We undertook this study to assess the accuracy of the clindamycin-erythromycin disk approximation test (D-test) for detection of inducible clindamycin resistance in Staphylococcus spp. One hundred sixty-three Staphylococcus aureus and 68 coagulase-negative Staphylococcus (CoNS) spp. which were erythromycin nonsusceptible but clindamycin susceptible were tested using the D-test performed at both 15-mm and 22-mm disk separations and compared with genotyping as the "gold standard." The rate of inducible clindamycin resistance was 96.3% for S. aureus and 33.8% for CoNS spp. The sensitivities of the D-tests performed at 15 mm and 22 mm were 100% and 87.7%, respectively, and specificities were 100% for both. The use of 22-mm disk separation for the D-test to detect inducible clindamycin resistance results in an unacceptably high very major error rate (12.3%). All isolates with false-negative results harbored the ermA gene, and the majority were methicillin-resistant Staphylococcus aureus. False-negative results were associated with smaller clindamycin zone sizes and double-edged zones. We recommend using a disk separation distance of </=15 mm. There is wide geographic variation in the rates of inducible clindamycin resistance, and each laboratory should determine the local rate before deciding whether to either perform the D-test routinely or else report that all erythromycin-resistant S. aureus isolates are also clindamycin resistant.