Correction of judgments of people of different ethnicities and sexes was explored by exposing participants to cues to the risk of making biased judgments (from explicit warnings to subtle hints). In three experiments, a three-way interaction was revealed, where the effect of a cue to bias varied as a function of both the ethnicity and sex of the target person. Some targets (White males) were generally rated less favorably when judges were reminded of bias, whereas other targets (Black males, Middle Eastern males and White females) were generally rated more favorably, indicating bidirectional correction. Finally, a normative account of the results was explored. In a pattern consistent with the experimental results, it was considered more important to avoid overrating White men than all other groups, and more important to avoid underrating all other groups than White men. The results are discussed in relation to theories of correction and intergroup bias.