Three experiments were conducted to determine whether an instructional set for visual imagery would facilitate reading comprehension of elementary school children. In the first experiment, 38 fifth graders randomly assigned to experimental and control groups were given the Durrell-Sullivan Reading Capacity and Achievement tests after the experimental group had been given instructions to visualize what they had heard and read. Raw scores revealed no significant differences between the groups as a whole, but the upper third of the experimental group was favored at the .05 level of significance. A second experiment with 40 randomly assigned fifth and sixth graders confirmed two hypotheses: (1) that when given a comprehension test, the upper third of the readers who had been instructed to visualize would do better than the readers who had just been told that comprehension was going to be tested, and (2) that the mean scores would be the same for both groups. A third experiment with 36 third and fourth graders duplicating the second experiment produced the same results. These findings indicate that teachers should try to help children develop the metacognitive skill of visual imagery as a strategy for improving comprehension. (JL)