Three groups of isolates, each comprising four isolates of Candida species, were selected for their diversity of susceptibilities to amphotericin B, flucytosine, or ketoconazole. The isolates were distributed in duplicate and in blinded fashion to three laboratories where a total of eight procedures were performed for each drug. From the decoded results, intralaboratory variability among replicate determinations was found usually to fall within a fourfold range. Interlaboratory variation, however, was 16-fold or greater for all isolates, ranging to 50,000-fold differences for some isolates. Relative susceptibilities of isolates within each method could be determined in 11 of the 24 drug-method combinations and agreed with the reference rank order in all but one instance. Our findings underscore the lack of agreement among laboratories for the susceptibility testing of yeasts but indicate that such differences could likely be resolved by standardization without loss of clinical value.Susceptibility testing of fungi, including yeasts such as Candida species, is unstandardized. Numerous broth and agar dilution methods and the application of these methods in specific laboratories under differing test conditions may produce significantly divergent test results (7). Little information has been published concerning differences in results produced by broth and agar dilution procedures. However, a recent interlaboratory comparison testing the reproducibility of a single broth dilution method found an unacceptable discordance of results (3). This lack of agreement undermines the interpretation of in vitro results and raises the question of which method correlates best with the results of treatment.It is possible for the susceptibility of an isolate, relative to others, to be similar even when there exist wide differences in endpoint results when tested under different conditions. That a consistent rank order of susceptibility might exist for yeasts has been suggested in previous reports (3,4,8,12). Unfortunately, in some of those studies identical results were produced by many isolates, and in others values were beyond the highest or lowest drug dilutions tested. Both circumstances restricted detailed determination of rank order. However, if relative sensitivities were constant, even by widely diverse methodology, then each method could be expected to produce equivalent correlations with clinical responses, and the standardization process would thus be significantly simplified.