%A John Voiklis %T A Thing Is What We Say It Is: Referential Communication and Indirect Category Learning %X This study investigates the interaction of referential communication and the structure of perceptual features on the joint processes of inventing a referential lexicon for novel objects and discovering the functional significance of those objects during an indirect category learning activity. During the learning task, participants worked either individually or as cooperative dyads to learn four combinations of orthogonal functional features--nutritive vs. not nutritive and destructive vs. not destructive--that defined four categories of fictional extra-terrestrial creatures. These categories were not specifically identified or labeled; rather, participants had to infer them indirectly as they predicted the functions. Also, these functionally defined categories exhibited a complex perceptual structure: a unidimensional (simple) rule predicted one function, while a family resemblance (complex) sub-structure predicted the other function. The function-learning task yielded function prediction data. In addition, each learner worked individually to sort the creatures (pre- and post-function learning) and to predict their functions in an individual function prediction posttest that also yielded selective attention data. Together, the prediction data, sort data, and selective attention data supported three a priori hypotheses. Referential communication generates conceptual homogeneity (H3) and enhances indirect category learning (H1), though simple rules are learned earlier and better than complex relationships (H2). In explaining the learning advantages observed among dyadic learners, I argue that referential communication may highlight attention to relationships between features (perceptual and functional) and actions as well as render such relationships more memorable. Moreover, communication may foster greater motivation among collaborators and may allow them to take advantage of the differing expectations and heuristics each collaborator brings to the task. In explaining the simplicity advantages observed among dyadic learners, I argue that referential communication may provide explicit "rules" for otherwise implicit (and perhaps more difficult) judgements. Dyads appear to have established reference to simple rules earlier than they established reference to complex rules; thus, they could explicitly (and perhaps more easily) learn the simple rule earlier than the complex rule. Finally, in explaining the conceptual homogeneity between and within dyads, I consider whether communication pushes "public" conceptualizations and publicly-formed "private" conceptualizations towards a limited range of widely shareable conceptual structures. %D 2008 %K communication language-games categorization decision-making coordination joint-action %I Columbia University %L cogprints6286