Validity of Systematic Behavioral Analysis in Detecting High-Stakes Deception

Abstract The aim of this study was to expand the ecological validity in the area of human deception detection by examining empirical field data of Behavioral Analysis for Veracity Assessment (BAVA) which was a technique developed to assess the veracity of the statements made by the criminal suspects or victims during the course of the investigation. We predicted that deception occurring in high-stakes situations would be correctly detected by BAVA analysts. To examine the validity of the BAVA analysts’ decisions in the absence of ground truth, we compared their lie-truth decisions with the final court judgments and estimated the accuracy of BAVA based on three methods (Spencer’s (2007) method, latent class model, and overwhelming evidence). The BAVA decision of the defendant’s guilt or innocence was highly consistent with the court’s judgment (agreement rate = 87.4%, N = 222), and the accuracy of BAVA was estimated to be between 0.86 and 0.93. In comparison with previous findings that the human ability to detect deception is mediocre, the results of this study suggest that high-stakes lies can be detected with high accuracy. Theoretical and practical implications for improving detection accuracy are discussed.