Russian exhibits a rich pattern of phonological vowel reduction, by which some vowel contrasts are neutralized in unstressed syllables. Recent work in phonology suggests a mechanism by which phonetic vowel reduction - compression of the overall vowel space due to target undershoot - might lead to patterns like Russian. Presenting acoustic data from 9 speakers of Russian, we use Euclidean distance measures, measures of F1-F0 and F2-F1, and Bayesian classification to provide a basic picture of how the overall vowel space, as well as the distribution of vowels, change as stress is reduced. We are particularly interested in whether contraction of the vowel space in unstressed positions is primarily due to raising, and in whether contrasting pairs of vowels are evenly spaced within and across contexts. Our results provide qualified support for the first hypothesis, but largely do not support the hypothesis of equal spacing, in particular across contexts. Of additional interest, we find that some impressionistically described neutralizations are incomplete.

This study presents EMA (electromagnetic articulography) data on articulation of the vowel /a/ at different prosodic boundaries in French. Three speakers of metropolitan French produced utterances containing the vowel /a/, preceded by /t/ and followed by one of six consonants /b d g f s S/ (three stops and three fricatives), with different prosodic boundaries intervening between the /a/ and the six different consonants. The prosodic boundaries investigated are the Utterance, the Intonational phrase, the Accentual phrase, and the Word. Data for the Tongue Tip, Tongue Body, and Jaw are presented. The articulatory data presented here were recorded at the same time as the acoustic data presented in Tabain [J. Acoust. Soc. Am. 113, 516-531 (2003)]. Analyses show that there is a strong effect on peak displacement of the vowel according to the prosodic hierarchy, with the stronger prosodic boundaries inducing a much lower Tongue Body and Jaw position than the weaker prosodic boundaries. Durations of both the opening movement into and the closing movement out of the vowel are also affected. Peak velocity of the articulatory movements is also examined, and, contrary to results for phrase-final lengthening, it is found that peak velocity of the opening movement into the vowel tends to increase with the higher prosodic boundaries, together with the increased magnitude of the movement between the consonant and the vowel. Results for the closing movement out of the vowel and into the consonant are not so clear. Since one speaker shows evidence of utterance-level articulatory declension, it is suggested that the competing constraints of articulatory declension and prosodic effects might explain some previous results on phrase-final lengthening.