Symposium 5: Mental Representations in the Brain

evaluating the predictions they make concerning overshadowing, the role of similarity in discrimination learning, and the effectiveness of retroactive interference. Taken together, the results that will be described lend more support to configural than elemental theories of learning. Belgium In what sense can mental representations be implicit? Does it even makesense to ask the question? Today, several constrasting perspectives on the implicit/explicit distinction have been proposed. They offer very different answers to each of three central questions about the nature of knowledge: What is conscious? What is represented? What is causally efficacious? In this talk, I review these proposals in the context of implicit leaming research. For some authors, human learning is always accompanied by consciousness, whereas for others, learning may occur without concomitant changes in subjective experience. I argue that most current perspectives on this debate are rooted in the classical metaphor of mind as a symbol-processing sytem. I further argue that this perspective makes it impossible to understand implicit learning and implicit cognition, short of (1) denying the possibility of its existence (i.e., all causally efficacious representations are necessarily conscious to some extent), or (2) assuming the existence of an unconscious system ("a zombie") that is just as powerful as the conscious system, only minus consciousness. In contrast, I suggest that dynamical approaches, such as instantiated by connectionist models, make it clear how representations can be implicit in the senses of (1) being capable of influencing behavior in a way that does not require access or interpretation, and of (2) conveying information about task-relevant stimulus dimensions in a way that does not require explicit tokening of such dimensions. From this perspective thus, while awareness often accompanies learning, it is not always necessary for it to occur.