A meta-analysis was undertaken to reexamine near- and far-transfer effects following working-memory training and to consider potential moderators more systematically. Forty-seven studies with 65 group comparisons were included in the meta-analysis. Results showed near-transfer effects to short-term and working-memory skills that were sustained at follow-up with effect sizes ranging from g = 0.37 to g = 0.72 for immediate transfer and g = 0.22 to g = 0.78 for long-term transfer. Far-transfer effects to other cognitive skills were small, limited to nonverbal (g = 0.14) and verbal (g = 0.16) ability and not sustained at follow-up. Several moderators (e.g., duration of training sessions, supervision during training) had an influence on transfer effects, including far-transfer effects. We present principles for how best to improve working memory through training in the narrow-task paradigm and conjecture how best to improve basic cognitive functions in complex activity contexts.