Many clinical studies of colorectal adenomatous polyps rely on endoscopic estimation of polyp size. To examine the reliability of such measurements, we conducted a study using artificial polyps in an endoscopy teaching model. Eight experienced endoscopists estimated the size of 13 polyps in two separate sessions 2 wk apart. Endoscopic estimates of polyp size tended to be significantly lower than the true polyp size for all polyps and all endoscopists at both sessions. We also found a statistically significant difference in the magnitude of the underestimation between the first and second session (p < 0.0001). At the first session, polyps tended to be estimated at 64% of their true size, and at the second session, the estimates tended to be at 77% of the actual polyp size. We estimate the magnitude of the variation in polyp measurements due to individual polyps, endoscopist, and examination session, and discuss the impact these sources of variation have in planning of clinical trials.