Abstract This article describes the development of a theoretically-based instrument to measure aversive reactions to computers. Previous efforts to assess computer averse states are reviewed, and it is argued that a self-report instrument could be useful in detecting negative reactions to computers. The Computer Aversion Scale (CAVS) is based on three expectancy categories (efficacy, outcome, and reinforcement expectations) derived from social learning theory. Data analyses demonstrated moderate to high reliability and validity estimates for CAVS total score and three expectancy subscales. In addition, a study using a modified CAVS was shown to predict poor performance on a computer-based test of cognitive abilities and to moderate a relationship between mood and memory measures. Implications of CAVS findings for psychological assessment via computer are discussed.