This paper examines the distribution of schwa-initial versus schwa-less allomorphs in specific areas of Dutch derivational morphology, with particular reference to Shannon's (1991) argument that schwa epenthesis is motivated by the need to improve poor syllable contacts. The present paper makes two basic claims. First, although optimization of syllable contacts may play a role in certain segmental constraints on the distribution of schwa, its main motivation is prosodic. That is to say, schwa-less suffix allomorphs follow a branching foot. If the foot is non-branching, schwa is inserted as a linking element to make it branch. Furthermore, it is also argued that monosyllabic feet can be branching, so that “heavy” syllables of this type pattern in the same way as two syllables. The second claim made here is that these prosodic conditions on schwa allomorphy are best formulated not in terms of a rule but rather as a schema in the sense of Bybee and Slobin 1982.I would like to thank Michael Wardle and Ricardo Bermudez-Otero for their very helpful advice and comments. I am also grateful to the following native speakers of Dutch for their cooperation: Arachne van der Eijk, Roel Vismans, Henriette Louwerse, Christy Irosemito, Willemien Reckman, Maarten Dijkstra, and Philip van der Eijk. Thanks also to an anonymous JGL reader for valuable feedback.
