Vowel quality and word stress in native and non-native English

In a recent study [Fokes and Bond, J. Acoust. Soc. Am. Suppl. 1, 79, S27 (1986)], we reported that non-native speakers of English lengthened unstressed syllables and shortened stressed syllables of three and four syllable words, such as confession and combination, in comparison to native English speakers' patterns. While all speakers were not consistent in their use of either fundamental frequency or amplitude in differentiating stressed from unstressed syllables, the non-native speakers were much more variable. In an attempt to clarify differences in speech patterns between these two groups of speakers, we measured the first two formants of the first and second syllable of the same set of words as produced in isolation and in sentence context. The F1 and F2 values of the native English speakers were as expected and differed little from isolation to sentence context. The non-native English speakers varied F1 and F2 from isolation to sentence context but with no discernible pattern. The formant measurements will be discussed both in terms of intended and perceived vowels.