The relative sensitivities of and interrelations between different measurements of diastolic function were studied in 50 patients with left ventricular hypertrophy diagnosed on anatomical grounds. Isovolumic relaxation time, the interval from minimum cavity dimension to mitral valve opening and relative dimension increase during this period, and the peak rate of dimension increase and wall thinning during rapid ventricular filling were measured by digitised M mode echocardiography. The relative heights of peak early diastolic and atrial velocities (a/E) and the time for decline of early diastolic velocity to half its peak value (velocity half time) were measured on continuous wave and pulsed Doppler and the relative height of the "a" wave was measured by apexcardiogram. All sets of values except those of the interval from minimum dimension to mitral opening were unimodally distributed, and all differed significantly from those in 20 age matched controls. The relative height of the "a" wave on the apexcardiogram (90% values were abnormal) was the most sensitive method of studying left ventricular diastolic function and peak rate of dimension increase was the least sensitive. Though none of the correlations was high, there were individual associations between peak rate of dimension increase, a/E, peak wall thinning rate, and velocity half time, and independently between delay in mitral valve opening and dimension change during this period. Other values seemed to be independent of one another, suggesting a different physiological basis. It is concluded that these various abnormal values do not reflect a single underlying disturbance of diastolic function. There are at least four possible discrete abnormalities: prolongation of isovolumic relaxation; incoordination during isovolumic relaxation; reduced rate of rapid filling; and an increase in the relative amplitude of the "a" wave probably caused by increased passive stiffness. These may be present singly or in combination in any patient.