Pick up any recent policy paper on women’s participation in science and you will find assurances that gender diversity enhances knowledge outcomes. Universities and science-policy stakeholders, including the European Commission and the US National Institutes of Health, readily subscribe to this argument (1⇓–3). But is there, in fact, a gender-diversity dividend in science? The data suggest that there is. Under the right conditions, teams may benefit from various types of diversity, including scientific discipline, work experience, gender, ethnicity, and nationality. In this paper, we highlight gender diversity (Fig. 1). Guided by key research findings, we propose the following “mechanisms for innovation” specifying why gender diversity matters for scientific discovery and what managers should do to maximize its benefits (Fig. 2). Encouraging greater diversity is not only the right thing to do: it allows scientific organizations to derive an “innovation dividend” that leads to smarter, more creative teams, hence opening the door to new discoveries. Fig. 1. When it comes to science collaborations, there’s ample data to suggest that gender diversity pays a substantial research and productivity dividend. Image courtesy of Dave Cutler (artist). Well-run, well-performing research teams have become increasingly crucial to the success of modern scientific investigations. Already, experimental research points to positive links between gender diversity and collective problem solving. In a study of group performance, Anita Woolley et al. (4) randomly assigned 699 participants to teams of varying sizes and asked them to solve a set of both simple and complicated tasks (e.g., visual puzzles, brainstorming, making collective moral judgments, and negotiating over limited resources). Through these experiments, the authors found evidence of a collective intelligence factor that predicts group performance better than the IQ of individual group members. Key components of this factor include the group members’ social perceptiveness and parity in conversational turn-taking. Furthermore, … [↵][1]1To whom correspondence should be addressed. Email: mwn{at}stanford.edu. [1]: #xref-corresp-1-1