The selection of a group of patients based on a high value of a clinical or biological parameter leads to the finding of a tendency to a reduction of this value when remeasured, known as "regression towards the mean". The amplitude of this phenomenon is greater when the selected subjects are far from normal values and the intra-individual variability of the parameter under consideration is very high. Measurement of left ventricular mass is very affected by this statistical phenomenon. The authors undertook a prospective study to analyse the components of variability of repeated echocardiographic measurements of left ventricular mass and to quantify the expected effect of regression towards the mean in the follow-up of patients with left ventricular hypertrophy. Twenty-five randomly chosen subjects underwent 2 echocardiographic examinations at 2 week intervals: at each visit, the patient had two recordings and each recording was measured twice by the same "blinded" operator. Variance analysis showed intra-individual variability represented 30% of total variability, comprising only 2% for the measurements and 28% for the recordings and the visits. The importance of regression towards the mean was calculated with respect to the initial value of the left ventricular mass index: for example, when the left ventricular mass index was 150 g/m2, a spontaneous regression of 18 g/m2 can be expected at the next measurement. This phenomenon should be taken into consideration in the interpretation of longitudinal echocardiographic studies.