Non-Stationary Response Distribution: A Telltale Sign of the Dynamics of Category Rating

Abstract:

A category-rating experiment reveals a markedly non-stationary response distribution: the average response gradually increases and the response variance decreases in the course of 450 trials even though the stimulus distribution remains fixed and uniform throughout. These dynamic effects constrain the theories of judgment and category rating. In particular, they invalidate the widespread assumption of static response policy and suggest a dynamic categorization system driven jointly by environmental statistics and explicit strategies. ANCHOR -- a memory-based model of category rating and absolute identification -- accounts quantitatively for these and various other results (Petrov & Anderson, 2000, 2003, 2005). A small set of anchors stored in memory compete to match the target stimulus magnitude. Two incremental learning mechanisms update the locations and base-level activations of these anchors, thereby changing the stimulus-response mapping continuously. A slight asymmetry in an explicit correction strategy explains the upward drift in the average response levels.

Poster (pdf) ANCHOR Project Software