OBJECTIVE --To examine rates of heterosexual transmission of human immunodeficiency virus (HIV) and associated risk factors and to determine the relative efficiency of female-to-male and male-to-female transmission. METHODS --Survey of infected individuals and their heterosexual partners recruited since 1985. METHODS --Participants were recruited from various HIV counseling and testing sites throughout California but were generally interviewed and tested in their homes. METHODS --Data from 379 couples at entry to the study are reported: 72 male partners of infected women and 307 female partners of infected men. The infected index case had a well-established source of risk; couples were eliminated if the direction of transmission could not be established. The majority of couples were monogamous since 1978, white, and in their 30s. Most partners did not know their serostatus at entry into the study. METHODS --HIV serostatus in the exposed sexual partner. RESULTS --We observed one probable instance (1%) of female-to-male transmission compared with 20% transmission rates in the female partners of infected men. All couples were sampled in the same way. Male index cases were more likely to be symptomatic than female index cases. CONCLUSIONS --The odds of male-to-female transmission were significantly greater than female-to-male transmission. The one case of female-to-male transmission was unique in that the couple reported numerous unprotected sexual contacts and noted several instances of vaginal and penile bleeding during intercourse.