The authors present a method for defining the inception of perimenopause that is based on self-reported data. The study sample (n = 1,550) was obtained from a 5-year longitudinal study of 2,569 Massachusetts women aged 45-55 years that began in 1981. The definition was derived from the ability of responses to questions concerning timing of the last menstrual period, menstrual regularity, and presence of menopausal symptoms (hot flashes/sweats) to predict menopause 3 years later. The two items that best defined the inception of perimenopause were 3-11 months of amenorrhea and increased menstrual irregularity for those without amenorrhea. This definition can easily be used in large epidemiologic investigations.