Grammar at the Borderline: A Case Study of P as a Lexical Category Rose-Marie Déchaine University of British Columbia 1. Interfaces are borderlines What is an interface? Talk of interfaces arises when one is confronted with some boundary phenomenon that sits at a point of contact between two domains. In my area of study, namely syntax, people talk about the “syntax-phonology interface”, the “syntax-morphology interface”, or the “syntax-semantics interface”. So an interface can be understood as a borderline between two domains. This metaphor of interfaces as borderlines reminds us of the liminal aspects of language that sit either at or across the borderline. In turn, this means that understanding what an interface requires understanding what we have in mind when we talk about “domain”. It is a tenet of modern linguistics to partition language phenomena into distinct domains of description, including phonetics, phonology, morphology, syntax, semantics, information structure, pragmatics, and so on. 1 Each of these domains is in turn defined by a set of well-formedness conditions that are usually assumed to be discrete. The relative robustness of these domains of description is taken to reflect the extent to which the well- formedness conditions define a natural class of objects, e.g. phonetic objects, phonological objects, syntactic objects, semantic objects. In this way, well-formedness conditions acquire the status of boundary conditions, and are (implicitly or explicitly) invoked to determine whether such-and-such a phenomenon is phonetic, phonological, morphological, syntactic, semantic, or otherwise. While it is a useful fiction to assume that these partitions are discrete, it is evident that they are not. A large number of events that linguists analyze have an ambivalent or ambiguous status with respect to these traditional domains of description. It is in the face of such ambivalent and ambiguous data that linguists talk of interfaces. This interface talk simultaneously invokes discrete domains of descriptions and draws attention to the borderlines that both separate and join them. Such borderline phenomena are conceived of in two different ways. Assuming an orthodox set of domains of description—Phonetics, Phonology, Morphology, Syntax, Semantics—and a cyclic mapping from one to other, then an interface is anything that lies at the borderline of these domains, as in Figure 1 (next page). This cyclic model yields the following set of interfaces: Phonetics-Phonology (Browman & Goldstein 1986, 1989), Phonology-Morphology (morpho-phonology, Kiparsky 1981), Morphology-Syntax (morpho-syntax, Anderson 1982, Baker 1988), Syntax-Semantics (Diesing 1992, Kratzer 2003), Semantics-Phonetics (Fónagy 1983, Bolinger 1989). Starting with the same domains of description, another view of the borderline starts with the idea that meaning (qua Semantics) has a privileged role, and that other domains of description map onto Semantics. This is illustrated in Figure 2 (next page). Such a model yields interfaces such as Phonetic-Semantics, Phonology-Semantics (Selkirk 1984), Morphology-Semantics (Cinque 1999, Rice 2000,), and Syntax-Semantics. Although the geometry of these two approaches is very different, they both conceive of interfaces are borderlines between domains of description. Thus, these models present alternative, rather than contradictory, points of view. Showing this requires that we deconstruct the domains of description: upon close examination they themselves turn out to be conventionalized borderlines. 1 This research is supported by SSHRC 410-96-1445 (Standard Research Grant, R.-M. Déchaine, PI), 412-03- 1003 (MCRI, A.M. DiSciullo, PI). Thanks to S. Armoskaite, C. Cook, R. Canac-Marquis, J. Muehlbauer, A. Neeleman, K. Szendroi, E. Vatikiotis-Bateson, M. Wiltschko and the WCCFL 24 audience for discussion. © 2005 Rose-Marie Déchaine. Proceedings of the 24th West Coast Conference on Formal Linguistics, ed. John Alderete et al., 1-18. Somerville, MA: Cascadilla Proceedings Project.