Abstract: The diagnosis method by Classification And Regression Tree (CART) is gaining more importance every year, specially in Rheumatology by the efforts of the American College of Rheumatology. The classification tree not only provides an accurate tool for diagnosis, it gives a better insight of the manifestations of the disease and the value of different criteria for the diagnosis. In this study we analyzed 1992 Behcet's patients (BP) and 1506 control patients (CP). Patients and controls were each divided into 2 groups. One group was the learning sample and the other was the cross validation sample. The classification tree was constructed using the learning sample. Results were validated by testing them on the cross validation sample. Four trees were constructed, they all gave the same results (less than 0.5% difference in accuracy). The simplest one was chosen to be presented here. The sensitivity is 97.3%, the specificity is 94.2%, and the accuracy is 96%. The cross validation gave a sensitivity of 97.2%, a specificity of 93.9%, and an accuracy of 95.8%. The diagnosis tree is self explanatory. As seen there are 5 terminal branches which are classified as BD and 4 terminal branches which are non BD patients. The number of BD and control patients (in parenthesis) are indicated for each terminal branch.