An Evaluation of Performance-Based Tests Designed to Predict Success in Primary Flight Training

Abstract

In 1986, the Naval Aerospace Medical Research Laboratory (NAMRL) completed the development of a performance assessment test battery designed to assess cognitive abilities, higher-order processes, psychomotor skills, time-sharing ability, and personality traits. This automated performance-based test battery was intended to augment the Navy's paper-and-pencil selection tests for aviators. The main purpose of this study was to evaluate, in an hierarchical multiple regression model, the constituent tests comprising our performance-based test battery. To further such a goal, this paper presents a statistical assessment of all the tests in the battery when they enter as variables in a regression equation to predict success in primary flight training. Our analysis revealed that derived scores from three tests, (a) Absolute Difference-Horizontal Tracking (ADHT), (b) Complex Visual Information Processing (CVT), and (c) a Risk-Taking Task (RISK), were generally equivalent in predicting success. The derived scores from the Manikin, Baddeley, and Psychomotor/Dichotic Listening Task tests did not account for significant variance. In addition, the linear regression models were not improved by adding the variables of other test sets when the model already included one significant test set. In contrast, interactions of college major and accession source with derived scores of the three significant test sets contributed significant amounts of variability when added to the model. These results appear to indicate differential validity of these selection tests.