Patterns of variation and change in English schwa

Schwa has been somewhat neglected in studies of spontaneous speech. This thesis addresses this gap by providing a detailed investigation of acoustic variation in schwa according to a number of variables. Large amounts of spontaneous speech are analysed from two different varieties of English; Derby, UK (Milroy et al, 1996), and New Zealand English (Gordon et al, 2007). The work in this thesis contributes to the understanding of unstressed vowels in English in a number of ways. It is shown that schwa is distinct from /ɪ/ in unstressed syllables, although I also show that the way in which these vowel qualities are distributed differs amongst speakers. In addition, an extensive methodological analysis is presented, which explores the way that automated measurements of unstressed vowels can be filtered in order to make them suitable for analysis. The thesis also contributes to debates about whether schwa has a phonetic target (cf. Browman and Goldstein, 1992; Flemming, 2009). A phonetic distinction between schwas that occur before a pause and before a consonant is found. Schwas before a pause are lower vowels and are also less variable in backness. Conversely, many pre-consonantal schwas are quite high in realisation, and vary widely in backness. However, when variation within schwa is considered, the evidence clearly points towards it having a phonetic target. Variation in schwa is explored according to its formant trajectory, its length, and also according to speaker year of birth. Clear evidence of schwa moving towards a phonetic target as it gets longer is found. Longer schwas are overall lower vowels, and also less variable in backness. Evidence of schwa undergoing change over time is also provided, with schwa having undergone substantial lowering in New Zealand English. Overall, the findings in this thesis provide clear evidence that schwa has a phonetic target.