This meta-analytic study provides a systematic statistical synthesis of the effects of output tasks on second or foreign incidental vocabulary learning. A total of 12 studies were included in this meta-analysis. Five mediator variables were examined: design quality, types of output task, time on task, genres of text, and text-target word ratios. Results show that language learners who completed an output task outperformed those who only read a text. Results also support the involvement load hypothesis: Language learners who performed a task with a higher degree of involvement load, gained more vocabulary. Studies with high and medium levels of design quality were more likely to detect statistically significant differences among groups with different output tasks, compared to studies with a low level of design quality. Furthermore, results indicate that time on task had positive effects on vocabulary learning. Learners who read a combination of expository and narrative text gained more vocabulary than those who only read either an expository or narrative text. Learners who read a text with text-target word ratios of less than or equal to 2% did not learn significantly more vocabulary than those who read a text with a ratio of 2% to 5%. (Contains 5 tables and 1 figure.)