The objective of this study was to determine the efficacy of using milk estradiol-17beta (E(2)) sampled during regular milking periods as a predictor of estrus in dairy cows. Twenty-three primiparous Holstein cows received radiotelemetric transmitters on day 16 of the estrous cycle (day 0 = estrus) for continuous monitoring of behavioral estrus. Milk and blood samples were collected every 12h at each milking, from day 18 of the estrous cycle until the fourth milking after the onset of estrus, for radioimmunoassay of E(2). Onset of estrus was indicated by the first standing event identified by radiotelemetry. Ultrasound examination of ovaries was conducted daily in a subset of cows (n = 17) from day 18 until ovulation was confirmed. Statistical analysis involved utilization of Pearson correlation to observe any association of mean plasma and milk E(2) concentrations. Intervals from highest measured plasma and highest measured milk E(2) until the first expression of standing behavior and intervals from highest measured plasma and highest measured milk E(2) until ovulation were compared using Student's t-test. Repeated measures were utilized to evaluate the effect of day and time on milk E(2) concentration. Chi-square procedures were utilized to detect differences in actual time of onset of estrus compared to the expectation of random time of onset of estrus throughout the 24 h day. Mean duration from highest measured milk E(2) until onset of standing behavior was 21+/-3.7h and until ovulation was 46.7+/-5.3h. Mean duration from onset of standing behavior until ovulation was 26.4+/-4.2h. Mean milk E(2) concentrations increased (P<0.01) from the fourth milking period before estrus to the milking period immediately before the first expression of standing behavior, followed by a reduction (P<0.01) in mean concentrations at the milking immediately after the onset of estrus.