Adding illustrations to texts might improve students' science achievement. This meta-analysis of 121 effect sizes from 63 studies of 7,621 students across five decades determines both the overall effect and moderators that account for differences across studies. Our random-effects model shows a positive effect of adding illustrations to texts on science achievement (g = 0.232, k = 121, 95% CI [0.166, 0.298]). Added illustrations showed a stronger effect in (a) countries with higher national incomes, (b) risk accepting cultures, (c) short-term orientation cultures, (d) higher grade levels; (e) organizational illustrations, with effects tapering in explanative, representational, and incongruent ones; (f) flow or pictorial diagrams, (g) black-and-white illustrations; (h) cause-and-effect or comparison texts; or (i) tests of understanding and transfer. Test media (paper vs. computer) did not moderate this effect. These findings suggest that teachers can wisely choose and add illustrations to text to aid their students' learning.