«The acquisition of word-final clusters in French * Katherine Demuth Brown University Margaret Kehoe University of Hamburg Address for correspondence: ...»
Running Head: Word-Final Clusters
The acquisition of word-final clusters in French *
University of Hamburg
Address for correspondence:
Cognitive & Linguistic Sciences
Brown University, Box 1978
Providence, RI 02906
Tel: (401) 863-1053
Fax: (401) 863-2255
Word-Final Clusters 2
The structure of French syllables has long been controversial, particularly with respect to the status of word-final consonants. Some researchers suggest that word-final consonants are syllabified as codas, whereas others propose that these are onsets of empty-headed syllables.
This raises questions regarding the nature of syllabic representations in children’s developing French, a topic that has received little attention. This study examines 2-year-olds’ elicited productions of word-final obstruent-liquid (OL) clusters, and compares these with the acquisition of word-initial OL clusters. The acquisition of singleton word-final consonants is also discussed. Although word-final clusters are acquired later than both word-initial clusters and word-final singletons, the error patterns are the same, with earlier acquisition and preservation of obstruents. A few children exhibit final vowel epenthesis, raising the possibility that some French-speaking children may syllabify word-final consonants as onsets. The paper concludes with a discussion of the cross-linguistic implications of these findings, identifying several areas for further research.
1. Introduction The syllabic status of French word-final consonants has long been controversial, reflecting different theoretical positions in the field (e.g., Hammond, 1999; Harris, 1994; Kaye, 1990;
Kaye, Lowenstamm, & Vergnaud, 1990). Some researchers suggest that French can have a bipositional (bimoraic) rhyme, where a singleton coda consonant is permitted (e.g., Bouchard, 1980, Dell, 1995; Féry, 2003, Rialland, 1994, Scullen, 1997, Tranel, 1987). This structure is shown in (1a). Others agree that the French rhyme is maximally bipositional, but that the second rhyme slot can only be occupied by a sonorant consonant. Plénat’s (1987) justification for this proposal comes from the fact that sonorant consonants tend to be ‘fixed’, whereas stops can be ‘latent’, showing surface alternation. Still others have proposed that all Word-Final Clusters 3 French singleton word-final consonants are syllabified as onsets of empty-headed syllables regardless of sonority (Charette, 1991; Nikièma, 1999), as illustrated by the structure in (1b).
How, then, is a learner to determine the syllable structure of singleton word-final consonants in French?
(1) Representation of word-final singleton consonants for lexical item botte‘boot’
The linguistic facts of French provide support for both the ‘coda’ and ‘onset’ representational approaches. On the one hand, French vowel distribution supports a coda approach to French syllable structure. In many dialects of French, tense vowels typically appear in open syllables and lax (mid) vowels in closed syllables. This tendency, known as règles de position, suggests that at least some types of word-final consonants function as codas (e.g., Dell 1973; Selkirk 1972; Tranel 1995; also Féry, 2003 and others). On the other hand, stress assignment and final schwa epenthesis provide some support for an onset approach to word-final consonants in French. Stress assignment is predictable, the final full vowel receiving stress regardless of the presence of the word-final consonant. In other words, stress assignment in French is independent of syllable weight. In addition, French speakers often epenthesize a vowel (schwa or the mid front unrounded vowel ) following production of a word-final consonant. Both of these factors suggest that singleton word-final
how the syllable structure of French is learned, and if a better understanding of the acquisition process might also shed light on the syllable structure of French.
The issue becomes more complex when one considers the representational status of wordfinal consonant clusters. Consider the Sonority Sequencing Principle (SSP), which prohibits increases in sonority from the nucleus to the edges of the syllable (2), and the Sonority Hierarchy (Clements 1990) in (3), which treats liquids as being more sonorous than obstruents (fricatives and stops).
(2) Sonority Sequencing Principle (SSP) The level of sonority must not increase from the nucleus to the edges of the syllable.
(3) Sonority Hierarchy:
Thus, a word-final obstruent-liquid (OL) sequence that violates the SSP, such as in the word lettre / / ‘letter’, will be phonologically marked if it is contained within the coda of a syllable. However, some researchers do assume that word-final OL sequences appear as codas (Tranel, 1987). Support for this position comes from several dialects of French (e.g., Québec French, Haitian French, and even standard French), where word-final OL clusters may undergo liquid reduction (Casagrande, 1984; Côté, 2000; 2004; Charette, 1991;
Nikièma, 1999; van Oostendorp, 2000), a process that would not be expected if these were onsets. This is shown in (4) (examples are taken from Québec French (Côté, 2004: 9)).
One explanation for this type of reduction process is the SSP, which seeks to avoid a rise in sonority within the coda. Perhaps learners may also delete such liquids, leaving the less sonorant consonant as the coda. However, the fact that the liquid can delete in Québec
immediately following the vowel may benefit from cues present in the vocalic transitions that make it perceptually stronger (Côté, 2000; 2004). Liquids are often also devoiced (cf. Féry,
2003) and this may further enhance a tendency for reduction. In addition, deletion of the most sonorous element of a consonant cluster is a common pattern in cross-linguistic and developmental grammars (Barlow, 1997; Gnanadesikan, 2004), leading us to expect that the same might be true for French.
Another approach which also assumes a monosyllabic representation of OL clusters links the entire word-final cluster directly to the prosodic word node, creating an extrasyllabic consonant cluster, thereby avoiding violations of the SSP altogether (Rialland, 1994).
Although the perceptual issues would be the same here, the representational issues are quite different, and it is not clear what one might predict learners would do. On one hand, the lack of SSP violations would be a plus. On the other hand, perhaps ‘extraprosodic’ elements are more marked, and would be dispreferred by learners (though English word-final obstruent +/s/ clusters, which are often assumed to contain an extraprosodic appendix, are some of the earliest acquired (Kirk & Demuth, 2005)).
In contrast to these coda and other monosyllabic accounts, researchers working within the general framework of Government Phonology propose that word-final OL sequences in French are syllabified as complex onsets of empty-headed syllables (Charette, 1991; Dell, 1995; Féry, 2003; Nikièma, 1999). Evidence for this form of syllabification comes from the patterning of long vowels in Québec French. A vowel may be lengthened or diphthongized before an OL cluster (5a) and before a singleton word-final consonant (5b) but not before a liquid-obstruent (LO) cluster (5c). Charette (1991) interprets this pattern as revealing that neither OL clusters nor singleton word-final consonants are syllabified within a branching rhyme, but rather form onsets of empty-headed syllables. The LO cluster receives a different
onset of an empty-headed syllable. This is consistent with phonotactic restrictions in French:
a LO cluster is not an acceptable onset cluster word-initially in French, so it is unlikely to be one word-finally. Perhaps learners are sensitive to the distribution of long vowels and following consonants early in the learning process, and develop a syllabification algorithm similar to that proposed by Charette (1991).
(5) Vowel lengthening in Québec French (Charette, 1991: 124)
Additional evidence for an onset syllabification of OL clusters includes the fact that, as already noted for singletons, word-final OL clusters are often produced with a final epenthetic schwa (e.g., lettre - ‘letter’). (Note, however, that this would also be consistent with the obstruent being syllabified as a coda, and the liquid being syllabified as an onset (O.L) (see below for further discussion)). Féry (2003) supports an onset analysis of word-final OL clusters in terms of optimality-theoretic constraints (Prince & Smolensky, 1993), maximizing the onset to the empty-headed syllable. Perhaps children take a similar approach. The representation of word-final OL clusters prosodified as either the coda or as the onset to an empty-headed syllable is given in (6a) and (6b) respectively (the former is adapted from Tranel (1987: 136)).
(6) Representation of word-final OL clusters in French - lettre ‘letter’
Apart from a syllable-final and a syllable-initial representation of OL sequences, some accounts of the French sound system suggest that the final liquid in OL clusters is syllabic (Casagrande, 1984). Although the phonetic realization of the final liquid may appear to be syllabified, it differs greatly from the realization of syllabic liquids in other languages, where the syllabic consonant occupies the nucleus of the syllable. This is shown for the English word letter in (7a). However, learners might break the cluster into two parts, syllabifying the obstruent as a coda, and the liquid as an onset (O.L) (7b).
(7) Possible syllabifications of final OL sequences.
a. Syllabic liquid: English letter b. Coda+Onset: French lettre ‘letter’
The purpose of this study was therefore to examine the acquisition of French word-final consonant clusters. Given the ongoing debate regarding the syllabic representation of both French word-final consonants and consonant clusters, it may be difficult for learners to
these structures. We might also expect to find individual variation, with learners showing different patterns of development. Of course, determining the syllabic structures children assume may be an impossible task if the facts of the adult language can be interpreted in several different ways. Nonetheless, we hope that by examining the development of French word-final clusters, we might better understand how these are acquired, and perhaps also shed light on their structure. Before proceeding to the study, we first review the literature on the acquisition of singleton word-final consonants in French, and then turn to a discussion of the acquisition of clusters more generally.
2. The acquisition of singleton word-final consonants in French As illustrated in the discussion above, many of the issues relating to the acquisition of French singleton word-final consonants and word-final clusters are the same. It has been proposed that both can be realized within the syllable coda, or that both can be represented as onsets to empty-headed syllables. It has also been proposed that some singleton word-final consonants (e.g., sonorants) can be syllabified as codas, but that others (e.g., obstruents) must be syllabified as onsets to empty-headed syllables (Plénat, 1987). If this were the case we might expect to find a different pattern of acquisition for word-final singleton obstruents than for word-final singleton nasals and liquids.
Goad and Brannen (2003) suggest that onsets to empty-headed syllables are less marked than codas, and will therefore be acquired earlier. Furthermore, they suggest that the apparent early acquisition of codas, even in languages like English, is due to the fact that the codas are actually onsets to empty-headed syllables. As support for their claim they point to the fact that English-learning children often follow a target word-final consonant with a vowel (CVC CVCV), or exhibit a strange release on the final consonant (e.g., aspiration). They take both as evidence that learners treat these consonants as onsets (though see Demuth, Culbertson &
French. Note, however, that it is also possible that onsets to empty-headed syllables may be acquired later than canonical onsets due to the fact that the best onsets are followed by a vowel (cf. Demuth, 1995, Gnanadesikan, 2004). That is, contrary to proposals by Goad and Brannen (2003), it is possible that onsets to empty-headed syllables are more marked than codas.
In this section we review experimental findings regarding French-speaking children’s acquisition of singleton word-final consonants, providing a baseline for assessing the acquisition of word-final clusters. The children in both studies lived in Lyon, France. Thus, any differences in the results are not due to dialectal variation.
The singleton word-final consonant study was conducted with 15 children (9 girls and 6 boys) ranging in age from 1;8-2;9 (mean age 2;2) (Hilaire-Debove & Kehoe, 2004). Children were asked to identify pictures of 29 familiar words commonly found in French-speaking children’s vocabularies (17 monosyllables and 12 disyllables). Sessions lasted approximately 30 minutes, with multiple productions of the same word included for analysis. Most wordfinal consonants were target-like, thus a word-final consonant was classified as correctly produced if a consonant was produced. Overall, the children were 85% correct at producing word-final consonants in a target-like manner, though the older children tended to produce more coda consonants than the younger children. Performance was best on the least sonorant stop consonants (90%), very good on fricatives (85%) and nasals (87%), and less good on the more sonorant liquid consonants (77%) The production of word-final liquids was much better for the older children, suggesting an effect of segmental development, especially for /r/.