Unlike most psychiatric diagnoses, posttraumatic stress disorder (PTSD) is defined in relation to a potentially etiologic event (the traumatic "stressor criterion") that is fundamental to its conceptualization. The diagnosis of PTSD thus inherently depends on two separate but confounded processes: exposure to trauma and development of a specific pattern of symptoms that appear following the trauma. Attempts to define the range of trauma exposure inherent in the diagnosis of PTSD have generated controversy, as reflected in successive revisions of the criterion from DSM-III onward. It is still not established whether or not there are specific types of traumatic events and levels of exposure to them that are associated with a syndrome that is cohesive in clinical characteristics, biological correlates, familial patterns, and longitudinal diagnostic stability. On the other hand, the symptomatic description of PTSD is becoming more clear. Of three categories of symptoms associated with PTSD--intrusive memories, avoidance and numbing, and hyperarousal--avoidance and numbing appear to be the most specific for identification of PTSD. Research is now poised to answer questions about the relevance of traumatic events based on their relationship to symptomatic outcome. The authors recommend that future research begin with existing diagnostic criteria, testing and further refining them in accordance with the classic Robins and Guze strategy for validation of psychiatric diagnoses. In this process, diligent adherence to the criteria under examination is paramount to successful PTSD research, and changes in criteria are driven by empirical data rather than theory. Collaborations among trauma research biologists, epidemiologists, and nosologists to map the correspondence between the clinical and biological indicators of psychopathology are necessary to advance validation and further understanding of PTSD.