Effect sizes obtained from 39 experiments were used to evaluate the predictions of the basic tasks model of graphical efficacy. This model predicts that performance will be attenuated with graphical displays as a function of the particular specifier, or visual dimension, used to code data values. In this review the basic tasks model predicted performance more accurately than did Tufte's data-ink principle. In addition, variability in effect sizes across studies revealed that the model was more successful at predicting performance in local (focusing) tasks than in global information synthesis tasks. Furthermore, the model was better at predicting performance in tasks requiring the use of physically present rather than remembered graphs. Further differences in effect sizes resulted from variability in the exact specifiers used in experimental graphs. Minimal differences were obtained among graphs that used position, length, or angle as specifiers. However, graphs that used area or volume to represent quantitative values were associated with consistently worse performance than found with other formats.