Summary No previous controlled studies of ovarian germ cell tumours have been reported; however the tumour is similar to germ cell testicular cancer in terms of histology, age-specific incidence rates (i.e. highest rates in young adulthood), and secular trends of increasing incidence. The investigation was designed to determine if maternal hormonal factors which have been found to increase the risk of testis cancer in male offspring are also risk factors for the ovarian tumour. The analysis is based on 73 cases diagnosed before age 35 and 138 age-race matched controls. The cases were identified by tumour registries in Los Angeles and Seattle Ovarian germ cell cancer is quite rare and no previous controlled epidemiologic studies have been reported. The similarity of these tumours to germ cell testicular cancer in terms of histology (Teilum, 1976), age-specific incidence rates highest in the young adult age range (Walker et al., 1984; Weiss, 1982), and secular trends of increasing incidence (Walker et al., 1984) Angeles case series and 15 were from Seattle.For each case an attempt was made to identify two ageand race-matched controls using friends or neighbours so as to approximate a socioeconomic match. For cases under age 18 at diagnosis, the controls were selected from friends or neighbours of the case in the year prior to diagnosis. For cases over age 18 at diagnosis, controls were selected from among the case's friends or neighbours at the time of graduation from high school, in order to approximate the social class of the case's mother rather than of the case herself.The procedure for obtaining the potential friend control names was to ask the case mother for a list of first names of friends of the case from the appropriate time period at the beginning of the interview and to have her rank them according to closeness in friendship and age to the case. Then at the end of the interview, permission was requested to contact these friends, beginning with the first two ranked friends who were within two years of the case's age and of the same race.Neighbourhood controls were sought when two friend controls could not be identified. The neighbourhood controls were selected by following a specified walking pattern in the blocks around the case's residence. Residents of each unit were surveyed until a race and age (±5 years) match was located.Once the control was selected, permission was sought to interview her mother. At least one control mother interview was completed for 73 of the 74 cases whose mothers were interviewed, and two or more were obtained for 54 cases. For 20 cases, only neighbourhood control mothers were used, and in 8 cases both friend and neighbourhood control mothers were used. In total, 138 control mothers were interviewed, 93 from friends and 45 from neighbourhood residents. Three potential friend controls and 10 qualifying neighbourhood controls refused to participate.