Two anomalies continue to confound researchers and science teacher educators. First, new science teachers are quick to discard the pedagogy and practices that they learn in their teacher education programs in favor of a traditional, didactic approach to teaching science. Second, a discrepancy exists at all stages of science teachers’ careers between a science teacher's self‐reported pedagogical preferences and the behavior they exhibit during classroom observations. Previous attempts to resolve these anomalies draw upon models of teacher cognition that often prioritize beliefs or knowledge. In this paper, we resolve these anomalies and further elaborate the relationship between science teacher education and classroom practice via a goal‐driven model of cognition. Our model of goal‐driven cognition takes as its central premise that a science teacher's instructional practice is an attempt to satisfy one or more of their goals. While knowledge and beliefs play an important role in cognition, a teacher's goals are ultimately the mental constructs that engage the motivational and behavioral system, leading to action. In this paper, we first specify our goal‐driven model of cognition, and detail the role played by goals, beliefs and knowledge, and contextual information in cognition and action. When then compare our model to other theoretical approaches in research on science teacher cognition, showing how our model is distinct from these approaches. Next, we return to the two anomalies mentioned above, showing how a goal‐driven model resolves each. We also contrast our resolution with how the anomalies are resolved via a belief or knowledge driven model. We then provide implications of our theory for science teacher education, with respect to both course work and field experiences. Finally, we provide directions for future research, paying particular attention to ways to provide evidence in support of this theoretical approach. © 2017 Wiley Periodicals, Inc. J Res Sci Teach 54:713–736, 2017