We analyzed our experience with 64 infants with esophageal atresia (EA) and tracheoesophageal fistula (TEF), to determine the possibility of prediction and prevention of anastomotic complications (leak, stricture, and recurrent TEF). In most of the infants, the anatomical level of the fistula was documented preoperatively by bronchoscopy. The level of the fistula, in turn, correlated with the esophageal anatomy at thoracotomy, ie, carinal fistulas had a wide gap between esophageal pouches, whereas midtracheal or cervical fistulas had a minimal gap. Major anastomotic complications were defined as leak requiring reoperation, symptomatic strictures requiring four or more dilatations, or a recurrent TEF. The complication rates wre: leak (major and minor), 21%; major stricture, 15%; and recurrent TEF, 5%. Major complications occurred in 42% (11/26) of infants with wide gaps, compared with 8% (3/36) of infants with minimal gaps. Route of repair (transpleural or retropleural) made no difference in incidence of anastomotic complications. No infant died of an anastomotic complication. Survival was 100% for Waterston A and B infants, 83% for Waterston C, and 90% overall. Severe gastroesophageal reflux, requiring Nissen fundoplication, was more common among infants with wide gaps than those with minimal gaps (32% v 3%). The most important pathogenetic factor, present in 79% (11/14) of major anastomotic complications, was anastomotic tension, determined by the gap between esophageal pouches, and predicted by preoperative bronchoscopy. Thus the bronchoscopic finding of a carinal fistula signals the need for technical measures that may limit anastomotic morbidity, such as myotomy, patching the anastomosis, retropleural approach, or delayed repair. Assuming precise technique and gentle handling of tissues, the anatomy of the anomaly determines the anastomotic morbidity of EA and TEF.