The objectives of the study were (1) to determine the interobserver variation in interpretation of abdominal radiographs in children with clinically suspected intussusception, and (2) to determine the diagnostic value of abdominal radiographs in these patients. One hundred and eighty-two plain abdominal radiographic examinations (AXR) performed in children with clinically suspected intussusception were reviewed blind to the clinical history and findings of air enema. The presence or absence of nine AXR signs relevant to intussusception was documented. Each AXR was categorized as equivocal, positive or negative for intussusception, with the aim of achieving no false negatives. Interobserver variation in the identification of AXR signs and radiologic diagnoses was calculated using the kappa statistic for 60 cases assessed independently by three observers. Using the findings of air enema as gold standard, the prevalence of AXR signs in all patients with (60) and without (122) intussusception was determined and their diagnostic values calculated. The best observer agreement was for the presence of sparse small bowel gas (supine, k = 0.68) and the worst for the presence of cecal gas (erect, k = 0.18). All three observers agreed intussusception to be present or absent in only 7 of 60 cases and the majority agreement was equivocal in more than half. Overall agreement between observers for the diagnosis of intussusception was k = 0.30. The best positive predictors of intussusception were the soft tissue mass and sparse large bowel gas, with likelihood ratios of 3.9 and 2.5. Cecal feces predicted against intussusception, likelihood ratio 0.11. AXR was equivocal in 53%, positive in 21% and negative in 26%. Where a firm radiographic diagnosis was made, the diagnostic accuracy of AXR was 84%.(ABSTRACT TRUNCATED AT 250 WORDS)