The relationship between computer anxiety and computer performance is examined using a self-efficacy framework. A novel database searching task was demonstrated to 50 participants using two procedures (namely, accessing the data tables directly and constructing look-up tables). Levels of computer anxiety, prior experience and perceptions of self-efficacy were recorded. The results indicate that computer anxiety directly influences the number of correct responses obtained whilst self-efficacy determines how the task is attempted. Less anxious subjects obtained more correct responses and subjects with higher perceptions of self-efficacy used more look-up tables. The results indicate that self-efficacy theory can account for around half the variance in computer performance and that how a task is attempted should be assessed in addition to accuracy and speed of performance.