Sixty-three endometrial biopsies were dated histologically by using the standard criteria on two separate occasions by the same observer. Overall, it was found that exact agreement occurred in 15 (24%), but disagreement of more than 2 days occurred in 6 (10%). The proportion of exact agreement in the first half of the luteal phase (32%) was found to be significantly higher (P less than 0.05) than that in the second half of the luteal phase (9%). In a separate part of the study, 27 women had two endometrial biopsies, each performed in a separate cycle. The within-subject between-cycle variation of the results of endometrial dating (exact agreement: 4%, disagreement of more than 2 days: 41%) was found to be significantly different from intraobserver variation (P less than 0.01 for both). The amount of intraobserver variation suggests that the traditional dating criteria are not precise enough to quantify corpus luteum function in the second half of the luteal phase, whereas the amount of within-subject between-cycle variation implies that the result of endometrial dating in one cycle cannot be used reliably to predict that of another cycle.