Although both broth microdilution (BMD) and disk diffusion (DD) are listed by NCCLS as acceptable methods for testing Acinetobacter spp. for antimicrobial susceptibility, few studies have compared the results generated by the two methods. We tested 196 isolates of Acinetobacter spp. from nine U.S. hospitals and from the Centers for Disease Control culture collection by using BMD and DD and clinically appropriate antimicrobial agents. Categorical results for amikacin, ciprofloxacin, gatifloxacin, gentamicin, imipenem, levofloxacin, meropenem, tobramycin, and trimethoprim-sulfamethoxazole were comparable for the two methods: there was only one very major (VM) error, with tobramycin, and only one major (M) error, with meropenem, when DD results were compared with BMD results. However, VM errors were frequent with the beta-lactams and beta-lactam-beta-lactam inhibitor combinations, while M errors were often observed with tetracyclines. For BMD, tests frequently exhibited subtle growth patterns that were difficult to interpret, especially for beta-lactams. If subtle growth (i.e., granular, small button, or "starry" growth) was considered positive, error rates between BMD and DD were unacceptably high for ampicillin-sulbactam (VM error, 9.8%; minor [m] error, 16.1%), piperacillin (VM error, 5.7%; m error, 13.5%), piperacillin-tazobactam (VM error, 9.3%; m error, 12.9%), ceftazidime (VM error, 6.2%; m error, 11.4%), cefepime (VM error, 6.2%; m error, 13.0%), cefotaxime (m error, 21.2%), ceftriaxone (m error, 23.3%), tetracycline (M error, 11.4%; m error, 32.1%), and doxycycline (M error, 2.6%). When subtle growth patterns were ignored, the agreement still did not achieve acceptable levels. To determine if the problems with BMD testing occurred in other laboratories, we sent frozen BMD panels containing beta-lactam drugs and nine isolates to six labs with experience in performing BMD and DD. Among these laboratories, cefepime MICs ranged from < or =8 to > or =32 microg/ml for four of the nine strains, confirming the problem in interpreting BMD results. Discrepancies between the categorical interpretations of BMD and DD tests were noted primarily with cefepime and piperacillin, for which the BMD results were typically more resistant. Clinical laboratories should be aware of these discrepancies. At present, there are no data to indicate which method provides more clinically relevant information.