Constraining the phonology-phonetics interfaces: with exemplification from Spanish and Italian dialects (1995)

In this thesis, I address the interrelated issues of phonological underspecification and phonetic specification. Working within the model of Modified Contrastive Specification, I provide evidence for underspecification of height features from the phonological and phonetic patterning of vowels in Iberian Spanish and Italian dialects. I argue that (a) the feature [high] is present in inventories only when it is required to appear in order to mark a contrast between high and non-high vowels in the non-low region, and (b) that the mid vowels [e] and [o] are underlyingly underspecified for height features. Given the evidence for underspecification, I explore the relationship between the underspecified phonological representations and more fully specified phonetic representations. To do so, I first present data showing that phonologically underspecified vowels display predictable patterns of phonetic variation. I then derive such patterns from the phonetic rule of enhancement, a rule which inserts non-contrastive features in order to more fully specify underspecified representations. The main goal of this thesis, then, is to argue that the interface between underspecified phonological representations and maximally specified phonetic representations is highly constrained. I demonstrate that there are a limited number of ways to interpret phonologically underspecified representations. I also show that the interpretation of phonetic data--i.e. phonemicization--is similarly constrained, arguing in particular for the possibility of phonemicization without reference to rules