Formal and functional aspects of Russian vowel reduction

In Contemporary Standard Russian, the stressed vowel inventory /i,e,a,o,u/ is reduced to [i, ́,u] in unstressed syllables. (The precise realization of “ ́” depends on whether the relevant vowel is pretonic or not, among other things.) Examples of the resulting alternations are shown in (1). After palatalized consonants, only [i,u] occur. Though vowel reduction in Russian has been very well studied (e.g., Avanesov 1956, Halle 1959), it largely eluded formal analysis within generative phonology for many years. In fact, mainstream generative phonology has not dealt well with vowel reduction in general. Recent work suggests a new fruitful approach to vowel reduction, one that roots the formal analysis in functional phonetic constraints (Crosswhite 2001, Flemming 2002). Flemming’s approach makes the following assumptions. First, in unstressed syllables there is a tendency to undershoot the low jaw position needed for lower vowels like [a], because this requires a relatively large articulatory movement in little time. (This presupposes that unstressed vowels are short.) Second, because of the raising of the vowel space ‘floor’ that results, vowel phonemes become more perceptually crowded. Finally, in order to maintain the required perceptual distance among vowels, neutralization occurs, leading to fewer vowel contrasts. The idea is illustrated in (2), where “∆” indicates a required perceptual distance between vowels. A recent phonetic study of Russian lends support to the claims about vowel duration, jaw raising, and to a lesser extent, distances between vowels (Padgett and Tabain 2003). In this talk I provide a formal analysis of vowel reduction in Russian along the lines of Flemming (2002), cast within the framework of Optimality Theory (Prince and Smolensky 1993). The account explains the central facts of Russian reduction with considerable success. It therefore provides support for the view that explanation in phonology requires appeal to constraints that make reference to the perceptual distance between contrasting forms, and to levels of articulatory difficulty, a notion advocated by much recent work. However, there are some challenges for the account as well. Russian vowel reduction appears to be derivationally opaque in two ways. First, though both [i] and [u] are allowed after palatalized consonants, underlying /o/ reduces to [i], not [u]. This is a problem for a theory based on output constraints, and input-output faithfulness, alone. A possible solution invokes an intermediate step in the derivation: /o/ Æ ́ Æ [i]. Interestingly, however, the phonetic data of Padgett and Tabain raises