Previous studies of program visualization have generally failed to provide convincing support for the benefits of algorithm animation in promoting the understanding of computations. This paper presents a study in which the use of program visualization resulted in significantly better understanding of a distributed computation. Understanding was measured in terms of the number of correct responses to questions about the algorithm. The environment used in this study differs from that of previous studies in a number of aspects; it combines the use of distributed algorithm visualization, 3-D visualization, and legends. In addition, the design of both the experiment and animation focused on reducing cognitive noise.