One-hundred and five Candida albicans isolates from various anatomic sites of 28 patients, obtained at the onset of two consecutive episodes of well-documented recurrent vulvovaginitis, were typed by methods relying on physiologic or genomic markers. The isolates represented a wide variety of types, and neither a single biotype nor genotype was associated with recurrent vaginitis or a particular body site. Patients generally carried similar strains at various anatomic sites that persisted over time. Genomic methods indicated an 86% rate of relapse, which suggested that most recurrent vaginal infections are of endogenous origin. A similar evaluation with biotyping methods was inconclusive because of a lack of reproducibility, resulting from clonal variation or switching, and difficulties in establishing the number of phenotypic tests necessary to distinguish between identical and different strains. Therefore, Southern hybridization was considered the ideal reference method to study the epidemiology of C. albicans infections.