Past experimental studies have documented a positive effect of observability on prosocial behavior. However, little is known about spillover effects on subsequent, unobserved prosocial actions. This paper studies the dynamic effect of observability on prosocial behavior. We hypothesize a twofold positive effect. First, in accordance with previous literature, people should act more prosocially when being observed. Second, this increased level of prosociality should motivate an ongoing elevated altruistic attitude, in accordance with the concept of altruistic capital formation. We test our predictions by running two experiments in which subjects make a first donation decision either observably or anonymously. Subsequently, all subjects face a second anonymous donation decision. In general, we observe high rates of altruistic behavior. However, we find only weak positive effects of observability on first-stage prosocial behavior and no effects on second-stage prosocial behavior.