From April 15 through June 15, 1999, a total of 660 yeast isolates were collected from 22 hospitals in Taiwan to investigate factors determining the accuracy of yeast identification. The germ tube test was the method most frequently used by hospitals for yeast identification, followed by the API-32C, cornmeal agar window test, and assimilation method. All of the submitted isolates were re-speciated in the National Health Research Institutes laboratory. The frequencies of inconsistent identification of isolates between hospitals and the National Health Research Institutes laboratory varied with the location and the type of hospital. The sensitivity and specificity of the germ tube test were 95% and 98.6%, respectively. This study showed that hospitals using the germ tube test as the first step in yeast identification had fewer inconsistent identifications of isolates than those using other methods. The VITEK Yeast Biochemical Card and API-32C had a sensitivity of 92.6% and 98.3%, respectively. No single method consistently identified all yeast isolates. Thus, every laboratory should have at least 2 methods available for yeast identification.