ABSTRACT
Within the literature, a number of cases from various languages have been adduced which are incompatible with the hypothesis that phonological derivations are strictly Markovian in character and which argue for the necessity of global rules in phonology. that there are two such rules within the phonological component of Dakota, specifically VELAR PALATALIZATION and E-DELETION. Notably, both these rules require access to the same fact about the derivational history of their input forms, viz. whether an e in the input was present in the underlying form or was derived through the application of ABLAUT. The two Dakota cases of global rules conform to neither Kiparsky's nor Dinnsen's proposed constraints, however. Moreover, a re-examination of the Klamath data which Kisseberth argues requires analysis in terms of a global rule of VOWEL SHORTENING reveals that Kiparsky's proposal fails to adequately characterize the application of the Klamath global rule in the very same way as it fails for Dakota.
