Zoque Verb Compounds

Linguists have long sought to understand the tie between verb meaning and the number and types of arguments the verb projects to the syntax (see 1). While subcategorization frames and selectional restrictions describe the constraints on the range of arguments that may appear with verbs, these devices are obvious bandaid solutions that call for deeper explanation. 1. English verb argument structures a. Subcategorization constraints i. The stick broke. ii. * The lamb sheared. iii. * The farmer put the cow. b. Selectional restrictions i. Lucy tore her blouse. ii. ? Larry tore the sissors. iii. ? The cloth shattered. Hale and Keyser (1985) proposed decomposing verb meanings at a level of conceptual structure. Semantic decomposition allows the theory to pick out semantic features that play a role in constraining verb argument structure. The divergent behaviors of the English verbs break and cut in the causative alternation (2) provide one well known example. 2. The causative alternation in English a. Moe broke the window. The window broke. b. Curly cut the paper. * The paper cut. Hale and Keyser tied the syntactic behavior of break and cut to the feature of SURFACE CONTACT that is part of the meaning of cut, but not break. This feature requires the expression of an external agent (a knife welding murderer for example) that is not demanded for verbs that refer to simple changes of state like break. Steven Pinker (1989) as well as Beth Levin and Malka Rappaport Hovav (1996) have developed this approach by noting that English verbs describing a particular manner of action (e.g. cut, shell, waddle, slouch) have more constraints on their argument structures and do not participate in the causative alternation. The basic difficulty in evaluating semantic determinants of verb argument structure is that semantic features are only as restrictive as our semantic theory happens to be. Intuitions about word meaning are notoriously opaque (Higginbothim 1989; Putnam 1993), and intuitions about the meanings of semantic features are even more suspect (Fodor 1975). In the absence of a robust