Usage-based linguistic theory argues that experience is the fundamental organizing principle of language. Linguistic representations are extracted from – and continuously tuned by – probabilistic features of language use. Much psycholinguistic evidence supports this argument, particularly in the domain of lexical processing. For example, how a word is distributed across its various lexical and morphological contexts influences how quickly it is recognized and produced in isolation. Fewer studies have explored how syntactic distributions affect lexical processing, and of these, even fewer have adopted comprehensive, abstract measurements of syntax. In this dissertation, I present several new information-theoretic tools for measuring the syntactic distributions of words based on the Dependency Grammar formalism. This formalism allows me to contrast two independent dimensions of syntactic structure: hierarchical status and word order. Further, I provide a new method for teasing apart information bound to syntactic and lexical contexts. I compute these measures for nouns based on two large corpora of English.
These measures are correlated with behavior in several contexts. First, I re-analyze the noun-based trials of two previously published databases of visual lexical decision response time data, one simple and the other primed. I then turn to production, reporting two picture-naming studies. In the first, participants produce nouns in isolation. This task consitutes a stong attack on the hypothesis that syntactic distributions affect noun production; at least on its face, it does not require participants to access syntactic information in order to successfully complete the task. In a follow up, participants were asked to name the images using a syntactic frame (the + NAME). This task should promote syntactic access, increasing the likelihood that prior syntactic distributions should play a role. Finally, I test whether children are senstive to these syntactic distributions (based on adult speech) as they begin to produce nouns in syntactic contexts for the first time using a large, densely sampled longitudinal corpus of child speech.
Results show that isolated noun processing is affected by prior syntactic distributions in both comprehension and production. However, the specific nature of these effects differs across modalities, and in production, as a function of whether the nouns were produced in isolation or within a syntactic frame. The measures also predict the age at which nouns first emerge in the speech of children.
