A major new project involving 865 researchers has assessed the replicability of 21,000 social and behavioral science studies. Results published in Nature aim to strengthen trust in scientific evidence.