Students in a mathematics classroom are motivated to do many things, not only the ones we expect them to do. In order to understand student behaviour in classrooms we need to increase our understanding of what motivation is and how it is regulated. Two issues relevant to a critique of mainstream motivation research need consideration: (a) the importance of the unconscious in motivation and (b) focusing on motivational states and processes rather than traits. In the present paper, motivation is conceptualised as a potential to direct behaviour through the mechanisms that control emotion. As a potential, motivation cannot be directly observed. It is observable only as it manifests itself in affect and cognition, for example as beliefs, values and emotional reactions. This potential is structured through needs and goals. Based on this view of motivation and the author's earlier studies, three aspects of motivation regulation are discussed. Primarily, goals are derived from needs: in learning situations, the psychological needs for autonomy, competence, and social belonging are the most significant determinants of goal choices. As a second aspect, this view accepts the influence of students' beliefs about the accessibility of different goals. As a third aspect, the influence of automatic emotional reactions for goal regulation will be discussed. The case of Frank will be used 1) to illustrate how motivation can be inferred from different kinds of data and 2) as an example of how conflicting goals lead to non-straightforward self-regulation.