A total of 17 radiographs from seven babies with clinical evidence of necrotizing enterocolitis were examined by six pediatric radiologits and four neonatologists from the Southern California area. The participants were asked to interpret each radiograph as to the presence or absence of necrotizing enterocolitis and to indicate whether pneumoperitoneum was present. A substantial degree of interobserver variability was observed, with only one of the 17 films yielding universal agreement. We propose that this interobserver variability in interpreting abdominal films of neonates with clinically suspected necrotizing enterocolitis may in fact contribute to the reported difference in incidence of this disease among centers.