Previous studies have failed to differentiate adenomatous from nonadenomatous diminutive polyps according to their gross macroscopic features at endoscopy. We prospectively evaluated the prevalence, distribution, and predictive value of a recently described morphologic feature-the "disappearing phenomenon"-in diminutive polyps of the distal 20 cm of the large bowel by studying 218 polyps in 90 consecutive patients. Disappearance was graded as complete, incomplete, or absent. Overall, complete disappearance was noted in 93 (43.1%) polyps, with a significantly higher prevalence in the middle and lower rectum (p < .05) and among smaller, paler, and smooth-surface polyps (p < .001). Incomplete disappearance was detected in both nonadenomas (23.1%) and adenomas (15.8%), but, more importantly, complete disappearance occurred in none of the 19 observed adenomas compared with 93 of 199 nonadenomas. Multiple logistic regression analysis revealed that disappearance was the strongest predictor (p < .001) of nonadenomatous histology among considered morphologic criteria. When complete disappearance was used to predict histologic type of diminutive polyps, its sensitivity was 100% and its specificity was 46.7%. In conclusion, the disappearing phenomenon represents a reliable visual marker for identifying nonadenomatous rectal and rectosigmoidal diminutive polyps at endoscopy. Diminutive polyps that disappear completely upon insufflation are invariably nonadenomatous and should not require endoscopic biopsy or removal.