Between 1 January 1975 and 31 December 1982, 111 patients with atrioventricular septal defect underwent surgical repair. Of these, 43 had the complete, 11 the intermediate, and 57 the partial form. The postoperative mortality rate was 37%, 9%, and 6% respectively. To determine which factors were independently responsible for the operative risk multivariate analysis of the surgical mortality was applied simultaneously to all three forms of the malformation. The form of defect, although strongly influencing the natural history and clinical presentation, was not by itself an operative risk factor. The risk was related primarily to failure to obtain a well functioning atrioventricular valve, to the presence of left ventricular dominance, to the degree of pulmonary vascular resistance, and, finally, to the technique of reconstructing a two leaflet left atrioventricular valve. The small size of the patient was also a significant incremental risk factor, but if the other factors were not unfavorable good results could be achieved in small infants with mortality rates less than 10%. Multivariate analysis showed that severe postoperative left atrioventricular valve malfunction was related to the technique used to reconstruct a "normal" two leaflet left atrioventricular valve. These findings support the policy of reconstructing the left atrioventricular valve as a three leaflet valve. Nevertheless, the implicit beneficial effect of this technique has not as yet proved to be statistically significant.