Background: The purpose of this study was to evaluate the accuracy and reproducibility of measurements obtained with a new 3-dimensional stereoscopic endoscopy system in vitro and in vivo. Methods: The system includes an electronic stereoscopic endoscope and a personal computer. After correction of barrel distortion by an image-processing technique, 3-dimensional data are computed with triangulation. Initially, graph paper and models of gastric lesions, both elevated and depressed, were used as objects. For evaluating accuracy and reproducibility, the absolute error (|E|) and the coefficient of variation (CV) were calculated: |E| = average value of absolute value (measured value − true value)/true value. Then, the sizes of 50 gastric lesions were measured by 2 endoscopists, one experienced and the other inexperienced, and the findings were compared with measurements made radiographically. Results: The maximum for |E| and for the CV for the distance on the graph paper were, respectively, 5.6% and 7.0%. The |E| and CV for the diameter and for the height of the elevated model were, respectively, 5.9% and 5.7% for the former and, respectively, 16.8% and 16.5% for the latter. The |E| and CV for the area of the depressed model were, respectively, 9.0% and 1.6%. The correlation coefficients between the sizes of lesions measured by stereoscopic endoscopy and radiographically were 0.91: 99% CI [0.81, 1.00], and 0.98: 99% CI [0.85, 1.00], respectively, for the experienced and inexperienced endoscopists. Conclusions: The new stereoscopic endoscopy system is useful for measuring gastric lesion size in clinical practice.