Towards clinical trials of lie detection with fMRI

Abstract

Recent reports of successful fMRI-based discrimination between lie and truth in single subjects raised the interest of prospective users and a public concern about the potential scope of this technology. The increased scrutiny highlighted the lack of controlled “real life”, i.e. prospective clinical trials of this technology that conform to the common standards of medical device development. The ethics of conducting such trials given the paucity of data on fMRI-based lie detection has also been questioned. To probe the potential issues of translating the laboratory research into practice, we conducted a case study in which we adapted the standard Guilty Knowledge Test (GKT), a well-established model of producing deception, to the common scenario of lying on a resumé. The task consisted of questions about pertinent items on the subject's resumé, three of which could be independently verified as truth (KNOWN) and three that could not be verified and were thus termed UNKNOWN. The subject had an incentive to lie on all UNKNOWN items, and on debriefing confirmed that he had done so. Data was preprocessed, masked with a priori regions of interest, thresholded, and qualitatively evaluated for consistency with the previously reported prefronto-parietal Lie > Truth pattern. Deceptive responses to two out of the three UNKNOWN items were associated with the predicted prefronto-parietal fMRI pattern. In the third UNKNOWN this pattern was absent, and instead, increased limbic (amygdala and hippocampus) response was observed. Based on published prefronto-parietal Lie response pattern, only the first two items could be categorized as Lie. If confirmed, this demonstration of amygdala and hippocampus activation in a Lie > Truth contrast illustrates the need to integrate the limbic system and its emotional and cognitive correlates into the existing model of deception. Our experiment suggests an approach to a naturalistic scenario and the research questions that need to be answered in order to set the stage for prospective clinical trials of fMRI-based lie detection.

Acknowledgements

This work was supported in part by a grant from the US Army Office of Research (ARO). Parts of this report appeared in The Washington Post (Garreau, 2006).