This paper proposes that social psychologists have much to gain by exploring connectionist or parallel distributed processing models of mental representation and process, and also much to contribute to connectionist theory in return. In contrast to traditional symbolic theories, connectionist models require consideration of two levels: one involving many simple processing units that send activation signals over connections, and a higher level at which the representation of concepts (as distributed patterns of activation) and information processing, learning, and memory can be described. Connectionist models naturally offer many properties emphasized in existing social psychological theories: they can operate like schemas to fill in typical values for input information, reconstruct memories based on many sources of accessible knowledge rather than by retrieving static representations, operate with flexible and context-sensitive concepts, and compute by satisfying numerous constraints in parallel. They also make new, mostly untested, empirical predictions in many cases. The paper reviews critiques and open questions regarding connectionist models, and concludes that the contributions of our field, perhaps particularly to the understanding of cognition-motivation interactions, may be important for the future development of connectionist models that can integrate psychology as a whole.