Why.Dont.We.Catch.Liars

Social Research, Vol. 63, No. 3, 801-817 (Fall 1996) Paul Ekman’s research is supported by a Research Scientist Award from the National Institute of Mental Health (MH06092). Why Don’t We Catch Liars? Paul Ekman Our research (Ekman and O’Sullivan, 1990; Frank and Ekman, forthcoming; Ekman, Frank, and O’Sullivan, forth​coming) suggests that most people cannot tell from demeanor whether someone is lying or telling the truth. Such poor performance is not because lies are told flawlessly. Most liars make mistakes which could be detected but usually are missed. Both perpetrating a lie and detecting a lie, in most people, seem to be poorly developed skills. In this article, I consider six explanations for why most of us do not catch liars from demeanor. I will first explain how I distinguish lying from other forms of deceit, and then discuss the evidence which suggests that people are such poor lie catchers. The intent of the liar is one of the two criteria I (Ekman, [1985] 1992) use to distinguish lies from other kinds of deception. The liar deliberately chooses to mislead the target. Liars may actually tell the truth, but that is not their intent. And truthful people may provide false information—bad advice from a stock broker—but that is not their intent. The~ liar has a choice; the liar could chose not to lie. We are all tempted to lie, but we do not always do so. Lying is not irresistible; it is, by my definition, a conscious, considered choice. I do recognize that lying can become a habit and then performed with little consideration, but, at least initially, all such habits began as considered choices about whether or not to do so. Presumably, a pathological liar is compelled to lie and by my definition, therefore, is not a liar. The second criterion for distinguishing lies from other deceptions is that the target is not notified about the liar’s intention to mislead. A magician is not a liar by this criterion, but Uri Geller is a liar, since Geller claimed his tricks were not magic. An actor is not a liar, but an impostor is. Let the buyer beware is one example of an explicit warning that products or services may not be what they are presented to be. (Of course, that warning does not appear in advertisements, nearly all of which are designed to convey the opposite message.) Poker is still another situation in which the rules of the game sanction and notify the players that deception will occur, and, therefore, one cannot consider bluffing to be a lie. Sometimes notification of an intention to mislead is implicit in the framing, to use Goffman’s (1974) term, of the situation. In real estate transactions, the potential buyer is implicitly notified that the seller’s asking price is not the actual price the seller would accept. Various forms of politeness are other instances in which the nature of the situation notifies the target that the truth may not be spoken. The host would not properly’ scrutinize the dinner guest to determine if the guest’s claim to have enjoyed the evening is true anymore than the aunt should worry whether the nephew is lying when he says that he appreciated being given a tie for Christmas. Deception is expected; even if the target might suspect that the truth is not being told, it is improper to question it. Only certain types of deception may be allowable: the poker player cannot use marked cards; the home seller cannot conceal a known defect. In courtship, it is ambiguous whether the parties should expect truthfulness. The saying “all’s fair in love and war” would seem to warn lovers not to believe all they are told. Recent public opinion polls suggest that lies that downplay the number of previous sexual partners are common among college-aged adults. Yet I expect that lovers want to believe in the truthfulness of their lover. Many popular songs testify to the betrayal felt when lies are discovered (although some do warn that lies may be expected). Romantic love requires collusive efforts to develop and maintain myths about each other and the nature of the relationship. I differ from Bok (1982), who only considers false statements to be lies. I (Ekman, [1985] 1992) argued that one can falsify without words, and one need not falsify, verbally or nonverbally, to lie. Concealment is just as much a lie as falsification, if there is an expectation that information will be revealed. When filling out a job application that asks for a listing of all previous employment, omitting the one from which one was fired would be a concealment lie, for there is an obligation to reveal. In personal relationships it is not always so clear cut, and the liar, once discovered, and the target of the lie may disagree about whether or not an obligation to reveal the concealed information was in force. Concealment and falsification are different techniques for accomplishing the same objective. The issue is the motive, not the technique employed to accomplish it. If the motive is to mislead, then the choice between falsifying or concealing is simply a matter of which technique will work better in a given instance. Elsewhere (Ekman, [198531992) I have explained why most liars would prefer to conceal rather than falsify if the situation will allow it and also described some other techniques for implementing a lie. Now let us consider what we know about how well people can detect lies from demeanor. The evidence that most people do poorly in catching lies comes from the following type of experiment. Students are recruited to lie or tell the truth about something which usually does not matter much to them. It has no relevance to their past or their expected future life. Sometimes in a weak (in my judgment) attempt to motivate them, they are told it is important to be able to lie, or that smart or successful people succeed in this task. Videotapes of their behavior are shown to other students who are asked to identify who is lying and who is telling the truth. Typically, most of those trying to catch the liars perform at chance or just slightly better than chance. Our (Ekman and Friesen, 1974; Ekman, Frank, and O’Sullivan, forthcoming) research has differed in a number of ways. We have tried to make the lies relevant to their lives and to set the stakes for success or failure as high as we could. We attempted this for two reasons. It is only in high stake lies that emotions about lying (fear, guilt, excitement, or what I have called duping delight) are likely to be aroused and betray the lie. It is not just the leakage of these strong emotions which provide behavioral clues to deceit, but these strong emotions also max’ disruPt the liar’s cognitive processing and result in evasive, implausible, and stumbling accounts. A second reason for studying high stake lies is that these are the lies with which society is most concerned. In one of our experimental scenarios, we examined how well nurses could conceal the negative emotions they felt when witnessing films showing amputations and burns. They were highly motivated to succeed in this lie, because they thought our experiment offered them the opportunity to develop a skill they would need to use when confronting just such upsetting scenes in their future work. In another of our scenarios, the subjects had a chance to take and keep $50 if they could convince the interrogator they’ had not taken the money. Those subjects who did not take the money could earn $10 if the interrogator believed them when they’ said they had not taken the $50. In our last scenario, we first identified the social issues the subjects felt most strongly about, and then asked them to describe that opinion honestly (and earn $10 if believed) or claim to have the opposite of their true opinion (and earn $50 if believed). In our most recent work we gave some of our subjects the choice as to whether to lie or tell the truth, as people have in real life. There are many reasons why some people choose not to lie, one of which is their own knowledge, based on past experience, that they are almost always caught. Including in the sample of liars such terrible liars, people who would not choose to lie unless forced to do so by the experimenter, could inflate the detection rate. In virtually all previous research, on either interpersonal deception or polygraph lie detection, subjects were not given the choice as to whether to lie or be truthful. One exception is the study of the polygraph by Ginton, Daie, Elaad, and Ben-Shakhar (1982), in which they were able to know which policemen had cheated on a test for eligibility for promotion; Stiff, Corman, Knizek, and Snider (1994) in a similar fashion knew which students cheated on a quiz. Bradley (1988) also allowed subjects to choose whether to lie or tell the truth in a polygraph study. Another unique feature of our recent experiments is that we told the subjects that they would be punished, and it was a considerable punishment, if the interrogator judged them to be lying. Both the truthful person mistakenly judged to be lying and the liar who was detected would receive the same punishment. Thus, for the first time in research on lying, both the truthful person and the liar might be afraid—of being disbelieved if telling the truth, of being caught if lying. If it is only the liar who might be afraid of being accused of lying, it is too easy’ for the lie catcher and not relevant to most of real life. And if neither liar nor the truthful person fear punishment, it should have little relevance to the lies that occur in the criminal justice world or in national security, let alone in marital disputes, parent-child conflicts, and so on. Although our recent experiments can claim to have more ecological validity than our older studies, or than most of the literature on either interpersonal deceit or polygraph lie detection, the findings about detectability were not much different. Most of those who saw the videotapes and made their judgments operated at a chance level or only slightly better than chance. Before proceeding to consider why people do so poorly as lie-catchers, let us consider some limitations of our research which could have led us to underestimate the ability to detect lies from demeanor. For the most part, the observers who judged who was lying and who was telling the truth had no vital interest at stake in achieving accuracy. They were not offered higher pay if they were more accurate. And catching liars was not intrinsically rewarding, for most of these people did not make a living catching liars. This limitation was addressed in our (Ekman and O’Sullivan, 1991) study and work by other research groups (Kraut and Poe, 1980; DePaulo and Pfeifer, 1986) which did study professionals concerned with catching liars. We found that customs officials, policemen, trial court judges, F.B.I., C.I.A., B.A.T.F., D.E.A., forensic psychiatrists, and trial lawyers were not much better than chance. Perhaps accuracy would be higher if those making the judgments had been able to ask the questions, rather than being passive observers. I cannot rule this out, although I doubt it would be so. The requirement to formulate questions might well detract from the ability to process the information provided by the person being judged. It is for this reason that in many interrogations one person asks the questions while another sits passively considering the suspects responses. It would be interesting to have professional interrogators ask the questions in our experiments and then determine if those who see the videotapes generated were more accurate than has been so far found. Our observers were not familiar with those whom they judged, and it might be argued that such familiarity would benefit accuracy. There are, of course, many situations in which judgments about lying are made without any prior familiarity with the person being evaluated, and our experi​ments at least are relevant to those instances. But I doubt that familiarity always benefits lie detection. While it should provide the basis for discounting idiosyncratic behaviors, it may do so at a cost. We tend to become invested in our friendships and work relationships, and the wish to preserve them may lead us to develop blindness to behaviors which could disrupt them. Trust makes one vulnerable to being misled, as usual levels of wariness are reduced and the benefit of the doubt is routinely given. Involvement in a relationship can lead also to confidence in one’s ability to detect deception (Sillars and Scott, 1983), and such confidence may itself make one more vulnerable (Levine and McCornack, 1992). Familiarity should be an unmitigated benefit only when it is with a person one has had reason to distrust, and about whom one has acquired knowledge of how and when they betray the relationship. In our experiments, the observers were only shown a few min​utes of each interview before being required to make their judg​ment. But longer samples may not necessarily benefit lie detec​tion. We did do one study in which the samples shown were twice as long, and accuracy did not improve. Furthermore, we know from the behavioral measurements we have done that there are clues to deceit in these shorter samples. Nevertheless, we cannot rule out this limitation. If people were given much longer sam​ples to judge, of an hour or two, accuracy might improve. A critic might also have wondered if accuracy was so poor because there were few behavioral clues to deceit, but, as I have just mentioned, that is not the case in our experiments. Measurements we and our collaborators have done of the facial movements, voice, and speech show that high levels of accuracy are possible—over 80 percent correct classifications of who is lying and who is telling the truth. While those measurements required slowed motion replays, we also know that accurate judgments are possible just by viewing the videotapes at real time. A small percent of those we have studied have reached 80 percent or better accuracy, and they have done so in judging more than one scenario, so it is unlikely their accuracy was a fluke. And we have found a few occupational groups which as a group were highly accurate. The United States Secret Service were highly accurate on the emotion lie; none of them scored at or below chance, and a third were above 80 percent accurate. Interrogators specially selected for their known skills and given a week of training showed similar accuracy on the opinion lie. Although the stakes in the lies we studied were much higher than in other research on lying, certainly they were not as high as they are in many criminal or national security cases. Perhaps if the stakes were much higher, the videotapes would have contained many obvious signs of deceit, resulting in much higher accuracy. I cannot argue against that possibility but, as I just described, there were some occupational groups who were accurate when judging our videotapes. The question remains why were all these other groups not accurate. The information is there, and it can be detected by some but not by most. Before considering why the overwhelming majority of people do poorly, consider one more feature of our experiments which probably benefited accuracy and may have led us to over- rather than under-estimate accuracy. In all of our recent studies we have told our observers that between 40 and 60 percent of the people they will see are lying. Initially we did not give this instruction, and found that a group of policemen judged everyone they saw on the videotape as lying, later explaining that everyone lies, especially to the police. Knowing the base rate of lies is an advantage people do not always have, and should enhance lie detection. I have more to say about this later. Granting that our evidence is not conclusive, nevertheless, our videotapes do contain behavioral clues to deceit, which some people can recognize accurately but most do not. For the purpose of this discussion, let us consider this evidence as suggesting that in actual life most people, the overwhelming majority of people, do not detect high stake lies from demeanor. The question I pose is why not, why can we not all do better at this? It is not that we do not care. Public opinion polls time and again show that honesty is among the top five characteristics people want in a leader, friend, or lover. And the world of entertainment is full of stories, films, and songs which describe the tragic consequences of betrayal. My first explanation of why we may be such poor lie catchers is that we are not prepared by our evolutionary history to be either very good lie catchers or lie perpetrators.1 I suspect that our ancestral environment was not one in which there were many opportunities to lie and get away with it, and the costs for being caught in a lie might have been severe. If this speculation is correct, there would not have been any selection for those people who were unusually adept in catching or perpetrating lies. The fossil record does not tell us much about social life, so one must speculate about what life as hunter-gatherers might have been like. I add to that my experience thirty years ago working in what was then a stone-age preliterate culture in what is now called Papua New Guinea. There were no rooms with doors, little privacy in this group living, small village, in which everyone knew and saw everyone else every day. Lies would most often be betrayed by the target or someone else observing actions which contradicted the lie or by other physical evidence. Adultery was an activity which lying often attempted to conceal in the village where I lived. Such lies were uncovered not by reading the betrayer’s demeanor when proclaiming fidelity, but by stumbling over him or her in the bush. Perhaps lies about beliefs, emotions, and plans could have better avoided detection in such an environment.2 But some of those lies would eventually lead to one or another action, and then my argument about how hard it is to conceal or falsify actions in a setting in which there is no privacy would apply. In a society in which an individual’s survival depended on cooperative efforts with other members of their village, the reputational loss for being caught in a high stake lie might well be deadly. No one might cooperate with someone known to have engaged in serious lies. One could not change spouses, jobs, or villages with any ease. Cheney and Seyfarth (1990), in their chapter on animal deception, make very similar points. An important constraint against lying . . . arises from a species social structure. Animals that live in stable social groups face special problems in any attempt at deceptive communication. . . . Among socially living animals deceptive signals will probably have to be more subtle and occur at lower frequencies if they are to go undetected. Equally important, if animals live in social groups in which some degree of cooperation is essential for survival, the need for cooperation can reduce the rate at which unreliable signals are given (1990, p. 189). To have had some special skill in detecting (or perpetrating for that matter) lies would not have had much adaptive value in such circumstances. Serious, high stake lies probably did not occur that often because of limited opportunity and high costs. When lies were suspected or uncovered, it was probably not by judgments of demeanor. (Note I have focused just on intra-group lies; certainly lies might between groups, and their costs and detection could be quite different).3 While there are altruistic lies, my discussion has dealt with less friendly lies, lies that occur when one person gains an advantage, often at the cost of the target of the lie. When the advantage is gained by violating a rule or expectation, we call that cheating. Lies sometimes may be required to accomplish the cheating activity, and lies are always required to conceal h