New meta-analysis on stereotype threat finds effect sizes are compromised by publication bias, much smaller when incentives (e.g., financial) are present, and smaller in high-stakes testing scenarios (which most approximate real world contexts).