There has been little interaction between speech recognition and linguistic phonology, due to the different aims of the two fields. It is proposed here that information about phonological processes can be of use in recognition. This paper compares two phonological rule components in a system having limited dialect normalization, one illustrating context-free rules and one making use of context sensitivity. It is argued that a context-free set of phonological rules is inadequate to deal with phonological processes in natural language.