Semantic Composition: Implementing and Lexicalizing

Paul Pietroski, University of Maryland, College Park
Friday, January 25, 11am

From an I-language perspective, the languages that children naturally acquire are /implemented procedures /that associate instructions to articulatory/perceptual systems with instructions to conceptual/intentional systems, in accord with constraints of Universal Gramamar. If we take this perspective seriously, with regard to semantic composition, we are led to ask which composition operations are plausibly invoked and implemented when the relevant instructions to conceptual/intentional systems are issued and executed. And if the process of acquiring an I-language involves lexicalization of concepts already available, we are led to ask which composition operations are presupposed, for purposes of determining the "semantic type" of a word that lexicalizes a given concept. I'll review a range of theoretical considerations and empirical facts, many familiar, that favor the following idea: the primary composition operation is predicate-conjunction (as opposed to function-application or predicate-saturation); and so open-class lexical items are understood as expressions of type [e, t], as if they uniformly indicated monadic concepts, despite the prevalence of singular and polyadic concepts (of types [e], [e, [e, t]], [e, [e, [e, t]]], e). If this is correct, it suggests that the language faculty plays an active role in "systematizing" human concepts, by forcing us to form monadic analogs of preexisting concepts; where the monadic analogs exhibit a common "format" that permits rapid and efficient conjunction, given an operation of existential closure and a few thematic roles associated with closed-class functional items or dedicated grammatical relations. For purposes of this talk, I'll set aside issues concerning quantification and other topics discussed in /Events and Semantic Architecture/, in order to focus on issues of lexicalization and implementation.