Phonetic evidence for phonological structure in syllabification

It has been widely observed that codas are marked vis-à-vis onsets. In adult languages, coda markedness manifests itself in two ways: some languages do not tolerate codas at all (e.g. Cayuvava (Key 1961), Senufo (Clements and Keyser 1983)), while virtually all languages that do permit codas place restrictions on what this position can license. Coda markedness has consequences for the syllabification of consonants at the right edge of words as well. While one view is that final consonants are always syllabified as codas, there are many languages for which this analysis is not motivated. In the 1980s, final consonants were commonly designated as extraprosodic in these types of languages (Steriade 1982; Borowsky 1986; Itô 1986). More recently, it has been observed that there is a striking parallel between right-edge extraprosodic consonants and onsets in many languages; accordingly, extraprosodic consonants have been interpreted as onsets of empty-headed syllables by a number of scholars (see e.g. Giegerich 1985; Kaye 1990; McCarthy and Prince 1990; Charette 1991; Piggott 1991, 1999; Rice 1992; Harris 1994). The marked status of codas is observed in child language as well. Early grammars show a strong preference for CV syllables, independent of the constraints of the target language (see e.g. Jakobson 1941/68; Ingram 1978; Fikkert 1994). When word-final consonants ultimately emerge, the predominant view is that they are syllabified as codas (e.g. Fikkert 1994; Demuth and Fee 1995; Stemberger 1996). In this paper, we argue against this view. We propose that right-edge consonants are first syllabified as onsets – specifically, as onset-nuclear sequences – regardless of their status in the language being acquired. The principal evidence in support of this view comes from the phonetic properties that characterize early obstruent-final CVC forms: for the final consonant, the presence of aspiration (final release), length, and homorganic nasal release; and for the medial vowel, the presence of length and post-vocalic pause. The data on which we base our analysis come from five English-speaking children between the ages of 18 and 26 months, shortly after the point when CVC forms emerge. We begin in §2 by laying out our assumptions about language acquisition and about the representations that we assume for segment structure and syllable structure. We turn in §3 to demonstrate how our views on syllable structure yield a three-way typology for the