Abstract
Most current perceptual learning (PL) theories neglect several critical components of the cognitive architecture, including the necessary mechanisms for categorization, decision making, and top-down control. The resulting gap is evident in the mainstream models of PL, which can account for the gradual improvement in a fixed stimulus environment but not for the intricate patterns of specificity and transfer from one environment to another. The perceptual categorization (PC) literature has identified mechanisms that can fill this gap. The emerging consensus in this literature is that human category learning is mediated by multiple distinct (but partially overlapping) systems. One system is explicit, involves verbal rules, working memory, and executive attention. This system supports greater generalization to novel stimuli and tasks. Another system is implicit, learns stimulus-response associations via reinforcement learning, but generalizes relatively poorly. We propose a Dual Process Model of PL (Dimple) that integrates the influential selective reweighting model of PL (Petrov, Dosher, & Lu, 2005, Psychological Review) with the influential COVIS theory of PC (Ashby et al, 1998, Psychological Review). The selective reweighting model maps naturally onto the implicit system in COVIS. The innovation in Dimple lies in the explicit system, which operates on intermediate-level representations that give separate, controlled access to individual stimulus attributes such as orientation and spatial frequency (Olzak & Thomas, 1990, VR). Dimple also has a working memory layer that maintains and adjusts the current decision boundary. Top-down selection of spatial locations and stimulus dimensions is based on the normalization model of attention (Reynolds & Heeger, 2009, Neuron). The implicit system determines the fine-tuned performance after prolonged training in a given environment, whereas the explicit system supports much of the generalization to novel stimuli and tasks. Both are necessary to account for the full pattern of specificity, transfer, and various dual-training effects (Zhang et al, 2010, J. Neuroscience).
Meeting abstract presented at VSS 2012