To evaluate the association between 14 demographic variables and the loss of virginity in a specific sample of junior high school students in Chicago. Nine hundred seventy-six students in nine Chicago junior high schools, sixth through eighth grades, were given an anonymous behavior survey (the noncognitive assessment survey). Two separate logistic regression equations were used to determine the relative relationships of the demographic variables to self-reported virginity loss. Five variables were significantly associated with virginity loss in both regression equations. In rank order, they were gender, ethnic group, pubertal status, suicidal ideation, and sibling number (adjusted odds ratio 13.3, 4.57, 3.38, 1.93, and 1.24, respectively.) Nine variables did not have a consistent relation with early sexual activity: church attendance, religious affiliation, grade average, housing status, marital status of natural parents, self-esteem, sex education knowledge, school attendance, and chronologic age. These results call into question two widely held assumptions that form the foundation of many teen pregnancy prevention efforts. First, although many believe that sex education courses can affect behavior, we found no link, either positive or negative, between knowledge of reproductive biology and age of first intercourse. Second, self-esteem level was not associated with age of first intercourse. The variables that did seem related to early sexual activity do not lend themselves to easy manipulation. Our findings suggest that current school-based efforts to alter teen pregnancy rates and sexual behavior are unlikely to succeed.