OBJECTIVE In one of the earlier hypotheses of the etiology of breast cancer (Korenman's 'oestrogen window' hypothesis (1981)), it was assumed that women with a later age at menopause have a longer period with irregular cycles preceding menopause than women with an early menopause. This assumption was tested in a prospective study. METHODS Subjects were 628 women, born between 1932 and 1941, who had participated in a breast cancer screening project in Utrecht, The Netherlands (the DOM-project) in 1982-1985, and who were still menstruating at that time. The women had filled out a questionnaire and a menstrual calendar every 2 years to determine their age at menopause prospectively. The women had not used oral contraceptives or medicines for menopausal complaints and had reached natural menopause by 1992. The median of the mean menstrual cycle length per woman and the median of the standard deviation of the mean menstrual cycle length per woman were plotted against number of years prior to menopause in three categories of age at menopause (44-49; 50-54; 55-59). RESULTS During the 9 years prior to menopause, women with a late age at menopause have a somewhat higher mean menstrual cycle length than women with a younger age at menopause (P = 0.0008). Cycle length variability in the 9 years prior to menopause is not statistically significantly different between the three categories of age at menopause (P = 0.16). CONCLUSIONS The assumption that women with a late age at menopause have a longer period with irregular cycles than women with an earlier age at menopause was not corroborated by our results.