Job satisfaction has enjoyed a long history of investigation by many scholars in a great diversity of situations. Few studies have been done, however, in the public sector. Earlier studies were restricted in terms of theoretical approach and also methodologically flawed when they ignored the measurement problems in their analysis. The major purposes of this study have been to present a refined model of job satisfaction and to test it with more caution on measurement problems. The structural equations model with latent variables is the approach used to estimate the model. The data were from the Federal Employee Attitude Survey. These data were analyzed by using a computer program called LISREL. The results confirm the findings reported by the job design school. That is, individuals with high growth need strength tend to have positive perceptions about their jobs and consequently are satisfied with them. This study also suggests the direct and indirect effects of goal setting. Race, age, job status, and job longevity are found to have low effects upon job satisfaction. All these findings support the previous research. However, this study indicates the insignificant effects of feedback, sex, and educational level upon job satisfaction. The rejection of sex and educational level confirms the previous research. The rejection of feedback is surprising, however, since it contrasts with the consistent reports on the importance of feedback as a significant determinant of job satisfaction. Overall, the model proposed in this study can explain over 70 percent of variation in job satisfaction, which is considerably high. A major contribution of this study is the introduction of the structural equations model with latent variables as a standard approach to develop and test models in organization research. This study also has important ramifications for practitioners at various levels. Generally, it reminds them of the important role of individual characteristics and goal setting as significant determinants of job satisfaction. All in all, this study serves as a beginning of a very long step attempted to understand job satisfaction. The refined model proposed should be replicated to test whether it can generalize to other samples in both the private and public sectors. / Ph. D.