|Title:||Silent Onsets? An optimality-theoretic approach to French h aspire words|
|Authors:||Christoph Gabriel, Trudel Meisenburg|
|Comment:||Presented as a poster at OCP2, Tromso¸ (Norway), Jan. 20-22, 2005|
|Abstract:||The aim of this paper is to examine the possible contributions of Optimality Theory (OT) to one of the classical problems of French phonology: the phenomenon of h aspire words (Fouche 1959, Klein 1963, Rothe 1978, Encreve 1988). These are vowel initial words that do not provoke any of the typical prosodic processes of French such as Enchainement (syllabification of a fixed final consonant into the empty onset position of the following word (1)), Liaison (surfacing of a latent final consonant into the onset position of the following word (2)) and Elision (deletion of certain vowels in pre-vocalic position (3)).
(1) sept amis [sE.ta.mi] 'seven friends' vs. sept Hongrois [sEt.O~gRwa] 'seven Hungarians'
(2) les amis [le.za.mi] 'the friends' vs. les Hongrois [le.O~gRwa] 'the Hungarians'
(3) le [l@] + ami -> l'ami [la.mi] 'the friend' vs. le Hongrois [l@.O~gRwa] 'the Hungarian'
H aspire words even require the final latent schwa of a preceding word to be realized:
(4) une souris [yn.su.Ri] 'a mouse' vs. une hausse [y.n@.os] 'a rise'
This provides strong evidence for the assumption that the beginning of h aspire words is granted a special protection while usual vowel initial words have their left edges masked by the above mentioned processes. Such a protection, however, causes other problems: the absence of Elision and the non-surfacing of latent consonants as well as the surfacing of latent schwas provoke an increase of hiatuses (V.V), which are rather dispreferred in French. Furthermore, the non-application of Enchainement yields the structure C.V (i.e. a syllable with a coda immediately followed by a syllable without an onset), a constellation that implies a con-joint violation of the basic universal constraints on syllable structure ONSET and NOCODA. The question arises of how such problematic cases are dealt with phonetically.
In order to answer this question we examined data from an experiment conducted in May 2004 with 12 native speakers of French. The analysis reveals that, following a fixed coda consonant, the beginning of an h aspire word is crucially delayed, thus leaving a pause between the coda consonant and the initial vowel, the latter sometimes being characterized by the phenomenon of creaky voice: sept Hongrois [sEt._?O~gRwa]. Void in some cases, this pause is often filled with phonetic material caused by constrictions of the glottis and/or a glottal stop - sounds that in French are usually limited to cases of special emphasis. Speakers thus tend to create some sort of 'silent' onset position in order to avoid the problematic C.V constellation. In our data such 'silent onsets' also appear after a preceding vowel (tout Hongrois [tu._?O~gRwa] 'every Hungarian'), although the constellation V.V is tolerated in French. Such hiatuses are even created when underlying floating schwas surface before h aspire words as required by the norm (une hausse [y.n@.os]). Here again our data display variation: Instead of realizing the schwa, some speakers have the coda consonant followed by a neat interruption and a glottal stop (une hausse [y.n@_?os]), while others combine both strategies (une hausse [yn._?os]). Besides these cases where the specific character of h aspire words is maintained by the creation of a 'silent onset' and/or through hiatus, another frequent strategy consists in treating the items in question as usual vowel initial words: The final (fixed or latent) consonant of the preceding word consequently resyllabifies into the onset position (une hausse [y.nos], tout Hongrois [tu.tO~gRwa]).
In order to account for the phonetic cues we observed in the beginning of h aspire words, we propose an underlying initial segment /~?/ that is responsible for the special behavior of the items in question and surfaces as a (more or less) 'silent onset'. As the non-treatment of h aspire words as such cannot systematically be interpreted as a function of a certain speech style, the segment (/~?/) is assumed to be absent from the input of speakers who pronounce them like usual vowel initial words. For speakers who switch between these two main realizations, double input forms have to be admitted. The main dimension of variation encountered here thus constitutes a case of so-called pseudo-optionality (Mueller 1999, Anttila 2002): it is kept out of the grammar and left to the lexicon, an approach that is in sharp contrast to Tranel / del Gobbo's 2002 proposal where word specific alignment constraints are adopted. To account for the other types of variation mentioned above, we discuss and evaluate Anttila's 1997, 2002 stratified grammars model and Boersma/Hayes' 2001 concept of stochastic OT.
Anttila, Arto (1997). 'Deriving variation from grammar.' - In: Hinskens, Frans / van Houten, Roeland / Wetzels, Leo (eds.): Variation, Change, and Phonological Theory. - Amsterdam: Benjamins, 35-68.
Anttila, Arto (2002). 'Variation and Phonological Theory.' - In: Chambers, Jack K. / Trudgill, Peter / Schilling-Estes, Natalie (eds.): The Handbook of Language Variation and Change. - Malden, Mass.: Blackwell, 206-243.
Boersma, Paul / Hayes, Bruce (2001). 'Empirical Tests of the Gradual Learning Algorithm.' - In: Linguistic Inquiry 32, 45-86.
Encreve, Pierre (1988). La liaison avec et sans enchainement. Phonologie tridimensionelle et usages du francais. - Paris: Seuil.
Fouche, Pierre (21959). Traite de prononciation francaise. - Paris: Klincksieck.
Klein, Hans-Wilhelm (1963). Phonetik und Phonologie des Franzoesischen. - Muenchen: Hueber.
Mueller, Gereon (1999). 'Optionality in optimality-theoretic syntax.' - In: Glot International 4:5, 3-8.
Rothe, Wolfgang (1978). Phonologie des Franzoesischen. - Berlin: Schmidt.
Tranel, Bernard / del Gobbo, Francesca (2002). 'Local Conjunction in Italian and French Phonology.' - In: Wiltshire, Caroline R. / Camps, Joaquim (eds.): Romance Phonology and Variation. - Amsterdam: Benjamins, 191-218.