Predictions of Brain Damage from an Aphasia Screening Test; An Application of Discriminant Functions and a Comparison with a Non-Linear Method of Analysis

Abstract

Under linear discriminant function analysis of four cerebral-damage criterion groups (including non-damage), arranged in seven two-way comparisons, the aphasia screening test gave over-all percentages of correct prediction of 262 Ss' actual brain-damage classifications, ranging between 77.5% and 93.4%. An earlier analysis of the same Ss' responses to the aphasia test, organized in terms of four mutually exclusive and exhaustive, non-linear rules for classifying each S to a single damage category (control, left, right, and diffuse or bilateral), produced percentages of correct prediction within virtually the same range as did the discriminant function analysis, but with a different pattern of accuracy of prediction. The results of the aphasia test discriminant function analysis were almost completely congruent with those obtained from the same kind of analysis applied to 23 entirely different test variables from the Wechsler, Halstead, and Trail Making tests. The implications of such methods as the discriminant function, for the production of new understanding of the effects of brain damage on human behavior, are discussed.