We performed a retrospective analysis to ascertain how accurately women who believe that they have regular menstrual cycles estimate the length of their actual cycles. Data were extracted from a chart review of subjects from three different studies of barrier contraceptives. Subjects were between 18 and 40 years of age and reported "regular" prestudy menstrual cycles with a consistent cycle length between 21 and 35 days. Participants prospectively recorded their menses for the up to 30 weeks. Each subject's estimated cycle length was compared to the average of her actual cycle lengths and the range and variability in each individual's cycle length was calculated. A total of 786 cycles from 130 women who recorded 4 or more cycles were analyzed. The averages of the participants' estimated cycle lengths was similar to the prospective averages of their actual cycle lengths (29.0 +/- 2.7 days vs. 29.1 +/- 3.5 days, respectively, p = 0.8). Forty-six percent of all subjects had a cycle range of 7 days or more, and 20% had a cycle range of 14 days or more. The average length of menses was 5.2 +/- 1.0 days. When evaluating only women with cycle lengths from 21 to 35 days, the average length of menses was positively associated with the average actual cycle length (p = 0.04). Although the average of a woman's menstrual cycles compares favorably to her impression of her cycle length, the variability in menstrual cycle lengths is significant. This variation may have clinical impact on contraceptive practice, contraceptive research studies and pregnancy-related care.