Abstract
A reanalysis of Bond and DePaulo’s meta-analysis of deception detection accuracy from the perspective of truth-default theory is reported, focusing on truth bias, the veracity effect, and the implications of the ubiquitous 50%–50% base rates used in primary experiments. Unlike Bond and DePaulo, we examine the relationships among truth bias, the veracity effect, and overall accuracy providing new insights from old data. Truth bias is substantially positively correlated (r = .88) with accuracy for truths, negatively correlated (r = −.88) with accuracy for lies, uncorrelated with overall accuracy (r < .03), and functionally isomorphic with the veracity effect. When accuracies for truths and lies are reported separately, the results may reflect truth bias, message veracity, or both. Substantially improved overall accuracy would be expected with more realistic base rates and levels of truth bias. The reanalyzed data highlight insights from truth-default theory and suggest that the 54% accuracy claim needs to be contextualized.
Get full access to this article
View all access options for this article.
