Are reaction times obtained during fMRI scanning reliable and valid measures of behavior?

Assuming that behavior observed during functional magnetic resonance imaging (fMRI) is comparable with behavior outside the scanner appears to be a basic tenet in cognitive neuroscience. Nevertheless, this assumption has rarely been tested directly. Here, we examined the reliability and validity of speeded performance during fMRI scanning by having the same 30 participants perform a battery of five reaction time (RT) tasks in two separate fMRI sessions and a standard laboratory (i.e., outside-scanner) session. Medium-to-high intra-class correlations between the three sessions showed that individual RT differences were conserved across sessions. Thus, for the range of tasks used, test-retest reliability and criterion validity of performance during scanning were satisfactory. Further, the pattern of between-task relations did not change within the scanner, attesting to the construct validity of performance measurements during scanning. In some tasks, however, RTs obtained from fMRI conditions were significantly shorter than those observed under normal laboratory conditions. In summary, RTs obtained during fMRI scanning appear to be largely reliable and valid measures of behavior. The observed RT speed-up during scanning might reflect task-specific interactions with a slightly different neuro-cognitive state, indicating some limits to generalizing brain-behavior relations observed with fMRI. These findings encourage further efforts in fMRI research to establish the external validity of within-scanner task performance.