Colonoscopy is used in the differential diagnosis of inflammatory bowel disease but its accuracy and the "weight" of the various endoscopic signs have not been assessed. In a prospective study 357 patients with 606 colonoscopies, in whom the endoscopic appearances were those of ulcerative colitis, Crohn's colitis, or indeterminate colitis, were followed-up for an average period of 22 mo. A final, definite, endoscopy-independent diagnosis was reached by means of autopsy, surgery, or histology on biopsy in 71% of patients. Accuracy of colonoscopy was 89%, with 4% errors and 7% indeterminate diagnoses. Errors were more frequent in severe inflammatory activity (9%). The most useful endoscopic features in this differential diagnosis were discontinuous involvement, anal lesions, and cobblestoning of mucosa for Crohn's disease, and erosions or microulcers and granularity for ulcerative colitis. After selecting the endoscopic features with best predictive value, an "endoscopic score" was calculated by means of "likelihood ratios."