Methods were developed to assess potential adverse effects of semiconductor fabrication work in a prospective study of reproductive health. All women aged 18-44 years who worked in seven silicon-wafer fabrication sites in five companies and a frequency-matched sample of women in nonfabrication jobs were included. Among 3,480 selected for screening, 2,639 (75%) completed a self-administered questionnaire to identify women at risk of pregnancy. Among the 739 (28%) eligible women, 481 (65%) completed baseline interviews and 402 (83.6%) completed at least one menstrual cycle of follow-up by providing daily diaries and daily urine samples. Menstrual cycle characteristics were assessed from questionnaires and diaries. Urine samples were assayed for reproductive hormones to identify conceptions. The usual cycle length recalled at baseline was 28 days. The mean cycle length (MCL) recorded in diaries was 29 days, with greater dispersion than at baseline. The median of the MCL from diary data was 28 days for women reporting regular cycles at baseline but 34 days for women reporting irregular cycles at baseline, and the median standard deviation in cycle length per woman was 2.5 days and 7.5 days, respectively. The prospective method, while expensive and labor intensive, showed good compliance. Nevertheless, recall also provided reasonably accurate estimates and distinguished women with regular and irregular cycles.