A total of 49 type and neotype isolates and 32 clinical isolates of the anamorph genus Candida and related teleomorph genera were obtained from different culture collections and clinical laboratories. Isolates were subjected to two phenotypic methods of identification, Vitek yeast biochemical card (YBC) and API ID 32C, both based on carbohydrate assimilation, and one genotypic method, PCR fingerprinting, based on the detection of DNA polymorphisms between minisatellite-specific sequences with the primer M13 (5' GAGGGTGGCGGTTCT 3'). The correct identification of a strain at the Centraalbureau voor Schimmelcultures was used as the gold standard for the identification of an isolate. When the study was restricted to species included in the respective biochemical databases, the Vitek YBC and API ID 32C systems performed adequately with positive identification rates of 87.3 and 76.8%, respectively. When uncommon species were added to the study, several of which are not included in the databases, the identification efficiencies were 76.5 and 77.5%, respectively. By comparison, all isolates were correctly identified by PCR fingerprinting, with 63 reference species profiles in the databank. Sufficient polymorphisms among the total set of banding patterns were observed, with adequate similarity in the major patterns obtained from a given species, to allow each isolate to be assigned unambiguously to a particular species. In addition, variations in minor bands allowed for differentiation to the strain level. PCR fingerprinting was found to be rapid, reproducible, and more cost-effective than either biochemical approach. Our results provide reference laboratories with an improved identification method for yeasts based on genotypic rather than phenotypic markers.