The military healthcare system serves personnel from diverse ethnic and demographic backgrounds, who face diagnoses that at one level are equalizers: coronary disease is coronary disease, breast cancer is breast cancer. Yet the differing expression of disease in individuals from different backgrounds, and individual patient experience of disease as a particular illness vary enormously, and thus so do interactions between patients and providers in any given encounter. Clinicians, in fact, vary greatly in their understanding of individual and cultural variability issues, as traditional training in these areas relies on a system of apprenticeship and exposure over time to multiple populations. Consequently, providers are often thrown into situations where clinical communications falter through inadequate direct experience, making patient diversity a critical variable in the encounter outcome. Recent research suggests that experiential training virtual environments can compensate for the randomness and length of the traditional approaches to learning clinical communication skills relating to individual and cultural differences. A training system based on this approach has been developed to improve skills in delivering culturally sensitive care to African-American women with breast cancer. The system, called TEACH (Training to Enable/Achieve Culturally Sensitive Healthcare), was developed to enhance healthcare provider skills in delivering culturally sensitive care to African-American women with breast cancer. The system uses a population of virtual patients who incorporate underlying models of differing individual and sub-cultural beliefs about breast cancer that can affect the patient's communication with the clinician as well as the patient's approach to treatment. Users (clinicians or medical students) interact with these synthetic patients at virtual clinical encounters representing different stages of the disease progression. The cognitive and cultural models that drive the synthetic patient behavior are discussed, along with the instructional model and (generalizable) system design and architecture.