PURPOSE: To develop and test a model of community-based education. METHOD: In 1995, after developing a hypothesized causal model for community-based education, the authors collected data for 106 students at the University of Gezira, the Sudan, who had participated over three summers in an interdisciplinary field training research and rural development course. The students rated each other on leadership, interaction with the community, subject-matter contributions, and effort. Teaching staff assessed the readiness of the community to collaborate. The students' achievement was measured by short essays measuring knowledge, supervisors' assessments of the students' performances in the community, the community's observations of the students' activities, the community's satisfaction, and a group-produced report evaluated by faculty. The effect of the students' activities on the community was measured by comparing baseline and post-intervention community health data. The students also indicated their levels of interest in the community's problems. The authors analyzed the resulting covariances using structural-equations modeling. RESULTS: After minor adaptations, the model fitted the data reasonably well. The path coefficients were quite high, particularly among the peer ratings. Leadership had a potent effect on the outcome measures, as did, to a lesser extent, the readiness of the community to collaborate with the students. CONCLUSION: This study was the first reported attempt to test a model of community-based education. Although the fit of the data to the model in the study was reasonable, further study is needed to unearth additional important elements of community-based education. This article also discusses methodologic shortcomings of the present study, such as a possible “halo effect” in the peer ratings and the retrospective nature of many of the measurements.