Prolactin, a hormone indispensable for milk secretion, has been shown to enhance the development and growth of mammary tumors in rodents; however, its importance in human breast cancer is uncertain. Serum prolactin levels are known to fluctuate considerably under normal conditions, and lack of precision in the hormone measurements may have contributed to the largely negative findings in humans to date. The purpose of this study was to investigate the reliability of prolactin measurements in women using stored serum from an ongoing prospective study of breast cancer. Separate groups of postmenopausal and premenopausal women who donated multiple blood samples at approximately 1-year intervals were studied. The reliability of a single log prolactin determination, as measured by the intraclass correlation coefficient, was 0.76 for the postmenopausal women (95% confidence interval, 0.66-0.85) and 0.48 for the premenopausal women (95% confidence interval, 0.31-0.62). These findings suggest that a single measurement is sufficient to characterize the serum prolactin level of postmenopausal women for epidemiological research. For premenopausal women, however, multiple samples are desirable. Controlling for phase of the menstrual cycle does not appear to substantially improve the reliability of premenopausal measurements.