I recall watching a television show just as microexpressions were receiving pop-science attention which claimed that tests against videotaped gold standards showed that the only people who were reliably able to distinguish liars from truth-tellers at a rate above chance were either people explicitly trained to detect microexpressions or professional spies.