The Distribution and Comprehension of Implicit Arguments

A core question in linguistics is how expressions are built from functions and their arguments, e.g. verbs and nouns. Generally all of a function’s arguments must be pronounced. Certain arguments may remain implicit (i.e., unpronounced), but only as long as they are recoverable from prior discourse. This project investigates how this recovery procedure is similar to that of determining the antecedent of a pronoun, which is likewise a discourse-dependent process. Investigation will proceed along two lines: computational modeling and psycholinguistic experimentation. In the computational modeling component, we will annotate a large-scale corpus of conversational text for pronoun and implicit argument antecedents, leading to distributional analysis of the differences between these two expressions and computational models to capture these differences. In the psycholinguistic component, we will investigate to what extent people’s cognitive processes (specifically, attentional focus and memory-sensitive behavior) during comprehension of implicit arguments and pronouns are distinct. One important effect we will investigate is how prior discourse context may bias a comprehender towards the recovery of particular values for a given implicit argument. Both components will target implicit spatial and experiential perspectives (e.g., “delicious” [to whom?]), an understudied area in the linguistic sciences.