Between September 1984 and June 1985, a total of 75 adolescent girls, 35 vegetarians residing in a Seventh-Day Adventist school and 40 nonvegetarians residing in a private non-Adventist boarding school, underwent measurement of their plasma hormone levels in the follicular and luteal phase of their menstrual cycles as well as dietary intake measured by 3-day food records, medical history, height, and weight. There were no significant differences between vegetarians and nonvegetarians in average age of the girls, weight, body mass index, age at menarche, years since the onset of menstruation, or percentage of girls with ovulatory cycles. Vegetarian girls had significantly higher levels of log follicular estradiol [2.00 +/- 0.27 (SD) versus 1.85 +/- 0.27 pg/ml, P less than or equal to 0.05] and luteal dehydroepiandrosterone sulfate (DHS) (1.88 +/- 0.71 versus 1.45 +/- 0.80 microgram/ml, P less than or equal to 0.05) than nonvegetarian girls. Follicular DHS was higher in vegetarians than in nonvegetarians (1.72 +/- 0.79 versus 1.45 +/- 0.95 microgram/ml), but the difference was not significant. The differences in follicular and luteal DHS, but not the difference in log estradiol, were significant (P less than or equal to 0.05) after controlling for ovulation, smoking, and alcohol intake with multivariable regression analysis. There were no significant differences in testosterone or in percentage free estradiol levels between vegetarians and nonvegetarians. Smoking was significantly associated with follicular and luteal DHS and with percentage free follicular estradiol, while alcohol use was significantly and inversely associated with percentage free follicular estradiol after controlling for other variables. The implications for breast cancer risk are discussed.