Cognitive psychology has provided valuable theoretical perspectives on learning mathematics. Based on the metaphor of the mind as an information processing device, educators and psychologists have developed detailed models of competence in a variety of areas of mathematical skill and understanding. Unquestionably, these models are an asset in thinking about the curriculum we want our students to follow. But any psychological paradigm has aspects of learning and knowledge that it accounts for well, and others that it accounts for less well. For instance, the paradigm of cognitive science gives us valuable models of the knowledge we want our students to acquire; but in picturing the mind as a computational device it reduces us to conceiving of learning in individualist terms. It is less useful in helping us develop effective learning communities in our classrooms. In this paper I review some of the significant accomplishments of cognitive psychology for mathematics education, and some of the directions that situated cognition theorists are taking in trying to understand knowing and learning in terms that blend individual and social perspectives.