Reaction time (RT) is often used in the assessment of patients with traumatic brain injury (TBI), presumably because it reflects either information processing speed or attentional capacity. To clarify this distinction, we examined behavioral RT and the within-subject variability of RT as they relate to electrophysiological measures of attention and information processing. These include the P300 latency, which reflects stimulus evaluation time, P300 amplitude, which reflects attentional allocation, and the preresponse component of the contingent negative variation (CNV), which reflects sustained attention. We found that the latency and variability in behavioral RT were not correlated with the latency or variability of the P300, suggesting that stimulus evaluation time is not a major contributor to RT and its variability in this paradigm. However, among normal controls, RT was related to P300 amplitude, and therefore to attentional allocation. For the TBI subjects, it was the variability, not the speed, of RT that was related to P300 amplitude and to the preresponse component of the CNV. These data suggest that, while in normal controls RT reflects attentional allocation, among TBI subjects it is the variability in RT that is sensitive to the ability to allocate and sustain attention. (JINS, 1997, 3, 95–107.)