The development of a regression model for the diagnosis of carpal tunnel syndrome (CTS) was examined. The subjects, consisting of 28 CTS cases and 34 age and sex matched controls, completed a medical questionnaire and underwent clinical examination. Computerized axial tomography, nerve conduction velocity, and vibratory threshold measurements were obtained. From these measures, a total of 48 variables were identified. Of these variables, 39 were grouped into eight clusters: size, Raynaud's symptoms, median nerve motor function, ulnar nerve motor function, median nerve sensory function, ulnar nerve sensory function, carpal canal area, and vibration threshold. The remaining nine variables were excluded from subsequent analysis because of low response rates or poor correlation with CTS. The eight clusters were then subjected to principal components analysis. Multiple regression analysis was conducted with the first principal component from each cluster. Because of high correlations with the presence of CTS, Raynaud's symptoms, vibration threshold, carpal tunnel area, and median nerve function components were entered into the regression equation. The presence of Raynaud's symptoms correlated positively with the diagnosis of CTS. Median motor and sensory functions correlated negatively with the diagnosis of CTS, indicating that CTS cases exhibited diminished median nerve function. Vibration thresholds were higher among CTS cases than among controls. The carpal canal area of CTS cases was larger than that of controls. Due to weak correlations with the presence of CTS, age, sex, alcohol use, and ulnar nerve function components were not entered into the regression equation. Based on backwards, stepwise logistics regression analysis, Raynaud's symptoms and median nerve motor function were the best predictors of CTS. The authors conclude that a regression model for the diagnosis of CTS can be developed from a series of variables representing nerve function.