While there is ample theoretical and empirical evidence detailing which conditions benefit learning from one's own errors, the evidence on learning from others' errors has not yet been synthesized. In this meta-analysis, we examine the overall impact of erroneous examples on learning and the effects of potential moderating variables based on a novel framework. Following the robust variance estimation method, we synthesized findings from 42 papers (177 effect sizes) comparing erroneous examples with correct examples or problem-solving in experimental studies. The results revealed a statistically significant but weak effect of erroneous examples on learning (g = .136). Further analysis indicated a statistically significant moderating effect of the design of error-explanation activities. Specifically, providing self-explanation prompts or instructional explanations enhanced learning from erroneous examples more than not providing any error explanations. Our findings draw attention to the design of error explanation activities as well as several areas for future research.