Frequently, indexes of systolic and diastolic left ventricular (LV) function obtained from radionuclide angiography are computed after the LV time-activity curve has been temporally smoothed. This smoothing process may introduce important systematic errors into the analysis. To assess this potential effect, high temporal resolution time-activity curves (20 ms/point) were obtained in 113 normal subjects, 175 patients with coronary artery disease and 171 patients with hypertrophic cardiomyopathy. The curves were then subjected to 0-, 3-, 5-, 7- and 9-point temporal smoothing. In each group, increased smoothing progressively and consistently underestimated ejection fraction by up to 5% (p less than 0.001) and peak ejection rate by up to 14% (p less than 0.001). A greater effect on peak filling rate was observed: 5-point and 9-point smoothing reduced peak filling rate by 10% and 23% in normal subjects, 3% and 10% in patients with coronary artery disease and 7% and 15%, respectively, in patients with hypertrophic cardiomyopathy (all p less than 0.001). These errors were compounded further when the same data obtained at lower temporal resolution (40 ms/point) were analyzed: 5-point and 9-point smoothing resulted in underestimation of peak filling rate by 20% and 46% in normal subjects, 13% and 43% in coronary artery disease and 16% and 34% in hypertrophic cardiomyopathy. The underestimation was not uniform, and its magnitude varied considerably among individuals in each of the 3 groups. Thus, smoothing of LV time-activity curves may result in significant systematic errors in computation of indexes of LV systolic and diastolic function, especially in data with poor temporal resolution. These concepts apply to other imaging methods, such as magnetic resonance imaging and cine-computed tomography, that assess LV function from the LV volume curve. Although ejection fraction is affected only mildly by these errors, both peak filling rate and peak ejection rate may be severely underestimated.