To evaluate the effects of pericardial effusion on the ECG, we compared clinical, echocardiographic and ECG findings in 459 patients. The prevalence of echocardiographic effusion ranged from 1% (1/79) among normal subjects, to 28% (32/114) among patients with valvular disease, 30% (27/90) in patients with hypertension, and 86% (18/21) in patients with pericardial disease. No relationship existed between left ventricular function and the prevalence of effusion, but a strong inverse relationship was found between LV function and effusion size (r = -0.63, p less than 0.01). Small and moderate sized effusions had a progressive damping effect on ECG voltage, displacing the regression lines between Sokolow -Lyon voltage and left ventricular mass downward by 1.2 and 4.4 mm respectively. Standard ECG criteria for low voltage (leads I, II, III each less than 0.5 mV, or V1 to V6 each less than 1.0 mV) were extremely insensitive for detection of effusions (12%), although highly specific (94%). Other ECG criteria which improved sensitivity resulted in an unacceptably high prevalence of false-positive diagnoses of pericardial effusion. Thus, echocardiographic effusions occur in only 1% of normal subjects but in more than 25% of patients with hemodynamic loading conditions, with a strong relationship between worsening left ventricular function and increasing effusion size. In contrast to the close relationship between echocardiographic pericardial effusions and clinical findings, low electrocardiographic QRS voltage is a weak predictor of the presence of pericardial effusion.