Recent research reported that general mental ability (GMA) predicted counterproductive work behavior (CWB), whereas some previous studies failed to find such a relationship. We tested occupational homogeneity of the sample and criterion measurement as two potential explanations for these inconsistencies. Study 1 replicated major design features of one previous study, which found no GMA–CWB relation in a heterogeneous sample, with occupationally homogeneous groups. Results confirmed previous null findings, indicating no effect of sample homogeneity. In Study 2, using a controlled laboratory setting, GMA was again unrelated to self-reported CWB, but partially predicted observed CWB negatively. Combined findings suggest that GMA is consistently unrelated to CWB self-reports but may predict objectively measured CWB independently of the likelihood of being caught. Copyright © 2009 John Wiley & Sons, Ltd.