Ji Y. Son (jys@indiana.edu) Robert L. Goldstone (rgoldsto@indiana.edu)
Department of Psychology, Indiana University 1101 E. 10th Street, Bloomington, IN 47401 Abstract
Two experiments examined the role of relational language on analogical transfer. Participants were taught Signal Detection Theory (SDT) embedded in a doctor story. In the experimental condition, relational words accompanied the story. Relational words that shared superficial similarity with the contextual elements facilitated transfer. Without the shared semantics, relational words were detrimental to transfer performance. A computational model lends a more structured perspective on how language changes cognition.
Introduction
Most people simply expect learning words to lead to learning concepts because we use l anguage to communicate ideas. Evidence of this intuition is found in the emphasis on vocabulary and “bolded words ” in educational settings. Lexical categorization is considered equivalent to possessing certain concepts. Indeed, language has played a significant role in the development of human thought not only as a means of communication but also as a tool that augments our computational abilities (Clark, 1998; Gentner, 2003). Language dovetails with other cognitive functions well becaus e it can act as a filter for complex perceptual experience by directing our attention to particul ar similarities. Relational language is a particularly powerful illustration of these qualities of language since relations are not as perceptually salient or stable as objects (Gentner, 2003). For our purposes, objects are loosely defined as things that are immediately process ed and relations exist between such objects making them more cognitively demanding. A simple example comes from a study on four-year-olds who show sensitivity to higher-order relations such as symmetry after being exposed to triads like xXx that were labeled “even” (Kotovsky &