Six theoretical models of social support in relation to perceived job stress, burnout, and health were tested both cross-sectionally and longitudinally. Participants at Time 1 were state correctional officers (N = 262) who completed questionnaires in which multiple indicators of each construct were assessed. Time 2 participants (N = 177) were those officers from the Time 1 sample who completed the questionnaire again three months later. Structural equation analyses revealed that only one of the six models was supported by cross-sectional results. In this model a direct negative relationship between the WORKPLACE SOCIAL SUPPORT and BURNOUT latent variables was specified, along with direct, positive relationships between the JOB STRESS and BURNOUT latent variables and the BURNOUT and POOR HEALTH latent variables. However, none of the six models was supported by the longitudinal results. Thus, cross-sectional results were consistent with a model in which social support on the job influences positive health only through its direct and negative effect on burnout symptoms, but such causal connections were not validated longitudinally.