New rules for quality assurance provoked a comparison of effects of two approaches used concurrently for 14 years. In an incidental approach, a multidisciplinary conference reviewed all postoperative complications as they occurred and attributed each to one of six causes. Remedies were instituted and data were filed. In a statistical approach, death and complication rates were computed annually and compared with previous years' rates and with rates reported to Congress as national norms. Statistics suggested acceptable quality in each specialty but calculations were tedious and differences achieved significance too rarely or too slowly to identify problems, protect patients, and improve care. The incidental approach was popular and produced immediate improvements in patient care. Conferees attributed one half of complications to errors. Frequent acknowledgment of susceptibility to error may contribute to the safety and quality shown by our statistics.