Insufficient attention to tasks can result in slips of action as automatic, unintended action sequences are triggered inappropriately. Such slips arise in part from deficits in sustained attention, which are particularly likely to happen following frontal lobe and white matter damage in traumatic brain injury (TBI). We present a reliable laboratory paradigm that elicits such slips of action and demonstrates high correlations between the severity of brain damage and relative-reported everyday attention failures in a group of 34 TBI patients. We also demonstrate significant correlations between self- and informant-reported everyday attentional failures and performance on this paradigm in a group of 75 normal controls. The paradigm (the Sustained Attention to Response Task-SART) involves the withholding of key presses to rare (one in nine) targets. Performance on the SART correlates significantly with performance on tests of sustained attention, but not other types of attention, supporting the view that this is indeed a measure of sustained attention. We also show that errors (false presses) on the SART can be predicted by a significant shortening of reaction times in the immediately preceding responses, supporting the view that these errors are a result of 'drift' of controlled processing into automatic responding consequent on impaired sustained attention to task. We also report a highly significant correlation of -0.58 between SART performance and Glasgow Coma Scale Scores in the TBI group.