The present paper discusses several complications which arise in applying Dollard and Miller's analysis of conflict to the study of psychotherapy. Discussed are problems associated with recent theorizing on the nonunitary nature of fear; considerations relevant to the question of why individuals exposed to partial cues for anxiety do not show extinction of the anxiety, generalization of extinction, and increased approach behavior; the role of interpersonal feedback in maintaining neuorotic behavior; and the issue of changing goals as therapy proceeds.