This article proposes that patterns of phonological contrast should be added to the list of factors that influence sound change. It adopts a hierarchically determined model of contrast that allows for a constrained degree of crosslinguistic variation in contrastive feature specifications. The predictions of this model are tested against a database comprising the set of vowel changes in the Algonquian languages. The model reveals striking commonalities in the underlying sources of these changes and straightforwardly predicts the previously unrecognized patterning of the languages into two groups: (i) those in which */ɛ/ tends to merge with */i/ and palatalization is triggered by */i/, and (ii) those in which */ɛ/ tends to merge with */a/ and palatalization is triggered by */ɛ/. In addition to providing a new argument for the relevance of contrast to phonology, the model also gives us a way to import traditional philological findings into a framework that brings them to bear on theoretical questions.