While the Wisconsin Card Sort Test is a popular procedure used to evaluate executive functions, the test may take extended times to administer, especially in impaired clients. This has led to interest in the development of a short form of the test. Axelrod, Paolo, and Abraham (1997) earlier failed in an attempt to develop a short form of the Wisconsin Card Sort Test (WCST). The authors suggested that a regression-based approach might work better than their predictive model. In a sample of 145 consecutive subjects referred for neuropsychological testing, linear and nonlinear regression formulas were generated to predict the full 128 card version of the test from 32, 64, and 96 item versions of the test. Overall, the data obtained strongly suggests that multiple regression equations can reliably predict WCST whole test scores from a 64 or a 96-item version of the test, but not a 32-item version. The results indicate that the 64-item version is a clinically sound substitute for the entire test.