This investigation had two purposes. First, the study was designed to confirm the factor structure of the Health Belief Model Inventory (HBMI) among employees of a medical center. Second, a causal model was developed and tested using HBMI components to predict exercise behavior;Data were collected from 511 employees of a regional medical center. Analysis of the data included (1) confirmatory factor analysis of the proposed model including benefits, barriers, susceptibility, social influences, cues to action, and perceived physical ability. (2) structural equation modeling of the relationship between these factors and exercise behavior;Confirmatory factor analysis of the HBMI revealed relatively stable factor structure for benefits, susceptibility, social influences, cues to action, and perceived physical ability. Barriers was somewhat less unidimensional and, although the goodness of fit was improved by trimming some items, it continued to be the poorest fitting factor in the model;The structural equation modeling of exercise behavior used the HBMI constructs as predictors and minutes per week and weeks of exercise as indicators in a multiple indicators-multiple causes analysis. Susceptibility was negatively associated with exercise behavior while Cues to Action and Perceived Physical Ability were positively associated. Benefits, Barriers, and Social Influences were nonsignificant in their contributions to the model;When high exertion subjects were compared with low exertion subjects, some differences emerged. The high exertion group displayed negative path coefficients for Susceptibility and positive coefficients for Cues to Action and Perceived Physical Ability. The low exertion group displayed negative path coefficients for Susceptibility and Social Influences and a positive coefficient for Cues to Action;Results of this study substantiate the addition of Perceived Physical Ability as a component of the model, at least among high exertion subjects. This model's overall ability to predict exercise behavior was substantially less than that achieved in a previous application of the HBMI instrument used in this study.