A cross-linguistic study of phonotactics and vowel length

Date of Completion

January 1995


Language, Linguistics




This study seeks to account for vowel shortening and the distribution of vowel length within a constraint-based framework in which all parts of the lexical representation are within the domain of constraints. I demonstrate that by taking into account languages' differing phonotactic constraints, we can account for languages' differing environments for vowel shortening. In addition, I argue that children do not need negative evidence to acquire phonotactic constraints because they automatically build such syllable structure or output constraints in the process of learning how to vocalize.^ I begin with Michaels' (1989) syllable structure theory and expand it to allow an X-bar structure for coda. With this I integrate Clements' (1992) sonority theory to formulate a theory of phonotactics. I motivate coda phonotactics for English based on the distribution of consonants in the context of both short and long vowels, and then derive vowel shortening based on stress requirements and phonotactics. I also demonstrate that the constraint-based framework correctly predicts that there is more than one possible output for a given input, thus accounting for a group that remained exceptions in previous analyses. In addition, I present evidence from the acquisition literature that the past tense of keep/kept type verbs may be memorized rather than derived.^ To test the cross-linguistic ability of my approach, I examine three additional languages exhibiting vowel shortening. For Icelandic, I propose a phonotactic constraint on stressed syllables that accounts for both the distribution of vowel length and consonant clusters, as well as vowel shortening. For Turkish, which allows some coda consonant clusters but never has long vowels in closed syllables, I propose a phonotactic constraint on coda that allows the well-formed clusters and also triggers vowel shortening. Clements' sonority Dispersion Principle encounters some difficulty with Turkish clusters, and I propose a solution which takes into account the salience of consonants in the context of other consonants. Finally, for Yawelmani I argue that, due to reasons of economy, vowel shortening is the preferred structure saving strategy over vowel epenthesis because the latter requires specification of unspecified features. ^