We compared contrast angiography with three techniques of quantitating valvular regurgitation from radionuclide ventriculograms in 70 patients: 45 with documented regurgitation graded 1-4+, and 25 without regurgitation. The radionuclide "regurgitant index" (ratio of L to R ventricular stroke counts) was measured from fixed end-diastolic regions of interest (method A), from separate end-diastolic and end-systolic regions of interest (method B), and from a "stroke-volume image" (method C). Sensitivities for detecting 1+ or more regurgitation were: method A = 57.8%, method B = 37.8% and method C = 62.2%. Sensitivities for detecting 2+ or more regurgitation were: method A = 74.2%, method B = 54.8%, and method C = 77.4%. All methods are greater than 97% specific. Interobserver coefficients of variability were: method A = 9.1%, method B = 19.2%, and method C = 5.4%. The sensitivity of each method was improved when left-ventricular ejection fractions were greater than 0.35. No method consistently differentiated between 2+, 3+, and 4+ valvular regurgitation.