When visualization tools utilized in computer programming education have been evaluated empirically, the results have remained controversial. To address this problem, we have developed a model of short-term effects of program animation, and used it in a series of experiments. In the current experiment, we varied visual representation of an animation tool and the type of students' engagement. Results of the current experiment analyzed together with the results from the earlier experiments provided support for the hypothesis that what a student does plays a more central role in the usefulness of a visualization than representation used by the tool. Moreover, the levels of engagement as they are generally used in the research literature seem not to be the best possible indicators of the effectiveness of a visualization. (Contains 11 tables and 4 figures.)