It is well established that gesture facilitates learning, but understanding the best way to harness gesture and "how" gesture helps learners are still open questions. Here, we consider one of the properties that may make gesture a powerful teaching tool: its temporal alignment with spoken language. Previous work shows that the simultaneity of speech and gesture matters when children receive instruction from a teacher (Congdon et al., 2017). In Study 1, we ask whether simultaneity also matters when children themselves are the ones who produce speech and gesture strategies. Third-graders (N = 75) were taught to produce one strategy in speech and one strategy in gesture for correctly solving mathematical equivalence problems; they were told to produce these strategies either simultaneously (S + G) or sequentially (S[right arrow]G; G[right arrow]S) during a training session. Learning was assessed immediately after training, at a 24-h follow-up, and at a 4-week follow-up. Children showed evidence of learning and retention across all three conditions. Study 2 was conducted to explore whether it was the special relationship between speech and gesture that helped children learn. Third-graders (N = 87) were taught an action strategy instead of a gesture strategy; all other aspects of the design were the same. Children again learned across all three conditions. But only children who produced "simultaneous" speech and action retained what they had learned at the follow-up sessions. Results have implications for why gesture is beneficial to learners and, taken in relation to previous literature, reveal differences in the mechanisms by which doing versus seeing gesture facilitates learning.