Traumatic brain injury (TBI) is the most common cause of brain damage, resulting in long-term disability. The "reserve" construct has been proposed to account for the reported mismatch between brain damage and its clinical expression. Although numerous studies have used various measures thought to reflect this construct, few studies have examined its underlying structure in clinical populations, and no studies have systematically studied this construct in TBI. In the present study, structural equation modeling technique was used to evaluate several models hypothesized to represent cognitive reserve (CR) in TBI. A broad range of data typically reported in the literature as representing CR was collected from 89 individuals who sustained moderate-to-severe TBI. Analyses revealed a best fitting model that consisted of three separate factors representing premorbid intelligence, socioeconomic status and leisure activity, with distinct pattern of associations among the three factors. Findings provide empirical support for the notion of a multi-factorial CR and suggest a coherent framework for further investigation.