Computer self-efficacy (CSE) has been successfully used in many studies as a significant predictor of individual performance and attitudes. However there have been ongoing criticisms concerning selfefficacy’s conceptual foundations as well as its measurement. In the IS field, there have been a number of studies in which the relationship between CSE and consequent behaviors or attitudes was weak or nonexistent. Reasons given for such results include problems associated with the construct of general selfefficacy, and more recently the mismatching of scope (general or specific) between self-efficacy and its predicted outcomes. Little research exists in the IS field to ascertain which instrument is appropriate given the circumstances of the study or to determine if scope plays a significant role in predictability. This study examines three CSE instruments, one general, one global, and one specific (in six domains) in their ability to predict specific and general outcomes (competence and attitudes). The CSE instruments are empirically compared using a sample of 310 ROTC Midshipmen from fourteen universities, using six common computing domains. Results suggest that instrument choice makes a significant difference in predictability and that the alignment in scope of CSE with outcomes is moderated by task mastery in that domain.