Maltese word-final singleton-geminate contrasts : an information-theoretic approach

Abstract

It is well established that cross-linguistically there is a strong tendency for phonological contrasts to be more robust towards the beginning of the word than the end. This is supported by the static distribution of sound patterns and by the patterning of sounds in phonological processes. Distributional evidence, such as that shown for Shona in Beckman (1997), reveals greater numbers of contrastive sounds at the beginning than at the end of the word. In Shona, this is manifested as a full set of five contrastive vowels across all vowel heights in word-initial syllables, with contrast more restricted in subsequent syllables. This reduction in contrast is consistent with the widely observed pattern for phonological contrasts to be neutralised at word endings. For example, word-final laryngeal neutralization is attested in many languages, including Maltese, where the voiced voiceless opposition in obstruents is reduced to voiceless word-finally. Similarly, reduction processes typically target segments towards the end of the word, as demonstrated by the well-known case of word-final /t, d/ deletion in American English (Guy, 1980; Neu, 1980; Bybee, 1995; Jurafsky et al., 2001).