Hiding and Detecting Deception
This chapter considers two broad questions: Can a deceiver be detected under conditions in which he or she is generally unaware of being suspected? Can a deceiver avoid being detected under those conditions? These questions largely rule out the use of polygraph tests, “truth ” serums, and other invasive techniques for detecting deception, and no attempt is made to review the large literature on those techniques. (Interested readers are referred to reviews of the literature and discussions of the complex issues surrounding the use of invasive lie-detection techniques; see, especially, Ben-Shakhar et al., 1982; Lykken, 1981, 1985; Raskin, 1987.)
The possibilities for detecting deception under noninvasive conditions depend either on the content of the supplied information or on nonverbal behaviors inadvertently “leaked” by the subject. The content of the supplied information may be suspect because it is inconsistent with other information supplied by the subject or because it is inconsistent with information from other sources. The nonverbal information may also be suspect because of divergences from expressions or postures seen in other situations or because of known deception clues observed by the detector. In this chapter we deal primarily with deception clues that can be detected from nonverbal behavior.
The key assumption underlying the hope for detecting deception from nonverbal behavior is that a person, when knowingly lying or deceiving, enters a psychological state that differs from the state he or she is in when telling the truth. These psychological states, the deceptive and the honest, are further assumed to affect both verbal and nonverbal channels
of communication. These channels, in turn, can be monitored by a detector's unaided perceptual processes. When one or more channels provide cues that “betray” the person's deceptions, researchers claim that leakage has occurred.
Within this framework, several possibilities arise. The simplest is that deception leaks in predictable ways for all people regardless of differences in experience or culture: the predictability would derive from a fixed set of communication cues, such as body movements, rather than facial expressions. More complicated are the possibilities that a variety of behavioral patterns are diagnostic, depending on a person's experience, training, and cultural background or the particular circumstances under which he or she is being observed. Moreover, it is possible that individuals are conscious about such cues and can manage the expressions or movements being monitored by an observer.
This chapter reviews ongoing research programs that focus on those issues: what has been learned from laboratory studies on deception and what needs to be learned through further experimental research. (The next chapter is more conceptual, reviewing various taxonomies and frameworks designed to broaden the concept of deception.) We also discuss the issue of cultural and other subpopulation differences in the interpretation and execution of deception in the context of generality of findings obtained in laboratory studies with American subjects.
PHYSICAL INDICATORS AND PSYCHOLOGICAL STATES
Although the problem of inferring specific psychological states from specific observed behaviors is quite complex, there is little doubt about the existence of linkages between internal states and external behaviors. A large research literature on nonverbal behaviors associated with emotions and intentions has made apparent the value of using such behaviors as windows on underlying psychological states.1 The focus of this section is more specialized: to draw on relevant parts of the large literature on nonverbal behaviors for insights and suggestions for research on nonverbal indicators of deception.
The Laboratory Paradigm
Most of the contemporary experimental research on deception involves the communication of false verbal statements made with intent to deceive. Researchers are interested in the cues sent by deceivers and the cues used by receivers to detect deception; the cues consist, for the most part, of nonverbal behaviors coded systematically from replays of videotaped simulated performances (speeches, interviews), usually en-
acted by college students. This situation represents a particular type of lie captured by Knapp and Comadena's definition (1979:271) as “the conscious alteration of information a person believes to be true in order to significantly change another's perceptions from what the deceiver thought they would be without the alteration.” It is less relevant to other forms of deception (as discussed in the next chapter).
The experiments have examined deception from the standpoint of a sender, who is perpetrating a lie or being honest, and from the standpoint of an observer, who attempts to detect whether a particular communication is deceptive or truthful. Experiments concerned primarily with senders ask subjects to convey false (or true) information, usually in response to questions from an interviewer who may be evaluating candidates for a job, interrogating suspects about a crime, or asking about issues in world affairs or local politics.
One series of studies asked subjects to take the role of the Soviet ambassador to the United States. In that role, subjects were interviewed for 15 minutes (in a “Meet the Press” format) about three then-current events: the Soviet invasion of Afghanistan, Soviet military activities in Cuba, and Soviet troops in Europe. Three experimental conditions consisted of misrepresenting actual Soviet policy (deception), avoiding direct answers to the questions being asked (evasion), and conveying actual Soviet policy as communicated to you (honesty). Other types of scenarios used in these experiments include car sales, making speeches about academic topics, conveying feelings about hospital scenes, and interactions stylized in the form of candid-camera situations. In all of these experiments, subjects' nonverbal behaviors are coded in each of several channels of communication: facial expressions, body movements, visual behavior, paralanguage (verbal utterances other than the content of the message). The coded behaviors are analyzed for differences by condition, which is usually a comparison of deceitful and honest communications.
Experiments concerned primarily with observers ask subjects to rate taped enactments of senders in a previous experiment or actors whose scripts were prepared by the experimenter. For example, in one experiment, observers were asked to rate the sales pitches made by salespersons with experience on incentive pay plans and automobile customers with experience in bargaining for trade-in allowances. In another experiment, observers rated the descriptions of student nurses' feelings about a film that was intended to arouse either positive or negative reactions; nurses in the deception condition were asked to conceal their actual feelings aroused by the negative film. In all of these experiments, observers report their perceptions of the communicators' deceptiveness, usually on a numerical scale that discriminates degrees of de-
ceptiveness or honesty. Observers are also asked to indicate the extent to which they are confident in making these judgments. A variety of other impressions of communicators have been assessed in experiments, including, for example, liking for the product, anxiety, confidence, suspicion, aggressiveness, and comfort. The ratings of deception are analyzed for accuracy and related to expressed confidence. They are also correlated with particular nonverbal channels or behaviors used in the enactment and with the other impressions rated by the observer. The results obtained from both types of experiments are reviewed in the following sections, focusing first on the nonverbal behaviors displayed by subjects and then on the accuracy of observers who attempt to detect the deception portrayed by others.
The relevance of nonverbal behaviors as clues to deception derives from the assumption that these behaviors, in particular, will leak information that a person is trying to hide. The rationale for this assumption is provided by Ekman and Friesen (1969) in terms of the variables of automaticity and controllability. These variables refer to the relationship between emotions and behavior. According to Ekman (1981:271): “When emotion is aroused certain changes occur in the face, body, and voice which can be considered automatic. ” Such automatic links present a problem for deceivers who attempt to control their nonverbal behavior during deception: they must override these links in order to avoid leakage. Some nonverbal cues are more likely to be controlled than others. Words and facial expressions have been found to be easier to control than body and tone of voice cues (Ekman, 1981; Zuckerman et al., 1981). Thus, deceivers should be most successful when using words, next most successful when using facial expressions, and least successful when using body movements and tone of voice cues. Indeed, according to the DePaulo et al. (1985) review of the evidence on detection, both body cues and tone of voice cues were found to be more revealing than facial cues. Less clear, however, are the results on verbal cues: these cues were found to be informative and revealing (see DePaulo et al., 1985, for an explanation).
Nonverbal cues that reveal deception to observers should also distinguish between the actual enactments of deceivers and truth tellers. In their early experiment on leakage, Ekman and Friesen's (1969) deceivers claimed that they attempted to disguise the face more than the body while perpetrating the deception. The discriminating cues provided by the body included decreasing use of hand illustrators (movements used for emphasis), increased use of self-adapters (movements directed to
ward one part of the body by another), and more postural position shifts for deceivers. More illustrator movements also were used by the deceivers (compared to truth tellers) in Mehrabian's (1972) study, while increased self-adapters and postural shifts were also found for the deceiving subjects in the McClintock and Hunt (1975) interview study. Further confirmative evidence was obtained by the Druckman et al. (1982) finding of coordinated body movements for deceivers: rocking, headshaking, trunk swivels, and nodding movements varied together. That study also found more fidgeting with objects for deceivers and more leg movements for evaders (indirect deception). Two studies reported differences in vocal cues: Mehrabian (1971) showed more speech errors as deviations from baseline data; DePaulo et al. (1980) found tone of voice differences between deceivers and truth tellers. Taken together, these findings suggest where to look for clues to deceptive intentions. They do not, however, provide evidence for the specific role played by psychological states or emotions.
The concept of leakage is based on a model that construes psychological states or emotions as variables that mediate the relationship between nonverbal behaviors and intentions (see Druckman et al., 1982: Figure 1.1). While demonstrating the importance of these intervening variables, the results reported by Druckman et al. (1982) also illustrate the complexity of the relationships. Subjects (interviewees playing the role of the Soviet ambassador to the United States) in each of three experimental conditions (honesty, evasion, and deception) indicated their feelings on a postinterview questionnaire consisting of 26 bipolar adjective scales. The ratings were analyzed for condition differences and for correlations with each of 25 nonverbal behaviors coded for all subjects. Three types of states seemed to capture the diverse relationships obtained among these variables: involvement, stress, and confidence. Honest and evasive subjects displayed behaviors whose meaning indicated involvement. Evasive and deceptive subjects displayed behaviors that correlated with feelings of stress or tension. Subjects in all three conditions showed behaviors related to feelings of confidence or effectiveness. This evidence suggests which emotions reflect primarily honesty (feelings of involvement), direct and indirect deception (stress), and role-playing, which is characteristic of all three conditions (confidence). Interestingly, evasive subjects showed relations between nonverbal behaviors and two emotions, one found primarily for honest subjects (involvement), the other found primarily for deceivers (stress). This result is consistent with the definitions of evaders as being neither completely honest nor completely deceptive.
The results also identified a subset of nonverbal behaviors that indicate either high or low involvement, stress, or confidence: for example,
looking at versus looking away from the interviewer as an indicator of high or low involvement for honest subjects or as an indicator of stress for evaders, shoulder shrugs versus trunk swivels as an indicator of high and low stress experienced by deceivers (see Druckman et al., 1982:Table 4.10). Supporting the concept of leakage, these findings highlight the role played by emotions or feelings in the process of making inferences from nonverbal behaviors to intentions. However, they also make evident some of the difficulties involved in the task of diagnosing deception or other intentions. In some cases, different nonverbal behaviors have similar meanings for different intentions, that is, they are functionally equivalent behaviors. In other cases, different nonverbal behaviors reflect different feelings for the same intention, as well as for different intentions. The challenge for researchers is to isolate the unique relationships between particular nonverbal behaviors and underlying feelings for particular intentions. These would be cues that diagnosticians could use with confidence; they have not yet been identified.
There seems to be a relationship between motivation to deceive and success at deceiving (DePaulo et al., 1985). The question asked is whether highly motivated deceivers are easier or harder to detect than their less well-motivated counterparts, that is, when the stakes for successful deception are either high or low. This question has implications for differences between professionals (higher stakes) and the amateurs used in most laboratory experiments (lower stakes). Results obtained from two experiments (summarized in DePaulo et al., 1985) are quite intriguing. When motivation to lie was high (e.g., deceivers were being monitored by a panel of their peers), nonverbal cues were more revealing than verbal clues; when motivation was low (e.g., deceivers were to play a “little” game), verbal cues were more revealing. The highly motivated deceivers were accurately detected by observers more often than the less highly motivated deceivers if the observers could see the nonverbal behaviors. When judges only had access to verbal statements, the highly motivated deceivers were quite successful (and more successful than their less motivated counterparts) in their deception. According to DePaulo et al. (1985:334): “the harder senders try to get away with their lies, the less successful they will be (when their nonverbal behaviors are showing).”
An explanation for these findings is that when emotionality is high, as it would be for high-stakes' performances, nonverbal leaks are more likely to occur. The highly motivated deceivers in the experiments of DePaulo and her colleagues showed more behavioral inhibition and ri-
gidity, as well as more negativity, than the deceivers who were less motivated. The motivated deceivers may think that they are using a safe strategy by not moving, which they equate with not revealing, but this strategy backfired for reasons understandable in terms of the concept of leakage. Based on an hydraulic model analogy, attempts to control one channel of communication (e.g., facial expressions) leads to excessive activity in another channel (e.g., body movements). A more direct test of this hypothesis would be a comparison of conditions in which subjects are told to control their faces or their bodies while deceiving; more body activity in the former condition and more facial activity in the latter (in comparison with a baseline control condition) would support this hypothesis. This experiment has been done, and analyses are in progress.2
An implication of this research is that people are betrayed by their nonverbal behaviors in situations in which they are most motivated not to be detected. This implication is especially relevant for the challenges confronting professional deceivers (such as spies). From the standpoint of detectors, it would seem appropriate to monitor the unattended channels, that is, either those channels under less control by the deceiver or those that he or she believes are not being observed. For nonprofessionals, whether or not the challenge for successful deception is high, nonverbal deception cues are likely to be derived from different emotions: insecurity and concern about the managed impressions being conveyed (due in part to a lack of practice), guilt or anxiety about a morally dubious act, tension caused by the complexity inherent in a deceptive performance. The nonverbal behaviors that derive from these feelings may well differ from those that are the result of leakage. However, deceptive contexts vary considerably on a number of dimensions, and these dimensions of the situation affect both the displays shown by senders and impressions made by observers.
It is clear from the research on detecting deception that people have trouble distinguishing between lies and truthful statements. For samples of college students, there is a discernible bias in favor of judging messages as truthful when the frequency of lies and truth is about 50-50 (DePaulo et al., 1985). People who are professional lie detectors, on the other hand, evince a “deception bias,” that is, they judge most (truthful and deceptive) messages as being deceptive (O'Sullivan et al., 1988); also see Kraut and Poe's (1980) study of customs inspectors, Hendershot and Hess' (1982) study of police detectives, and the Druckman et al. (1982) study of oil company executives. Both types of samples usually
do no better than chance in accurately detecting deception: college students judge actual truthfulness more accurately than actual deception, while professionals are somewhat more accurate at judging actual deception than actual truthfulness. Despite the poor accuracy, however, both college students and experts indicate high confidence in their judgments (Druckman et al., 1982), and confidence may actually increase with experience (DePaulo et al., 1985). Of interest, then, is the question of why it is so difficult to detect deception.
The weak correspondence between actual and perceived deception cues suggests that the folk wisdom about how deceivers are supposed to act (e.g., less gaze, smile less, postural shifts, longer response time) are not diagnostic of deception. They are, however, widely held cultural stereotypes, and they are easier to predict than the actual cues that are likely to be more sensitive to context (DePaulo et al., 1985; O'Sullivan et al., 1988). Recent results reported by Bond et al. (1990) suggest that the stereotypes may be cross-cultural: their American and Jordanian subjects used similar nonverbal cues in making judgments of deception, which were inaccurate, evincing what is known as a “demeanor bias” (judgments based on the target's apparent honesty or dishonesty rather than the perceiver's detection skills). Modeling indicates that the dimensions used by judges of deception were the well-known impressions of appearing apprehensive and undependable, often looking away from the camera, fidgeting with objects, and a relatively poor sense of humor (Druckman et al., 1982). Yet, these impressions were misleading. The connection between observed behavior and impressions, on the one hand, and between impressions of emotions and intentions on the other were inaccurate.
One proposed explanation for these results is that detectors in the experiments rarely get feedback about whether the sender was actually lying or telling the truth. Without feedback, detectors cannot correct their judgments and the cues used in making these decisions. Another proposed explanation is that detectors are often insufficiently suspicious about the possibility that deception might be occurring; without being alert to the possibility, detectors are unlikely to search for diagnostic cues. Neither of these explanations is supported by research, however. Feedback does not serve to revise detectors' theories; rather, they tend to select “evidence” that supports their initial judgment and, in fact, increased their confidence in the (inaccurate) judgments (DePaulo et al., 1985). Similarly, increased suspiciousness does not improve accuracy although it may serve to decrease confidence in judgments. Other proposed explanations for the difficulty of detecting deception (O'Sullivan et al., 1988) include: the task of detecting strangers (used in most studies) is more difficult than detecting friends; detectors confuse the
distinction between negative and positive emotions with lying versus telling the truth; some types of lies are easier to detect than others; and some liars may be better at lying than others. Each of these explanations is an empirical issue that can eventually be resolved by research; until then, it should be possible on the basis of what is known to improve accuracy even for detecting the lies of strangers in fleeting situations, for uncovering subtle lies, and for recognizing skilled liars.
Detecting another person's intentions is a learned skill. As with other skills, it has been shown to improve with age and training. A number of studies (summarized in DePaulo et al., 1985) have shown that during adolescence, perceivers become increasingly sensitive to issues of sincerity and deceit (e.g., Flavell, 1977). The studies also indicate that children do progressively better as they get older at distinguishing between direct deception, on the one hand, and exaggerations, omissions, or slight distortions on the other.
Sensitivity to nonverbal deception clues contribute to detection accuracy. Such sensitivity may be related to the nature of one's job experience. A recent experiment assessed the detection accuracy of 509 professionals from agencies and occupations in which lie detection is part of the job: the Central Intelligence Agency, the Federal Bureau of Investigation, the National Security Agency, police, and judges (Ekman and O'Sullivan, 1991). Their results show that accuracy was not related to occupation and actually decreased with age and experience for some occupational groups. Closer examination of the accurate detectors revealed, however, that the critical variable may be job function. People in jobs in which they must interact with others on a regular basis were better detectors than those in administrative positions; the former tended to be younger than the latter due at least in part to promotions of relatively older workers to administrative positions. In addition to job tasks, frequent social interaction helps people detect deceivers. Knowing what to look for—in the sense of distinguishing critical from noncritical cues—and how to process those cues helps more.
Improvement can be accelerated further by training that is designed specifically to take advantage of relevant information. Lie-detection success improves when subjects are prompted to pay special attention to tone of voice cues (DePaulo et al., 1982; see also Zuckerman et al., 1985). Success also improved for subjects who are given a technical briefing intended to separate the nonverbal signals from the nonverbal
“noise” (Druckman et al., 1982). In addition to specific nonverbal cues, subjects have been shown to do better when they pay attention to behaviors that are related to deceptiveness, such as ambivalence or tension (DePaulo et al., 1982).
It is clear that specialized training can make a difference. Less clear, however, are the specific dimensions of training likely to have the strongest impact. Three of the dimensions suggested are whether to focus on particular behaviors or psychological states (see above), to examine the difference between judging friends versus strangers (see O'Sullivan et al., 1988), and to compare different groups of detectors who vary in terms of their motivation and experience (see O'Sullivan et al., 1988). With regard to the difference between cues shown by deceiving friends and strangers, Buller and Aune (1987) found that friends attempt to control their arousal and negative affect cues more than strangers, although other cues were leaked by friends perhaps as a result of the masking.
The judgment heuristics used by observers are also interesting. An evaluation of alternative detection strategies indicate that providing observers with an information-processing approach to detection improves accuracy considerably. Referred to as “inference training,” the strategy consists of an organized method for processing information about several critical cues (see Druckman et al., 1982:Figure 5.1). Significant gains in accuracy were found for subjects who were trained to distinguish between relevant and irrelevant cues (a technical briefing) and presented a strategy for processing the relevant cues. Judgments about another 's intentions are based on both available information (observed cues) and the way the information is processed (judgment heuristic). An inappropriate strategy can lead to the wrong judgment, despite accurate information about the relevant cues.3
The gains obtained for inference training are consistent with research findings from cognitive social psychology (see, e.g., Nisbett and Ross, 1980). By taking account of combination rules and sequencing of behaviors, a processing strategy captures the concept of “display package,” that is, the way that nonverbal behaviors are displayed for particular emotions or intentions. To be useful, however, the strategy must process known deception cues. The research uncovered the relevant cues in a first stage, followed by training detectors to use them in arriving at decisions about actors whose displays were based on the earlier findings. This requirement poses a dilemma for detection in situations in which the cues are unknown. The detector must either rely on the possibility of universal cues or suspend judgment until he or she has had more experience with the suspected deceiver. But it is unlikely that there are universal cues (see discussion below on context), and suspend
ing judgment may not be viable under certain circumstances, e.g., a fleeting interaction with a suspected terrorist. Although there are no obvious nostrums to guide a detector in these situations, there may be useful ways to approach the problem (see next section on the role of analytical methodologies).
Further progress on the problem of detecting deception may depend on advances in methodology. One promising avenue is to apply the sophisticated approaches developed in other fields in which systematic diagnostic efforts have been made. Recent work in radiology provides a good example of decision aids that can determine the maximal accuracy provided by the cues. The aids address the general question: What are the relative weights of the possibly diagnostic clues (when their scale of values are merged to determine a probability of deception) that will maximize detection accuracy in each of the relevant deception contexts (see, e.g., Swets et al., 1991; Getty et al., 1988; Swets, 1988). Answering this question would go a long way toward a technique for training detectors.
Using Analytical Methodologies
The knowledge base for understanding the physical indicators of deception consists almost entirely of laboratory studies (reviewed above). A concern is whether, and to what extent, the results of those studies have relevance for other settings. Commenting on differences between the conditions of laboratory research and less formal observations outside the laboratory, Doob (1975) distinguishes between “professional ” and “lay” observers. Professionals systematically decode nonverbal behaviors, channel by channel, from replays of videotaped performances (usually enacted by college students). Lay observers gain impressions of psychological states from a large number of relatively fleeting expressions, body movements, or vocalizations seen in particular social or cultural settings. Based on these different sources of information and approaches, Doob suggests two possibilities: experimental results do not aid the task of deciphering clues in the real world, and lay observers cannot use a professional's systematic approach for inferring intentions from behavior (see Doob's [1975:115-135] discussion of nonverbal symbols).
Doob's grim outlook for applying laboratory results and deriving accurate inferences from informal observations is only one interpretation of the problem. Another interpretation provides a more optimistic outlook in three ways. First, experimental results can highlight what to look for in broader terms than specific nonverbal behaviors (e.g., abrupt changes, leaks), and some findings have widespread generality (e.g., micromomentary facial expressions for emotions, sources of detection
inaccuracies). Second, the approach of systematic observation with notation or coding rules, used in laboratory studies, can also be used for making informed observations outside a laboratory. Third, a lay observer has the advantage of viewing behavior within its context or cultural setting, helping to improve interpretations.
A lay observer can benefit from the laboratory-based research on deception clues. The research serves to direct an observer's attention to particular themes: for example, incongruities or inconsistencies between verbal and nonverbal behaviors for indicators of stress or irritability, tone-of-voice deviations from baseline observations for possible deception, and an increased intensity of behaviors in a variety of channels for commitment to policies. These themes call attention to relationships, particular nonverbal channels, or to the amount of feelings (see Rozelle et al., 1986, for an elaboration of these themes). Also, paying attention to qualities related to deception, such as ambivalence, indifference, and tension, has been shown to increase detection accuracy (DePaulo et al., 1982). Actual cues to deception reflect feelings of being less confident, more insecure, more self-conscious (due perhaps to a lack of practice at deceiving), and guilt since the act is morally dubious in most societies. This is useful information, even though it is based on averages across many subjects in different experimental situations. Intuitions about these and related qualities may improve lay observers ' detection accuracy.
Laboratory research has contributed techniques for analysis that are also useful in field settings. Notation systems for coding nonverbal behaviors enable an analyst, who must respond to time-sensitive requests, to identify a subset of behaviors for on-the-spot commentary. A government or corporate analyst is frequently asked to provide interpretations without the benefits of penetrating analysis, extensive video footage, or hindsight, and coding systems are especially useful under these conditions. They provide an analyst with a structure for focusing attention on relevant details: for example, behaviors that change quickly (micro-momentary expressions) or obviously (incongruities), and those that occur within the time frame of a statement (leaks). These are the kinds of observations that can be used for making inferences from limited data. They may also be useful for an analyst who has access to more data (extended videotape coverage) and is given more time to develop interpretations. Under these conditions, it is possible to perform systematic analyses that include charting trends over time, making comparisons, and developing profiles. Each of these tasks is aided by a coding system that encompasses a range of situations, purposes of appearances (speech, interview, casual interactions), and verbal statements, as well as types of displayed nonverbal behaviors (channels and specific movements or expressions). By categorizing videotapes in this manner, it
becomes feasible to test hypotheses about relationships, for example, between situations and observed behaviors. Analyzed over time, these relationships would highlight the influence of context (as varied situations) on nonverbal behavior.
Unlike laboratory researchers, field analysts (and lay observers) can develop interpretations based on observations of “subjects” in real world contexts. In essence, he or she can follow a subject, and, by so doing, record nonverbal behavior displays that occur under a variety of circumstances. These comparisons contribute to the development of profiles of coordinated movements that may change over time and situations. The coordinated movements can also be depicted in animated graphic displays (see Badler and Smoliar, 1979, for examples). Such displays illuminate “postural” or expression differences in the same people over time and also among people.4 When associated with specific events and context, the displays lead to the question: How are the feelings and intentions that are evoked by different situations represented in body movements? When compared with displays shown by people in other cultural settings, the coded observations raise the question: What are the contributions of culture to observed nonverbal behaviors?
CONTEXT AND CULTURE
There is no reason to expect that a particular set of nonverbal behaviors is associated with every type of deceiver and every type of deceit. As noted by DePaulo et al. (1985:343):
Deceptive contexts vary considerably in the emotions, motivations, and expectations that they engender in the deceivers, and they also impose different cognitive-processing demands. As these and other important dimensions of deceptive contexts vary so, too, may the specific behaviors that distinguish lies from truths.
In other words, the critical physical clues to deception are likely to be situation and culture specific. Behaviors observed in one setting are likely to generalize only to other similar settings. A more complete understanding of the nonverbal indicators of intentions would result from frameworks that match behaviors to situations, that is, research-based taxonomies of situations and behaviors. Without such a framework, it is difficult to know the extent to which findings obtained from experiments with college students or amateurs apply also to people in occupations in which deception is assumed to be practiced and that people would like to detect (see, e.g., Doob, 1975). Referred to as a context-relevant approach to research (Druckman et al., 1982; Mahoney and Druckman, 1975), these observations are captured also by Ekman's (1978) concept of display rules.
Defined originally with regard to facial expressions, display rules are culturally learned “rules” for managing expressions. They serve to modify expressions in keeping with the demands of the social situation and may also reflect the cognitive processing that precedes the observed expression. Whether display rules are the primary determinant of observed nonverbal behaviors is an empirical issue that is captured in terms of the debate between universal and culturally specific bases for expressions. This debate has been the basis for a large number of cross-cultural experiments on facial expressions by Ekman and his colleagues. While not resolving the debate clearly in favor of the one or other explanation, their interpretation of the findings suggests that while the distinctive appearance of the face may be universal—that is, unique muscle movements are made for each primary emotion—the particular expression that appears in certain circumstances varies with culture (see below). However, these experiments have not investigated the role of culture in other channels of nonverbal expression. Nor did they focus specifically on cues to deception.
Culture and Facial Expressions
In spite of the fact that most literature on nonverbal behavior notes the importance of context and agrees that expressions of senders and theories of detectors are conditioned by rather specific environments, there is very little evidence that demonstrates such relationships. The number of cross-cultural studies is small, and there are practically no studies on cultural similarities and differences in deceiving or detecting deception. One issue, however, has received considerable attention by Ekman and his collaborators, namely, the debate between universal and culturally specific facial expressions.
Channel-by-channel reviews of nonverbal behavior studies indicate uneven attention to cultural factors (Harper et al., 1978; Druckman et al., 1982). The reported studies on kinesics (body language) have been largely descriptive, consisting of naturalistic observations and informal commentary (e.g., Argyle, 1975). One study of cultural differences in gazing behavior is probably the most extensive to date on that topic (Watson, 1970): Arabs, Latins, and Southern Europeans focused their gaze on the eyes of their partner, while Asians, Indian-Pakistanis, and Northern Europeans showed peripheral gaze or no gaze at all. Several cross-cultural studies on proxemics (the use of space) support Hall's (1966) observation that behaviors related to space are quite sensitive to cultural variation: for example, Watson and Graves (1966) found ex-
pected differences between Arabs and Americans in their space behavior and body orientations (see also Sommer, 1959; Little, 1968; Shuter, 1976; Baxter, 1970). In addition, studies have shown cultural differences in touching behavior (Watson and Graves, 1966; Jourard, 1966; Shuter, 1976).
The issue of universal nonverbal expressions versus culturally determined expressions has been probed in a series of studies on facial expressions. Defining the issue in terms of “relative importance,” Ekman developed an approach that allows for both universal and culturally specific sources for expressions of emotion (see, e.g., Ekman and O'Sullivan, 1988). According to this “theory,” eliciting events vary from culture to culture but the particular facial muscle movements triggered when a given emotion is elicited are universal. (Blends of emotions may be subject to more cultural variability than the primary emotions.) That is, although the underlying physiology for the primary emotions may be universal, the actual expression elicited is subject to cultural and situation-determined display rules. Display rules serve to control an expression or to modify or mask certain expressions that would be socially inappropriate or would reveal deception. Evidence relevant to this theory is presented in a series of experiments reported by Ekman and his colleagues.
One series of experiments explored the extent to which respondents from different cultures agreed on the facial configuration signifying contempt as opposed to anger or disgust (Ekman and Friesen, 1986; Ekman and Heider, 1988). They found that the expression is widely recognized as contempt (75 percent agreement across cultures) just as the “anger expression” was seen as signifying anger (74 percent agreement) and the “disgust picture” showed disgust (73 percent agreement). Less clear, however, is the explanation for such universality: agreement could be the result of either species-specific learning or biological evolution. If the result of learning, the expressions would be characteristic only of humans; if the result of evolution, they should also be observed in other species, such as primates. A resolution of this issue depends on demonstrations of cross-species similarities in the underlying physiology. Such a demonstration has not yet been presented.
In a particularly ambitious cross-cultural study (Ekman et al., 1987), respondents from ten cultures were asked to identify facial expressions as indicators of one or more of six emotions—happiness, surprise, sadness, fear, disgust, anger. Three pictures were selected for each of the six emotions. High levels of agreement were found for judgments of the strongest emotion, the second strongest emotion, and the relative strength of two different expressions of the same emotion in each set of pictures. The majority of the observers in every culture judge the emotions as
predicted. The authors claim that these results are strong support for the view of universal facial expressions (although not for the species-specific learning versus evolutionary explanations of such agreement). However, several aspects of their research design may have biased the results in favor of high agreement: their procedures serve to take the expressions out of a social context, eliminate simultaneous speech and body cues, freeze the expression in a photograph, force special attention to it, and ask for judgments from detached uninvolved observers. Further studies are needed to observe how facial expressions change in different social contexts. Following Doob' s (1975) advice, these studies should embed the expression in textual material in order to capture the content as well as the tone of the emotion. Of interest, then, is the question: Would similar levels of agreement be obtained for expressions viewed as part of a cultural context?
The issue of universality versus cultural specificity has not been addressed systematically in relation to other nonverbal channels. However, there is indication of considerable variation in the display of emblems—physical acts that can fully take the place of words. Ekman et al. (1984) report regional, national, and intranational (urban versus rural) variation among respondents in their survey of emblems and indicate that a geographical dictionary of emblems would be a useful compilation for researchers as well as travelers. Such work would raise a question of interest: Does the source of emblems differ from facial expressions with regard to the issue of universality versus cultural specificity?
Finally, in terms of the central theme of this chapter, we located only two teams of investigators who have been recently exploring cultural influences on deceptive enactments. Comparing Chinese experimental liars with truth tellers, Cody et al. (1989) report some similarities and some differences to findings reported in earlier studies with American subjects (see also Yi Chao, 1987). Like their American counterparts, Chinese liars found lying to be a more difficult challenge than telling the truth: they experienced difficulties in communicating detailed and thorough answers to the questions that required effort. (See DePaulo et al., 1985, for a review of the research with American subjects.) Overall, however, few nonverbal behavioral cues distinguished the Chinese liars from truth tellers: no differences were found in either arousal or visual cues, although some discrimination was found in the auditory channel. These findings may be explained in part by culturally learned constraints against the display of negative feelings. Both the liars and truth tellers were brief in communicating negative feelings; they smiled frequently and suppressed body and hand movements. While speech errors were related to deceit, many other paralinguistic variables related more strongly to question difficulty.
The researchers suggest that future cross-cultural research assess the more subtle cues to arousal and anxiety that are more sensitive to cultural influences, such as types of smiling, symmetry of expressions, and tone of voice. Another study shows that vocal stress may be a sensitive indicator of emotions associated with deception, at least for the male Chinese subjects in their sample (O'Hair et al., 1989). One implication of these findings is that different display rules operate in different societies, resulting in culturally specific expressions, some of which are manifest as deception cues. The mode of expression of nonverbal behaviors in various channels differs even if the same type of behavior is displayed. When stated as an hypothesis, this issue provides a basis for further research that would elucidate the cultural dimensions of deception clues.
Bond and his colleagues (1990) videotaped Americans and Jordanians while telling lies and truths about persons they liked or disliked. Similarities and differences in nonverbal displays between the cultures indicate some support for both culturally specific expressions and for universal expressions. They found similarities between the cultures on such nonverbal behaviors as smiling, head movements, blinking, and hand gestures. Differences were found on eye contact, movements per minute, and filled pauses: Jordanians displayed more of each of these behaviors. However, only one behavior, filled pauses, distinguished between the liars and truth tellers among the Jordanian subjects only: Jordanians displayed more filled pauses when lying than when telling the truth. It is little wonder then that the detectors from both cultures were inaccurate: very few nonverbal behaviors could be used as reliable cues to deception.
Three kinds of conclusions are suggested by this review of research related to deception. The evidence underscores the relevance of nonverbal behaviors as indicators of emotions and intentions, although it raises additional issues that can only be resolved by further research. It also suggests some strategies that can be used to detect intentions in real-world environments.
It is clear that a number of nonverbal behaviors are associated with deception. Particularly revealing are body movements and tone of voice changes, although a number of other behaviors also indicate deception in certain situations. Nonverbal behaviors reflect feelings or emotions more directly than specific intentions to deceive. Emotions —psychological states—can be considered as intervening variables, relating to observed behaviors and to inferred intentions. Nonverbal expressions
are influenced by culture and context, but some components of emotional expression may be universal and independent of cultural overlays.
Motivated liars are easier to detect than nonmotivated liars. When emotionality is high, as in the case of tasks with high stakes associated with outcomes, nonverbal behaviors can be especially revealing. The revealing behaviors are likely to be those in a liar's unattended channels, which are leaked as a result of behavioral inhibition or rigidity.
Detectors, both amateurs and experts, are generally inaccurate despite high levels of confidence in their judgments. Neither feedback nor enhanced suspiciousness seems to improve their accuracy, but training them and providing them with a plan for processing information can increase accuracy of detection. The increased accuracy is more likely to occur if the discriminating cues to deception are known, and processing strategies aid the process of weighing and combining cues for judgments. In the absence of known relevant cues, whether to rely on possible universal or situation-specific cues remains an unresolved research issue.
Many questions are raised by the tantalizing information on deception and nonverbal behaviors. What nonverbal behaviors are associated with forms of deception that are not considered to be direct lies? What are the specific connections between nonverbal behaviors, emotions, and intentions as these occur in different situations? What are the implications of conscious control of facial expressions or body movements? There is a need to develop a framework that enables an analyst to match behaviors to situations in order to specify more precisely the contextual basis for behavior. What are the effects of varying types of lies and types of liars on the accuracy of detection? What are the dimensions of detection accuracy? There is a need to develop analytical strategies that aid the task of inferring intentions from behavior when the discriminating cues are not known. Cross-cultural studies of nonverbal behaviors associated with deception are needed. Little is known about the culturally specific display rules for every channel of communication, not only those manifest in facial expressions. On this topic, the gap between scientific evidence and anecdotal reporting is especially large. Researchers could take advantage of the large number of foreign students currently attending United States universities as a source for cross-cultural comparisons.
In spite of the number and breadth of unanswered questions, laboratory research does suggest particular aspects of behavior that may reveal intentions, for example, relationships between incongruities in verbal and nonverbal behaviors and stress, deviations from baseline observations of tone-of-voice as an indicator of possible deception, and an increased overall intensity of behaviors for extent of commitment to poli-
cies. Techniques developed for laboratory research may also be useful. They provide a structured and systematic approach to analysis for detecting deception in a variety of situations. The techniques are especially useful to an analyst who has access to videotaped interactions or who can interact with a subject over long periods of time. It is important not to rely solely on the judgments of “experts” even if they claim to be highly confident in those judgments. Expert-based judgments should be verified with other data gathered and processed in a systematic manner.
It should be possible to learn to distinguish between the universal and culturally specific expressions of emotions. The primary emotions may evoke similar expressions from one culture to another, and combinations of emotions seem to be subject to display rules, with which detectors can become familiar.
Leakage cues are likely to be evident in a highly motivated liar. A professional may be more likely to show signs of deception in unattended channels than an experimental subject, who operates in an environment in which the stakes for getting caught are relatively low.
1. Comprehensive reviews of this evidence can be found in a number of sources. Harper et al. (1978) provide an ambitious summary of the literature organized by channel of communication, that is, paralanguage, facial expression, kinesics or body language, visual behavior, and proxemics or the study of space. An edited volume by Siegman and Feldstein (1978) reviews the evidence on specialized topics by the key researchers in those fields, for example, Dittmann on the role of bodily communication, Ekman on facial expression, Exline on visual interaction, Hess on pupillary behavior, and Patterson on the role of space. Druckman et al. (1982) present a channel-by-channel review of studies organized in terms of the functions of information processing and impression management; the same authors provided a more recent review appearing in Hargie's (1986) Handbook of Communication Skills. Influential frameworks for conceptualizing the diverse functions served by nonverbal behavior are those devised by Ekman and Friesen (1969), Birdwhistle (1970), Dittmann (1972), Mehrabian (1972), and Argyle (1975). To keep abreast of current developments in the field, readers can consult recent issues of the Journal of Nonverbal Behavior.
2. Results obtained from a preliminary analysis are instructive. Subjects participated in a main interview and a postinterview session; in the later, they were told to “fool the interviewer by controlling their facial expressions.” In both segments, the same issue was discussed. Fifteen nonverbal behaviors (facial expressions and body movements) were coded for both the main interview and the postinterview sessions. Contrary to the leakage hypothesis, the difference scores indicated less movement on most indicators in the postinterview session. Following the experimenter's instructions, subjects spent considerably more time looking at the interviewer and less time looking elsewhere during the postinterview session. They also showed less activity on other indicators, the only exceptions being head movements and frowns, where they showed somewhat more movement during the postinterview session. Although this finding contradicts the idea of leakage, it does suggest a deception clue for deceivers who consciously control their facial expressions. The clue is less overall animation, which has also
been found to discriminate among deceivers, evaders, and truth tellers (Druckman et al., 1982:137) and to distinguish between deceivers sensitive to detector's possible suspicions about their deception and those not sensitive to suspicions in the Buller et al. (1989) study. It also replicates the DePaulo et al. (1985) finding of behavioral inhibition for motivated liars discussed above. More definitive results await the analyses of data which have been obtained from a larger sample of subjects, some of whom were told to “control their face, ” while others were told to “control their body.”
3. Another problem for detectors is posed by skillful deceivers. Skill at deceiving may be developed through on-the-job experience or by specific training programs. Studies by Ekman and Friesen (1974) and Alker (1976) suggest that experience in a profession (nursing, politics) may enhance a person's skill at deceiving. DePaulo et al. (1985) discuss training with regard to both effective and ineffective procedures. Appropriate training for would-be deceivers may be games where feedback is direct, unambiguous, and immediate. Inappropriate preparation may be repeated rehearsals of the future enactment. Careful planning may make a person appear less spontaneous and more tense, leading the observer to conclude that he or she is being deceptive. These suggestions apply more generally to the art of impression management, including those forms of impressions that are less morally dubious than deception.
4. The laboratory research addresses the question of what to look for and how to measure it. While contributing analytical strategies as discussed in this section, it would be facile to suggest that the findings from laboratory studies also contribute to the meaning of displays observed in other settings. The meaning of measured displays must be established in the setting of interest.
Alker, H.A. 1976 Mystification and Deception in Presidential Press Conferences. Paper presented at the meeting of the American Psychological Association , Washington, D.C.
Argyle, M. 1975 Bodily Communication. New York: International Universities Press.
Badler, N.I., and S.W. Smoliar 1979 Digital representations of human movement. Computing Surveys 11:19-38.
Baxter, J.C. 1970 Interpersonal spacing in natural settings. Sociometry 33:444-456.
Ben-Shakhar, G., I. Lieblich, and M. Bar-Hillel 1982 An evaluation of polygraphers' judgments: a review from a decision theoretic perspective. Journal of Applied Psychology 67:701-713.
Birdwhistle, R.L. 1970 Kinesics and Context. Philadelphia, Pa.: University of Pennsylvania Press.
Bond, C.F., Jr., A. Omar, A. Mahmoud, and R.N. Bonser 1990 Lie detection across cultures. Journal of Nonverbal Behavior 14:189-204.
Buller, D.B., and R.K. Aune 1987 Nonverbal cues to deception among intimates, friends, and strangers . Journal of Nonverbal Behavior 11:269-290.
Buller, D.B., J. Comstock, R.K. Aune, and K.D. Strzyzewski 1989 The effect of probing on deceivers and truthtellers. Journal of Nonverbal Behavior 13:155-169.
Cody, M.J., W. Lee, and E. Yi Chao 1989 Telling lies: correlates of deception among Chinese. In J.P. Forgas and J.M. Innes, eds., Recent Advances in Social Psychology: An International Perspective. New York: Elsevier Science Publishers.
DePaulo, B.M., M. Zuckerman, and R. Rosenthal 1980 Detecting deception: modality effects. In L. Wheeler, ed., Review of Personality and Social Psychology, Vol. 1. Beverly Hills, Calif.: Sage Publications.
DePaulo, B.M., R. Rosenthal, C.R. Green, and J. Rosenkrantz 1982 Diagnosing deceptive and mixed messages from verbal and nonverbal clues. Journal of Experimental Social Psychology 18:433-446.
DePaulo, B.M., J.I. Stone, and G.D. Lassiter 1985 Deceiving and detecting deceit. In B.R. Schlenker, ed., The Self and Social Life. New York: McGraw-Hill.
Dittmann, A.T. 1972 Interpersonal Messages of Emotion. New York: Springer.
Doob, L.W. 1975 Pathways to People. New Haven, Conn.: Yale University Press.
Druckman, D., R.M. Rozelle, and J.C. Baxter 1982 Nonverbal Communication: Survey, Theory, and Research. Beverly Hills, Calif.: Sage Publications.
Ekman, P. 1978 Facial expression. In A.W. Siegman and S. Feldstein, eds., Nonverbal Behavior and Communication. Hillsdale, N.J.: Erlbaum.
1981 Mistakes when deceiving. Annals of the New York Academy of Sciences 364:269-278.
Ekman, P., and W.V. Friesen 1969 Nonverbal leakage and clues to deception. Psychiatry 32:88-106.
1974 Detecting deception from the body or face. Journal of Personality and Social Psychology 29:288-298.
1986 A new pan-cultural facial expression of emotion. Motivation and Emotion 10:159-168.
Ekman, P., and K.G. Heider 1988 The universality of a contempt expression: a replication. Motivation and Emotion 12:303-308.
Ekman, P., and M. O'Sullivan 1988 The role of context on interpreting facial expression: comment on Russell and Fehr (1987). Journal of Experimental Psychology, General 117:86-88.
1991 Who can catch a liar? Unpublished manuscript, Human Interaction Laboratory, University of California, San Francisco.
Ekman, P., W.V. Friesen, and J. Bear 1984 The international language of gestures. Psychology Today May:64-69.
Ekman, P., W.V. Friesen, M. O'Sullivan, A. Chan, I. Diacoyanni-Tarlatzis, K. Heider, R. Krause, W. Ayhrans LeCompte, T. Pitcairn, P.E. Ricci-Bitti, K. Scherer, M. Tomita, and A. Tzavards 1987 Universals and cultural differences in the judgments of facial expressions of emotion. Journal of Personality and Social Psychology 53:712-717.
Flavell, J.H. 1977 Cognitive Development. Englewood Cliffs, N.J.: Prentice-Hall.
Getty, D.J., R.M. Pickett, C.J. D'Orsi, and J.A. Swets 1988 Enhanced interpretation of diagnostic images. Investigative Radiology 23:240-252.
Hall, E.T. 1966 The Hidden Dimension. Garden City, N.Y.: Doubleday.
Hargie, O. 1986 A Handbook of Communication Skills. Dover, N.H.: Croom Helm, Ltd.
Harper, R.G., A.N. Wiens, and J.D. Metarazzo 1978 Nonverbal Communication: The State of the Art. New York: John Wiley.
Hendershot, J., and A.K. Hess 1982 Detecting Deception: The Effects of Training and Socialization Levels on Verbal and Nonverbal Cue Utilization and Detection Accuracy. Unpublished manuscript, Department of Psychology, Auburn University.
Jourard, S.M. 1966 An exploratory study of body-accessibility. British Journal of Social and Clinical Psychology 5:221-231.
Knapp, M.L., and M.E. Comadena 1979 Telling it like it isn't: a review of theory and research on deceptive communications. Human Communication Research 5:270-285.
Kraut, R.E., and D. Poe 1980 Behavioral roots of person perception: the deception judgments of customs inspectors and laymen. Journal of Personality and Social Psychology 39:784-798.
Little, K.B. 1968 Cultural variations in social schemata. Journal of Personality and Social Psychology 7:63-72.
Lykken, D.T. 1981 A Tremor in the Blood: Uses and Abuses of the Lie Detector. New York: McGraw-Hill.
1985 The probity of the polygraph. In S.M. Kassin and L.S. Wrightsman, eds., The Psychology of Evidence and Trial Procedure. Beverly Hills, Calif.: Sage Publications.
Mahoney, R., and D. Druckman 1975 Simulation, experimentation, and context: dimensions of design and inference. Simulation and Games 6:235-270.
McClintock, C.C., and R.G. Hunt 1975 Nonverbal indicators of affect and deception. Journal of Applied Social Psychology 1:54-67.
Mehrabian, A. 1971 Nonverbal betrayal of feeling. Journal of Experimental Research in Personality 5:64-73.
1972 Nonverbal Communication. Chicago: Aldine.
Nisbett, R., and L. Ross 1980 Human Inference: Strategies and Shortcomings of Social Judgment. Englewood Cliffs, N.J.: Prentice-Hall.
O'Hair, D., M.J. Cody, X. Wang, and E. Yi Chao 1989 Vocal Stress and Deception Detection Among Chinese. Paper presented at the annual meetings of the Western Speech Communication Association, Spokane, Wash.
O'Sullivan, M., P. Ekman, and W.V. Friesen 1988 The effect of comparisons on detecting deceit. Journal of Nonverbal Behavior 12:203-215.
Raskin, D.C. 1987 Does science support polygraph testing? In A. Gale, ed., The Polygraph Test: Lies, Truth, and Science. Beverly Hills, Calif.: Sage Publications.
Rozelle, R.M., D. Druckman, and J.C. Baxter 1986 Nonverbal communication. In O. Hargie, ed., A Handbook of Communication Skills. Dover, N.H.: Croom Helm, Ltd.
Shuter, P. 1976 Proxemics and tactility in Latin America. Journal of Communication 26:46-52.
Siegman, A.W., and S. Feldstein, eds. 1978 Nonverbal Behavior and Communication. Hillsdale, N.J.: Lawrence Erlbaum.
Sommer, R. 1959 Studies in personal space. Sociometry 22:247-260.
Swets, J.A. 1988 Measuring the accuracy of diagnostic systems. Science 240:1285-1293.
Swets, J.A., D.J. Getty, R.M. Dickett, C.J. D'Orsi, S.E. Seltzer, and B.J. McNeil 1991 Enhancing and evaluating diagnostic accuracy. Medical Decision Making 11:9-18.
Watson, O.M. 1970 Proxemic Behavior: A Cross-Cultural Study. The Hague, Netherlands: Mouton.
Watson, O.M., and T.D. Graves 1966 Quantitative research in proxemic behavior. American Anthropologist 68:971-985.
Yi Chao, E. 1987 Correlates of Deceit: A Cross-Cultural Examination. Unpublished doctoral dissertation, Department of Communication, University of Southern California.
Zuckerman, M., D.T. Larrance, N.H. Spiegel, and R. Klorman 1981 Controlling nonverbal clues: facial expressions and tone of voice . Journal of Experimental Social Psychology 17:506-524.
Zuckerman, M., R. Koestner, and A.O. Alton 1984 Learning to detect deception. Journal of Personality and Social Psychology 46:519-528.
Zuckerman, M., R. Koestner, and M.J. Colella 1985 Learning to detect deception from three communication channels. Journal of Nonverbal Behavior 9:188-194.