Coordination class theory is used to explain college students' judgments about animated depictions of moving objects. diSessa's coordination class theory models a "concept" as a complex knowledge system that can reliably determine a particular type of information in widely varying situations. In the experiment described here, fifty individually interviewed college students judged the realism of two sets of computer animations depicting balls rolling on a pair of tracks. The judgments of students from an introductory physics class were strongly affected by the number of balls depicted (one or two), but the judgments of students from an educational psychology class were not. Coordination analysis of interview transcripts supports the interpretation that physics students' developing physics knowledge led them to consistently miss or ignore some observations that the other students consistently paid attention to. The analysis highlights the context sensitivity and potential fragility of coordination systems, and leads to the conclusion that students' developing knowledge systems might not necessarily result in consistently improving performance.