In this meeting, we'll briefly review some of the points in Lidz, Gleiman, & Gleitman (2001) "Kidz in the 'Hood" (assigned reading for acquisition lab students) in preparation for discussing White, Hacquard & Lidz (in prep) "Projecting Attitudes." The abstract for the latter paper can be found below.

This paper explores the granularity with which a word's semantic properties are specified in its syntactic distribution, taking propositional attitude verbs (PAVs), such as "think" and "want," as a case study. Three behavioral experiments aimed at quantifying the relationship between semantic properties and syntactic distribution are reported. Experiment 1 gathers a measure of various PAVs syntactic distribution using an acceptability judgment task. Experiments 2 and 3 gather a measure of semantic similarity among those same PAVs using a generalized discrimination task and a likert scale task, respectively. Analysis of the experimental results suggests the existence of fine-grained semantic information in PAV syntactic distribution. In the course of this analysis, two computational models are developed to uncover the particular relationships between semantic properties and aspects of syntactic distribution. The first model, which directly implements the linguist's notion of projection, extracts words' discrete distributional features from their syntactic distributions. The second model maps these distributional features into the semantic similarity judgments using the notion of a kernel. Two kinds of kernels are considered: a linear kernel and a diffusion (exponential) kernel. The diffusion kernel, which can be viewed as the discrete version of the gaussian kernel, provides the best fit to the generalized discrimination task data, suggesting the existence of a symbolic perceptual magnet effect. The paper concludes by showing how to convert the first model into a model of verb-learning.