The most frequently used computational models in social psychology are probably various kinds of connectionist models, such as constraint satisfaction networks, feedforward pattern associators with delta-rule learning, and multilayer recurrent networks with learning. The chapter begins with work on causal learning, causal reasoning, and impression formation. A large number of central phenomena in social psychology can be captured by a fairly simple feedback or recurrent network with learning. Important findings on causal learning, causal reasoning, individual and group impression formation, and attitude change can all be captured within the same basic architecture. This suggests that we might be close to being able to provide an integrated theory or account of a wide range of social psychological phenomena. It also suggests that underlying the apparent high degree of complexity of social and personality phenomena may be more fundamental simplicity.