The dominant model of behavioral learning may not apply to organizations because it assumes that the people involved agree in their outcome assessments of actions and have relatively equal power to engage in joint learning. We relax these assumptions of consensus and power balance in order to apply the model to organizational (as opposed to individual) learning. We examine what happens when parties from different organizational units and levels engage jointly in learning from recurrent events. We examine behavioral learning as recurrent cycles of action–outcome assessment–response on similar tasks (i.e., recurrent events) over time. We observe different patterns of organizational learning (adaptive and dialectical) and nonlearning (persistent and compulsory behavior) in our 8-year real-time field study of the recurrent events while integrating a large health-care system. These four processes are interdependent and explained by just two contingencies: (1) agreement on outcome assessments and (2) power balance among the parties engaged in actions. These findings expand our repertoire to include processes that heretofore have not been considered in the organizational learning literature and explore how they change with different degrees of agreement and power among parties; that is, how an imposition of power by more powerful parties leads to nonlearning, while an openness to conflict and an empowerment approach enables organizational learning.