Controlling dynamic processes is a task that is increasingly important in many areas. A possible visualization technique for time-dependent events is animation. However, there are to our knowledge no data available on how different animation speeds affect the performance of human decision makers when trying to discover complex patterns in animations. An experiment was therefore deigned using a cue probability learning paradigm. The results indicate that speed is indeed affecting performance, at least in the more challenging condition we used. Some implications of these results are discussed.