A causal model of nurses' job satisfaction was tested using longitudinal analysis of 13 causal determinants and five correlates measured at Time 1, and job satisfaction measured at Time 2. Data were collected from 370 registered nurses at five hospitals using questionnaires mailed eight months apart. Four different models were analyzed using the LISREL maximum likelihood procedure to estimate the path coefficients. Variables reaching statistically significant levels included, in order of importance, routinization, promotional opportunity, distributive justice, age, day shift, workload, kinship responsibility, and opportunity for jobs outside the employing hospital. With the prior level of job satisfaction controlled, only the effects of day shift remained significant.