The aversive racism framework (S. L. Gaertner & J. F. Dovidio, 1986) suggests that bias against Blacks is most likely to be expressed by Whites when it can be explained or justified along non-racial grounds. The present experiment adopted a 2 (Evidence: admissible vs. inadmissible) × 2 (Defendant Race: White vs. Black) between subjects design, asking White participants, whose self-reported prejudice was assessed, to judge a legal case. As predicted, increased guilt ratings and longer sentencing recommendations were forwarded for the Black (vs. White) defendant only when DNA evidence linking the defendant to the crime had previously been ruled inadmissible. This result was not qualified by self-report racial attitudes. The implications for evidence inadmissibility in interracial contexts are considered, along with the repercussions of finding experimental evidence of aversive racism outside of North America. Copyright © 2005 John Wiley & Sons, Ltd.