Modifications of the Farnsworth-Munsell 100-Huetest (e.g., selection of new fixed-reference caps from within the test) have been proposed, with little or no theoretical justification or experimental verification. Predictions based on theoretical considerations of the underlying nature of the test and verified by experimental measurements on subjects with known color defects demonstrate that (1) modification can destroy the very nature of the test; (2) modification can alter axis determination and there-tore, potentially, the diagnosis; (3) the resulting test scores cannot be compared reliably to established norms; and (4) accurate predictions of test performance can be made from theoretical considerations.