Deception involves various factors that influence behaviours (Kraut, 1980). Three theoretical perspectives (the emotional, cognitive effort and attempted behavioural control) are often used to explain the relationship between deception and behavior. Depaulo, Lindsay et al’s (2003) meta-analysis of cues to deception summarized 158 indicators of cues and 1338 estimates from previous studies, compared to truth tellers, liars tend to have more tense voice, less pleasant looking faces, more lip pressing, show more word and sentence repetitions and speech errors. The results provided general support for the theoretical perspectives, but most of them only show a weak relationship with deception. With regard to nonverbal behavior, there is no single behavior that all liars show each time they lie, and no theoretical perspective that directly exams nonverbal behavior correlates with deception (DePaulo, Stone, & Lassiter, 1985; Vrij, 2000).
There are a number of physiological ways to determine whether a person is lying or not, some of them involved the use of machines, such as lie detectors; Nonverbal cues to deception vary from person to person, depending on the interpersonal differences (Ekman, 1972; Feldman & Phillipot, 1993). Considering all these facts, this review is limited to nonverbal behavior cues to deception that can be detected by humans without the aid of any special equipment. It is also limited to studies of adults. However, the theoretical perspectives on nonverbal cues to deception will be summarized, the accuracy rate of detecting lies through all different types of studies will be outlined, and the reason for why only a few and rather weak relationship between nonverbal behavior and deception.
Theories of deception
Ekman and Friesen (1969) published the first influential theoretical statement about cues to deception. They described two categories of cues: leakage cues (the nonverbal act reveals a message otherwise being concealed) and deception cues (the nonverbal act suggest that deception is occurring but doesn’t reveal the concealed message). Ekman and Friesen (1969, 1974) proposed that certain aspects of nonverbal behaviour are less controllable than others. They noted that these parts of the body such as hands and legs are rarely used in conscious communication. Hence, these parts emit more leakage than the face, because deceivers pay less attention to controlling the body. However certain paralinguistic aspects of speech, such as tone, pause and pitch, may be less controllable than either face or the body. Therefore, they are more likely to leakage (Scherer, 1986).
Ekman and Friesen (1974) hold the belief that deception is associated with an increase in body movements, but research evidence showed a decrease in hand, foot and movements in actual deception (Vrij, 1995; Ekman, 1989; Depaulo, 1992). Zuckerman, Koestner and Driver (1981) found that people hold many incorrect beliefs about those specific cues to deception. People from different countries and with different professional background share beliefs about how liars behave and what they say. Therefore, measuring beliefs about cues to deception might predict which cues people use in actual lie deception (Granhag & Hartwig, 2004). Forrest, Feldman and Tyler (2004) found that people with more accurate beliefs about cues to deception make better lie detectors than people with less accurate beliefs. The explanation for showing a decrease in movements might be due to stress, and behave nervously when they lie.
Three theoretical perspectives that are particularly important for predicting and explain the nonverbal deceptive behaviour are emotional reactions, cognitive effort and attempted behavioural control (Zuckerman, Depaulo & Rosenthal, 1981; Vrij, 2000). Although research has revealed there is no typical deceptive behaviour, some behaviours are more likely to occur during deception, for example, liars waited longer before giving an answer than truth tellers (Depaulo, Lindsay et al, 2003; Sporer & Schwandt; 2006). In the emotional approach, to the extent that telling a lie is most commonly associated with three different emotions: guilt, fear and excitement (Ekman, 1989), liars experience guilt about lying as well as fear about being caught, more often than truth tellers, or they may feel excited because of the opportunity to fool others. Zuckerman et al, (1981) suggested liars might fidget more than truth tellers and make less eye contact. They also suggest fear may result in increased physiological arousal, and this might lead to an increased in cues such as an increased speech hesitations (mm’s and er’s) and speech errors. Excitement may result in behaviour sighs of joy, e.g. an increase in movements and in smiling. The cognitive effort approach assumes that it is cognitively difficult to formulate a lie consistent with what others already know. Liars will be more inclined than truth tellers to monitor and control their demeanour to make them look honest (Depaulo & Kirkendol, 1989). There are evidences to suggest that people engaged in cognitive complex tasks make more speech hesitation, longer response latencies, more gaze aversion and fewer hand and arm movements (Ekman & Friesen, 1972). The attempted behavioural control approach assumes liars are stressed so that may attempt to control their behaviour in order to avoid giving nonverbal cues to deception (Buller & Burgoon, 1996; Ekman, 1989). For example, people hold the beliefs that liars tend more often to make speech hesitations, errors so that they will deliberately try to avoid making such behaviours. Nevertheless, this controlled behaviour may appear planned, rehearsed and lacking in spontaneity. However, some behaviours are not easy to control, because they are linked to strong felt emotion or high stress, such nonverbal cues may betray a lie (Ekman, 1985, 2001). Darwin (1872) suggested that muscles that are difficult to activate voluntarily might escape efforts to inhibit expression, revealing true feelings. Ekman, Roper & Hager (1980) found that fewer than 25% of their subjects were able to deliberately produce several facial actions. Thus, nonverbal behaviours may nonetheless be the best source of emotional leakage of deception (Ekman & Friesen 1969, 1975).
All three of these processes could occur simultaneously. The occurrences of these processes should depend on the type of a lie. Emotional cues (e.g. nervous behaviours) are more likely to occur in high stake lies (Ekman & Frank, 1997). Liars have to think hard when the lie is complex, therefore, cognitive load are more likely to occur in complicated lies than in easy lies (McCornack, 1997). Attempting to control behaviours may often occur in motivated liars, who actively address their behaviour and try to appear credible (Depaulo &Kirkendol, 1989). However, Depaulo, Lindsay et, al,’s (2003) self-presentation perspectives point out that truth tellers may also experience emotion or cognitive load. That is, they may also show nonverbal cues associated with emotion or cognitive load. Because both truth tellers and liars could face negative consequence if they are not be believed (Ofshe & Leo, 1997). Moreover, truth tellers are less likely to think that it is important to make a convincing impression on others. Therefore, they could probably show behaviour that appears suspicious (Vrij & Mann, 2001). However, those theoretical perspectives may lead to opposite behaviours, for example, the emotional and cognitive load predicted an increase in speech hesitation and speech errors, whereas, the attempted control approach predicted that liars will try to control those speech disturbances in order to make them look honest, and therefore their speech will sound fluent and smooth.
Buller and Burgoon’s (1996) interpersonal deception theory suggested that deceptive behaviour may not only be influenced by psychological variables such as emotion or motivation but also interpersonal communicative process. It emphasized that while managing their emotions and displaying credible nonverbal behaviour simultaneously, they may also monitor targets’ behaviours and make the appropriately turn-taking. Therefore, Buller and Burgoon (1996) predicted that deceivers in interactive contexts will display increasing fluency, smooth turn-taking and composure during the face to face communication. Based on the Zuckerman et al’s (1981) three perspectives, they further proposed two type of deceivers’ behaviours: strategic behaviours (intentional behaviours and plans to protect liars’ image and avoiding relationship problems)?? non-strategic leakage (unintentional leakage such as physiological arousal and nervousness). As a result, motivated liar used more strategic behaviours than instrumentally motivated liars.
Accuracy in detecting deception by observing its behaviour correlates
In research studies of nonverbal cues to deception, observers are typically given video footage or sometimes audiotapes from a number of people who are either lying or telling the truth. Average accuracy in detecting deceit has rarely been above 60%, and some groups have done even worse than chance, where 50 % is chance level (Kraut, 1980; Depaulo, Zuckerman & Rosenthal; 1980; Vrij & Graham; 1997). Similarly accuracy rates are reported by Vrij (2000), who reviewed a series of 39 studies investigating laypersons accuracy at detecting deception and found the overall accuracy rate of 56.6%.The accuracy range in most studies is the 54% to 56%, and in none of the experiment either lower than 30% or higher than 64%. The observers showed higher accuracy rate for truthful statement (67%) compared to deceptive statement (44%), which indicated that observers are more likely to consider that messages are truthful than deceptive. It has been suggested that because most participants were lay persons, who are more often confronted with truthful than with deceptive statements in daily life (O’Sullivan, Ekman, & Friesen, 1988). However, these findings may not be applicable to the performance of professional detectors, such as police officers. Because they are more practiced, this may increase their detection accuracy.
Ekman and O’Sullivan (1991) examined 509 professional lie catchers to detect deceit, including law-enforcement personnel, such as members of the Secret Service, police officers, as well as college students and working adults. The results showed that certain group do better than others, it has shown that members of Secret Service (64% accuracy rate), Central Intelligence Agency (73% accuracy rates), and Sheriffs (67% accuracy rates) were better lie detector than other groups. The findings suggest that lie catchers used different information than did the inaccurate ones. They relied on more varied behaviours, and emphasized nonverbal more than verbal ones. The finding also showed that accuracy in identifying micro-expression test was correlated with overall accuracy. According to this assumption, researchers (Zuckman, Koestner & Alton, 1984; Kohnken, 1987) provided information to observers based on behavioural measurements and trained them in recognizing micro-expression. They also reported this benefit only in judging by person they had received training. However, one reason for why even professional lie detectors showed low accuracy rates is probably because they don’t know where to look at and hold the false cues to detect deceit. For example, most of police officers in this experiment used gaze aversion as a useful tool to detect deceit, but results showed is significant negatively correlated with use of gaze aversion.
It has been argued that the disappointing accurate rates may due to the artificial laboratorial setting, and the negative consequences of getting caught are not high enough for liars to exhibit clear deceptive cues to observers. Mann, Vrij and Bull (2002) conducted the first real life study to examine police officers’ skills when they detect lies and truths that are told in real life situations. There were two limitations (suspect and police officers spoke different languages; they were of different nationalities). Mann, Vrij and Bull (2004) re-did the experiment and overcame the limitations in previous study. They exposed British police officers to fragments of videotaped real life police interviews with English-speaking suspects and asked them to detect truths and lies told by these suspects during their interviews. The result showed the accuracy rates for lies (66.16%) and truth (63.61%). The accuracy rates were higher than those found in laboratorial researches, and it also showed that accuracy was significantly positively correlated with perceived interview experience and negatively correlated with the use of stereotypical cues, e.g. gaze aversion. However, the deception real life studies that have been published are often of poor quality, most of the researchers were not able to obtain video footage, establish the ground truth satisfactorily, and fail to select comparable truth (Depaulo & Friedman, 1998; Vrij, Mann, & Bull, 2006).
Some researchers have raised the stakes in laboratory experiment by offering money when they get away with their lies or telling them that being a good liar is an important indicator of success in a future career (Vrij, 1995; Ekman & Friesen, 1974). However, the most impressive experiment with even higher stakes is conducted by Frank and Ekman (1997). Participants were given the opportunity to steal US $50. If they could convince the interviewer that they had not taken the money, they could keep all of it. If they fail to convince the interviewer and being judged as lying, they had to give the US $50 back and also lost their US 10$ participation fee, and some participants faced an additional punishment, they were told that they could have to sit on a cold metal chair inside a cramped, darkened room ominously labelled XXX, where they would have to endure anything from 10 to 40 randomly sequenced 110-decibel starting blasts of white noise over the course of one hour. The results showed that ability to detect high stake lies generalizes across high-stake situations and is most likely to produce emotional cues. However, a study like this raises ethical issues, and the punishments are never realistic. Therefore, it may not be possible to exam high-stakes lies in laboratory experiments.
Reasons for the weak relationship between nonverbal cues and deception
The theoretical perspective, outlined above, already predicted that research would reveal only a few relatively reliable nonverbal behaviour indicators of deception (DePaulo, Stone, & Lassiter, 1985; Ekman, 1992; Vrij, 2000). Many explanations have been emerged for this limited accuracy, and several of these reasons will be discussed.
One explanation for not finding infallible cues is that the scoring systems used to measure them are not detailed enough. Research on detecting accuracy showed that many nonverbal behaviours such as gaze aversion, fidgeting, and speech pause are commonly related to deception, but they are not valid cues (Vrij & Semin, 1996). For example, frequency and duration of pauses combine were not related to deception. But it appeared liars pause longer but no more often than truth tellers when examining them separately. Also, Smith and Clark (1993) found that ‘ums’ indicates higher cognitive load than ‘uhs’. Therefore examining them separately may be indicative of lying. But research (Davis, Markus et al, 2005) indicates that ‘ums’ and ‘uhs’ are associated with truthful answers rather than false ones. The distinction between cues accepted as invalid and cues validated through research is important.
Ekman, Friesen and O’Sullivan (1988) found that liars make more false smiles than truth tellers, and truth tellers make more felt smiles. If the distinction between false and felt smiles is not made, the results would have shown that truth tellers smile as frequently as liars. Some researchers also failed to notice some specific movements liars make. Ekman and Friesen (1969, 1972) made a distinction between three hand movement categories: emblems (gestures with meaning, such as thumb up), illustrators, and self-adaptors. It has been investigated In Depaulo’s meta-analysis, and showed that emblems were not a diagnostic cue to deceit (Depaulo, Lindsay et al, 2003). However, Researcher may also failed to notice some others cues that are subtle and hard to detect by untrained observers (Davis, Markus et al, 2006). For example, micro-expressions of emotions are present only for a short period time, within 1/25 of a second.
The second explanation could be that different people show different nonverbal cues to deceit. The nonverbal cues to deception people display may be influenced by their personality. The empirical evidences show that extraverts display different and few clues to deception than introverts (Riggio & Friedman, 1983; Siegman & Reynolds, 1983). It might be reasoned that introverts feel more uncomfortable in social interactions than extraverts.
For people high in Machiavellianism, lying is a normal and acceptable way of achieving their goals, and they typically feel less guilt then others while lying, and they also don’t find lying too cognitively complicated (Kashy & Depaulo, 1996). It was also found that Machiavellians are more likely to engage in strategic self-presentation to influence others. Exline, Thibsuy et al, (1970) found people high in Machiavellianism kept more eye contact when lying than those in low Machiavellianism. People who find themselves good at acting also find lying easier (Gozna et al, 2001), and less intelligent people find it harder to lie (Ekman & Frank, 1993; Vrij & Mann, 2001).
According to Buller and Burgoon’s Interpersonal Deception Theory, people who are high in Public Self-Conscious, e.g. politicians, try particularly hard to control their behaviour to make a good impression, such as exhibiting less eye contact or avoiding speech pause (Baumeister, 1984; Gallaher; 1992), but the findings couldn’t be replicated in a deception experiment (Vrij, Edward, & Bull, 2001).
However, the nonverbal cues to deception can also be affected by ethnic origins and gender differences. For ethic origins or culture, as already discussed, people of different ethnicities hold different beliefs about cues to deception. Research (Sitton &Griffin, 1981; Vrij &Winkel, 1991) compared cues to deception in participants of different ethnic group, and no differences between ethnic groups were found. The difference in behaviour displayed by people from different ethnic origins may be the reason to lead to errors in lie detection, e.g. looking conversation partners in the eye. However, most of deception studies have been carried out in Western countries, and may lack cross-culture. Researchers rarely report gender differences in their detection experiment, because gender differences between truth tellers and liars are unlikely to occur, and there are no theoretical reasons why differences may occur in many situations (Hall, 2006).
Although researches show no diagnostic cue to deception, a pattern emerges when the combination of cues is taken into account. Frank and Ekman (1997) found up to 80% of truths and lies could be detected when a trained observers paid attention to micro-expression, but 86% of truths and lies could be detected on the basis of a combination of micro-expression and the tone of voices (Ekman & O’Sullivan, 1991). It’s also supported by Vrij, Edward et al. (2002), found that accuracy rate 70.6% (truths) and 84.6% (lies) at detecting deceit when combines four nonverbal cues (illustrators, hesitations, latency period, and hand/finger movements). Other studies (Davis, Markus et al. 2005; Vrij, Akehurst et al, 2004) also showed high accuracy rates, which is the 71% to 78% range when researchers investigated a combination of behavioural cues. However, researchers suggest that more accuracy lie detection can be made if a combination cues is examined rather than each of cues individually.
Summary and Conclusion
This review focused on three major issues: (1) theoretical cues that associated with deception; (2) accuracy of lie detection based on nonverbal cues ;( 3) reasons for weak relationship and low accuracy between nonverbal behaviors and deception.
The theoretical perspectives discussed above make clear that the relationship between nonverbal behavior and deception is complicated. In an extension of leakage hypothesis, outlined above (Ekman & Friesen, 1969), three theoretical perspectives (the emotional, cognitive load and attempted control approach) are believed that can influence a liar’s nonverbal behaviors (Zuckerman, Depaulo & Rosenthal, 1981). Deceivers may experience specific affects (e.g. guilt), they may be engaging in a more complicated processing, and they may also try to control their nonverbal displays. Depaulo, Lindsay et al.’s (2003) meta-analysis of cues to deception provided general support for the theoretical perspectives discussed previously that cues associated with those three states may be most likely the indicators of deception. None of these approaches claim that the presence of these signs necessarily indicates deception. There are no theoretical perspectives that predict the diagnostic nonverbal cues that may only occur when people lie.
Research show that human ability to detect deception simply by observing the sender’s behavior is poor, by showing accuracy rate vary from 45% to 60%. The professional lie detectors (e.g. police officers) also seem to perform poorly, none of the experiments showed accuracy rate higher than 75%. It has been argued that observers in experiments lack of skill and have false beliefs about cues to deception. Research has shown that observers improve their skills in detecting deceit if they received some information about relationship between nonverbal behavior and deception. Moreover, there are a number of methodological concerns. Most of the experiments are laboratory studies. Participants do not choose to lie but instructed to do by experimenter, and the stakes are never really high, but the deception filed studies also show the problem with video footage, ground truth, and comparable truths.
Research showed that examining a combination of cues provided a high hit rate in accurately detecting whether someone was lying or telling the truth.
Many explanations for few nonverbal cues to deception, most of the cues are invalid. It may be the result of inadequate scoring systems. It will be easier to find the agnostic cues to deception if observers examine nonverbal response in more detail. It may further make a difference if the frequency of occurrence or the duration of each deliberate behavior can be measured. For example, the onset time, offset time of micro-expression. Furthermore, group differences should also be taken into account, e.g. Machiavellianism, self-monitoring. The gender differences are also necessary to distinguish, but there is no theoretical evidence to support any of the hypotheses (Hall, 1980). More valuable information can be found by using a cluster of nonverbal cues, it can also explain why behaviors such as tension, ambivalence, pupil dilation, and lip pressing are emerged as cues to deceit. The situational factors may also affect to draw the conclusion, such as high or low stakes, motivation of liars. Therefore, observers need to consider the individual differences and circumstances by examining their changes in behavior under similar situation. The scope of review is limited to adults, the effectiveness of deceiving may be markedly different in children.
For future research on the cues to deception, more filed studies should be examined in order to have high stakes, and should also focus on the way that a person is interviewed in the real life in terms of the ‘Interpersonal Deception Theory’. A number of questions also needed to be addressed. For example, which behaviors should be clustered? Can the results of a combination of cues generalize across different situations?