Lie detection with contingent negative variation

Topographies of contingent negative variation (CNV) were recorded in a paradigm of delayed response with feedback for three kinds of faces: familiar, strange and target. Subjects made responses to the faces according to whether the faces were familiar or not, but also, gave deliberately deceptive responses to target faces to 'cheat the computer'. Subjects were told that the computer could judge whether they were being honest or not. For each trial of the experiment, if subjects cheated the computer successfully and their responses were judged as honest and they were given a reward, otherwise they received a penalty. In this simulated lie detection test, CNV exhibited more negative shifts for target than those for non-target (familiar and strange). These differences could be accounted for by subjects' motivation and uncertainty about passing the test. With the results of further paired t-tests between target and non-target faces at each electrode, CNV was demonstrated as a reliable indicator for lie detection. In addition, vector length was used to capture global CNV information and was found to be a very good indicator, even better than the CNV information at the individual electrode sites. (C) 2003 Elsevier B.V. All rights reserved.