Educators designing procedures for identifying students as learning disabled are faced with serious technical difficulties. A student with a learning disability is one who "has a severe discrepancey between achievement and intellectual ability" in one or more designated areas. The methods of quantifying achievement, intellectual ability, and their difference together with the nature and degree of difference specified as "severe" determine, in large part, the number and characteristics of students identified as learning disabled. Three types of approaches are currently used for the quantification of severe discrepancies: absolute difference methods, expected achievement methods, and proportional difference methods. Each approach may be utilized with a variety of statistical models. The most complex models adjust the quantification and standard for many characteristics of the student and of the tests. The three approaches are reviewed with reference to some of the commonly used statistical techniques. The effects of test reliability, regression to the mean, test intercorrelations, multiple comparisons, Type I and Type II error rates, and the choice of student and test characteristics for which adjustments in the characteristics of students identified as learning disabled are considered. The advantages and disadvantages of the model proposed by the state of California are discussed. (Author/CM)