Windows to the Soul? Deliberate Eye Contact as a Cue to Deceit
Although people overwhelmingly believe that liars avoid eye contact, meta-analyses of deception literature have shown a non-significant relationship between gaze and deception. In the present experiment we measured eye movements in an innovative way. We coded the extent to which interviewees deliberately made eye contact with the interviewer. Liars take their credibility less for granted than truth tellers. They therefore may have a greater desire to be convincing and hence more inclined to monitor the interviewer to determine whether they seem to be being believed. We therefore hypothesized that liars would give more appearance of deliberately making eye contact than truth tellers (a relationship which opposes the stereotypical belief that liars look away). A total of 338 passengers at an international airport told the truth or lied about their forthcoming trip. As well as the deliberate eye contact variable, we coded the amount of time the interviewees looked away from the interviewer (e.g., gaze aversion), which is typically examined in deception research. Liars displayed more deliberate eye contact than truth tellers, whereas the amount of gaze aversion did not differ between truth tellers and liars.
People’s Beliefs About the Relationship Between Gaze and Deception
Both laypersons and professional lie catchers overwhelmingly expect liars to exhibit gaze aversion (“liars look away”) (Strömwall et al. 2004; Taylor and Hick 2007; The Global Deception Team 2006; Vrij 2004, 2008a; Vrij et al. 2006a, 2010). For example, when Mann et al. (2004) asked 99 British police officers how they could tell when someone is lying, 73 % of them reported that they believed that liars avert their gaze. The second most frequently mentioned cue, body movements, was mentioned by only 25 % of participants. Charles Bond headed an ambitious ‘beliefs about cues to deception’ project that was published under the name ‘The Global Deception Team’. Comprising an international team of researchers from 58 countries, each researcher collected data from 20 male and 20 female adult residents of their country. The participants were asked to write down their response to the question: “How can you tell when people are lying?” Respondents mentioned 103 different cues, of which nine were given by more than 15 % of participants. Of particular note was the notion that a liar cannot maintain eye contact, and this belief was expressed by 64 % of the participants; gaze aversion was the most frequently mentioned belief in 51 out of 58 countries.
When people overwhelmingly say that liars avert their gaze, it does not mean that they overwhelmingly rely on gaze aversion when they attempt to detect deceit. For example, Vrij (1993) correlated the behaviors displayed by videotaped liars and truth tellers (gaze behavior, smiling, different types of movements, stutters, etc.) with the veracity judgments made by Dutch police detectives who observed those videotapes. The gaze patterns displayed by the liars and truth tellers did not predict the police detectives’ veracity judgments in this particular study, whereas smiling (people who smiled less were perceived as more suspicious) and movements (people who moved their arms and hands more were perceived as more suspicious) did. In a meta-analysis of such studies, Hartwig and Bond (2011) found a correlation of r = .27 between averting gaze and veracity judgments (people who avert their gaze are perceived as more suspicious). This significant correlation was somewhat lower than some other correlations. The cues that had the strongest relationship with veracity judgments were incompetence (r = .54) and ambivalence (r = .51). People who appear incompetent and/or ambivalent are judged as deceptive.
The Actual Relationship Between Gaze and Deception
There are reasons to believe that liars would indeed be more inclined to look away than truth tellers. People often avert their gaze when they feel ashamed (DePaulo et al. 2003; Ekman 1985/2001; Mehrabian 1972), and they may feel ashamed when engaging in the reprehensible act of lying (DePaulo et al. 2003). In addition, people often avert their gaze when their cognitive load increases (Doherty-Sneddon et al. 2002; Doherty-Sneddon and Phelps 2005; Glenberg et al. 1998), and lying can be more mentally taxing than telling the truth (Vrij et al. 2006b, 2008, b, 2011).
There, are, however, also reasons to believe that liars are less inclined than truth tellers to exhibit gaze aversion. For example, people attempt to persuade others by looking them in the eyes (Kleinke 1986), which is something many of us learn to do from a young age, and liars are more inclined than truth tellers to try to appear convincing as they take their credibility less for granted (DePaulo et al. 2003; Kassin 2005, 2008a, b; Kassin et al. 2010; Kassin and Gudjonsson 2004; Kassin and Norwick 2004). In addition, because liars do not take credibility for granted, they may monitor the interviewer’s reactions more carefully in order to assess whether they appear to be getting away with their lie (Buller and Burgoon 1996; Schweitzer et al. 2002).
Meta-analyses of previous research have revealed a non-significant relationship between lying and the amount of time a liar looks at his or her target (DePaulo et al. 2003; Sporer and Schwandt 2007; Vrij 2008a). However, there are different implications to be drawn from a non-significant result in a meta-analysis depending on whether it is derived from a large body of null effects or, alternatively, from a large mixture of positive and negative effects that average to zero. Vrij (2008a) was aware of 45 studies that examined the relationship between deception and gaze and found a non-significant (‘null’) effect in 34 (75 %) of them. In other words, it seems that the non-significant effect in the meta-analyses was mainly caused by a large body of null effects. However, a mixture of positive and negative effects appeared in the remaining 11 studies. The finding that liars look away more than truth tellers was observed in six studies (Buller and Aune 1987; Buller et al. 1989; Hocking and Leathers 1980; Kalma et al. 1996; Knapp et al. 1974; Miller et al. 1983), whereas in the remaining five, the finding that liars look into the eyes of conversation partners more than truth tellers was found in five studies (Bond et al. 1985; Granhag and Strömwall 2002; Griffin and Oppenheimer 2006; Riggio and Friedman 1983; Sitton and Griffin 1981).
Deliberate Eye Contact Versus Traditional Eye Contact
The fact that 75 % of the studies obtained a null effect for gaze aversion could perhaps be explained as follows: The liar is caught in a battle of the desire to avoid making eye contact, but a need to do so in order to appear convincing, the end result is a similar amount of eye contact as would be exhibited when truth telling. This explanation does not necessarily mean that the eyes cannot reveal that someone is lying. It could be that gaze behavior can reveal deception if it is measured differently and more in alignment with the theoretical rationales presented above as to why liars would attempt to engage in eye contact. This implies that eye contact needs to be measured in a more subtle way than the gross calculation of time typically used in deception research. It has been argued that nonverbal cues of deception are often not found because they are not measured precisely enough (Bull 2009; Vrij 2006b, 2008a).
In the present experiment we examined this and determined whether the participants made ‘deliberate eye contact’ with the interviewers, that is, whether the participant made the impression of deliberately attempting to seek eye contact with the interviewer. Two of the factors mentioned above, that liars are more inclined (a) to appear convincing and (b) to monitor the interviewer, would contribute to deliberate eye contact.
The ‘deliberate eye contact’ measure differs from ‘eye contact’ as defined by deception and communication researchers. Eye contact is purely measured by the number of seconds that the participant looks into the eyes of the interviewer, whereas deliberate eye contact is not. It can occur when someone looks into the eyes just a fraction longer than what would normally be expected in such a situation. This short, but vital, prolonged eye contact probably remains unnoticed when the total number of seconds that someone looks into the eyes of the interviewer during the entire interview is calculated. Deliberate eye contact is more subtle than gaze aversion but also somewhat more subjective. However, while subjective, such eye contact is often unmistakeable. As in attraction, when a person holds the eye of the object of their desire for just a little longer than normal, such eye contact while brief is often noted by the receiver as deliberate. However, in an interview situation this is most likely interpreted (as intended) as a sign of candor. Deliberate eye contact probably grasps theoretical concepts better than eye contact. For example, if liars are more inclined to monitor the interviewer they would probably like to do so in a subtle way. We hypothesized that liars would exhibit more deliberate eye contact than truth tellers (Hypothesis 1). Although we measured it for comparison, based on previous research, we expected the most likely outcome to be that no differences would occur between liars and truth tellers for gaze aversion.
We carried out the experiment at a large international airport, as part of a large project. Lie detection in an airport is important, as it is a setting that is often targeted by terrorists. Most investigative (forensic) deception research to date has attempted to resemble (simulated) police—suspect interviews, but there are of course, many other forensic settings in which lie detection is relevant, and an airport is an increasingly good example. In addition, it gave us the opportunity to recruit participants from all over the world. Differences in gaze aversion between groups of participants may be expected as people from different ethnic backgrounds or cultures often display different behavioral patterns (Matsumoto 2006). For example, Caucasians tend to look conversation partners in the eye more than African Americans (Fugita et al. 1974; Ickes 1984; Johnson 2006a, b; LaFrance and Mayo 1976; Smith 1983). Group differences in gaze behavior can be important in daily life in that some groups may be falsely identified as liars because they automatically exhibit more of a behavior that is associated with deception. That is, if people think that liars cannot maintain eye contact, then groups of people who tend to look away more are more likely to be identified as liars. Our experiment, however, does not examine the effect that gaze behavior has on observers, rather we examined the actual differences in gaze behavior between truth tellers and liars. There is no theoretical reason as to why gaze aversion would emerge as a cue to deceit in one group of participants and not in another, and the same applies to deliberate eye contact. The desire to be convincing and the inclination to monitor the interviewer should occur across cultures, and we therefore do not expect a Veracity X Ethnicity interaction effect to occur for deliberate eye contact.
Gaze aversion as a function of ethnicity
Gaze aversion (percentages)
Deliberate eye contact
1 North Europe
2 East Europe
3 South Europe
4 Central Asia
5 South and East Asia
8 South America
9 West Africa
10 East Africa
Where are you going to fly to today?
How would you describe the main purpose of your trip?
These questions were asked to establish the ground truth in the experiment. All participants (truth tellers and liars) were asked to reveal their true destination during the actual interview, and truth tellers were also asked to reveal the true purpose of their trip. When we compared the answers during the actual interview with the answers given to the experimenter, we found that all interviewees honestly reported their destination, and that all truth tellers honestly reported the true purpose of their trip while all liars, as asked, told a different story. The experimenter wrote down the answers and asked the participants whether they had made the trip they were going to make previously.
The participants were allocated randomly to the truth/lie condition. A total of 177 passengers were asked to tell the truth while answering every single question during the interview, whereas the remaining 161 participants were asked to tell the truth about the destination they were flying but to lie about the purpose of their trip. (All participants who were asked to lie were happy to do so.) The experimenter then asked the participant whether s/he needed preparation time. If the participant expressed a need to prepare (significantly more liars [N = 8] than truth tellers [N = 1] did so, χ2 = 6.31, p < .05), the experimenter gave them as much time as they wanted by asking them to return when they were ready to be interviewed, at which point the experimenter took the participant to one of the two female interviewers. The interviews were conducted in separate small interview/search rooms just off the main departure hall area. The interviewer introduced herself, invited the participant to take a seat, and conducted the interview. The interview was recorded with a video-camera, and the distance between the camera and participant was approximately three meters. The interviewer was not on screen, and the participant could be seen from the head to about mid-calf.
The interview protocol consisted of 16 open-ended questions. Although unscripted follow-up questions directed at the participants’ answers would have been desirable and possible, the interviewers were instructed not to do this to ensure optimum experimental control and were instructed to follow the standardized 16 questions protocol. In other words, the interview conformed to a typical investigative interview except that follow-up questions were not asked. Broadly speaking, the questions focused on the reasons for making the trip (‘What would you say is the main purpose of your trip?’) and planned activities during it (‘Please describe in detail what you are going to do at your destination’). After the interview, the interviewer told the interviewee whether or not she believed him/her. The interviewee completed a post-interview questionnaire in which we checked again that participants had followed their lying/truth telling instruction, and was given £10, thanked and debriefed. Allocation to one of the two interviewers was random, and both interviewers were allocated a mixture of truth tellers and liars. The interviews lasted on average just under 5 min (M = 486.61 s, SD = 163.10), with truthful interviews (M = 509.96 s, SD = 180.09) being longer than deceptive (M = 460.96 s, SD = 138.16) interviews, F(1, 336) = 7.76, p = < .01, η2 = .02.
Duration of gaze aversion was coded with the Observer software package. It was defined as ‘the percentage of time looking away from the interviewer’ and we have coded this behavior many times before (Mann et al. 2002; Vrij 2006a; Vrij et al. 1997, 2004, 1996, 2001a, b, 2000, 2008b, 2010; Vrij and Winkel 1991, 1992). A rater, blind to the veracity status of the clips, coded the clips. A second rater, also blind to the veracity status of the clips, coded 52 clips (15 % of the sample) for reliability purposes. The inter-rater reliability (r = .81) was high.
A third rater, blind to the veracity status of the clips, coded the deliberate eye contact variable on a 5-point Likert scale ranging from (1) does not occur at all to (5) occurs very strongly. Deliberate eye contact was defined as where the interviewee clearly makes a point of looking into the eyes of the interviewer (particularly towards the beginning of the interview where, perhaps, it might occur to him or her to do so). The eye contact may not be sustained, may even be below an average amount in total, but must look slightly out of sync in some way. For example perhaps it is held for an unnaturally long time while answering a particular question. Or perhaps the participant looks uncomfortable doing it (most people for example look at their conversation partner more when listening than when actually talking so perhaps instead the person seems to be making a deliberate attempt to hold eye contact instead of looking away). For reliability purposes, fourth and fifth raters coded deliberate eye contact for 78 clips (23 % of the sample). The inter-rater reliability between the three coders combined was satisfactory (Cronbach’s alpha = .85). The inter-rater reliability between third and fourth rater was r = .78, between third and fifth rater was r = .60, and between fourth and fifth rater was r = .64. Because this cue is more subjective than the gaze aversion cue, we used two rather than one inter-rater reliability coder. We conclude therefore that the variable can be reliably coded.
Data were analyzed with two separate 2 (veracity) × 10 (ethnicity) analyses of variance with the two different ways of measuring eye movements (gaze aversion and deliberate eye contact) as dependent variables. The gaze aversion analysis revealed a significant ethnicity effect, F(1, 318) = 3.02, p < .01, η2 = .08. Tukey Post hoc tests did not reveal many differences between the ten ethnicity groups, as Table 1 shows. In fact, there were only two significant differences. West European and Central Asian participants displayed relatively little gaze aversion but their percentages differed significantly only from the East European participants. As we reported in the Introduction, it has been found in the past that in daily interactions, Caucasian (white) participants display less gaze aversion than Afro-Caribbean (black) participants. In alignment with this, West European participants displayed less gaze aversion (11.94 % of the time) than East African participants (20.94 %), but the difference was not significant (p = .13). For the purpose of this article, the veracity main effect and Veracity X Ethnicity interaction effect are more relevant. Neither the main effect, F(1, 318) = .30, ns, η2 = .00, nor the interaction effect, F(1, 318) = .58, ns, η2 = .02, were significant. Liars did not exhibit more (or less) gaze aversion than truth tellers overall, or amongst any ethnic denomination.
The deliberate eye contact analysis revealed a significant veracity effect, F(1, 318) = 20.24, p < .01, η2 = .06. As predicted in Hypothesis 1, truth tellers displayed less deliberate eye contact (M = 2.81, SD = .88) than liars (M = 3.25, SD = .92), d = .49. Neither the ethnicity main effect, F(1, 318) = 1.24, ns, η2 = .03, nor the Veracity X Ethnicity interaction-effect, F(1, 318) = .82, ns, η2 = .02, were significant.
In the present experiment we examined eye contact in an innovative way. We measured whether participants gave the impression of having made deliberate eye contact with their interviewers. As predicted, liars made more deliberate eye contact than truth tellers. We also examined eye contact in the traditional way by measuring the amount of time the interviewees averted their gaze from the interviewer. Like so many researchers in the past, we found no difference between liars and truth tellers in gaze aversion. One reason why deliberate eye contact was diagnostic to deceit where gaze aversion was not is that the measurement of deliberate eye contact is more subtle than the gross measurement of gaze aversion. As we reported earlier, it has been argued that nonverbal cues of deception are often not found because they are not measured precisely enough (Bull 2009; Vrij 2006b, 2008a).
Importantly, the occurrence of deliberate eye contact should not be used as a stop-rule. That is, a single instance of deliberate eye contact should not be used to infer deception. Instead, it is a matter of degree; what the present paper shows is that liars give the impression of making more deliberate eye contact than truth tellers.
Although cultural differences emerged in gaze aversion, they were not large. Although previous research showed differences in gaze behavior between Caucasian and African people (see Introduction), we could not replicate this finding. The findings revealed that the West European participants did not display much gaze aversion, yet the East African participants did not display much gaze aversion either. A possible explanation for this is that our sample of participants, being air passengers from or through the UK, were relatively ‘Westernized’ resulting perhaps in relatively ‘Western’ behaviors. There is some evidence for this. Aside from the fact that all participants had at least a rudimentary grasp of English, we asked participants whether they had made the forthcoming trip before and 26 out of 28 East Africans reported that they had.
Strengths of the present experiment include that participants were members of the general population who told the truth and lied in a real life setting about a realistic event. This makes this experiment different from most deception experiments where participants are typically university students who talk about issues that do not really relate to them. However, such a real life setting has its limitations, including that it is impossible to raise the stakes. We, however, believe that our findings will hold in high stakes settings. Liars’ desire to be convincing and their inclination to monitor interviewers are likely to increase as the stakes get higher, and differences in deliberate eye contact between truth tellers and liars are therefore likely to become more, rather than less, pronounced in high-stakes settings. One could argue that the desire to be convincing and the inclination to monitor interviewers may also increase in truth tellers when the stakes get higher, but that would balance out between truth teller and liars, resulting in similar findings to those reported in the present experiment. In their experiment examining the behavior of participants in very high-stakes settings (suspects of serious crimes in their police interviews) Mann et al. (2002) found similar eye contact patterns as in previous low-stakes studies, i.e. no differences in length of gaze aversion. The only reason why our findings would not hold in high stakes settings is if higher stakes increased the desire to be convincing and the inclination to monitor interviewers more in truth tellers than in liars. We do not see any valid theoretical reason why this would be the case.
Another effect of our paradigm is that it is likely that some liars had made a similar trip in the past and thus had a past experience to rely upon on which to construct a lie. We believe that this increases the ecological validity of the experiment as this is what liars often do in real life. When telling their lies, rather than fabricating an entire story, they relate to an event they have actually experienced, albeit at a different time than when they claim they have (Roach 2010; Vrij 2008a).
The finding that deliberate eye contact emerged a diagnostic cue to deceit makes it stand out in nonverbal deception research, as not many behavioral cues appear to be diagnostic. A further benefit of this cue above other behavioral cues is that it is quite unique in that it can be coded instantly during the interview in real time. Possibly the diagnosticity of the deliberate eye contact cue can be enhanced if it is measured within-subjects rather than between-subjects as we did in the present experiment. Within-subjects lie detection tools have more potential than between-subjects tools because they control for the typically large individual differences in people’s behavior and speech. This could be examined in future research. Although we focused on deliberate eye contact in this article, it does not mean that we recommend that security personnel at airports or other professionals singly focus on deliberate eye contact. Lie detectors should focus on a variety of cues, and, given that deception research has shown that verbal cues are typically more diagnostic to deceit than nonverbal cues, focusing on verbal cues appears to be most promising (DePaulo et al. 2003; Sporer and Schwandt 2006, 2007; Vrij 2008a, b). Perhaps deliberate eye contact could be incorporated in predominantly verbal lie detection tools, and there are at least two attractive aspects of doing this. First, deliberate eye contact is a cue that liars display. In contrast, the vast majority of verbal cues that discriminate between truth tellers and liars according to research, are cues that truth tellers display (and liars avoid). By definition, a balanced search for cues to deceit and truth should result in higher diagnosticity than an unbalanced search for cues to either deceit or truth. Second, the elicitation of verbal cues highly depends on the quality and timing of the questions that are asked; see Vrij et al. (2010) for examples of good questioning techniques. High quality questioning does not come naturally and is a skill that must be learned and practiced. In addition, many efficient interview protocols can only be used in specific settings. We expect deliberate eye contact to be less sensitive to the quality of the interview protocol; and it can be examined in every setting.