Prior beliefs that contradict data may interfere with the ability to draw accurate conclusions; however, evidence shows that through engaged activity individuals may learn new information. In Experiment 1, undergraduates performed or read about two physics experiments involving a ball and ramp. The first experiment was consistent with most people's prior beliefs, while the second was inconsistent with most people's prior beliefs. Participants' predictions, experimentation adequacy, conclusions and ability to generalize knowledge were measured to determine the effects of prior belief bias depending on whether participants conducted or read about the experiment. In Experiment 2, a structured hands-on condition was included and developmental trends across adolescence and adulthood were examined. In both Experiments, reading about a belief-inconsistent experiment led to improved conclusion accuracy and ability to generalize knowledge as compared to performing the experiment. This was the case when experiments were performed well and after a 12-week delay for adolescents. Copyright © 2009 John Wiley & Sons, Ltd.