One hundred fourteen coronary stenoses were quantified before and after percutaneous transluminal coronary angioplasty (PTCA) using a semi-automated digital system. The values obtained were considered as standard for comparison with visual estimation by the PTCA operator as well as by independent consensus-reading. The measured percent stenosis was 62.7 +/- 13.7% before and 27.7 +/- 12.4% after angioplasty. Before PTCA, the operator consistently overestimated stenosis severity (87.8 +/- 8.5%, P < 0.0001) and consensus-reading reduced but did not eliminate this overestimation (78.0 +/- 12.3%, P < 0.05). The error in visual estimation was inversely correlated with the measured degree of stenosis: coefficients were -0.79 (P < 0.0001) and -0.51 (P < 0.0001) for operator and consensus-readers, respectively. After PTCA, the operator underestimated the residual stenosis (21.2 +/- 9.9%, P < 0.0001) but there was no systematic bias by consensus-reading (29.4 +/- 12.0%, NS). Again the error in visual estimation was inversely correlated with the measured degree of residual stenosis: coefficients were -0.76 (P < 0.0001) and -0.58 (P < 0.0001) for operator and consensus-reading, respectively. In conclusion, the operator overestimates lesion severity before and underestimates moderate residual stenoses after PTCA, a problem only partially corrected by independent consensus-readers.