Cues or Contexts in Feature Licensing Constraints

Steriade (1993, 1995, 1996) has proposed that the distribution of phonological contrasts is explained, not by a principle of Prosodic Licensing (Itô 1986, Goldsmith 1990) (licensing of features only in certain prosodic positions, e.g. onset position), but rather by the presence of sufficiently audible phonetic cues in the relevant contexts that signal these contrasts. For example, Steriade (1996) observes that contrastive glottalization in sonorants is frequently restricted to post-sonorant position (e.g. Yokuts), while in obstruents it is restricted to pre-sonorant position (e.g. Kashaya), where the relevant adjacent sonorant is voiced. In sonorants, glottalization is realized as creaky voice, typically beginning during the preceding segment; whereas in obstruents, it is realized as ejection, i.e. a loud release burst, followed by a long lag before the onset of voicing.1 Let us assume that creaky voice is incompatible with obstruents, plausibly for aerodynamic reasons. Therefore, in post-obstruent or word-initial position (or following a voiceless sonorant, if such exist in the language), glottalization lacks the ancillary cue of (at least partially) creaky voicing during the preceding segment. For ejective obstruents, on the other hand, in word-final position, or preceding a voiceless segment, the cue of long voice onset time is absent; in pre-obstruent position, the release burst is obscured by the following stop closure or fricative noise; moreover, even if the following obstruent is voiced, the typically stronger "spontaneous voicing" in a sonorant makes for a clearer voice onset time cue in pre-sonorant than pre-obstruent position. This analysis can be elegantly formalized in terms of a set of constraints expressing the requirement of sufficient cues: "if [+constricted, +son] then [+partial creak] in the preceding segment"; and "if [+constricted, -son] then [+long voice onset time] and [+ejective burst]."2 That is, we treat the contrastive feature and its licensing cues alike as phonological features, and formally express the dependencies directly between them. Call this Direct Licensing.