From Phonological Rules to the Person Case Constraint. Monovalent vs. Bivalent Features in Grammar

In phonology, segmental content has been predominantly represented in terms of binary features. Although binary features may provide an elegant description of some segmental contrasts, it is far from clear that speaker/hearer’s knowledge about segments is organized in a binary way, as we illustrated with specific reference to vocalic alternations (metaphony etc.). The debate about binarity in phonology has a potential parallel in morphosyntax. While syntactic categories (N, V, v, T etc.) are monovalent, a model like Distributed Morphology depends on standard generative phonology for a number of formal properties, including the adoption of binary features. Thus 1st and 2nd persons are [+participant] while 3rd person is the absence of such properties, namely [-participant]. We argue that this is not the most economical set of assumptions, specifically in the explanation of the syntactic generalization known as the Person Case Constraint (PCC). For both phonology and morphology, we show that the inherent richness of binary features leads to formal and conceptual problems, such as the fact that atomic segments or lexical items have as complex a feature matrix as non-atomic ones.