Explaining Kashaya Infixation Eugene Buckley (University of Pennsylvania) One of the most persuasive types of evidence in favor of the ranked surface constraints of Optimality Theory is the natural and intuitive account they provide for prosodically motivated infixation (Prince and Smolensky 1993). A classic example is Tagalog -um-, which occurs after the initial consonant(s) of a word in order to avoid codas (e.g. s-um-ulat) -- in OT terms, NoCoda >> AlignL-um). I show that the considerably more complex facts of infixation in Kashaya (Pomoan: N. California) must also be explained in terms of surface well-formedness, including the nature of the features in the coda. The optimal position of the affix depends on independent featural processes, a fact captured by comparing alternative surface representations. The Kashaya 'plural act' verb marker has several allomorphs (Oswalt 1961, Buckley 1994). Vowel-initial allomorphs are always suffixed (e.g. c'ihwin-at-, di?k'ol-aq-). But the allomorph -ta- occurs either as a suffix or infix depending on the final consonant of the root: when the verb ends in /l, n, c'/ it is suffixed (ditan-ta-); when the final consonant is /m, q, c/ it is infixed (bilaqha-ta-m-). A partial generalization is suffixation after a coronal, otherwise infixation; but the plain palato- alveolar /c/ -- like glottalized /c'/, clearly a coronal in the language -- patterns with the noncoronals (infixed duqa-ta-c- versus suffixed dayec'-ta-); more on this below. The basic contrast is directly due to the harmonic scale coronal coda > noncoronal coda: in ditan-ta- the coda is favored coronal; in *bilaqham-ta- the coda is dispreferred labial, forcing infixation. Specifically, *Dor]syl >> *Lab]syl >> *Cor]syl (Smolensky 1993). Unlike Tagalog, what matters in Kashaya is not whether a coda is created, but rather the content of that coda. Of course, vowel-initial -at, -aq can be suffixed because no codas at all are created. Why then doesn't duqa-ta-c- pattern with the other coronals? The alternative candidate *duqac-ta- ought to be optimal based on the other data, but by an independent process, a /c/ before a coronal loses its place features (*/c/-Cor), with lengthening of the preceding vowel (cf. s'uwac-ti --> s'uwa:ti). Thus the real alternative is *duqa:-ta-, with loss of the segment in violation of Max. Since Max >> AlignR-Plural, we find infixation. Another allomorph, -t-, is always infixed before a final consonant; but due to independent debuccalization, a /t/ which occurs before another coronal surfaces as [h], as in dahyut-ti --> dahyuhti. So there are three situations for this allomorph: suffix [t] after a vowel (qawa-t-); infix [t] before a noncoronal (phane-t-m-); and infix [h] before a coronal (ke-h-l-). Infixation in phane-t-m- is expected since it avoids the labial coda in *phanem-t-. But debuccalization -- triggered by a general restriction against adjacent coronals (OCP-Cor) -- is necessary to explain why infixation also occurs in ke-h-l-. If we had to choose between kel-t- and ke-t-l- "before" debuccalization, we would wrongly choose well-aligned *kel-t-, since the two forms both present coronal clusters. It is crucial, then, that the actual surface forms be compared. Under this analysis, there is no need to treat infixed -ta- and -t-/-h- as special among the plural acts: AlignR can have the same relatively low ranking for all allomorphs. The fact that only a subset of forms is found infixed follows from their prosodic and segmental shape, as outlined above. Vowel-initial -at and -aq must be suffixed to achieve the best syllable structure (cf. *c'ihwi-at-n-), the inverse of Tagalog. Similarly, the allomorph -m, which occurs only with coronal-final stems, is always a suffix (ba?t'il-m-): infixation would gratuitously create a disfavored noncoronal coda (*ba?t'i-m-l-). No stipulation of position for particular allomorphs is necessary -- an enormous improvement over the complex statements found descriptively (e.g. Oswalt 1961). While lexically idiosyncratic choice between, say, -ta- and -t- is unavoidable, this analysis derives the position of the chosen allomorph from a single alignment constraint for the entire suffixal class, plus general phonological principles. Since the outcome of featural processes must be taken into account in positioning the affix, the data strongly support the use of phonological and morphological constraints which refer to surface representations. Selected References Buckley, Eugene. 1994. Theoretical aspects of Kashaya phonology and morphology. CSLI, Stanford University. McCarthy, John, and Alan Prince. 1995. Faithfulness and reduplicative identity. Univ of Massachusetts Occasional Papers in Linguistics 18. Oswalt, Robert L. 1961. A Kashaya grammar (Southwestern Pomo). Dissertation, UC Berkeley. Prince, Alan, and Paul Smolensky. 1993. Optimality Theory: constraint interaction in generative grammar. Ms., Rutgers University and University of Colorado at Boulder. Smolensky, Paul. 1993. Harmony, markedness, and phonological activity. ROW-1.