This paper describes a non-invasive, self-measured procedure by which the precancerous breast can be distinguished from the normal breast. The method involves wearing a specially designed thermometric brassiere for 90 min each evening at home through one menstrual cycle. Profiles of progesterone through the cycle, obtained from daily saliva sampling, and determination of the steroid content by radiommunoassay, are made to allow the status and calendar date timing of the luteal phase to be established. Thus, cycles can be synchronised across subjects. In this study, two types of breast were compared: 50 normal breasts and 41 age-matched precancerous breasts. Differences between the groups were striking in terms of amplitude, phasing and average temperature during the luteal heat cycle. When these parameters and others were used as predictors in a linear discrimination and/or neural net analysis, a sensitivity and specificity of >90% was achieved.