On affix allomorphy and syllable counting

1. Introduction At issue in this paper is the interaction between prosody and morphology; specifically, the way in which metrical foot structure conditions allomorphs of affixes. I will propose an analysis of various affix allomorphies in Estonian, which are apparently conditioned by the number of syllables in the base to which the allomorphs attach. As I will show, it is not syllable number that matters, but rather the foot parsing of the complete base-plus-affix combination. This analysis will produce evidence for the relevance of metrical feet in morphological phenomena, and for a constraint-based (rather than rule-based) model of the morphology-phonology interface. It has been observed for various languages that allomorphs may be conditioned by prosodic structure of the base, such that one allomorph occurs with bases of a certain prosodic type, while the other allomorph is in complementary distribution, occurring with bases of all other prosodic types. Three familiar examples are sensitivity to (i) syllable structure (e.g. a distinction between bases ending in a consonant vs. those ending in a vowel), (ii) syllable count (e.g. a distinction between even-numbered vs. odd-numbered bases), and (iii) to stress (e.g. a distinction between bases ending in a stressed syllable vs. those ending in an unstressed syllable). Complementary distribution of prosody-dependent allomorphs is attested in two types that I will refer to as fully-conditioned and partially-conditioned. In the former, complementary distribution follows completely from prosodic principles (e.g. the avoidance of ill-formed syllables), without any need for supplementary morphological statements. In partially-conditioned allomorphy, prosody accounts for the distribution of only one allomorph, while the distribution of the other allomorph must be due to morphological principles (e.g. a preference for one allomorph over another). An example may serve to clarify the notion of partially-conditioned allomorphy. In Djabugay (Patz 1991), the genitive is marked by either: The choice of the proper allomorph is partially predictable from CV shape, in particular from the way in which it syllabifies with the base into a PrWd. Djabugay has no syllables that end in a consonant cluster, and therefore the single consonant allomorph-n cannot be attached to a base that is itself consonant-final: (2) /-n/ In short, allomorphy never produces ill-formed syllables. Notice that prosody predicts the complementary distribution of allomophs only partially, since syllabification alone is insufficient to explain why the allomorph /-Nun/ cannot be adjoined to vowel-final bases. That