The generalization of conditioned startle responses from known to unknown lies

Throughout history, there has always been a need to find out whether people are telling the truth. Classical deception detection methods, such as polygraph-based techniques, have so far failed to accurately and reliably detect deception, as they are limited in various aspects. Therefore, results are susceptible to manipulation. In the current study, we attempt to improve lie detection using a classical conditioning procedure with startle response as an outcome variable. Thirty-six participants were asked to report true and false sentences (10 each) before the test procedure. We knew the truth value of only 50% of the sentences (the value of the other 50% was revealed after the experiment was over). Aversive conditioning was used, i.e. participants were presented with the unconditioned stimulus (air blast of 5 bar, 50 ms; contingency of 75%) when uttering known lies. There was a significant difference in participants' startle reaction to false statements compared to truths, both in the known lies category and, more importantly, in the unknown lies category. We recommend further investigation of this phenomenon by changing the conditioning parameters (duration, contingency) in order to optimize this promising method and achieve a higher level of accuracy.