Featural representations of the New Zealand English short front vowels

The realisation of the short front vowels in NZE differs from British or American English (RP/AE) and is well documented (cf. Bauer, 1986). However, two crucial questions arise: (1) Do the acoustically different vowel realisations also involve different underlying representations as compared to RP/AE? (2) What is the precise nature of these representations? Previous investigations either suggest that most NZE speakers have the same vowel phonemes as British English speakers (cf. Bauer, 1986) or that the NZE vowel change reflects a re-distribution of phonetically detailed exemplars in the mental lexicon (cf. Warren, Hay, & Thomas, to appear). This paper proposes that NZE has in fact different vowel phonemes than RP/AE. In particular, the original three-way height contrast of the short front vowels developed into a twoway height contrast in NZE (cf. Hawkins, 1976). The underlying properties of these vowels is described in the framework of the Featurally Underspecified Lexicon (Lahiri & Reetz, 2002), in which lexical entries consist of phonological feature representations. The differing underlying vowel representations are accessed via a particular feature matching algorithm which allows variation to be resolved in one, possibly underspecified, representation, and not by storing multiple exemplars that interact with pre-lexical biases. Two behavioural experiments provide evidence for the claims that NZE vowels are represented in terms of phonological features and that NZE uses different vowel phonemes than AE. Experiment 1 conducted with NZE subjects using the indirect semantic priming technique showed that both bet and bit primes could activate the semantic relative of bat (club). The fact that bet and bit similarly primed club suggests that bat, through which the meaning of club was retrieved, is not specified for height in NZE such that the vowel in bet (with a phonetically high vowel) and the vowel in bit (with a mid vowel) could activate bat (and thereby, club) to the same degree. The results represent a challenge for exemplar-models which would expect priming differences to be based on the acoustic distance of prime and target vowels or on the frequency distributions of the stimuli. However, no such correlations could be found. Experiment 2 used the same material with speakers of AE. The results differed crucially in that the primes with a high vowel did not activate club through bat. This was predicted on the basis of a mismatch between a high and a low vowel and showed that speakers of AE have a different representation of bat words (with a low vowel) as compared to NZE speakers (with a non-low vowel).