The validity of recall of early menstrual characteristics is of interest because of their putative role in the etiology of breast cancer and other diseases. A retrospective follow-up of the Newton Girls Study (1965-1975) provided an opportunity to assess the accuracy and precision of recall of several early menstrual characteristics. In 1998-1999, 57 percent of the original 793 Newton Girls Study participants completed a mailed questionnaire to assess the accuracy of recall for age and body size at menarche, usual cycle length during the first 2 years, and age at regularity. Recalled and original age at menarche were highly correlated (r = 0.79, p < 0.001). The body mass index percentile at menarche was well correlated with recalled body size at menarche (r = 0.61, p < 0.001), but with some evidence of systematic bias. Overall, a woman's recall of menarcheal age and body size was better than recall of cycle length and occurrence of regularity. The failure to identify certain menstrual characteristics as exposures for subsequent disease may reflect limitations in the accuracy and precision of the recalled measures.