Magnetic resonance imaging (MRI) is considered to be a highly sensitive modality for visualizing white matter abnormalities. Estimations of its specificity are far less positive. However, diagnostic specificity depends upon both the inherent qualities of MRI and on the quality of image interpretation. Systematic and detailed analysis of many image elements, and substantial prior experience improve the quality of image interpretation and thus improve diagnostic specificity. The present study has been set up to develop a pattern recognition system which combines sensitivity and specificity, systematic analysis of image elements and prior experience. This pattern recognition is based on the data of 277 patients with white matter disorders referred for MRI. The information was stored in a data base and computer analyzed. Twenty-two MRI patterns were discerned in as many disease categories. The frequency of occurrence of each MRI abnormality was assessed per disease category to establish the pattern of abnormalities characteristic for each separate disease category. The pattern recognition program was also written so that: (a) when fed data about MRI abnormalities observed in a new case, the computer produces a differential diagnosis with probabilities and 95% confidence intervals for each differential diagnosis; (b) specific data on the MRI findings of new cases could be added to the data base to improve the experience and accuracy of the program. This program for pattern recognition of abnormalities in the MR images of white matter disorders enhances the specificity of image interpretation and provides a wonderful aid for teaching purposes.