Pictures are commonly used to represent problems. However, it is unclear how the addition of pictures affects students' problem-solving performance. The multimedia effect in problem solving describes the phenomenon whereby an individual's problem-solving performance is enhanced when equivalent pictures are added to illustrate or replace part of the problem text. Using meta-analytic techniques, this study sought to determine the overall size of the multimedia effect in problem solving and the possible boundary conditions (k = 51; N = 38,987; Range n = 10 -- 31,842; Median n = 63). The results showed a significant small-to-medium multimedia effect size on response accuracy (Hedges's g = 0.32) and a significant medium-to-large multimedia effect size on students' response certainty (Hedges's g = 0.74), but no significant multimedia effect on response time. The results for the effects of decorative pictures were not sufficient for a reliable interpretation. Representational (Hedges's g = 0.24) and organizational (Hedges's g = 0.52) pictures had a significant and positive impact on response accuracy, but informational or multiple pictures across studies did not have a significant aggregate effect on an individual's response accuracy. These findings suggest that the multimedia effect in problem solving is diverse and limited by multiple boundary conditions. Further primary studies are needed to further investigate the multimedia effect in problem solving.