Empty elements in French : what for? Marie-Hélène Côté • Université d'Ottawa

Analyses of schwa, liaison and aspiré in French have relied heavily on empty elements and have had a major impact more generally on the development of a typology of "defective" segments. My aim in this talk is to evaluate how exactly emptiness contributes to accounts of these three phenomena. I conclude that they are more insightfully understood without use of defective segments and that syllabic constituents, which are crucial elements of the defective segment analyses, play no role. I advocate a sequential approach on the segmental level coupled with a more structured lexicon.

Liaison consonants, schwa and h aspiré involve: - floating segments with respect to the skeleton and/or the , i.e. segments with no or only partial suprasegmental structure (for liaison consonants as floating consonants, see e.g. Clements & Keyser 1983, Booij 1983-84, Encrevé 1988, Prunet 1992, Tranel 1995, Davis 2000; for schwas as floating vowels, see e.g. Tranel 1987) - empty syllabic positions, without featural content (for schwas as empty nuclei, see e.g. Anderson 1982, Charette 1991; for h aspiré words as beginning with an empty onset position, see e.g. Encrevé 1988, Sung 1989, Girard & Lyche 1997).

All these analyses share two basic characteristics: 1) The defective structure serves to establish lexical distinctions: a. between variable schwa and other stable vowels (/œ/ and /ø/ in particular), e.g. pelouse 'lawn' [p(œ)luz] vs. jeunesse 'youth' [ZønEs] *[ZnEs] b. between variable liaison consonants and fixed final consonants e.g. grand 'small' [grã(t)] vs. trente 'thirty' [trãt] *[trã] c. between words beginning with h aspiré, which block liaison, vowel , suppletion and schwa omission (leaving enchaînement aside), and true vowel-inital words, which trigger these processes. h aspiré words normal vowel-initial words e.g. liaison des hibous 'owls' [deibu] des idiots 'idiots' [dezidjo] elision la honte 'the shame' [laç)t] l'onde 'the wave' [lç)d] suppletion vieux hibou 'old owl' [vjøibu] vieil idiot 'old idiot' [vjEjidjo] schwa de honte 'of shame' [døçt) ] d'onde 'of wave' [dç)d]

2) The processes involved (schwa omission, liaison, elision, suppletion, enchaînement) all crucially depend on prosodic constituents below the word, especially syllabic constituents. The distribution of schwa is determined by syllable well-formedness; liaison, elision, suppletion and enchaînement apply before onsetless .

The necessity of defective elements for our understanding of schwa, liaison and h aspiré would be challenged if it can be shown that one or both of the above characteristics is unmotivated. More specifically, I argue that: 1) the behavior of schwa is not adequately predicted by syllable well-formedness; 2) the variable pronunciation of liaison consonants is insightfully understood as deriving from and/or lexical allomorphy, rather than defective representations; 3) h aspiré is equivalently represented with syllable boundaries rather than empty onsets.

1) Schwa: The distribution of schwa cannot be predicted by syllable well-formedness. a. Schwa omission often yields consonant clusters that cannot be syllabified, according to the patterns attested in the language. Extending the range of permissible syllables or allowing for extrasyllabic consonants results in a circular analysis (or one that cannot avoid brute force). (1) a. il faut que je fasse tout ‘I have to do everything’ [ilfok_ß_fastu] b. tu fais que te moucher ‘you only blow your nose’ *[tyf´k_t_muße] (2) a. une chemise ‘a shirt’ [ynß_miz] b. une demande ‘a request’ *[ynd_mãd]

Sequences of three consonants are allowed in (1a) and (2a), but not (1b) and (2b), where at least one schwa must be pronouced in the underlined positions. Yet none of the medial consonants in (1-2) may form a complex coda with the preceding segment or a complex onset with the following one that is otherwise attested in French. b. Schwa is more often omitted in the context VC__ than in VCC__, e.g. la demande 'the request' [ladmãd] vs. une demande 'a request' [yndœmãd]. Yet speakers usually syllabify la demande [la.dmãd], not *[lad.mãd], which is the only syllabification predicted by syllable- based accounts. c. La distribution of schwa depends on segmental, morphological and prosodic factors (above the word), without reference to the syllable. The segmental generalizations are interpreted in terms of perceptual cues and sonority level (Côté 2000).

2) Liaison: (Obligatory) liaison consonants are normally epenthetic (only prenominal adjectives of the type bon 'good' involve suppletion between the liaison and non-liaison forms /bç), bçn/, where liaison consonants correspond to fixed final consonants (Tranel 1990)). Liaison as epenthesis has been advocated by Klausenburger (1974) and Tranel (1981); this analysis receives further support from additional types of data: a. Acquisition: Children appear to go through three stages during the acquisition of liaison. Liaison consonants are first encoded as inital consonants of the following word; they appear to later detach from that position but nothing suggests that they attach to the preceding word. In particular, while confusion between liaison consonants and fixed initial consonants is frequent, confusion between liaison consonants and fixed final consonants is never observed (Chevrot & Fayol 2001, Dugua 2002, Nardy 2003) b. Liaison at a distance: Adults often produce forms like ben mieux isolé 'much better insulated' [be) mjO nizçle], where liaison [n] is triggered by ben but is spearated from it by another word. This is highly unexpected if [n] is belongs to ben. c. Liaison consonants vs. consonants used in derivation: Floating consonant analyses assume that liaison consonants in prenominal adjectives are the same as those used to form the feminine or in derivation, e.g. the liaison [z] associated with gros 'big' is derived from the same consonant as the [s] used in the feminine grosse 'big-fem' and the derived noun grosseur 'size'. Yet there is no necessary correspondance between the consonants used in liaison and derivation, e.g. gros arbre [grotarbr] with [t] instead of [z] is frequent, but mistakes with the [s] of grosse and grosseur never occur.

3) H aspiré: H aspiré words are adequately represented with an initial syllable boundary, which eliminates reference to syllabic constituents. Empty onsets do not necessarily account for the obligatoriness of schwa in a preceding clitic, e.g. le hibou 'the owl' [lœibu] *[libu], while schwa remains optional before true consonant-initial words, e.g. le bureau [l(œ)byro]. The analysis of h aspiré is also made more complicated by the fact that h aspiré words often block some but not all of the processes associated with h aspiré. Typically, we observe the following hierarchy in the ease of application of processes: enchaînement > suppletion > schwa omission in clitics > liaison in [t] > liaison in [z] and [n]. Such a hierarchy cannot be obtained with a simply structural account (but is compatible with an optimality-theoretic one). References

Anderson, Stephen R. 1982. “The analysis of French shwa, or: How to get something for nothing”, Language 58 (3): 534-573. Booij, Gert. 1983-84. "French C/Ø alternations, extrasyllabicity and Lexical Phonology", The Linguistic review 3 : 181-207. Charette, Monik (1991) Conditions on phonological government, Cambridge: Cambridge University Press. Chevrot, Jean-Pierre & Michel Fayol. 2001. "Acquisition of French liaison and related child errors", in Margareta Almgren et al. (eds.) Research on child language acquisition, Somerville, MA : Cascadilla Press, 760-774. Clements, George N. & Samuel Jay Keyser. 1983. CV phonology: a generative theory of the syllable, Cambridge, MA: MIT Press. Côté, Marie-Hélène. 2000 Consonant cluster phonotactics: a perceptual approach, Ph.D. dissertation, MIT. Davis, Julia L. 2000. French liaison : a case study of the syntax-phonology interface, Thèse de doctorat, University of Indiana, Bloomington. Dugua, Céline. 2002. Liaison et segmentation du lexique en français : vers un scénario développemental, Mémoire de D.E.A., Université Grenoble 3. Encrevé, Pierre. 1988. La liaison avec et sans enchaînement, Paris : Seuil. Girard, Francine & Chantal Lyche. 1997. Phonétique et phonologie du français. 3rd ed. Oslo: Universtitetsforlaget. Klausenburger, Jürgen. 1974. "Rule inversion, opacity, conspiracy : French liaison and elision", Lingua 34 (2-3) : 167-179. Nardy, Aurélie. 2003. Production et jugement d'acceptabilité entre 2 et 6 ans: aspects psycholinguistiques et sociolinguistiqueq de l'acquisition des liaisons, Mémoire de D.E.A., Université Grenoble 3. Prunet, Jean-François. 1992. Spreading and locality domains in phonology, New York: Garland. Sung, Kuo-ming. 1989. “H-aspiré in French”. Romance linguistics & literature review 2: 1- 21. Tranel, Bernard. 1981. Concreteness in generative phonology : evidence from French, Berkeley, CA : University of California Press. Tranel, Bernard. 1987. “French schwa and nonlinear phonology”, Linguistics 25, 5 (291): 845-866. Tranel, Bernard. 1990. "On suppletion and French liaison", Probus 2 (2) : 169-208. Tranel, Bernard. 1995. "French Final Consonants and Nonlinear Phonology", in Jacques Durand & Marie-Anne Hintze (eds.) : morae, syllables, words, special issue of Lingua 95: 131-167.