Conventional and actuarial methods to detect response distortion on the basic personality inventory

Abstract

Response distortion remains a significant issue in the assessment of psychopathology. Here four groups of psychiatric patients, each of 40 people, were asked to respond honestly or to distort their presentation as either worse, better, or a "normal" pattern of responses to the Basic Personality Inventory (BPI; Jackson, 1989); only those cases showing acceptable consistency in responding ("person reliability") were analysed. Performance of the conventional cutting points on the BPI validity scales were compared with results from linear discriminant analysis calculated for the patients and from those variables selected previously for university students by Helmes and Holden (1986). Preliminary analyses showed that the "good" and "normal" groups could not be separated; the normal group was therefore not included in subsequent analyses. Results showed better classification results for discriminant functions than for the use of the standard BPI validity measures. Contrary to findings using multiple regression, heuristic weights were the least accurate. Implications for applications of the BPI involving suspected response distortion are discussed, along with the unexpected finding of poor performance of heuristic classification functions.