Repeated measurement of ovarian steroids in saliva could provide an advantage in studies estimating long-term sex steroid exposure in premenopausal women, by reducing the measurement error associated with collection of serum or urine samples. We previously reported on characteristics of ultrasensitive RIAs adapted for extraction-free measurement of estradiol (E2) and progesterone (PG) in saliva. The purpose of the present study was to evaluate the consistency of E2 and PG levels in saliva in the same women across menstrual cycles, and to compare this with the variation observed between women. We also evaluated the effect of altering the number of consecutive daily samples considered and the method for locating a particular cycle day in relation to ovulation (day 0). Study participants included 12 healthy women who provided daily saliva samples for two consecutive, ovulatory menstrual cycles. A single midluteal serum sample was collected 7-8 days after detection of a luteinizing hormone (LH) peak in urine. We plotted individual cycle profiles and computed intraclass correlation coefficients (ICC) for various definitions of peak and cumulative daily hormone level. For peak PG, determined as the maximal running 3-day mean, ICC was 0.68. For cumulative PG, based on 8 consecutive cycle days (+2 to +9), ICCs were 0.72-0.76 when reverse dating LH peak or rise in salivary PG determined day 0. For E2, ICCs ranged from 0.74 to 0.79 by various dating methods for the 5 preovulatory days (-4-0), and from 0.85 to 0.92 for the 15 days about the center of the cycle (-6 to +8). With exclusion of just the first 5 days of the cycle, the ICC for E2 was 0.91. For both E2 and PG, selection of 5 or 7 days for the estimation of the midluteal mean level provided separation of within and between subject variance that was comparable with a LH-timed serum sample. These results indicate that daily saliva samples can be combined to clarify the interindividual differences in E2 and PG levels in premenopausal women, and that these interindividual differences may be greater than previously imagined.