Objective. To evaluate the accuracy of arthroscopic measurement of full-thickness (International Cartilage Repair Society grades 3-4) cartilage lesions of the knee. Design. In 33 consecutive arthroscopically evaluated cartilage lesion knees the lesion size was routinely estimated. At the final treatment, a knee arthrotomy was performed, and the lesion was reestimated. The 2 estimates were then compared by paired t test and the area estimated by knee arthrotomy was used as the gold standard. Results. The mean area of the cartilage lesion estimated by arthroscopic surgery was 3.5 cm2 (standard deviation [SD] = 1.7) whereas in the open surgery the mean area was 3.2 cm2 (SD = 1.5). The mean difference was 0.3 cm2, indicating a tendency toward overestimation by the arthroscopic evaluation, but the difference was not significant (P = 0.09). Additional MRI assessment of the area demonstrated a mean area of 2.3 cm2 (SD = 1.6). This underestimated the size of the lesion by 1.2 cm2 (SD = 0.8; P = 0.015). Conclusion. Knee arthroscopic examination estimated a cartilage full-thickness lesion with an error of less than 25 % in the majority of the patients.