Between December 1, 1993, and December 1, 1996, we tested 4,411 isolates of Enterococcus sp. at gentamicin concentrations of 500 micrograms/mL and 2000 micrograms/mL using agar dilution to phenotypically categorize them into 3 groups: those with a MIC < or = 500 micrograms/mL (n = 3,132; 71%); a MIC > 500, but < or = 2000 micrograms/mL (n = 441; 10%); and those with a MIC > 2000 micrograms/mL (n = 838; 19%). Ten unique strains of each phenotype were tested to determine which gentamicin concentration was the best in vitro predictor of synergy with ampicillin. Testing was done by a time-kill method using clinically achievable levels of ampicillin and gentamicin. We found that for the gentamicin MIC < or = 500 micrograms/mL group, 7 of 10 isolates demonstrated synergy with ampicillin as manifested by a > or = 2 log10 increase in killing versus the effect of ampicillin alone (at 1/2 the MIC for ampicillin). In the group sensitive to a gentamicin MIC range between > 500 and < or = 2,000 micrograms/mL, none of the 10 isolates demonstrated synergy. Absence of synergy was also found in the group resistant to 2,000 micrograms/mL of gentamicin. Assessment of eight additional enterococcal isolates with reduced sensitivity to ampicillin (MIC from 32-256 micrograms/mL) found no correlation between gentamicin sensitivity at 500 micrograms/mL and any in vitro test for synergy, nor with clinical therapeutic outcome. Gentamicin at 2 micrograms/mL combined with ampicillin was as effective in enhancing killing as a higher level of 4 micrograms/mL. These findings validate the current NCCLS guideline for predicting synergistic activity against enterococci in strains with usual susceptibility to ampicillin, and suggest that a therapeutic level less than maximal recommended dosing is sufficient when using gentamicin in this setting.