Gouskova, Maria. 2012. Unexceptional segments. Natural Language and Linguistic Theory 30:1, pp. 79-133.
A famous perennial problem in Slavic phonology is yers: vowels that idiosyncratically alternate with zero (e.g., [mox] vs. [mx-a] `moss (nom/gen sg)’ alongside [nos] vs. [nos-a] `nose (nom/gen sg)’). The widely accepted analysis of these “ghost vowels” is that they must be underlyingly marked as exceptional on a segment-by-segment basis. Moreover, usually they are assumed to be underlyingly representationally defective—either nonmoraic or lacking features (Kenstowicz and Rubach 1987, inter alia). In this paper, I revisit yers from a different perspective. Instead of treating the segments as special, I argue that exceptionality is a property of whole morphemes. This theory of exceptionality has many incarnations (Chomsky and Halle 1968 et seq.), but my version is formalized as Lexically Indexed Constraints in Optimality Theory: in any given language, a universal constraint can be indexed to individual morphemes in the lexicon and ranked in two different positions in the language’s hierarchy (Pater 2000, 2006). In Russian, the relevant indexed constraint is *Mid, which penalizes the peripheral mid vowels [e] and [o]. The general, non-indexed constraint is independently needed to explain vowel reduction in unstressed syllables. The indexed version explains why only mid vowels alternate with zero in Russian. This generalization about yer quality is lost in representational accounts, since any vowel can be labeled as non-moraic underlyingly. Another unsolved mystery about Russian yers is that only vowels in the final syllable of a morpheme can alternate with zero. This requires a phonological explanation—labeling only the alternating vowels as underlyingly special does not address the position problem.