Semantic Leakage Enables Lie Detection, but First-Person Pronouns and Verbosity Can Get in the Way of Detection
Abstract
We investigated the impact of linguistic cues and autistic traits on lie detection. Adult participants (N = 125) judged suspects' statements in a detective game. Untruthful statements were marked by semantic leakage. Literature indicates that liars use fewer first-person pronouns and mental-state terms than truth-tellers. We manipulated the untruthful statements for the presence/absence of these cues to test their effect on lie detection. The adults were 89% accurate in detecting lies. Mental-state terms did not affect accuracy, while presence of first-person pronouns hindered it. Having autistic traits did not influence lie detection. However, adults with higher autistic traits struggled to detect lies when these contained both a first-person pronoun and a mental-state term. Post-hoc analysis revealed lower lie detection accuracy for longer sentences. Our findings underscore the significance of semantic leakage in lie detection, with nuanced effects of linguistic cues on accuracy, particularly for adults with higher autistic traits.