez, Francisco (eds.), Linguistic Theory and Language Development in Hispanic Languages. Papers from the 5th Hispanic Linguistics Symposium and the 4th Conference on the Acquisition of Spanish and Portuguese. Somerville, Mass: Cascadilla Press, pp. 229-253. McCarthy, John; Prince, Alan (1986). «Prosodic Morphology». Ms., University of Massachusetts Amherst and Brandeis University. — (1990). «Foot and word in prosodic morphology: The Arabic broken plural». Natural Language & Linguistic Theory 8: 209-282. Mertens, Piet (1987). L’intonation du français. De la description linguistique à la reconnaissance automatique. Katholieke Universiteit Leuven, doctoral dissertation. Nespor, Marina; Vogel, Irene (1986). Prosodic Phonology. Dordrecht: Foris. Ota, Mits (1999). Phonological Theory and the Acquisition of Prosodic Structure: Evidence from Child Japanese. Georgetown University, doctoral dissertation. Ouellet, Marise; Thibault, Linda (1996). «L’allongement prétonique: un phénomène opportuniste?». In: Dolbec, Jean; Ouellet, Marise (eds.), Recherches en phonétique et en phonologie au Québec. Québec: CIRAL, pp. 47-61. Paradis, Claude; Deshaies, Denise (1990). «Rules of stress assignment in Québec French: Evidence from perceptual data». Language Variation and Change 2: 135-154. Paradis, Johanne (2001). «Do bilingual two-year-olds have separate phonological systems?» International Journal of Bilingualism 5: 19-38. Paradis, Johanne; Petitclerc, Sophie; Genesee, Fred (1997). «Word truncation in Frenchspeaking two year olds». In: Hughes, Elizabeth; Hughes, Mary; Greenhill, Annabel (eds.), Proceedings of the 21st Annual Boston University Conference on Language Development. Somerville, Mass: Cascadilla Press, pp. 441-452. Parsons, Jennifer (2005). «Acquisition of the Prosodic Word». Ms., Memorial University of Newfoundland. Pater, Joe (1997). «Minimal violation and phonological development». Language Acquisition 6: 201-253. Pater, Joe; Werle, Adam (2003). «Direction of assimilation in child consonant harmony». Canadian Journal of Linguistics 48: 385-408. Peters, Ann; Menn, Lise (1993). «False starts and filler syllables: ways to learn grammatical morphemes». Language 69: 742-777. Post, Brechtje (2003). «French phrasing and accentuation in different speaking styles». In: Grabe, Esther; Wright, David G.S. (eds.), Oxford University Working Papers in Linguistics, Philology and Phonetics 8: 69-83. Rose, Yvan (2000). Headedness and Prosodic Licensing in the L1 Acquisition of Phonology. McGill University, doctoral dissertation. — (2002). «Place specification and segmental distribution in the acquisition of wordfinal consonant syllabification». Canadian Journal of Linguistics 48: 409-435. Rose, Yvan; dos Santos, Christophe (2004). «Prosodic domains and the acquisition of French phonology». Paper presented at the Second Lisbon Meeting on Language Acquisition, June. Prosodic Structure in Child French: Evidence for the Foot CatJL 5, 2006 141 Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 141 Schwartz, Richard G.; Leonard, Larry B. (1982). «Do children pick and choose? An examination of phonological selection and avoidance». Journal of Child Language 9: 319-336. Scullen, Mary Ellen (1997). French Prosodic Morphology: A Unified Account. Bloomington: Indiana University Linguistics Club. Selkirk, Elisabeth O. (1984). Phonology and Syntax: The Relation between Sound and Structure. Cambridge, Mass: MIT Press. — (1986). «On derived domains in sentence phonology». Phonology 3: 371-405. — (1996). «The prosodic structure of function words». In: Morgan, James; Demuth, Katherine (eds.), Signal to Syntax: Bootstrapping from Speech to Grammar in Early Acquisition. Mahwah, NJ: Lawrence Erlbaum, pp. 187-213. Smolensky, Paul (1996). «The comprehension/production dilemma in child language». Linguistic Inquiry 27: 720-731. Stampe, David (1969). «The acquisition of phonetic representation». Chicago Linguistic Society 5: 433-444. Tremblay, Annie (in press). «Prosodic constraints on the production of grammatical morphemes in early French: The case of determiners.» In: Deen, Kamil Ud; Nomura, Jun; Schulz, Barbara; Schwartz, Bonnie (eds.), Proceedings of the Inaugural Conference on Generative Approaches to Language Acquisition-North America. University of Connecticut Occasional Papers in Linguistics 4. Veneziano, Edy; Sinclair, Hermine (2000). «The changing status of ‘filler syllables’ on the way to grammatical morphemes». Journal of Child Language 27: 461-500. Verluyten, S. Paul (1982). Investigation on French Prosodics and Metrics. Universiteit Antwerpen, doctoral dissertation. Vijver, Ruben van de (1998). The Iambic Issue: Iambs as a Result of Constraint Interaction. The Hague: Holland Academic Graphics. Walker, Douglas C. (1984). The Pronunciation of Canadian French. Ottawa: University of Ottawa Press. Wijnen, Frank; Krikhaar, Evelien; den Os, Els (1994). «The (non)realization of unstressed elements in children’s utterances: Evidence for a rhythmic constraint». Journal of Child Language 21: 59-83. 142 CatJL 5, 2006 Heather Goad; Meaghen Buckley Cat.Jour.Ling. 5 001-276 3/11/06 16:27 Página 142