The reliability of lie detection performance.

We examined whether individuals' ability to detect deception remained stable over time. In two sessions, held one week apart, university students viewed video clips of individuals and attempted to differentiate between the lie-tellers and truth-tellers. Overall, participants had difficulty detecting all types of deception. When viewing children answering yes-no questions about a transgression (Experiments 1 and 5), participants' performance was highly reliable. However, rating adults who provided truthful or fabricated accounts did not produce a significant alternate forms correlation (Experiment 2). This lack of reliability was not due to the types of deceivers (i.e., children versus adults) or interviews (i.e., closed-ended questions versus extended accounts) (Experiment 3). Finally, the type of deceptive scenario (naturalistic vs. experimentally-manipulated) could not account for differences in reliability (Experiment 4). Theoretical and legal implications are discussed.