Recognition of Concealed Information with Behavioral and Spectral Analyses

Abstract

Abstract : The main thesis of this project was that individuals who have concealed information can be detected using a combined behavioral and psychophysiological approach. This study examined the similarities and differences that characterize the behavioral and power spectra responses in truthful and deceptive subjects. Half of the subjects participated in a mock crime while the other half participated in a non-crime scenario. The participants responded during their session to words related and not related to the scenarios they enacted. Although the participants in the mock-crime group were instructed to deny knowledge of any words related to their scenario, the behavioral and spectral data demonstrate that they actually possess concealed information. Therefore, behavioral and spectral indices elicited by concealed information in deceptive subjects can reflect some aspects of deceit. A quasi-replication of the initial experiment was then performed to examine the nature of the power spectra effects. The same protocol was used, except that a block format was used for stimulus presentation and the analysis window for the power spectra was expanded and all frequencies between 13 and 50Hz were available for analysis rather than analyzing the peaks and valleys alone. The results showed that indeed there were significant Group by Frequency effects which indicated that the power spectra of deceivers differed from those of nondeceivers. However, there was neither an Electrode by Group nor an Electrode by Frequency by Group effect that would have reinforced the power spectra differences found in the initial experiment. Although the details of the power spectra effects are yet to be resolved with further study, both experiments described in this Final Report revealed interaction effects, with Group that support the conclusion that power spectra can potentially identify whether an individual is or is not concealing information from their Examiner.